From 3123432dcc2313899aff488e1fb5abcda6583c3a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Sun, 8 Sep 2013 10:45:10 +0200 Subject: [PATCH 1/7] cellPngDec Module Added Implemented 'cellPngDec', using 'sys_fs' to access the files and the 'lodepng' library to decode PNG buffers to Raw-pixel buffers that can be used as textures. --- lodepng/lodepng.cpp | 6267 +++++++++++++++++++++ lodepng/lodepng.h | 1705 ++++++ rpcs3.sln | 10 +- rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp | 233 + rpcs3/rpcs3.vcxproj | 1 + rpcs3/rpcs3.vcxproj.filters | 3 + 6 files changed, 8217 insertions(+), 2 deletions(-) create mode 100644 lodepng/lodepng.cpp create mode 100644 lodepng/lodepng.h create mode 100644 rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp diff --git a/lodepng/lodepng.cpp b/lodepng/lodepng.cpp new file mode 100644 index 0000000000..6fbd6acb4c --- /dev/null +++ b/lodepng/lodepng.cpp @@ -0,0 +1,6267 @@ +/* +LodePNG version 20130831 + +Copyright (c) 2005-2013 Lode Vandevenne + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software + in a product, an acknowledgment in the product documentation would be + appreciated but is not required. + + 2. Altered source versions must be plainly marked as such, and must not be + misrepresented as being the original software. + + 3. This notice may not be removed or altered from any source + distribution. +*/ + +/* +The manual and changelog are in the header file "lodepng.h" +Rename this file to lodepng.cpp to use it for C++, or to lodepng.c to use it for C. +*/ + +#include "lodepng.h" + +#include +#include + +#ifdef LODEPNG_COMPILE_CPP +#include +#endif /*LODEPNG_COMPILE_CPP*/ + +#define VERSION_STRING "20130831" + +/* +This source file is built up in the following large parts. The code sections +with the "LODEPNG_COMPILE_" #defines divide this up further in an intermixed way. +-Tools for C and common code for PNG and Zlib +-C Code for Zlib (huffman, deflate, ...) +-C Code for PNG (file format chunks, adam7, PNG filters, color conversions, ...) +-The C++ wrapper around all of the above +*/ + +/*The malloc, realloc and free functions defined here with "lodepng_" in front +of the name, so that you can easily change them to others related to your +platform if needed. Everything else in the code calls these. Pass +-DLODEPNG_NO_COMPILE_ALLOCATORS to the compiler, or comment out +#define LODEPNG_COMPILE_ALLOCATORS in the header, to disable the ones here and +define them in your own project's source files without needing to change +lodepng source code. Don't forget to remove "static" if you copypaste them +from here.*/ + +#ifdef LODEPNG_COMPILE_ALLOCATORS +static void* lodepng_malloc(size_t size) +{ + return malloc(size); +} + +static void* lodepng_realloc(void* ptr, size_t new_size) +{ + return realloc(ptr, new_size); +} + +static void lodepng_free(void* ptr) +{ + free(ptr); +} +#else /*LODEPNG_COMPILE_ALLOCATORS*/ +void* lodepng_malloc(size_t size); +void* lodepng_realloc(void* ptr, size_t new_size); +void lodepng_free(void* ptr); +#endif /*LODEPNG_COMPILE_ALLOCATORS*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // Tools for C, and common code for PNG and Zlib. // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/* +Often in case of an error a value is assigned to a variable and then it breaks +out of a loop (to go to the cleanup phase of a function). This macro does that. +It makes the error handling code shorter and more readable. + +Example: if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83); +*/ +#define CERROR_BREAK(errorvar, code)\ +{\ + errorvar = code;\ + break;\ +} + +/*version of CERROR_BREAK that assumes the common case where the error variable is named "error"*/ +#define ERROR_BREAK(code) CERROR_BREAK(error, code) + +/*Set error var to the error code, and return it.*/ +#define CERROR_RETURN_ERROR(errorvar, code)\ +{\ + errorvar = code;\ + return code;\ +} + +/*Try the code, if it returns error, also return the error.*/ +#define CERROR_TRY_RETURN(call)\ +{\ + unsigned error = call;\ + if(error) return error;\ +} + +/* +About uivector, ucvector and string: +-All of them wrap dynamic arrays or text strings in a similar way. +-LodePNG was originally written in C++. The vectors replace the std::vectors that were used in the C++ version. +-The string tools are made to avoid problems with compilers that declare things like strncat as deprecated. +-They're not used in the interface, only internally in this file as static functions. +-As with many other structs in this file, the init and cleanup functions serve as ctor and dtor. +*/ + +#ifdef LODEPNG_COMPILE_ZLIB +/*dynamic vector of unsigned ints*/ +typedef struct uivector +{ + unsigned* data; + size_t size; /*size in number of unsigned longs*/ + size_t allocsize; /*allocated size in bytes*/ +} uivector; + +static void uivector_cleanup(void* p) +{ + ((uivector*)p)->size = ((uivector*)p)->allocsize = 0; + lodepng_free(((uivector*)p)->data); + ((uivector*)p)->data = NULL; +} + +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned uivector_resize(uivector* p, size_t size) +{ + if(size * sizeof(unsigned) > p->allocsize) + { + size_t newsize = size * sizeof(unsigned) * 2; + void* data = lodepng_realloc(p->data, newsize); + if(data) + { + p->allocsize = newsize; + p->data = (unsigned*)data; + p->size = size; + } + else return 0; + } + else p->size = size; + return 1; +} + +/*resize and give all new elements the value*/ +static unsigned uivector_resizev(uivector* p, size_t size, unsigned value) +{ + size_t oldsize = p->size, i; + if(!uivector_resize(p, size)) return 0; + for(i = oldsize; i < size; i++) p->data[i] = value; + return 1; +} + +static void uivector_init(uivector* p) +{ + p->data = NULL; + p->size = p->allocsize = 0; +} + +#ifdef LODEPNG_COMPILE_ENCODER +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned uivector_push_back(uivector* p, unsigned c) +{ + if(!uivector_resize(p, p->size + 1)) return 0; + p->data[p->size - 1] = c; + return 1; +} + +/*copy q to p, returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned uivector_copy(uivector* p, const uivector* q) +{ + size_t i; + if(!uivector_resize(p, q->size)) return 0; + for(i = 0; i < q->size; i++) p->data[i] = q->data[i]; + return 1; +} + +static void uivector_swap(uivector* p, uivector* q) +{ + size_t tmp; + unsigned* tmpp; + tmp = p->size; p->size = q->size; q->size = tmp; + tmp = p->allocsize; p->allocsize = q->allocsize; q->allocsize = tmp; + tmpp = p->data; p->data = q->data; q->data = tmpp; +} +#endif /*LODEPNG_COMPILE_ENCODER*/ +#endif /*LODEPNG_COMPILE_ZLIB*/ + +/* /////////////////////////////////////////////////////////////////////////// */ + +/*dynamic vector of unsigned chars*/ +typedef struct ucvector +{ + unsigned char* data; + size_t size; /*used size*/ + size_t allocsize; /*allocated size*/ +} ucvector; + +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned ucvector_resize(ucvector* p, size_t size) +{ + if(size * sizeof(unsigned char) > p->allocsize) + { + size_t newsize = size * sizeof(unsigned char) * 2; + void* data = lodepng_realloc(p->data, newsize); + if(data) + { + p->allocsize = newsize; + p->data = (unsigned char*)data; + p->size = size; + } + else return 0; /*error: not enough memory*/ + } + else p->size = size; + return 1; +} + +#ifdef LODEPNG_COMPILE_PNG + +static void ucvector_cleanup(void* p) +{ + ((ucvector*)p)->size = ((ucvector*)p)->allocsize = 0; + lodepng_free(((ucvector*)p)->data); + ((ucvector*)p)->data = NULL; +} + +static void ucvector_init(ucvector* p) +{ + p->data = NULL; + p->size = p->allocsize = 0; +} + +#ifdef LODEPNG_COMPILE_DECODER +/*resize and give all new elements the value*/ +static unsigned ucvector_resizev(ucvector* p, size_t size, unsigned char value) +{ + size_t oldsize = p->size, i; + if(!ucvector_resize(p, size)) return 0; + for(i = oldsize; i < size; i++) p->data[i] = value; + return 1; +} +#endif /*LODEPNG_COMPILE_DECODER*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ZLIB +/*you can both convert from vector to buffer&size and vica versa. If you use +init_buffer to take over a buffer and size, it is not needed to use cleanup*/ +static void ucvector_init_buffer(ucvector* p, unsigned char* buffer, size_t size) +{ + p->data = buffer; + p->allocsize = p->size = size; +} +#endif /*LODEPNG_COMPILE_ZLIB*/ + +#if (defined(LODEPNG_COMPILE_PNG) && defined(LODEPNG_COMPILE_ANCILLARY_CHUNKS)) || defined(LODEPNG_COMPILE_ENCODER) +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned ucvector_push_back(ucvector* p, unsigned char c) +{ + if(!ucvector_resize(p, p->size + 1)) return 0; + p->data[p->size - 1] = c; + return 1; +} +#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ + + +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_PNG +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +/*returns 1 if success, 0 if failure ==> nothing done*/ +static unsigned string_resize(char** out, size_t size) +{ + char* data = (char*)lodepng_realloc(*out, size + 1); + if(data) + { + data[size] = 0; /*null termination char*/ + *out = data; + } + return data != 0; +} + +/*init a {char*, size_t} pair for use as string*/ +static void string_init(char** out) +{ + *out = NULL; + string_resize(out, 0); +} + +/*free the above pair again*/ +static void string_cleanup(char** out) +{ + lodepng_free(*out); + *out = NULL; +} + +static void string_set(char** out, const char* in) +{ + size_t insize = strlen(in), i = 0; + if(string_resize(out, insize)) + { + for(i = 0; i < insize; i++) + { + (*out)[i] = in[i]; + } + } +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +/* ////////////////////////////////////////////////////////////////////////// */ + +unsigned lodepng_read32bitInt(const unsigned char* buffer) +{ + return (buffer[0] << 24) | (buffer[1] << 16) | (buffer[2] << 8) | buffer[3]; +} + +#if defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER) +/*buffer must have at least 4 allocated bytes available*/ +static void lodepng_set32bitInt(unsigned char* buffer, unsigned value) +{ + buffer[0] = (unsigned char)((value >> 24) & 0xff); + buffer[1] = (unsigned char)((value >> 16) & 0xff); + buffer[2] = (unsigned char)((value >> 8) & 0xff); + buffer[3] = (unsigned char)((value ) & 0xff); +} +#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ + +#ifdef LODEPNG_COMPILE_ENCODER +static void lodepng_add32bitInt(ucvector* buffer, unsigned value) +{ + ucvector_resize(buffer, buffer->size + 4); /*todo: give error if resize failed*/ + lodepng_set32bitInt(&buffer->data[buffer->size - 4], value); +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / File IO / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_DISK + +unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename) +{ + FILE* file; + long size; + + /*provide some proper output values if error will happen*/ + *out = 0; + *outsize = 0; + + file = fopen(filename, "rb"); + if(!file) return 78; + + /*get filesize:*/ + fseek(file , 0 , SEEK_END); + size = ftell(file); + rewind(file); + + /*read contents of the file into the vector*/ + *outsize = 0; + *out = (unsigned char*)lodepng_malloc((size_t)size); + if(size && (*out)) (*outsize) = fread(*out, 1, (size_t)size, file); + + fclose(file); + if(!(*out) && size) return 83; /*the above malloc failed*/ + return 0; +} + +/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ +unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename) +{ + FILE* file; + file = fopen(filename, "wb" ); + if(!file) return 79; + fwrite((char*)buffer , 1 , buffersize, file); + fclose(file); + return 0; +} + +#endif /*LODEPNG_COMPILE_DISK*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // End of common code and tools. Begin of Zlib related code. // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_ZLIB +#ifdef LODEPNG_COMPILE_ENCODER +/*TODO: this ignores potential out of memory errors*/ +static void addBitToStream(size_t* bitpointer, ucvector* bitstream, unsigned char bit) +{ + /*add a new byte at the end*/ + if((*bitpointer) % 8 == 0) ucvector_push_back(bitstream, (unsigned char)0); + /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/ + (bitstream->data[bitstream->size - 1]) |= (bit << ((*bitpointer) & 0x7)); + (*bitpointer)++; +} + +static void addBitsToStream(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) +{ + size_t i; + for(i = 0; i < nbits; i++) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> i) & 1)); +} + +static void addBitsToStreamReversed(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) +{ + size_t i; + for(i = 0; i < nbits; i++) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> (nbits - 1 - i)) & 1)); +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_DECODER + +#define READBIT(bitpointer, bitstream) ((bitstream[bitpointer >> 3] >> (bitpointer & 0x7)) & (unsigned char)1) + +static unsigned char readBitFromStream(size_t* bitpointer, const unsigned char* bitstream) +{ + unsigned char result = (unsigned char)(READBIT(*bitpointer, bitstream)); + (*bitpointer)++; + return result; +} + +static unsigned readBitsFromStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) +{ + unsigned result = 0, i; + for(i = 0; i < nbits; i++) + { + result += ((unsigned)READBIT(*bitpointer, bitstream)) << i; + (*bitpointer)++; + } + return result; +} +#endif /*LODEPNG_COMPILE_DECODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Deflate - Huffman / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#define FIRST_LENGTH_CODE_INDEX 257 +#define LAST_LENGTH_CODE_INDEX 285 +/*256 literals, the end code, some length codes, and 2 unused codes*/ +#define NUM_DEFLATE_CODE_SYMBOLS 288 +/*the distance codes have their own symbols, 30 used, 2 unused*/ +#define NUM_DISTANCE_SYMBOLS 32 +/*the code length codes. 0-15: code lengths, 16: copy previous 3-6 times, 17: 3-10 zeros, 18: 11-138 zeros*/ +#define NUM_CODE_LENGTH_CODES 19 + +/*the base lengths represented by codes 257-285*/ +static const unsigned LENGTHBASE[29] + = {3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51, 59, + 67, 83, 99, 115, 131, 163, 195, 227, 258}; + +/*the extra bits used by codes 257-285 (added to base length)*/ +static const unsigned LENGTHEXTRA[29] + = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, + 4, 4, 4, 4, 5, 5, 5, 5, 0}; + +/*the base backwards distances (the bits of distance codes appear after length codes and use their own huffman tree)*/ +static const unsigned DISTANCEBASE[30] + = {1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513, + 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577}; + +/*the extra bits of backwards distances (added to base)*/ +static const unsigned DISTANCEEXTRA[30] + = {0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, + 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13}; + +/*the order in which "code length alphabet code lengths" are stored, out of this +the huffman tree of the dynamic huffman tree lengths is generated*/ +static const unsigned CLCL_ORDER[NUM_CODE_LENGTH_CODES] + = {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; + +/* ////////////////////////////////////////////////////////////////////////// */ + +/* +Huffman tree struct, containing multiple representations of the tree +*/ +typedef struct HuffmanTree +{ + unsigned* tree2d; + unsigned* tree1d; + unsigned* lengths; /*the lengths of the codes of the 1d-tree*/ + unsigned maxbitlen; /*maximum number of bits a single code can get*/ + unsigned numcodes; /*number of symbols in the alphabet = number of codes*/ +} HuffmanTree; + +/*function used for debug purposes to draw the tree in ascii art with C++*/ +/* +static void HuffmanTree_draw(HuffmanTree* tree) +{ + std::cout << "tree. length: " << tree->numcodes << " maxbitlen: " << tree->maxbitlen << std::endl; + for(size_t i = 0; i < tree->tree1d.size; i++) + { + if(tree->lengths.data[i]) + std::cout << i << " " << tree->tree1d.data[i] << " " << tree->lengths.data[i] << std::endl; + } + std::cout << std::endl; +}*/ + +static void HuffmanTree_init(HuffmanTree* tree) +{ + tree->tree2d = 0; + tree->tree1d = 0; + tree->lengths = 0; +} + +static void HuffmanTree_cleanup(HuffmanTree* tree) +{ + lodepng_free(tree->tree2d); + lodepng_free(tree->tree1d); + lodepng_free(tree->lengths); +} + +/*the tree representation used by the decoder. return value is error*/ +static unsigned HuffmanTree_make2DTree(HuffmanTree* tree) +{ + unsigned nodefilled = 0; /*up to which node it is filled*/ + unsigned treepos = 0; /*position in the tree (1 of the numcodes columns)*/ + unsigned n, i; + + tree->tree2d = (unsigned*)lodepng_malloc(tree->numcodes * 2 * sizeof(unsigned)); + if(!tree->tree2d) return 83; /*alloc fail*/ + + /* + convert tree1d[] to tree2d[][]. In the 2D array, a value of 32767 means + uninited, a value >= numcodes is an address to another bit, a value < numcodes + is a code. The 2 rows are the 2 possible bit values (0 or 1), there are as + many columns as codes - 1. + A good huffmann tree has N * 2 - 1 nodes, of which N - 1 are internal nodes. + Here, the internal nodes are stored (what their 0 and 1 option point to). + There is only memory for such good tree currently, if there are more nodes + (due to too long length codes), error 55 will happen + */ + for(n = 0; n < tree->numcodes * 2; n++) + { + tree->tree2d[n] = 32767; /*32767 here means the tree2d isn't filled there yet*/ + } + + for(n = 0; n < tree->numcodes; n++) /*the codes*/ + { + for(i = 0; i < tree->lengths[n]; i++) /*the bits for this code*/ + { + unsigned char bit = (unsigned char)((tree->tree1d[n] >> (tree->lengths[n] - i - 1)) & 1); + if(treepos > tree->numcodes - 2) return 55; /*oversubscribed, see comment in lodepng_error_text*/ + if(tree->tree2d[2 * treepos + bit] == 32767) /*not yet filled in*/ + { + if(i + 1 == tree->lengths[n]) /*last bit*/ + { + tree->tree2d[2 * treepos + bit] = n; /*put the current code in it*/ + treepos = 0; + } + else + { + /*put address of the next step in here, first that address has to be found of course + (it's just nodefilled + 1)...*/ + nodefilled++; + /*addresses encoded with numcodes added to it*/ + tree->tree2d[2 * treepos + bit] = nodefilled + tree->numcodes; + treepos = nodefilled; + } + } + else treepos = tree->tree2d[2 * treepos + bit] - tree->numcodes; + } + } + + for(n = 0; n < tree->numcodes * 2; n++) + { + if(tree->tree2d[n] == 32767) tree->tree2d[n] = 0; /*remove possible remaining 32767's*/ + } + + return 0; +} + +/* +Second step for the ...makeFromLengths and ...makeFromFrequencies functions. +numcodes, lengths and maxbitlen must already be filled in correctly. return +value is error. +*/ +static unsigned HuffmanTree_makeFromLengths2(HuffmanTree* tree) +{ + uivector blcount; + uivector nextcode; + unsigned bits, n, error = 0; + + uivector_init(&blcount); + uivector_init(&nextcode); + + tree->tree1d = (unsigned*)lodepng_malloc(tree->numcodes * sizeof(unsigned)); + if(!tree->tree1d) error = 83; /*alloc fail*/ + + if(!uivector_resizev(&blcount, tree->maxbitlen + 1, 0) + || !uivector_resizev(&nextcode, tree->maxbitlen + 1, 0)) + error = 83; /*alloc fail*/ + + if(!error) + { + /*step 1: count number of instances of each code length*/ + for(bits = 0; bits < tree->numcodes; bits++) blcount.data[tree->lengths[bits]]++; + /*step 2: generate the nextcode values*/ + for(bits = 1; bits <= tree->maxbitlen; bits++) + { + nextcode.data[bits] = (nextcode.data[bits - 1] + blcount.data[bits - 1]) << 1; + } + /*step 3: generate all the codes*/ + for(n = 0; n < tree->numcodes; n++) + { + if(tree->lengths[n] != 0) tree->tree1d[n] = nextcode.data[tree->lengths[n]]++; + } + } + + uivector_cleanup(&blcount); + uivector_cleanup(&nextcode); + + if(!error) return HuffmanTree_make2DTree(tree); + else return error; +} + +/* +given the code lengths (as stored in the PNG file), generate the tree as defined +by Deflate. maxbitlen is the maximum bits that a code in the tree can have. +return value is error. +*/ +static unsigned HuffmanTree_makeFromLengths(HuffmanTree* tree, const unsigned* bitlen, + size_t numcodes, unsigned maxbitlen) +{ + unsigned i; + tree->lengths = (unsigned*)lodepng_malloc(numcodes * sizeof(unsigned)); + if(!tree->lengths) return 83; /*alloc fail*/ + for(i = 0; i < numcodes; i++) tree->lengths[i] = bitlen[i]; + tree->numcodes = (unsigned)numcodes; /*number of symbols*/ + tree->maxbitlen = maxbitlen; + return HuffmanTree_makeFromLengths2(tree); +} + +#ifdef LODEPNG_COMPILE_ENCODER + +/* +A coin, this is the terminology used for the package-merge algorithm and the +coin collector's problem. This is used to generate the huffman tree. +A coin can be multiple coins (when they're merged) +*/ +typedef struct Coin +{ + uivector symbols; + float weight; /*the sum of all weights in this coin*/ +} Coin; + +static void coin_init(Coin* c) +{ + uivector_init(&c->symbols); +} + +/*argument c is void* so that this dtor can be given as function pointer to the vector resize function*/ +static void coin_cleanup(void* c) +{ + uivector_cleanup(&((Coin*)c)->symbols); +} + +static void coin_copy(Coin* c1, const Coin* c2) +{ + c1->weight = c2->weight; + uivector_copy(&c1->symbols, &c2->symbols); +} + +static void add_coins(Coin* c1, const Coin* c2) +{ + size_t i; + for(i = 0; i < c2->symbols.size; i++) uivector_push_back(&c1->symbols, c2->symbols.data[i]); + c1->weight += c2->weight; +} + +static void init_coins(Coin* coins, size_t num) +{ + size_t i; + for(i = 0; i < num; i++) coin_init(&coins[i]); +} + +static void cleanup_coins(Coin* coins, size_t num) +{ + size_t i; + for(i = 0; i < num; i++) coin_cleanup(&coins[i]); +} + +/* +This uses a simple combsort to sort the data. This function is not critical for +overall encoding speed and the data amount isn't that large. +*/ +static void sort_coins(Coin* data, size_t amount) +{ + size_t gap = amount; + unsigned char swapped = 0; + while((gap > 1) || swapped) + { + size_t i; + gap = (gap * 10) / 13; /*shrink factor 1.3*/ + if(gap == 9 || gap == 10) gap = 11; /*combsort11*/ + if(gap < 1) gap = 1; + swapped = 0; + for(i = 0; i < amount - gap; i++) + { + size_t j = i + gap; + if(data[j].weight < data[i].weight) + { + float temp = data[j].weight; data[j].weight = data[i].weight; data[i].weight = temp; + uivector_swap(&data[i].symbols, &data[j].symbols); + swapped = 1; + } + } + } +} + +static unsigned append_symbol_coins(Coin* coins, const unsigned* frequencies, unsigned numcodes, size_t sum) +{ + unsigned i; + unsigned j = 0; /*index of present symbols*/ + for(i = 0; i < numcodes; i++) + { + if(frequencies[i] != 0) /*only include symbols that are present*/ + { + coins[j].weight = frequencies[i] / (float)sum; + uivector_push_back(&coins[j].symbols, i); + j++; + } + } + return 0; +} + +unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, + size_t numcodes, unsigned maxbitlen) +{ + unsigned i, j; + size_t sum = 0, numpresent = 0; + unsigned error = 0; + Coin* coins; /*the coins of the currently calculated row*/ + Coin* prev_row; /*the previous row of coins*/ + unsigned numcoins; + unsigned coinmem; + + if(numcodes == 0) return 80; /*error: a tree of 0 symbols is not supposed to be made*/ + + for(i = 0; i < numcodes; i++) + { + if(frequencies[i] > 0) + { + numpresent++; + sum += frequencies[i]; + } + } + + for(i = 0; i < numcodes; i++) lengths[i] = 0; + + /*ensure at least two present symbols. There should be at least one symbol + according to RFC 1951 section 3.2.7. To decoders incorrectly require two. To + make these work as well ensure there are at least two symbols. The + Package-Merge code below also doesn't work correctly if there's only one + symbol, it'd give it the theoritical 0 bits but in practice zlib wants 1 bit*/ + if(numpresent == 0) + { + lengths[0] = lengths[1] = 1; /*note that for RFC 1951 section 3.2.7, only lengths[0] = 1 is needed*/ + } + else if(numpresent == 1) + { + for(i = 0; i < numcodes; i++) + { + if(frequencies[i]) + { + lengths[i] = 1; + lengths[i == 0 ? 1 : 0] = 1; + break; + } + } + } + else + { + /*Package-Merge algorithm represented by coin collector's problem + For every symbol, maxbitlen coins will be created*/ + + coinmem = numpresent * 2; /*max amount of coins needed with the current algo*/ + coins = (Coin*)lodepng_malloc(sizeof(Coin) * coinmem); + prev_row = (Coin*)lodepng_malloc(sizeof(Coin) * coinmem); + if(!coins || !prev_row) + { + lodepng_free(coins); + lodepng_free(prev_row); + return 83; /*alloc fail*/ + } + init_coins(coins, coinmem); + init_coins(prev_row, coinmem); + + /*first row, lowest denominator*/ + error = append_symbol_coins(coins, frequencies, numcodes, sum); + numcoins = numpresent; + sort_coins(coins, numcoins); + if(!error) + { + unsigned numprev = 0; + for(j = 1; j <= maxbitlen && !error; j++) /*each of the remaining rows*/ + { + unsigned tempnum; + Coin* tempcoins; + /*swap prev_row and coins, and their amounts*/ + tempcoins = prev_row; prev_row = coins; coins = tempcoins; + tempnum = numprev; numprev = numcoins; numcoins = tempnum; + + cleanup_coins(coins, numcoins); + init_coins(coins, numcoins); + + numcoins = 0; + + /*fill in the merged coins of the previous row*/ + for(i = 0; i + 1 < numprev; i += 2) + { + /*merge prev_row[i] and prev_row[i + 1] into new coin*/ + Coin* coin = &coins[numcoins++]; + coin_copy(coin, &prev_row[i]); + add_coins(coin, &prev_row[i + 1]); + } + /*fill in all the original symbols again*/ + if(j < maxbitlen) + { + error = append_symbol_coins(coins + numcoins, frequencies, numcodes, sum); + numcoins += numpresent; + } + sort_coins(coins, numcoins); + } + } + + if(!error) + { + /*calculate the lenghts of each symbol, as the amount of times a coin of each symbol is used*/ + for(i = 0; i < numpresent - 1; i++) + { + Coin* coin = &coins[i]; + for(j = 0; j < coin->symbols.size; j++) lengths[coin->symbols.data[j]]++; + } + } + + cleanup_coins(coins, coinmem); + lodepng_free(coins); + cleanup_coins(prev_row, coinmem); + lodepng_free(prev_row); + } + + return error; +} + +/*Create the Huffman tree given the symbol frequencies*/ +static unsigned HuffmanTree_makeFromFrequencies(HuffmanTree* tree, const unsigned* frequencies, + size_t mincodes, size_t numcodes, unsigned maxbitlen) +{ + unsigned error = 0; + while(!frequencies[numcodes - 1] && numcodes > mincodes) numcodes--; /*trim zeroes*/ + tree->maxbitlen = maxbitlen; + tree->numcodes = (unsigned)numcodes; /*number of symbols*/ + tree->lengths = (unsigned*)lodepng_realloc(tree->lengths, numcodes * sizeof(unsigned)); + if(!tree->lengths) return 83; /*alloc fail*/ + /*initialize all lengths to 0*/ + memset(tree->lengths, 0, numcodes * sizeof(unsigned)); + + error = lodepng_huffman_code_lengths(tree->lengths, frequencies, numcodes, maxbitlen); + if(!error) error = HuffmanTree_makeFromLengths2(tree); + return error; +} + +static unsigned HuffmanTree_getCode(const HuffmanTree* tree, unsigned index) +{ + return tree->tree1d[index]; +} + +static unsigned HuffmanTree_getLength(const HuffmanTree* tree, unsigned index) +{ + return tree->lengths[index]; +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/*get the literal and length code tree of a deflated block with fixed tree, as per the deflate specification*/ +static unsigned generateFixedLitLenTree(HuffmanTree* tree) +{ + unsigned i, error = 0; + unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); + if(!bitlen) return 83; /*alloc fail*/ + + /*288 possible codes: 0-255=literals, 256=endcode, 257-285=lengthcodes, 286-287=unused*/ + for(i = 0; i <= 143; i++) bitlen[i] = 8; + for(i = 144; i <= 255; i++) bitlen[i] = 9; + for(i = 256; i <= 279; i++) bitlen[i] = 7; + for(i = 280; i <= 287; i++) bitlen[i] = 8; + + error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DEFLATE_CODE_SYMBOLS, 15); + + lodepng_free(bitlen); + return error; +} + +/*get the distance code tree of a deflated block with fixed tree, as specified in the deflate specification*/ +static unsigned generateFixedDistanceTree(HuffmanTree* tree) +{ + unsigned i, error = 0; + unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); + if(!bitlen) return 83; /*alloc fail*/ + + /*there are 32 distance codes, but 30-31 are unused*/ + for(i = 0; i < NUM_DISTANCE_SYMBOLS; i++) bitlen[i] = 5; + error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DISTANCE_SYMBOLS, 15); + + lodepng_free(bitlen); + return error; +} + +#ifdef LODEPNG_COMPILE_DECODER + +/* +returns the code, or (unsigned)(-1) if error happened +inbitlength is the length of the complete buffer, in bits (so its byte length times 8) +*/ +static unsigned huffmanDecodeSymbol(const unsigned char* in, size_t* bp, + const HuffmanTree* codetree, size_t inbitlength) +{ + unsigned treepos = 0, ct; + for(;;) + { + if(*bp >= inbitlength) return (unsigned)(-1); /*error: end of input memory reached without endcode*/ + /* + decode the symbol from the tree. The "readBitFromStream" code is inlined in + the expression below because this is the biggest bottleneck while decoding + */ + ct = codetree->tree2d[(treepos << 1) + READBIT(*bp, in)]; + (*bp)++; + if(ct < codetree->numcodes) return ct; /*the symbol is decoded, return it*/ + else treepos = ct - codetree->numcodes; /*symbol not yet decoded, instead move tree position*/ + + if(treepos >= codetree->numcodes) return (unsigned)(-1); /*error: it appeared outside the codetree*/ + } +} +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_DECODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Inflator (Decompressor) / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*get the tree of a deflated block with fixed tree, as specified in the deflate specification*/ +static void getTreeInflateFixed(HuffmanTree* tree_ll, HuffmanTree* tree_d) +{ + /*TODO: check for out of memory errors*/ + generateFixedLitLenTree(tree_ll); + generateFixedDistanceTree(tree_d); +} + +/*get the tree of a deflated block with dynamic tree, the tree itself is also Huffman compressed with a known tree*/ +static unsigned getTreeInflateDynamic(HuffmanTree* tree_ll, HuffmanTree* tree_d, + const unsigned char* in, size_t* bp, size_t inlength) +{ + /*make sure that length values that aren't filled in will be 0, or a wrong tree will be generated*/ + unsigned error = 0; + unsigned n, HLIT, HDIST, HCLEN, i; + size_t inbitlength = inlength * 8; + + /*see comments in deflateDynamic for explanation of the context and these variables, it is analogous*/ + unsigned* bitlen_ll = 0; /*lit,len code lengths*/ + unsigned* bitlen_d = 0; /*dist code lengths*/ + /*code length code lengths ("clcl"), the bit lengths of the huffman tree used to compress bitlen_ll and bitlen_d*/ + unsigned* bitlen_cl = 0; + HuffmanTree tree_cl; /*the code tree for code length codes (the huffman tree for compressed huffman trees)*/ + + if((*bp) >> 3 >= inlength - 2) return 49; /*error: the bit pointer is or will go past the memory*/ + + /*number of literal/length codes + 257. Unlike the spec, the value 257 is added to it here already*/ + HLIT = readBitsFromStream(bp, in, 5) + 257; + /*number of distance codes. Unlike the spec, the value 1 is added to it here already*/ + HDIST = readBitsFromStream(bp, in, 5) + 1; + /*number of code length codes. Unlike the spec, the value 4 is added to it here already*/ + HCLEN = readBitsFromStream(bp, in, 4) + 4; + + HuffmanTree_init(&tree_cl); + + while(!error) + { + /*read the code length codes out of 3 * (amount of code length codes) bits*/ + + bitlen_cl = (unsigned*)lodepng_malloc(NUM_CODE_LENGTH_CODES * sizeof(unsigned)); + if(!bitlen_cl) ERROR_BREAK(83 /*alloc fail*/); + + for(i = 0; i < NUM_CODE_LENGTH_CODES; i++) + { + if(i < HCLEN) bitlen_cl[CLCL_ORDER[i]] = readBitsFromStream(bp, in, 3); + else bitlen_cl[CLCL_ORDER[i]] = 0; /*if not, it must stay 0*/ + } + + error = HuffmanTree_makeFromLengths(&tree_cl, bitlen_cl, NUM_CODE_LENGTH_CODES, 7); + if(error) break; + + /*now we can use this tree to read the lengths for the tree that this function will return*/ + bitlen_ll = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); + bitlen_d = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); + if(!bitlen_ll || !bitlen_d) ERROR_BREAK(83 /*alloc fail*/); + for(i = 0; i < NUM_DEFLATE_CODE_SYMBOLS; i++) bitlen_ll[i] = 0; + for(i = 0; i < NUM_DISTANCE_SYMBOLS; i++) bitlen_d[i] = 0; + + /*i is the current symbol we're reading in the part that contains the code lengths of lit/len and dist codes*/ + i = 0; + while(i < HLIT + HDIST) + { + unsigned code = huffmanDecodeSymbol(in, bp, &tree_cl, inbitlength); + if(code <= 15) /*a length code*/ + { + if(i < HLIT) bitlen_ll[i] = code; + else bitlen_d[i - HLIT] = code; + i++; + } + else if(code == 16) /*repeat previous*/ + { + unsigned replength = 3; /*read in the 2 bits that indicate repeat length (3-6)*/ + unsigned value; /*set value to the previous code*/ + + if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ + if (i == 0) ERROR_BREAK(54); /*can't repeat previous if i is 0*/ + + replength += readBitsFromStream(bp, in, 2); + + if(i < HLIT + 1) value = bitlen_ll[i - 1]; + else value = bitlen_d[i - HLIT - 1]; + /*repeat this value in the next lengths*/ + for(n = 0; n < replength; n++) + { + if(i >= HLIT + HDIST) ERROR_BREAK(13); /*error: i is larger than the amount of codes*/ + if(i < HLIT) bitlen_ll[i] = value; + else bitlen_d[i - HLIT] = value; + i++; + } + } + else if(code == 17) /*repeat "0" 3-10 times*/ + { + unsigned replength = 3; /*read in the bits that indicate repeat length*/ + if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ + + replength += readBitsFromStream(bp, in, 3); + + /*repeat this value in the next lengths*/ + for(n = 0; n < replength; n++) + { + if(i >= HLIT + HDIST) ERROR_BREAK(14); /*error: i is larger than the amount of codes*/ + + if(i < HLIT) bitlen_ll[i] = 0; + else bitlen_d[i - HLIT] = 0; + i++; + } + } + else if(code == 18) /*repeat "0" 11-138 times*/ + { + unsigned replength = 11; /*read in the bits that indicate repeat length*/ + if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ + + replength += readBitsFromStream(bp, in, 7); + + /*repeat this value in the next lengths*/ + for(n = 0; n < replength; n++) + { + if(i >= HLIT + HDIST) ERROR_BREAK(15); /*error: i is larger than the amount of codes*/ + + if(i < HLIT) bitlen_ll[i] = 0; + else bitlen_d[i - HLIT] = 0; + i++; + } + } + else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ + { + if(code == (unsigned)(-1)) + { + /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol + (10=no endcode, 11=wrong jump outside of tree)*/ + error = (*bp) > inbitlength ? 10 : 11; + } + else error = 16; /*unexisting code, this can never happen*/ + break; + } + } + if(error) break; + + if(bitlen_ll[256] == 0) ERROR_BREAK(64); /*the length of the end code 256 must be larger than 0*/ + + /*now we've finally got HLIT and HDIST, so generate the code trees, and the function is done*/ + error = HuffmanTree_makeFromLengths(tree_ll, bitlen_ll, NUM_DEFLATE_CODE_SYMBOLS, 15); + if(error) break; + error = HuffmanTree_makeFromLengths(tree_d, bitlen_d, NUM_DISTANCE_SYMBOLS, 15); + + break; /*end of error-while*/ + } + + lodepng_free(bitlen_cl); + lodepng_free(bitlen_ll); + lodepng_free(bitlen_d); + HuffmanTree_cleanup(&tree_cl); + + return error; +} + +/*inflate a block with dynamic of fixed Huffman tree*/ +static unsigned inflateHuffmanBlock(ucvector* out, const unsigned char* in, size_t* bp, + size_t* pos, size_t inlength, unsigned btype) +{ + unsigned error = 0; + HuffmanTree tree_ll; /*the huffman tree for literal and length codes*/ + HuffmanTree tree_d; /*the huffman tree for distance codes*/ + size_t inbitlength = inlength * 8; + + HuffmanTree_init(&tree_ll); + HuffmanTree_init(&tree_d); + + if(btype == 1) getTreeInflateFixed(&tree_ll, &tree_d); + else if(btype == 2) error = getTreeInflateDynamic(&tree_ll, &tree_d, in, bp, inlength); + + while(!error) /*decode all symbols until end reached, breaks at end code*/ + { + /*code_ll is literal, length or end code*/ + unsigned code_ll = huffmanDecodeSymbol(in, bp, &tree_ll, inbitlength); + if(code_ll <= 255) /*literal symbol*/ + { + if((*pos) >= out->size) + { + /*reserve more room at once*/ + if(!ucvector_resize(out, ((*pos) + 1) * 2)) ERROR_BREAK(83 /*alloc fail*/); + } + out->data[(*pos)] = (unsigned char)(code_ll); + (*pos)++; + } + else if(code_ll >= FIRST_LENGTH_CODE_INDEX && code_ll <= LAST_LENGTH_CODE_INDEX) /*length code*/ + { + unsigned code_d, distance; + unsigned numextrabits_l, numextrabits_d; /*extra bits for length and distance*/ + size_t start, forward, backward, length; + + /*part 1: get length base*/ + length = LENGTHBASE[code_ll - FIRST_LENGTH_CODE_INDEX]; + + /*part 2: get extra bits and add the value of that to length*/ + numextrabits_l = LENGTHEXTRA[code_ll - FIRST_LENGTH_CODE_INDEX]; + if(*bp >= inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ + length += readBitsFromStream(bp, in, numextrabits_l); + + /*part 3: get distance code*/ + code_d = huffmanDecodeSymbol(in, bp, &tree_d, inbitlength); + if(code_d > 29) + { + if(code_ll == (unsigned)(-1)) /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ + { + /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol + (10=no endcode, 11=wrong jump outside of tree)*/ + error = (*bp) > inlength * 8 ? 10 : 11; + } + else error = 18; /*error: invalid distance code (30-31 are never used)*/ + break; + } + distance = DISTANCEBASE[code_d]; + + /*part 4: get extra bits from distance*/ + numextrabits_d = DISTANCEEXTRA[code_d]; + if(*bp >= inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ + + distance += readBitsFromStream(bp, in, numextrabits_d); + + /*part 5: fill in all the out[n] values based on the length and dist*/ + start = (*pos); + if(distance > start) ERROR_BREAK(52); /*too long backward distance*/ + backward = start - distance; + if((*pos) + length >= out->size) + { + /*reserve more room at once*/ + if(!ucvector_resize(out, ((*pos) + length) * 2)) ERROR_BREAK(83 /*alloc fail*/); + } + + for(forward = 0; forward < length; forward++) + { + out->data[(*pos)] = out->data[backward]; + (*pos)++; + backward++; + if(backward >= start) backward = start - distance; + } + } + else if(code_ll == 256) + { + break; /*end code, break the loop*/ + } + else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ + { + /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol + (10=no endcode, 11=wrong jump outside of tree)*/ + error = (*bp) > inlength * 8 ? 10 : 11; + break; + } + } + + HuffmanTree_cleanup(&tree_ll); + HuffmanTree_cleanup(&tree_d); + + return error; +} + +static unsigned inflateNoCompression(ucvector* out, const unsigned char* in, size_t* bp, size_t* pos, size_t inlength) +{ + /*go to first boundary of byte*/ + size_t p; + unsigned LEN, NLEN, n, error = 0; + while(((*bp) & 0x7) != 0) (*bp)++; + p = (*bp) / 8; /*byte position*/ + + /*read LEN (2 bytes) and NLEN (2 bytes)*/ + if(p >= inlength - 4) return 52; /*error, bit pointer will jump past memory*/ + LEN = in[p] + 256 * in[p + 1]; p += 2; + NLEN = in[p] + 256 * in[p + 1]; p += 2; + + /*check if 16-bit NLEN is really the one's complement of LEN*/ + if(LEN + NLEN != 65535) return 21; /*error: NLEN is not one's complement of LEN*/ + + if((*pos) + LEN >= out->size) + { + if(!ucvector_resize(out, (*pos) + LEN)) return 83; /*alloc fail*/ + } + + /*read the literal data: LEN bytes are now stored in the out buffer*/ + if(p + LEN > inlength) return 23; /*error: reading outside of in buffer*/ + for(n = 0; n < LEN; n++) out->data[(*pos)++] = in[p++]; + + (*bp) = p * 8; + + return error; +} + +static unsigned lodepng_inflatev(ucvector* out, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings) +{ + /*bit pointer in the "in" data, current byte is bp >> 3, current bit is bp & 0x7 (from lsb to msb of the byte)*/ + size_t bp = 0; + unsigned BFINAL = 0; + size_t pos = 0; /*byte position in the out buffer*/ + + unsigned error = 0; + + (void)settings; + + while(!BFINAL) + { + unsigned BTYPE; + if(bp + 2 >= insize * 8) return 52; /*error, bit pointer will jump past memory*/ + BFINAL = readBitFromStream(&bp, in); + BTYPE = 1 * readBitFromStream(&bp, in); + BTYPE += 2 * readBitFromStream(&bp, in); + + if(BTYPE == 3) return 20; /*error: invalid BTYPE*/ + else if(BTYPE == 0) error = inflateNoCompression(out, in, &bp, &pos, insize); /*no compression*/ + else error = inflateHuffmanBlock(out, in, &bp, &pos, insize, BTYPE); /*compression, BTYPE 01 or 10*/ + + if(error) return error; + } + + /*Only now we know the true size of out, resize it to that*/ + if(!ucvector_resize(out, pos)) error = 83; /*alloc fail*/ + + return error; +} + +unsigned lodepng_inflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings) +{ + unsigned error; + ucvector v; + ucvector_init_buffer(&v, *out, *outsize); + error = lodepng_inflatev(&v, in, insize, settings); + *out = v.data; + *outsize = v.size; + return error; +} + +static unsigned inflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings) +{ + if(settings->custom_inflate) + { + return settings->custom_inflate(out, outsize, in, insize, settings); + } + else + { + return lodepng_inflate(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Deflator (Compressor) / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +static const size_t MAX_SUPPORTED_DEFLATE_LENGTH = 258; + +/*bitlen is the size in bits of the code*/ +static void addHuffmanSymbol(size_t* bp, ucvector* compressed, unsigned code, unsigned bitlen) +{ + addBitsToStreamReversed(bp, compressed, code, bitlen); +} + +/*search the index in the array, that has the largest value smaller than or equal to the given value, +given array must be sorted (if no value is smaller, it returns the size of the given array)*/ +static size_t searchCodeIndex(const unsigned* array, size_t array_size, size_t value) +{ + /*linear search implementation*/ + /*for(size_t i = 1; i < array_size; i++) if(array[i] > value) return i - 1; + return array_size - 1;*/ + + /*binary search implementation (not that much faster) (precondition: array_size > 0)*/ + size_t left = 1; + size_t right = array_size - 1; + while(left <= right) + { + size_t mid = (left + right) / 2; + if(array[mid] <= value) left = mid + 1; /*the value to find is more to the right*/ + else if(array[mid - 1] > value) right = mid - 1; /*the value to find is more to the left*/ + else return mid - 1; + } + return array_size - 1; +} + +static void addLengthDistance(uivector* values, size_t length, size_t distance) +{ + /*values in encoded vector are those used by deflate: + 0-255: literal bytes + 256: end + 257-285: length/distance pair (length code, followed by extra length bits, distance code, extra distance bits) + 286-287: invalid*/ + + unsigned length_code = (unsigned)searchCodeIndex(LENGTHBASE, 29, length); + unsigned extra_length = (unsigned)(length - LENGTHBASE[length_code]); + unsigned dist_code = (unsigned)searchCodeIndex(DISTANCEBASE, 30, distance); + unsigned extra_distance = (unsigned)(distance - DISTANCEBASE[dist_code]); + + uivector_push_back(values, length_code + FIRST_LENGTH_CODE_INDEX); + uivector_push_back(values, extra_length); + uivector_push_back(values, dist_code); + uivector_push_back(values, extra_distance); +} + +static const unsigned HASH_NUM_VALUES = 65536; +static const unsigned HASH_NUM_CHARACTERS = 3; +static const unsigned HASH_SHIFT = 2; +/* +The HASH_NUM_CHARACTERS value is used to make encoding faster by using longer +sequences to generate a hash value from the stream bytes. Setting it to 3 +gives exactly the same compression as the brute force method, since deflate's +run length encoding starts with lengths of 3. Setting it to higher values, +like 6, can make the encoding faster (not always though!), but will cause the +encoding to miss any length between 3 and this value, so that the compression +may be worse (but this can vary too depending on the image, sometimes it is +even a bit better instead). +The HASH_NUM_VALUES is the amount of unique possible hash values that +combinations of bytes can give, the higher it is the more memory is needed, but +if it's too low the advantage of hashing is gone. +*/ + +typedef struct Hash +{ + int* head; /*hash value to head circular pos*/ + int* val; /*circular pos to hash value*/ + /*circular pos to prev circular pos*/ + unsigned short* chain; + unsigned short* zeros; +} Hash; + +static unsigned hash_init(Hash* hash, unsigned windowsize) +{ + unsigned i; + hash->head = (int*)lodepng_malloc(sizeof(int) * HASH_NUM_VALUES); + hash->val = (int*)lodepng_malloc(sizeof(int) * windowsize); + hash->chain = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); + hash->zeros = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); + + if(!hash->head || !hash->val || !hash->chain || !hash->zeros) return 83; /*alloc fail*/ + + /*initialize hash table*/ + for(i = 0; i < HASH_NUM_VALUES; i++) hash->head[i] = -1; + for(i = 0; i < windowsize; i++) hash->val[i] = -1; + for(i = 0; i < windowsize; i++) hash->chain[i] = i; /*same value as index indicates uninitialized*/ + + return 0; +} + +static void hash_cleanup(Hash* hash) +{ + lodepng_free(hash->head); + lodepng_free(hash->val); + lodepng_free(hash->chain); + lodepng_free(hash->zeros); +} + +static unsigned getHash(const unsigned char* data, size_t size, size_t pos) +{ + unsigned result = 0; + size_t amount, i; + if(pos >= size) return 0; + amount = HASH_NUM_CHARACTERS; + if(pos + amount >= size) amount = size - pos; + for(i = 0; i < amount; i++) result ^= (data[pos + i] << (i * HASH_SHIFT)); + return result % HASH_NUM_VALUES; +} + +static unsigned countZeros(const unsigned char* data, size_t size, size_t pos) +{ + const unsigned char* start = data + pos; + const unsigned char* end = start + MAX_SUPPORTED_DEFLATE_LENGTH; + if(end > data + size) end = data + size; + data = start; + while (data != end && *data == 0) data++; + /*subtracting two addresses returned as 32-bit number (max value is MAX_SUPPORTED_DEFLATE_LENGTH)*/ + return (unsigned)(data - start); +} + +static void updateHashChain(Hash* hash, size_t pos, int hashval, unsigned windowsize) +{ + unsigned wpos = pos % windowsize; + hash->val[wpos] = hashval; + if(hash->head[hashval] != -1) hash->chain[wpos] = hash->head[hashval]; + hash->head[hashval] = wpos; +} + +/* +LZ77-encode the data. Return value is error code. The input are raw bytes, the output +is in the form of unsigned integers with codes representing for example literal bytes, or +length/distance pairs. +It uses a hash table technique to let it encode faster. When doing LZ77 encoding, a +sliding window (of windowsize) is used, and all past bytes in that window can be used as +the "dictionary". A brute force search through all possible distances would be slow, and +this hash technique is one out of several ways to speed this up. +*/ +static unsigned encodeLZ77(uivector* out, Hash* hash, + const unsigned char* in, size_t inpos, size_t insize, unsigned windowsize, + unsigned minmatch, unsigned nicematch, unsigned lazymatching) +{ + unsigned short numzeros = 0; + int usezeros = windowsize >= 8192; /*for small window size, the 'max chain length' optimization does a better job*/ + unsigned pos, i, error = 0; + /*for large window lengths, assume the user wants no compression loss. Otherwise, max hash chain length speedup.*/ + unsigned maxchainlength = windowsize >= 8192 ? windowsize : windowsize / 8; + unsigned maxlazymatch = windowsize >= 8192 ? MAX_SUPPORTED_DEFLATE_LENGTH : 64; + + if(!error) + { + unsigned offset; /*the offset represents the distance in LZ77 terminology*/ + unsigned length; + unsigned lazy = 0; + unsigned lazylength = 0, lazyoffset = 0; + unsigned hashval; + unsigned current_offset, current_length; + const unsigned char *lastptr, *foreptr, *backptr; + unsigned short hashpos, prevpos; + + for(pos = inpos; pos < insize; pos++) + { + size_t wpos = pos % windowsize; /*position for in 'circular' hash buffers*/ + + hashval = getHash(in, insize, pos); + updateHashChain(hash, pos, hashval, windowsize); + + if(usezeros && hashval == 0) + { + numzeros = countZeros(in, insize, pos); + hash->zeros[wpos] = numzeros; + } + + /*the length and offset found for the current position*/ + length = 0; + offset = 0; + + prevpos = hash->head[hashval]; + hashpos = hash->chain[prevpos]; + + lastptr = &in[insize < pos + MAX_SUPPORTED_DEFLATE_LENGTH ? insize : pos + MAX_SUPPORTED_DEFLATE_LENGTH]; + + /*search for the longest string*/ + if(hash->val[wpos] == (int)hashval) + { + unsigned chainlength = 0; + for(;;) + { + /*stop when went completely around the circular buffer*/ + if(prevpos < wpos && hashpos > prevpos && hashpos <= wpos) break; + if(prevpos > wpos && (hashpos <= wpos || hashpos > prevpos)) break; + if(chainlength++ >= maxchainlength) break; + + current_offset = hashpos <= wpos ? wpos - hashpos : wpos - hashpos + windowsize; + if(current_offset > 0) + { + /*test the next characters*/ + foreptr = &in[pos]; + backptr = &in[pos - current_offset]; + + /*common case in PNGs is lots of zeros. Quickly skip over them as a speedup*/ + if(usezeros && hashval == 0 && hash->val[hashpos] == 0 /*hashval[hashpos] may be out of date*/) + { + unsigned short skip = hash->zeros[hashpos]; + if(skip > numzeros) skip = numzeros; + backptr += skip; + foreptr += skip; + } + + /* multiple checks at once per array bounds check */ + while(foreptr != lastptr && *backptr == *foreptr) /*maximum supported length by deflate is max length*/ + { + ++backptr; + ++foreptr; + } + current_length = (unsigned)(foreptr - &in[pos]); + + if(current_length > length) + { + length = current_length; /*the longest length*/ + offset = current_offset; /*the offset that is related to this longest length*/ + /*jump out once a length of max length is found (speed gain)*/ + if(current_length >= nicematch || current_length == MAX_SUPPORTED_DEFLATE_LENGTH) break; + } + } + + if(hashpos == hash->chain[hashpos]) break; + + prevpos = hashpos; + hashpos = hash->chain[hashpos]; + } + } + + if(lazymatching) + { + if(!lazy && length >= 3 && length <= maxlazymatch && length < MAX_SUPPORTED_DEFLATE_LENGTH) + { + lazy = 1; + lazylength = length; + lazyoffset = offset; + continue; /*try the next byte*/ + } + if(lazy) + { + lazy = 0; + if(pos == 0) ERROR_BREAK(81); + if(length > lazylength + 1) + { + /*push the previous character as literal*/ + if(!uivector_push_back(out, in[pos - 1])) ERROR_BREAK(83 /*alloc fail*/); + } + else + { + length = lazylength; + offset = lazyoffset; + hash->head[hashval] = -1; /*the same hashchain update will be done, this ensures no wrong alteration*/ + pos--; + } + } + } + if(length >= 3 && offset > windowsize) ERROR_BREAK(86 /*too big (or overflown negative) offset*/); + + /**encode it as length/distance pair or literal value**/ + if(length < 3) /*only lengths of 3 or higher are supported as length/distance pair*/ + { + if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); + } + else if(length < minmatch || (length == 3 && offset > 4096)) + { + /*compensate for the fact that longer offsets have more extra bits, a + length of only 3 may be not worth it then*/ + if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); + } + else + { + addLengthDistance(out, length, offset); + for(i = 1; i < length; i++) + { + pos++; + hashval = getHash(in, insize, pos); + updateHashChain(hash, pos, hashval, windowsize); + if(usezeros && hashval == 0) + { + hash->zeros[pos % windowsize] = countZeros(in, insize, pos); + } + } + } + + } /*end of the loop through each character of input*/ + } /*end of "if(!error)"*/ + + return error; +} + +/* /////////////////////////////////////////////////////////////////////////// */ + +static unsigned deflateNoCompression(ucvector* out, const unsigned char* data, size_t datasize) +{ + /*non compressed deflate block data: 1 bit BFINAL,2 bits BTYPE,(5 bits): it jumps to start of next byte, + 2 bytes LEN, 2 bytes NLEN, LEN bytes literal DATA*/ + + size_t i, j, numdeflateblocks = (datasize + 65534) / 65535; + unsigned datapos = 0; + for(i = 0; i < numdeflateblocks; i++) + { + unsigned BFINAL, BTYPE, LEN, NLEN; + unsigned char firstbyte; + + BFINAL = (i == numdeflateblocks - 1); + BTYPE = 0; + + firstbyte = (unsigned char)(BFINAL + ((BTYPE & 1) << 1) + ((BTYPE & 2) << 1)); + ucvector_push_back(out, firstbyte); + + LEN = 65535; + if(datasize - datapos < 65535) LEN = (unsigned)datasize - datapos; + NLEN = 65535 - LEN; + + ucvector_push_back(out, (unsigned char)(LEN % 256)); + ucvector_push_back(out, (unsigned char)(LEN / 256)); + ucvector_push_back(out, (unsigned char)(NLEN % 256)); + ucvector_push_back(out, (unsigned char)(NLEN / 256)); + + /*Decompressed data*/ + for(j = 0; j < 65535 && datapos < datasize; j++) + { + ucvector_push_back(out, data[datapos++]); + } + } + + return 0; +} + +/* +write the lz77-encoded data, which has lit, len and dist codes, to compressed stream using huffman trees. +tree_ll: the tree for lit and len codes. +tree_d: the tree for distance codes. +*/ +static void writeLZ77data(size_t* bp, ucvector* out, const uivector* lz77_encoded, + const HuffmanTree* tree_ll, const HuffmanTree* tree_d) +{ + size_t i = 0; + for(i = 0; i < lz77_encoded->size; i++) + { + unsigned val = lz77_encoded->data[i]; + addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_ll, val), HuffmanTree_getLength(tree_ll, val)); + if(val > 256) /*for a length code, 3 more things have to be added*/ + { + unsigned length_index = val - FIRST_LENGTH_CODE_INDEX; + unsigned n_length_extra_bits = LENGTHEXTRA[length_index]; + unsigned length_extra_bits = lz77_encoded->data[++i]; + + unsigned distance_code = lz77_encoded->data[++i]; + + unsigned distance_index = distance_code; + unsigned n_distance_extra_bits = DISTANCEEXTRA[distance_index]; + unsigned distance_extra_bits = lz77_encoded->data[++i]; + + addBitsToStream(bp, out, length_extra_bits, n_length_extra_bits); + addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_d, distance_code), + HuffmanTree_getLength(tree_d, distance_code)); + addBitsToStream(bp, out, distance_extra_bits, n_distance_extra_bits); + } + } +} + +/*Deflate for a block of type "dynamic", that is, with freely, optimally, created huffman trees*/ +static unsigned deflateDynamic(ucvector* out, size_t* bp, Hash* hash, + const unsigned char* data, size_t datapos, size_t dataend, + const LodePNGCompressSettings* settings, int final) +{ + unsigned error = 0; + + /* + A block is compressed as follows: The PNG data is lz77 encoded, resulting in + literal bytes and length/distance pairs. This is then huffman compressed with + two huffman trees. One huffman tree is used for the lit and len values ("ll"), + another huffman tree is used for the dist values ("d"). These two trees are + stored using their code lengths, and to compress even more these code lengths + are also run-length encoded and huffman compressed. This gives a huffman tree + of code lengths "cl". The code lenghts used to describe this third tree are + the code length code lengths ("clcl"). + */ + + /*The lz77 encoded data, represented with integers since there will also be length and distance codes in it*/ + uivector lz77_encoded; + HuffmanTree tree_ll; /*tree for lit,len values*/ + HuffmanTree tree_d; /*tree for distance codes*/ + HuffmanTree tree_cl; /*tree for encoding the code lengths representing tree_ll and tree_d*/ + uivector frequencies_ll; /*frequency of lit,len codes*/ + uivector frequencies_d; /*frequency of dist codes*/ + uivector frequencies_cl; /*frequency of code length codes*/ + uivector bitlen_lld; /*lit,len,dist code lenghts (int bits), literally (without repeat codes).*/ + uivector bitlen_lld_e; /*bitlen_lld encoded with repeat codes (this is a rudemtary run length compression)*/ + /*bitlen_cl is the code length code lengths ("clcl"). The bit lengths of codes to represent tree_cl + (these are written as is in the file, it would be crazy to compress these using yet another huffman + tree that needs to be represented by yet another set of code lengths)*/ + uivector bitlen_cl; + size_t datasize = dataend - datapos; + + /* + Due to the huffman compression of huffman tree representations ("two levels"), there are some anologies: + bitlen_lld is to tree_cl what data is to tree_ll and tree_d. + bitlen_lld_e is to bitlen_lld what lz77_encoded is to data. + bitlen_cl is to bitlen_lld_e what bitlen_lld is to lz77_encoded. + */ + + unsigned BFINAL = final; + size_t numcodes_ll, numcodes_d, i; + unsigned HLIT, HDIST, HCLEN; + + uivector_init(&lz77_encoded); + HuffmanTree_init(&tree_ll); + HuffmanTree_init(&tree_d); + HuffmanTree_init(&tree_cl); + uivector_init(&frequencies_ll); + uivector_init(&frequencies_d); + uivector_init(&frequencies_cl); + uivector_init(&bitlen_lld); + uivector_init(&bitlen_lld_e); + uivector_init(&bitlen_cl); + + /*This while loop never loops due to a break at the end, it is here to + allow breaking out of it to the cleanup phase on error conditions.*/ + while(!error) + { + if(settings->use_lz77) + { + error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, + settings->minmatch, settings->nicematch, settings->lazymatching); + if(error) break; + } + else + { + if(!uivector_resize(&lz77_encoded, datasize)) ERROR_BREAK(83 /*alloc fail*/); + for(i = datapos; i < dataend; i++) lz77_encoded.data[i] = data[i]; /*no LZ77, but still will be Huffman compressed*/ + } + + if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83 /*alloc fail*/); + if(!uivector_resizev(&frequencies_d, 30, 0)) ERROR_BREAK(83 /*alloc fail*/); + + /*Count the frequencies of lit, len and dist codes*/ + for(i = 0; i < lz77_encoded.size; i++) + { + unsigned symbol = lz77_encoded.data[i]; + frequencies_ll.data[symbol]++; + if(symbol > 256) + { + unsigned dist = lz77_encoded.data[i + 2]; + frequencies_d.data[dist]++; + i += 3; + } + } + frequencies_ll.data[256] = 1; /*there will be exactly 1 end code, at the end of the block*/ + + /*Make both huffman trees, one for the lit and len codes, one for the dist codes*/ + error = HuffmanTree_makeFromFrequencies(&tree_ll, frequencies_ll.data, 257, frequencies_ll.size, 15); + if(error) break; + /*2, not 1, is chosen for mincodes: some buggy PNG decoders require at least 2 symbols in the dist tree*/ + error = HuffmanTree_makeFromFrequencies(&tree_d, frequencies_d.data, 2, frequencies_d.size, 15); + if(error) break; + + numcodes_ll = tree_ll.numcodes; if(numcodes_ll > 286) numcodes_ll = 286; + numcodes_d = tree_d.numcodes; if(numcodes_d > 30) numcodes_d = 30; + /*store the code lengths of both generated trees in bitlen_lld*/ + for(i = 0; i < numcodes_ll; i++) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_ll, (unsigned)i)); + for(i = 0; i < numcodes_d; i++) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_d, (unsigned)i)); + + /*run-length compress bitlen_ldd into bitlen_lld_e by using repeat codes 16 (copy length 3-6 times), + 17 (3-10 zeroes), 18 (11-138 zeroes)*/ + for(i = 0; i < (unsigned)bitlen_lld.size; i++) + { + unsigned j = 0; /*amount of repititions*/ + while(i + j + 1 < (unsigned)bitlen_lld.size && bitlen_lld.data[i + j + 1] == bitlen_lld.data[i]) j++; + + if(bitlen_lld.data[i] == 0 && j >= 2) /*repeat code for zeroes*/ + { + j++; /*include the first zero*/ + if(j <= 10) /*repeat code 17 supports max 10 zeroes*/ + { + uivector_push_back(&bitlen_lld_e, 17); + uivector_push_back(&bitlen_lld_e, j - 3); + } + else /*repeat code 18 supports max 138 zeroes*/ + { + if(j > 138) j = 138; + uivector_push_back(&bitlen_lld_e, 18); + uivector_push_back(&bitlen_lld_e, j - 11); + } + i += (j - 1); + } + else if(j >= 3) /*repeat code for value other than zero*/ + { + size_t k; + unsigned num = j / 6, rest = j % 6; + uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); + for(k = 0; k < num; k++) + { + uivector_push_back(&bitlen_lld_e, 16); + uivector_push_back(&bitlen_lld_e, 6 - 3); + } + if(rest >= 3) + { + uivector_push_back(&bitlen_lld_e, 16); + uivector_push_back(&bitlen_lld_e, rest - 3); + } + else j -= rest; + i += j; + } + else /*too short to benefit from repeat code*/ + { + uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); + } + } + + /*generate tree_cl, the huffmantree of huffmantrees*/ + + if(!uivector_resizev(&frequencies_cl, NUM_CODE_LENGTH_CODES, 0)) ERROR_BREAK(83 /*alloc fail*/); + for(i = 0; i < bitlen_lld_e.size; i++) + { + frequencies_cl.data[bitlen_lld_e.data[i]]++; + /*after a repeat code come the bits that specify the number of repetitions, + those don't need to be in the frequencies_cl calculation*/ + if(bitlen_lld_e.data[i] >= 16) i++; + } + + error = HuffmanTree_makeFromFrequencies(&tree_cl, frequencies_cl.data, + frequencies_cl.size, frequencies_cl.size, 7); + if(error) break; + + if(!uivector_resize(&bitlen_cl, tree_cl.numcodes)) ERROR_BREAK(83 /*alloc fail*/); + for(i = 0; i < tree_cl.numcodes; i++) + { + /*lenghts of code length tree is in the order as specified by deflate*/ + bitlen_cl.data[i] = HuffmanTree_getLength(&tree_cl, CLCL_ORDER[i]); + } + while(bitlen_cl.data[bitlen_cl.size - 1] == 0 && bitlen_cl.size > 4) + { + /*remove zeros at the end, but minimum size must be 4*/ + if(!uivector_resize(&bitlen_cl, bitlen_cl.size - 1)) ERROR_BREAK(83 /*alloc fail*/); + } + if(error) break; + + /* + Write everything into the output + + After the BFINAL and BTYPE, the dynamic block consists out of the following: + - 5 bits HLIT, 5 bits HDIST, 4 bits HCLEN + - (HCLEN+4)*3 bits code lengths of code length alphabet + - HLIT + 257 code lenghts of lit/length alphabet (encoded using the code length + alphabet, + possible repetition codes 16, 17, 18) + - HDIST + 1 code lengths of distance alphabet (encoded using the code length + alphabet, + possible repetition codes 16, 17, 18) + - compressed data + - 256 (end code) + */ + + /*Write block type*/ + addBitToStream(bp, out, BFINAL); + addBitToStream(bp, out, 0); /*first bit of BTYPE "dynamic"*/ + addBitToStream(bp, out, 1); /*second bit of BTYPE "dynamic"*/ + + /*write the HLIT, HDIST and HCLEN values*/ + HLIT = (unsigned)(numcodes_ll - 257); + HDIST = (unsigned)(numcodes_d - 1); + HCLEN = (unsigned)bitlen_cl.size - 4; + /*trim zeroes for HCLEN. HLIT and HDIST were already trimmed at tree creation*/ + while(!bitlen_cl.data[HCLEN + 4 - 1] && HCLEN > 0) HCLEN--; + addBitsToStream(bp, out, HLIT, 5); + addBitsToStream(bp, out, HDIST, 5); + addBitsToStream(bp, out, HCLEN, 4); + + /*write the code lenghts of the code length alphabet*/ + for(i = 0; i < HCLEN + 4; i++) addBitsToStream(bp, out, bitlen_cl.data[i], 3); + + /*write the lenghts of the lit/len AND the dist alphabet*/ + for(i = 0; i < bitlen_lld_e.size; i++) + { + addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_cl, bitlen_lld_e.data[i]), + HuffmanTree_getLength(&tree_cl, bitlen_lld_e.data[i])); + /*extra bits of repeat codes*/ + if(bitlen_lld_e.data[i] == 16) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 2); + else if(bitlen_lld_e.data[i] == 17) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 3); + else if(bitlen_lld_e.data[i] == 18) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 7); + } + + /*write the compressed data symbols*/ + writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); + /*error: the length of the end code 256 must be larger than 0*/ + if(HuffmanTree_getLength(&tree_ll, 256) == 0) ERROR_BREAK(64); + + /*write the end code*/ + addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); + + break; /*end of error-while*/ + } + + /*cleanup*/ + uivector_cleanup(&lz77_encoded); + HuffmanTree_cleanup(&tree_ll); + HuffmanTree_cleanup(&tree_d); + HuffmanTree_cleanup(&tree_cl); + uivector_cleanup(&frequencies_ll); + uivector_cleanup(&frequencies_d); + uivector_cleanup(&frequencies_cl); + uivector_cleanup(&bitlen_lld_e); + uivector_cleanup(&bitlen_lld); + uivector_cleanup(&bitlen_cl); + + return error; +} + +static unsigned deflateFixed(ucvector* out, size_t* bp, Hash* hash, + const unsigned char* data, + size_t datapos, size_t dataend, + const LodePNGCompressSettings* settings, int final) +{ + HuffmanTree tree_ll; /*tree for literal values and length codes*/ + HuffmanTree tree_d; /*tree for distance codes*/ + + unsigned BFINAL = final; + unsigned error = 0; + size_t i; + + HuffmanTree_init(&tree_ll); + HuffmanTree_init(&tree_d); + + generateFixedLitLenTree(&tree_ll); + generateFixedDistanceTree(&tree_d); + + addBitToStream(bp, out, BFINAL); + addBitToStream(bp, out, 1); /*first bit of BTYPE*/ + addBitToStream(bp, out, 0); /*second bit of BTYPE*/ + + if(settings->use_lz77) /*LZ77 encoded*/ + { + uivector lz77_encoded; + uivector_init(&lz77_encoded); + error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, + settings->minmatch, settings->nicematch, settings->lazymatching); + if(!error) writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); + uivector_cleanup(&lz77_encoded); + } + else /*no LZ77, but still will be Huffman compressed*/ + { + for(i = datapos; i < dataend; i++) + { + addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, data[i]), HuffmanTree_getLength(&tree_ll, data[i])); + } + } + /*add END code*/ + if(!error) addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); + + /*cleanup*/ + HuffmanTree_cleanup(&tree_ll); + HuffmanTree_cleanup(&tree_d); + + return error; +} + +static unsigned lodepng_deflatev(ucvector* out, const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings) +{ + unsigned error = 0; + size_t i, blocksize, numdeflateblocks; + size_t bp = 0; /*the bit pointer*/ + Hash hash; + + if(settings->btype > 2) return 61; + else if(settings->btype == 0) return deflateNoCompression(out, in, insize); + else if(settings->btype == 1) blocksize = insize; + else /*if(settings->btype == 2)*/ + { + blocksize = insize / 8 + 8; + if(blocksize < 65535) blocksize = 65535; + } + + numdeflateblocks = (insize + blocksize - 1) / blocksize; + if(numdeflateblocks == 0) numdeflateblocks = 1; + + error = hash_init(&hash, settings->windowsize); + if(error) return error; + + for(i = 0; i < numdeflateblocks && !error; i++) + { + int final = i == numdeflateblocks - 1; + size_t start = i * blocksize; + size_t end = start + blocksize; + if(end > insize) end = insize; + + if(settings->btype == 1) error = deflateFixed(out, &bp, &hash, in, start, end, settings, final); + else if(settings->btype == 2) error = deflateDynamic(out, &bp, &hash, in, start, end, settings, final); + } + + hash_cleanup(&hash); + + return error; +} + +unsigned lodepng_deflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings) +{ + unsigned error; + ucvector v; + ucvector_init_buffer(&v, *out, *outsize); + error = lodepng_deflatev(&v, in, insize, settings); + *out = v.data; + *outsize = v.size; + return error; +} + +static unsigned deflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings) +{ + if(settings->custom_deflate) + { + return settings->custom_deflate(out, outsize, in, insize, settings); + } + else + { + return lodepng_deflate(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Adler32 */ +/* ////////////////////////////////////////////////////////////////////////// */ + +static unsigned update_adler32(unsigned adler, const unsigned char* data, unsigned len) +{ + unsigned s1 = adler & 0xffff; + unsigned s2 = (adler >> 16) & 0xffff; + + while(len > 0) + { + /*at least 5550 sums can be done before the sums overflow, saving a lot of module divisions*/ + unsigned amount = len > 5550 ? 5550 : len; + len -= amount; + while(amount > 0) + { + s1 += (*data++); + s2 += s1; + amount--; + } + s1 %= 65521; + s2 %= 65521; + } + + return (s2 << 16) | s1; +} + +/*Return the adler32 of the bytes data[0..len-1]*/ +static unsigned adler32(const unsigned char* data, unsigned len) +{ + return update_adler32(1L, data, len); +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Zlib / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_DECODER + +unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGDecompressSettings* settings) +{ + unsigned error = 0; + unsigned CM, CINFO, FDICT; + + if(insize < 2) return 53; /*error, size of zlib data too small*/ + /*read information from zlib header*/ + if((in[0] * 256 + in[1]) % 31 != 0) + { + /*error: 256 * in[0] + in[1] must be a multiple of 31, the FCHECK value is supposed to be made that way*/ + return 24; + } + + CM = in[0] & 15; + CINFO = (in[0] >> 4) & 15; + /*FCHECK = in[1] & 31;*/ /*FCHECK is already tested above*/ + FDICT = (in[1] >> 5) & 1; + /*FLEVEL = (in[1] >> 6) & 3;*/ /*FLEVEL is not used here*/ + + if(CM != 8 || CINFO > 7) + { + /*error: only compression method 8: inflate with sliding window of 32k is supported by the PNG spec*/ + return 25; + } + if(FDICT != 0) + { + /*error: the specification of PNG says about the zlib stream: + "The additional flags shall not specify a preset dictionary."*/ + return 26; + } + + error = inflate(out, outsize, in + 2, insize - 2, settings); + if(error) return error; + + if(!settings->ignore_adler32) + { + unsigned ADLER32 = lodepng_read32bitInt(&in[insize - 4]); + unsigned checksum = adler32(*out, (unsigned)(*outsize)); + if(checksum != ADLER32) return 58; /*error, adler checksum not correct, data must be corrupted*/ + } + + return 0; /*no error*/ +} + +static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGDecompressSettings* settings) +{ + if(settings->custom_zlib) + return settings->custom_zlib(out, outsize, in, insize, settings); + else + return lodepng_zlib_decompress(out, outsize, in, insize, settings); +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER + +unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGCompressSettings* settings) +{ + /*initially, *out must be NULL and outsize 0, if you just give some random *out + that's pointing to a non allocated buffer, this'll crash*/ + ucvector outv; + size_t i; + unsigned error; + unsigned char* deflatedata = 0; + size_t deflatesize = 0; + + unsigned ADLER32; + /*zlib data: 1 byte CMF (CM+CINFO), 1 byte FLG, deflate data, 4 byte ADLER32 checksum of the Decompressed data*/ + unsigned CMF = 120; /*0b01111000: CM 8, CINFO 7. With CINFO 7, any window size up to 32768 can be used.*/ + unsigned FLEVEL = 0; + unsigned FDICT = 0; + unsigned CMFFLG = 256 * CMF + FDICT * 32 + FLEVEL * 64; + unsigned FCHECK = 31 - CMFFLG % 31; + CMFFLG += FCHECK; + + /*ucvector-controlled version of the output buffer, for dynamic array*/ + ucvector_init_buffer(&outv, *out, *outsize); + + ucvector_push_back(&outv, (unsigned char)(CMFFLG / 256)); + ucvector_push_back(&outv, (unsigned char)(CMFFLG % 256)); + + error = deflate(&deflatedata, &deflatesize, in, insize, settings); + + if(!error) + { + ADLER32 = adler32(in, (unsigned)insize); + for(i = 0; i < deflatesize; i++) ucvector_push_back(&outv, deflatedata[i]); + lodepng_free(deflatedata); + lodepng_add32bitInt(&outv, ADLER32); + } + + *out = outv.data; + *outsize = outv.size; + + return error; +} + +/* compress using the default or custom zlib function */ +static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGCompressSettings* settings) +{ + if(settings->custom_zlib) + { + return settings->custom_zlib(out, outsize, in, insize, settings); + } + else + { + return lodepng_zlib_compress(out, outsize, in, insize, settings); + } +} + +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#else /*no LODEPNG_COMPILE_ZLIB*/ + +#ifdef LODEPNG_COMPILE_DECODER +static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGDecompressSettings* settings) +{ + if (!settings->custom_zlib) return 87; /*no custom zlib function provided */ + return settings->custom_zlib(out, outsize, in, insize, settings); +} +#endif /*LODEPNG_COMPILE_DECODER*/ +#ifdef LODEPNG_COMPILE_ENCODER +static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, + size_t insize, const LodePNGCompressSettings* settings) +{ + if (!settings->custom_zlib) return 87; /*no custom zlib function provided */ + return settings->custom_zlib(out, outsize, in, insize, settings); +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#endif /*LODEPNG_COMPILE_ZLIB*/ + +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_ENCODER + +/*this is a good tradeoff between speed and compression ratio*/ +#define DEFAULT_WINDOWSIZE 2048 + +void lodepng_compress_settings_init(LodePNGCompressSettings* settings) +{ + /*compress with dynamic huffman tree (not in the mathematical sense, just not the predefined one)*/ + settings->btype = 2; + settings->use_lz77 = 1; + settings->windowsize = DEFAULT_WINDOWSIZE; + settings->minmatch = 3; + settings->nicematch = 128; + settings->lazymatching = 1; + + settings->custom_zlib = 0; + settings->custom_deflate = 0; + settings->custom_context = 0; +} + +const LodePNGCompressSettings lodepng_default_compress_settings = {2, 1, DEFAULT_WINDOWSIZE, 3, 128, 1, 0, 0, 0}; + + +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_DECODER + +void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings) +{ + settings->ignore_adler32 = 0; + + settings->custom_zlib = 0; + settings->custom_inflate = 0; + settings->custom_context = 0; +} + +const LodePNGDecompressSettings lodepng_default_decompress_settings = {0, 0, 0, 0}; + +#endif /*LODEPNG_COMPILE_DECODER*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // End of Zlib related code. Begin of PNG related code. // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + +#ifdef LODEPNG_COMPILE_PNG + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / CRC32 / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/* CRC polynomial: 0xedb88320 */ +static unsigned lodepng_crc32_table[256] = { + 0u, 1996959894u, 3993919788u, 2567524794u, 124634137u, 1886057615u, 3915621685u, 2657392035u, + 249268274u, 2044508324u, 3772115230u, 2547177864u, 162941995u, 2125561021u, 3887607047u, 2428444049u, + 498536548u, 1789927666u, 4089016648u, 2227061214u, 450548861u, 1843258603u, 4107580753u, 2211677639u, + 325883990u, 1684777152u, 4251122042u, 2321926636u, 335633487u, 1661365465u, 4195302755u, 2366115317u, + 997073096u, 1281953886u, 3579855332u, 2724688242u, 1006888145u, 1258607687u, 3524101629u, 2768942443u, + 901097722u, 1119000684u, 3686517206u, 2898065728u, 853044451u, 1172266101u, 3705015759u, 2882616665u, + 651767980u, 1373503546u, 3369554304u, 3218104598u, 565507253u, 1454621731u, 3485111705u, 3099436303u, + 671266974u, 1594198024u, 3322730930u, 2970347812u, 795835527u, 1483230225u, 3244367275u, 3060149565u, + 1994146192u, 31158534u, 2563907772u, 4023717930u, 1907459465u, 112637215u, 2680153253u, 3904427059u, + 2013776290u, 251722036u, 2517215374u, 3775830040u, 2137656763u, 141376813u, 2439277719u, 3865271297u, + 1802195444u, 476864866u, 2238001368u, 4066508878u, 1812370925u, 453092731u, 2181625025u, 4111451223u, + 1706088902u, 314042704u, 2344532202u, 4240017532u, 1658658271u, 366619977u, 2362670323u, 4224994405u, + 1303535960u, 984961486u, 2747007092u, 3569037538u, 1256170817u, 1037604311u, 2765210733u, 3554079995u, + 1131014506u, 879679996u, 2909243462u, 3663771856u, 1141124467u, 855842277u, 2852801631u, 3708648649u, + 1342533948u, 654459306u, 3188396048u, 3373015174u, 1466479909u, 544179635u, 3110523913u, 3462522015u, + 1591671054u, 702138776u, 2966460450u, 3352799412u, 1504918807u, 783551873u, 3082640443u, 3233442989u, + 3988292384u, 2596254646u, 62317068u, 1957810842u, 3939845945u, 2647816111u, 81470997u, 1943803523u, + 3814918930u, 2489596804u, 225274430u, 2053790376u, 3826175755u, 2466906013u, 167816743u, 2097651377u, + 4027552580u, 2265490386u, 503444072u, 1762050814u, 4150417245u, 2154129355u, 426522225u, 1852507879u, + 4275313526u, 2312317920u, 282753626u, 1742555852u, 4189708143u, 2394877945u, 397917763u, 1622183637u, + 3604390888u, 2714866558u, 953729732u, 1340076626u, 3518719985u, 2797360999u, 1068828381u, 1219638859u, + 3624741850u, 2936675148u, 906185462u, 1090812512u, 3747672003u, 2825379669u, 829329135u, 1181335161u, + 3412177804u, 3160834842u, 628085408u, 1382605366u, 3423369109u, 3138078467u, 570562233u, 1426400815u, + 3317316542u, 2998733608u, 733239954u, 1555261956u, 3268935591u, 3050360625u, 752459403u, 1541320221u, + 2607071920u, 3965973030u, 1969922972u, 40735498u, 2617837225u, 3943577151u, 1913087877u, 83908371u, + 2512341634u, 3803740692u, 2075208622u, 213261112u, 2463272603u, 3855990285u, 2094854071u, 198958881u, + 2262029012u, 4057260610u, 1759359992u, 534414190u, 2176718541u, 4139329115u, 1873836001u, 414664567u, + 2282248934u, 4279200368u, 1711684554u, 285281116u, 2405801727u, 4167216745u, 1634467795u, 376229701u, + 2685067896u, 3608007406u, 1308918612u, 956543938u, 2808555105u, 3495958263u, 1231636301u, 1047427035u, + 2932959818u, 3654703836u, 1088359270u, 936918000u, 2847714899u, 3736837829u, 1202900863u, 817233897u, + 3183342108u, 3401237130u, 1404277552u, 615818150u, 3134207493u, 3453421203u, 1423857449u, 601450431u, + 3009837614u, 3294710456u, 1567103746u, 711928724u, 3020668471u, 3272380065u, 1510334235u, 755167117u +}; + +/*Return the CRC of the bytes buf[0..len-1].*/ +unsigned lodepng_crc32(const unsigned char* buf, size_t len) +{ + unsigned c = 0xffffffffL; + size_t n; + + for(n = 0; n < len; n++) + { + c = lodepng_crc32_table[(c ^ buf[n]) & 0xff] ^ (c >> 8); + } + return c ^ 0xffffffffL; +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Reading and writing single bits and bytes from/to stream for LodePNG / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +static unsigned char readBitFromReversedStream(size_t* bitpointer, const unsigned char* bitstream) +{ + unsigned char result = (unsigned char)((bitstream[(*bitpointer) >> 3] >> (7 - ((*bitpointer) & 0x7))) & 1); + (*bitpointer)++; + return result; +} + +static unsigned readBitsFromReversedStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) +{ + unsigned result = 0; + size_t i; + for(i = nbits - 1; i < nbits; i--) + { + result += (unsigned)readBitFromReversedStream(bitpointer, bitstream) << i; + } + return result; +} + +#ifdef LODEPNG_COMPILE_DECODER +static void setBitOfReversedStream0(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) +{ + /*the current bit in bitstream must be 0 for this to work*/ + if(bit) + { + /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/ + bitstream[(*bitpointer) >> 3] |= (bit << (7 - ((*bitpointer) & 0x7))); + } + (*bitpointer)++; +} +#endif /*LODEPNG_COMPILE_DECODER*/ + +static void setBitOfReversedStream(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) +{ + /*the current bit in bitstream may be 0 or 1 for this to work*/ + if(bit == 0) bitstream[(*bitpointer) >> 3] &= (unsigned char)(~(1 << (7 - ((*bitpointer) & 0x7)))); + else bitstream[(*bitpointer) >> 3] |= (1 << (7 - ((*bitpointer) & 0x7))); + (*bitpointer)++; +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / PNG chunks / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +unsigned lodepng_chunk_length(const unsigned char* chunk) +{ + return lodepng_read32bitInt(&chunk[0]); +} + +void lodepng_chunk_type(char type[5], const unsigned char* chunk) +{ + unsigned i; + for(i = 0; i < 4; i++) type[i] = chunk[4 + i]; + type[4] = 0; /*null termination char*/ +} + +unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type) +{ + if(strlen(type) != 4) return 0; + return (chunk[4] == type[0] && chunk[5] == type[1] && chunk[6] == type[2] && chunk[7] == type[3]); +} + +unsigned char lodepng_chunk_ancillary(const unsigned char* chunk) +{ + return((chunk[4] & 32) != 0); +} + +unsigned char lodepng_chunk_private(const unsigned char* chunk) +{ + return((chunk[6] & 32) != 0); +} + +unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk) +{ + return((chunk[7] & 32) != 0); +} + +unsigned char* lodepng_chunk_data(unsigned char* chunk) +{ + return &chunk[8]; +} + +const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk) +{ + return &chunk[8]; +} + +unsigned lodepng_chunk_check_crc(const unsigned char* chunk) +{ + unsigned length = lodepng_chunk_length(chunk); + unsigned CRC = lodepng_read32bitInt(&chunk[length + 8]); + /*the CRC is taken of the data and the 4 chunk type letters, not the length*/ + unsigned checksum = lodepng_crc32(&chunk[4], length + 4); + if(CRC != checksum) return 1; + else return 0; +} + +void lodepng_chunk_generate_crc(unsigned char* chunk) +{ + unsigned length = lodepng_chunk_length(chunk); + unsigned CRC = lodepng_crc32(&chunk[4], length + 4); + lodepng_set32bitInt(chunk + 8 + length, CRC); +} + +unsigned char* lodepng_chunk_next(unsigned char* chunk) +{ + unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; + return &chunk[total_chunk_length]; +} + +const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk) +{ + unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; + return &chunk[total_chunk_length]; +} + +unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk) +{ + unsigned i; + unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; + unsigned char *chunk_start, *new_buffer; + size_t new_length = (*outlength) + total_chunk_length; + if(new_length < total_chunk_length || new_length < (*outlength)) return 77; /*integer overflow happened*/ + + new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); + if(!new_buffer) return 83; /*alloc fail*/ + (*out) = new_buffer; + (*outlength) = new_length; + chunk_start = &(*out)[new_length - total_chunk_length]; + + for(i = 0; i < total_chunk_length; i++) chunk_start[i] = chunk[i]; + + return 0; +} + +unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, + const char* type, const unsigned char* data) +{ + unsigned i; + unsigned char *chunk, *new_buffer; + size_t new_length = (*outlength) + length + 12; + if(new_length < length + 12 || new_length < (*outlength)) return 77; /*integer overflow happened*/ + new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); + if(!new_buffer) return 83; /*alloc fail*/ + (*out) = new_buffer; + (*outlength) = new_length; + chunk = &(*out)[(*outlength) - length - 12]; + + /*1: length*/ + lodepng_set32bitInt(chunk, (unsigned)length); + + /*2: chunk name (4 letters)*/ + chunk[4] = type[0]; + chunk[5] = type[1]; + chunk[6] = type[2]; + chunk[7] = type[3]; + + /*3: the data*/ + for(i = 0; i < length; i++) chunk[8 + i] = data[i]; + + /*4: CRC (of the chunkname characters and the data)*/ + lodepng_chunk_generate_crc(chunk); + + return 0; +} + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / Color types and such / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*return type is a LodePNG error code*/ +static unsigned checkColorValidity(LodePNGColorType colortype, unsigned bd) /*bd = bitdepth*/ +{ + switch(colortype) + { + case 0: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 || bd == 16)) return 37; break; /*grey*/ + case 2: if(!( bd == 8 || bd == 16)) return 37; break; /*RGB*/ + case 3: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 )) return 37; break; /*palette*/ + case 4: if(!( bd == 8 || bd == 16)) return 37; break; /*grey + alpha*/ + case 6: if(!( bd == 8 || bd == 16)) return 37; break; /*RGBA*/ + default: return 31; + } + return 0; /*allowed color type / bits combination*/ +} + +static unsigned getNumColorChannels(LodePNGColorType colortype) +{ + switch(colortype) + { + case 0: return 1; /*grey*/ + case 2: return 3; /*RGB*/ + case 3: return 1; /*palette*/ + case 4: return 2; /*grey + alpha*/ + case 6: return 4; /*RGBA*/ + } + return 0; /*unexisting color type*/ +} + +static unsigned lodepng_get_bpp_lct(LodePNGColorType colortype, unsigned bitdepth) +{ + /*bits per pixel is amount of channels * bits per channel*/ + return getNumColorChannels(colortype) * bitdepth; +} + +/* ////////////////////////////////////////////////////////////////////////// */ + +void lodepng_color_mode_init(LodePNGColorMode* info) +{ + info->key_defined = 0; + info->key_r = info->key_g = info->key_b = 0; + info->colortype = LCT_RGBA; + info->bitdepth = 8; + info->palette = 0; + info->palettesize = 0; +} + +void lodepng_color_mode_cleanup(LodePNGColorMode* info) +{ + lodepng_palette_clear(info); +} + +unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source) +{ + size_t i; + lodepng_color_mode_cleanup(dest); + *dest = *source; + if(source->palette) + { + dest->palette = (unsigned char*)lodepng_malloc(1024); + if(!dest->palette && source->palettesize) return 83; /*alloc fail*/ + for(i = 0; i < source->palettesize * 4; i++) dest->palette[i] = source->palette[i]; + } + return 0; +} + +static int lodepng_color_mode_equal(const LodePNGColorMode* a, const LodePNGColorMode* b) +{ + size_t i; + if(a->colortype != b->colortype) return 0; + if(a->bitdepth != b->bitdepth) return 0; + if(a->key_defined != b->key_defined) return 0; + if(a->key_defined) + { + if(a->key_r != b->key_r) return 0; + if(a->key_g != b->key_g) return 0; + if(a->key_b != b->key_b) return 0; + } + if(a->palettesize != b->palettesize) return 0; + for(i = 0; i < a->palettesize * 4; i++) + { + if(a->palette[i] != b->palette[i]) return 0; + } + return 1; +} + +void lodepng_palette_clear(LodePNGColorMode* info) +{ + if(info->palette) lodepng_free(info->palette); + info->palette = 0; + info->palettesize = 0; +} + +unsigned lodepng_palette_add(LodePNGColorMode* info, + unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + unsigned char* data; + /*the same resize technique as C++ std::vectors is used, and here it's made so that for a palette with + the max of 256 colors, it'll have the exact alloc size*/ + if(!info->palette) /*allocate palette if empty*/ + { + /*room for 256 colors with 4 bytes each*/ + data = (unsigned char*)lodepng_realloc(info->palette, 1024); + if(!data) return 83; /*alloc fail*/ + else info->palette = data; + } + info->palette[4 * info->palettesize + 0] = r; + info->palette[4 * info->palettesize + 1] = g; + info->palette[4 * info->palettesize + 2] = b; + info->palette[4 * info->palettesize + 3] = a; + info->palettesize++; + return 0; +} + +unsigned lodepng_get_bpp(const LodePNGColorMode* info) +{ + /*calculate bits per pixel out of colortype and bitdepth*/ + return lodepng_get_bpp_lct(info->colortype, info->bitdepth); +} + +unsigned lodepng_get_channels(const LodePNGColorMode* info) +{ + return getNumColorChannels(info->colortype); +} + +unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info) +{ + return info->colortype == LCT_GREY || info->colortype == LCT_GREY_ALPHA; +} + +unsigned lodepng_is_alpha_type(const LodePNGColorMode* info) +{ + return (info->colortype & 4) != 0; /*4 or 6*/ +} + +unsigned lodepng_is_palette_type(const LodePNGColorMode* info) +{ + return info->colortype == LCT_PALETTE; +} + +unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info) +{ + size_t i; + for(i = 0; i < info->palettesize; i++) + { + if(info->palette[i * 4 + 3] < 255) return 1; + } + return 0; +} + +unsigned lodepng_can_have_alpha(const LodePNGColorMode* info) +{ + return info->key_defined + || lodepng_is_alpha_type(info) + || lodepng_has_palette_alpha(info); +} + +size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color) +{ + return (w * h * lodepng_get_bpp(color) + 7) / 8; +} + +size_t lodepng_get_raw_size_lct(unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) +{ + return (w * h * lodepng_get_bpp_lct(colortype, bitdepth) + 7) / 8; +} + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + +static void LodePNGUnknownChunks_init(LodePNGInfo* info) +{ + unsigned i; + for(i = 0; i < 3; i++) info->unknown_chunks_data[i] = 0; + for(i = 0; i < 3; i++) info->unknown_chunks_size[i] = 0; +} + +static void LodePNGUnknownChunks_cleanup(LodePNGInfo* info) +{ + unsigned i; + for(i = 0; i < 3; i++) lodepng_free(info->unknown_chunks_data[i]); +} + +static unsigned LodePNGUnknownChunks_copy(LodePNGInfo* dest, const LodePNGInfo* src) +{ + unsigned i; + + LodePNGUnknownChunks_cleanup(dest); + + for(i = 0; i < 3; i++) + { + size_t j; + dest->unknown_chunks_size[i] = src->unknown_chunks_size[i]; + dest->unknown_chunks_data[i] = (unsigned char*)lodepng_malloc(src->unknown_chunks_size[i]); + if(!dest->unknown_chunks_data[i] && dest->unknown_chunks_size[i]) return 83; /*alloc fail*/ + for(j = 0; j < src->unknown_chunks_size[i]; j++) + { + dest->unknown_chunks_data[i][j] = src->unknown_chunks_data[i][j]; + } + } + + return 0; +} + +/******************************************************************************/ + +static void LodePNGText_init(LodePNGInfo* info) +{ + info->text_num = 0; + info->text_keys = NULL; + info->text_strings = NULL; +} + +static void LodePNGText_cleanup(LodePNGInfo* info) +{ + size_t i; + for(i = 0; i < info->text_num; i++) + { + string_cleanup(&info->text_keys[i]); + string_cleanup(&info->text_strings[i]); + } + lodepng_free(info->text_keys); + lodepng_free(info->text_strings); +} + +static unsigned LodePNGText_copy(LodePNGInfo* dest, const LodePNGInfo* source) +{ + size_t i = 0; + dest->text_keys = 0; + dest->text_strings = 0; + dest->text_num = 0; + for(i = 0; i < source->text_num; i++) + { + CERROR_TRY_RETURN(lodepng_add_text(dest, source->text_keys[i], source->text_strings[i])); + } + return 0; +} + +void lodepng_clear_text(LodePNGInfo* info) +{ + LodePNGText_cleanup(info); +} + +unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str) +{ + char** new_keys = (char**)(lodepng_realloc(info->text_keys, sizeof(char*) * (info->text_num + 1))); + char** new_strings = (char**)(lodepng_realloc(info->text_strings, sizeof(char*) * (info->text_num + 1))); + if(!new_keys || !new_strings) + { + lodepng_free(new_keys); + lodepng_free(new_strings); + return 83; /*alloc fail*/ + } + + info->text_num++; + info->text_keys = new_keys; + info->text_strings = new_strings; + + string_init(&info->text_keys[info->text_num - 1]); + string_set(&info->text_keys[info->text_num - 1], key); + + string_init(&info->text_strings[info->text_num - 1]); + string_set(&info->text_strings[info->text_num - 1], str); + + return 0; +} + +/******************************************************************************/ + +static void LodePNGIText_init(LodePNGInfo* info) +{ + info->itext_num = 0; + info->itext_keys = NULL; + info->itext_langtags = NULL; + info->itext_transkeys = NULL; + info->itext_strings = NULL; +} + +static void LodePNGIText_cleanup(LodePNGInfo* info) +{ + size_t i; + for(i = 0; i < info->itext_num; i++) + { + string_cleanup(&info->itext_keys[i]); + string_cleanup(&info->itext_langtags[i]); + string_cleanup(&info->itext_transkeys[i]); + string_cleanup(&info->itext_strings[i]); + } + lodepng_free(info->itext_keys); + lodepng_free(info->itext_langtags); + lodepng_free(info->itext_transkeys); + lodepng_free(info->itext_strings); +} + +static unsigned LodePNGIText_copy(LodePNGInfo* dest, const LodePNGInfo* source) +{ + size_t i = 0; + dest->itext_keys = 0; + dest->itext_langtags = 0; + dest->itext_transkeys = 0; + dest->itext_strings = 0; + dest->itext_num = 0; + for(i = 0; i < source->itext_num; i++) + { + CERROR_TRY_RETURN(lodepng_add_itext(dest, source->itext_keys[i], source->itext_langtags[i], + source->itext_transkeys[i], source->itext_strings[i])); + } + return 0; +} + +void lodepng_clear_itext(LodePNGInfo* info) +{ + LodePNGIText_cleanup(info); +} + +unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, + const char* transkey, const char* str) +{ + char** new_keys = (char**)(lodepng_realloc(info->itext_keys, sizeof(char*) * (info->itext_num + 1))); + char** new_langtags = (char**)(lodepng_realloc(info->itext_langtags, sizeof(char*) * (info->itext_num + 1))); + char** new_transkeys = (char**)(lodepng_realloc(info->itext_transkeys, sizeof(char*) * (info->itext_num + 1))); + char** new_strings = (char**)(lodepng_realloc(info->itext_strings, sizeof(char*) * (info->itext_num + 1))); + if(!new_keys || !new_langtags || !new_transkeys || !new_strings) + { + lodepng_free(new_keys); + lodepng_free(new_langtags); + lodepng_free(new_transkeys); + lodepng_free(new_strings); + return 83; /*alloc fail*/ + } + + info->itext_num++; + info->itext_keys = new_keys; + info->itext_langtags = new_langtags; + info->itext_transkeys = new_transkeys; + info->itext_strings = new_strings; + + string_init(&info->itext_keys[info->itext_num - 1]); + string_set(&info->itext_keys[info->itext_num - 1], key); + + string_init(&info->itext_langtags[info->itext_num - 1]); + string_set(&info->itext_langtags[info->itext_num - 1], langtag); + + string_init(&info->itext_transkeys[info->itext_num - 1]); + string_set(&info->itext_transkeys[info->itext_num - 1], transkey); + + string_init(&info->itext_strings[info->itext_num - 1]); + string_set(&info->itext_strings[info->itext_num - 1], str); + + return 0; +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +void lodepng_info_init(LodePNGInfo* info) +{ + lodepng_color_mode_init(&info->color); + info->interlace_method = 0; + info->compression_method = 0; + info->filter_method = 0; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + info->background_defined = 0; + info->background_r = info->background_g = info->background_b = 0; + + LodePNGText_init(info); + LodePNGIText_init(info); + + info->time_defined = 0; + info->phys_defined = 0; + + LodePNGUnknownChunks_init(info); +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} + +void lodepng_info_cleanup(LodePNGInfo* info) +{ + lodepng_color_mode_cleanup(&info->color); +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + LodePNGText_cleanup(info); + LodePNGIText_cleanup(info); + + LodePNGUnknownChunks_cleanup(info); +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} + +unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source) +{ + lodepng_info_cleanup(dest); + *dest = *source; + lodepng_color_mode_init(&dest->color); + CERROR_TRY_RETURN(lodepng_color_mode_copy(&dest->color, &source->color)); + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + CERROR_TRY_RETURN(LodePNGText_copy(dest, source)); + CERROR_TRY_RETURN(LodePNGIText_copy(dest, source)); + + LodePNGUnknownChunks_init(dest); + CERROR_TRY_RETURN(LodePNGUnknownChunks_copy(dest, source)); +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + return 0; +} + +void lodepng_info_swap(LodePNGInfo* a, LodePNGInfo* b) +{ + LodePNGInfo temp = *a; + *a = *b; + *b = temp; +} + +/* ////////////////////////////////////////////////////////////////////////// */ + +/*index: bitgroup index, bits: bitgroup size(1, 2 or 4, in: bitgroup value, out: octet array to add bits to*/ +static void addColorBits(unsigned char* out, size_t index, unsigned bits, unsigned in) +{ + /*p = the partial index in the byte, e.g. with 4 palettebits it is 0 for first half or 1 for second half*/ + unsigned p = index % (8 / bits); + in &= (1 << bits) - 1; /*filter out any other bits of the input value*/ + in = in << (bits * (8 / bits - p - 1)); + if(p == 0) out[index * bits / 8] = in; + else out[index * bits / 8] |= in; +} + +typedef struct ColorTree ColorTree; + +/* +One node of a color tree +This is the data structure used to count the number of unique colors and to get a palette +index for a color. It's like an octree, but because the alpha channel is used too, each +node has 16 instead of 8 children. +*/ +struct ColorTree +{ + ColorTree* children[16]; /*up to 16 pointers to ColorTree of next level*/ + int index; /*the payload. Only has a meaningful value if this is in the last level*/ +}; + +static void color_tree_init(ColorTree* tree) +{ + int i; + for(i = 0; i < 16; i++) tree->children[i] = 0; + tree->index = -1; +} + +static void color_tree_cleanup(ColorTree* tree) +{ + int i; + for(i = 0; i < 16; i++) + { + if(tree->children[i]) + { + color_tree_cleanup(tree->children[i]); + lodepng_free(tree->children[i]); + } + } +} + +/*returns -1 if color not present, its index otherwise*/ +static int color_tree_get(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + int bit = 0; + for(bit = 0; bit < 8; bit++) + { + int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); + if(!tree->children[i]) return -1; + else tree = tree->children[i]; + } + return tree ? tree->index : -1; +} + +#ifdef LODEPNG_COMPILE_ENCODER +static int color_tree_has(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + return color_tree_get(tree, r, g, b, a) >= 0; +} +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/*color is not allowed to already exist. +Index should be >= 0 (it's signed to be compatible with using -1 for "doesn't exist")*/ +static void color_tree_add(ColorTree* tree, + unsigned char r, unsigned char g, unsigned char b, unsigned char a, int index) +{ + int bit; + for(bit = 0; bit < 8; bit++) + { + int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); + if(!tree->children[i]) + { + tree->children[i] = (ColorTree*)lodepng_malloc(sizeof(ColorTree)); + color_tree_init(tree->children[i]); + } + tree = tree->children[i]; + } + tree->index = index; +} + +/*put a pixel, given its RGBA color, into image of any color type*/ +static unsigned rgba8ToPixel(unsigned char* out, size_t i, + const LodePNGColorMode* mode, ColorTree* tree /*for palette*/, + unsigned char r, unsigned char g, unsigned char b, unsigned char a) +{ + if(mode->colortype == LCT_GREY) + { + unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; + if(mode->bitdepth == 8) out[i] = grey; + else if(mode->bitdepth == 16) out[i * 2 + 0] = out[i * 2 + 1] = grey; + else + { + /*take the most significant bits of grey*/ + grey = (grey >> (8 - mode->bitdepth)) & ((1 << mode->bitdepth) - 1); + addColorBits(out, i, mode->bitdepth, grey); + } + } + else if(mode->colortype == LCT_RGB) + { + if(mode->bitdepth == 8) + { + out[i * 3 + 0] = r; + out[i * 3 + 1] = g; + out[i * 3 + 2] = b; + } + else + { + out[i * 6 + 0] = out[i * 6 + 1] = r; + out[i * 6 + 2] = out[i * 6 + 3] = g; + out[i * 6 + 4] = out[i * 6 + 5] = b; + } + } + else if(mode->colortype == LCT_PALETTE) + { + int index = color_tree_get(tree, r, g, b, a); + if(index < 0) return 82; /*color not in palette*/ + if(mode->bitdepth == 8) out[i] = index; + else addColorBits(out, i, mode->bitdepth, index); + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; + if(mode->bitdepth == 8) + { + out[i * 2 + 0] = grey; + out[i * 2 + 1] = a; + } + else if(mode->bitdepth == 16) + { + out[i * 4 + 0] = out[i * 4 + 1] = grey; + out[i * 4 + 2] = out[i * 4 + 3] = a; + } + } + else if(mode->colortype == LCT_RGBA) + { + if(mode->bitdepth == 8) + { + out[i * 4 + 0] = r; + out[i * 4 + 1] = g; + out[i * 4 + 2] = b; + out[i * 4 + 3] = a; + } + else + { + out[i * 8 + 0] = out[i * 8 + 1] = r; + out[i * 8 + 2] = out[i * 8 + 3] = g; + out[i * 8 + 4] = out[i * 8 + 5] = b; + out[i * 8 + 6] = out[i * 8 + 7] = a; + } + } + + return 0; /*no error*/ +} + +/*put a pixel, given its RGBA16 color, into image of any color 16-bitdepth type*/ +static unsigned rgba16ToPixel(unsigned char* out, size_t i, + const LodePNGColorMode* mode, + unsigned short r, unsigned short g, unsigned short b, unsigned short a) +{ + if(mode->bitdepth != 16) return 85; /*must be 16 for this function*/ + if(mode->colortype == LCT_GREY) + { + unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; + out[i * 2 + 0] = (grey >> 8) & 255; + out[i * 2 + 1] = grey & 255; + } + else if(mode->colortype == LCT_RGB) + { + out[i * 6 + 0] = (r >> 8) & 255; + out[i * 6 + 1] = r & 255; + out[i * 6 + 2] = (g >> 8) & 255; + out[i * 6 + 3] = g & 255; + out[i * 6 + 4] = (b >> 8) & 255; + out[i * 6 + 5] = b & 255; + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; + out[i * 4 + 0] = (grey >> 8) & 255; + out[i * 4 + 1] = grey & 255; + out[i * 4 + 2] = (a >> 8) & 255; + out[i * 4 + 3] = a & 255; + } + else if(mode->colortype == LCT_RGBA) + { + out[i * 8 + 0] = (r >> 8) & 255; + out[i * 8 + 1] = r & 255; + out[i * 8 + 2] = (g >> 8) & 255; + out[i * 8 + 3] = g & 255; + out[i * 8 + 4] = (b >> 8) & 255; + out[i * 8 + 5] = b & 255; + out[i * 8 + 6] = (a >> 8) & 255; + out[i * 8 + 7] = a & 255; + } + + return 0; /*no error*/ +} + +/*Get RGBA8 color of pixel with index i (y * width + x) from the raw image with given color type.*/ +static unsigned getPixelColorRGBA8(unsigned char* r, unsigned char* g, + unsigned char* b, unsigned char* a, + const unsigned char* in, size_t i, + const LodePNGColorMode* mode, + unsigned fix_png) +{ + if(mode->colortype == LCT_GREY) + { + if(mode->bitdepth == 8) + { + *r = *g = *b = in[i]; + if(mode->key_defined && *r == mode->key_r) *a = 0; + else *a = 255; + } + else if(mode->bitdepth == 16) + { + *r = *g = *b = in[i * 2 + 0]; + if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; + else *a = 255; + } + else + { + unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ + size_t j = i * mode->bitdepth; + unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); + *r = *g = *b = (value * 255) / highest; + if(mode->key_defined && value == mode->key_r) *a = 0; + else *a = 255; + } + } + else if(mode->colortype == LCT_RGB) + { + if(mode->bitdepth == 8) + { + *r = in[i * 3 + 0]; *g = in[i * 3 + 1]; *b = in[i * 3 + 2]; + if(mode->key_defined && *r == mode->key_r && *g == mode->key_g && *b == mode->key_b) *a = 0; + else *a = 255; + } + else + { + *r = in[i * 6 + 0]; + *g = in[i * 6 + 2]; + *b = in[i * 6 + 4]; + if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r + && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g + && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; + else *a = 255; + } + } + else if(mode->colortype == LCT_PALETTE) + { + unsigned index; + if(mode->bitdepth == 8) index = in[i]; + else + { + size_t j = i * mode->bitdepth; + index = readBitsFromReversedStream(&j, in, mode->bitdepth); + } + + if(index >= mode->palettesize) + { + /*This is an error according to the PNG spec, but fix_png can ignore it*/ + if(!fix_png) return (mode->bitdepth == 8 ? 46 : 47); /*index out of palette*/ + *r = *g = *b = 0; + *a = 255; + } + else + { + *r = mode->palette[index * 4 + 0]; + *g = mode->palette[index * 4 + 1]; + *b = mode->palette[index * 4 + 2]; + *a = mode->palette[index * 4 + 3]; + } + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + if(mode->bitdepth == 8) + { + *r = *g = *b = in[i * 2 + 0]; + *a = in[i * 2 + 1]; + } + else + { + *r = *g = *b = in[i * 4 + 0]; + *a = in[i * 4 + 2]; + } + } + else if(mode->colortype == LCT_RGBA) + { + if(mode->bitdepth == 8) + { + *r = in[i * 4 + 0]; + *g = in[i * 4 + 1]; + *b = in[i * 4 + 2]; + *a = in[i * 4 + 3]; + } + else + { + *r = in[i * 8 + 0]; + *g = in[i * 8 + 2]; + *b = in[i * 8 + 4]; + *a = in[i * 8 + 6]; + } + } + + return 0; /*no error*/ +} + +/*Similar to getPixelColorRGBA8, but with all the for loops inside of the color +mode test cases, optimized to convert the colors much faster, when converting +to RGBA or RGB with 8 bit per cannel. buffer must be RGBA or RGB output with +enough memory, if has_alpha is true the output is RGBA. mode has the color mode +of the input buffer.*/ +static unsigned getPixelColorsRGBA8(unsigned char* buffer, size_t numpixels, + unsigned has_alpha, const unsigned char* in, + const LodePNGColorMode* mode, + unsigned fix_png) +{ + unsigned num_channels = has_alpha ? 4 : 3; + size_t i; + if(mode->colortype == LCT_GREY) + { + if(mode->bitdepth == 8) + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i]; + if(has_alpha) buffer[3] = mode->key_defined && in[i] == mode->key_r ? 0 : 255; + } + } + else if(mode->bitdepth == 16) + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i * 2]; + if(has_alpha) buffer[3] = mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r ? 0 : 255; + } + } + else + { + unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ + size_t j = 0; + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); + buffer[0] = buffer[1] = buffer[2] = (value * 255) / highest; + if(has_alpha) buffer[3] = mode->key_defined && value == mode->key_r ? 0 : 255; + } + } + } + else if(mode->colortype == LCT_RGB) + { + if(mode->bitdepth == 8) + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = in[i * 3 + 0]; + buffer[1] = in[i * 3 + 1]; + buffer[2] = in[i * 3 + 2]; + if(has_alpha) buffer[3] = mode->key_defined && buffer[0] == mode->key_r + && buffer[1]== mode->key_g && buffer[2] == mode->key_b ? 0 : 255; + } + } + else + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = in[i * 6 + 0]; + buffer[1] = in[i * 6 + 2]; + buffer[2] = in[i * 6 + 4]; + if(has_alpha) buffer[3] = mode->key_defined + && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r + && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g + && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b ? 0 : 255; + } + } + } + else if(mode->colortype == LCT_PALETTE) + { + unsigned index; + size_t j = 0; + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + if(mode->bitdepth == 8) index = in[i]; + else index = readBitsFromReversedStream(&j, in, mode->bitdepth); + + if(index >= mode->palettesize) + { + /*This is an error according to the PNG spec, but fix_png can ignore it*/ + if(!fix_png) return (mode->bitdepth == 8 ? 46 : 47); /*index out of palette*/ + buffer[0] = buffer[1] = buffer[2] = 0; + if(has_alpha) buffer[3] = 255; + } + else + { + buffer[0] = mode->palette[index * 4 + 0]; + buffer[1] = mode->palette[index * 4 + 1]; + buffer[2] = mode->palette[index * 4 + 2]; + if(has_alpha) buffer[3] = mode->palette[index * 4 + 3]; + } + } + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + if(mode->bitdepth == 8) + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i * 2 + 0]; + if(has_alpha) buffer[3] = in[i * 2 + 1]; + } + } + else + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = buffer[1] = buffer[2] = in[i * 4 + 0]; + if(has_alpha) buffer[3] = in[i * 4 + 2]; + } + } + } + else if(mode->colortype == LCT_RGBA) + { + if(mode->bitdepth == 8) + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = in[i * 4 + 0]; + buffer[1] = in[i * 4 + 1]; + buffer[2] = in[i * 4 + 2]; + if(has_alpha) buffer[3] = in[i * 4 + 3]; + } + } + else + { + for(i = 0; i < numpixels; i++, buffer += num_channels) + { + buffer[0] = in[i * 8 + 0]; + buffer[1] = in[i * 8 + 2]; + buffer[2] = in[i * 8 + 4]; + if(has_alpha) buffer[3] = in[i * 8 + 6]; + } + } + } + + return 0; /*no error*/ +} + +/*Get RGBA16 color of pixel with index i (y * width + x) from the raw image with +given color type, but the given color type must be 16-bit itself.*/ +static unsigned getPixelColorRGBA16(unsigned short* r, unsigned short* g, unsigned short* b, unsigned short* a, + const unsigned char* in, size_t i, const LodePNGColorMode* mode) +{ + if(mode->bitdepth != 16) return 85; /*error: this function only supports 16-bit input*/ + + if(mode->colortype == LCT_GREY) + { + *r = *g = *b = 256 * in[i * 2 + 0] + in[i * 2 + 1]; + if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; + else *a = 65535; + } + else if(mode->colortype == LCT_RGB) + { + *r = 256 * in[i * 6 + 0] + in[i * 6 + 1]; + *g = 256 * in[i * 6 + 2] + in[i * 6 + 3]; + *b = 256 * in[i * 6 + 4] + in[i * 6 + 5]; + if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r + && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g + && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; + else *a = 65535; + } + else if(mode->colortype == LCT_GREY_ALPHA) + { + *r = *g = *b = 256 * in[i * 4 + 0] + in[i * 4 + 1]; + *a = 256 * in[i * 4 + 2] + in[i * 4 + 3]; + } + else if(mode->colortype == LCT_RGBA) + { + *r = 256 * in[i * 8 + 0] + in[i * 8 + 1]; + *g = 256 * in[i * 8 + 2] + in[i * 8 + 3]; + *b = 256 * in[i * 8 + 4] + in[i * 8 + 5]; + *a = 256 * in[i * 8 + 6] + in[i * 8 + 7]; + } + else return 85; /*error: this function only supports 16-bit input, not palettes*/ + + return 0; /*no error*/ +} + +/* +converts from any color type to 24-bit or 32-bit (later maybe more supported). return value = LodePNG error code +the out buffer must have (w * h * bpp + 7) / 8 bytes, where bpp is the bits per pixel of the output color type +(lodepng_get_bpp) for < 8 bpp images, there may _not_ be padding bits at the end of scanlines. +*/ +unsigned lodepng_convert(unsigned char* out, const unsigned char* in, + LodePNGColorMode* mode_out, LodePNGColorMode* mode_in, + unsigned w, unsigned h, unsigned fix_png) +{ + unsigned error = 0; + size_t i; + ColorTree tree; + size_t numpixels = w * h; + + if(lodepng_color_mode_equal(mode_out, mode_in)) + { + size_t numbytes = lodepng_get_raw_size(w, h, mode_in); + for(i = 0; i < numbytes; i++) out[i] = in[i]; + return error; + } + + if(mode_out->colortype == LCT_PALETTE) + { + size_t palsize = 1 << mode_out->bitdepth; + if(mode_out->palettesize < palsize) palsize = mode_out->palettesize; + color_tree_init(&tree); + for(i = 0; i < palsize; i++) + { + unsigned char* p = &mode_out->palette[i * 4]; + color_tree_add(&tree, p[0], p[1], p[2], p[3], i); + } + } + + if(mode_in->bitdepth == 16 && mode_out->bitdepth == 16) + { + for(i = 0; i < numpixels; i++) + { + unsigned short r = 0, g = 0, b = 0, a = 0; + error = getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); + if(error) break; + error = rgba16ToPixel(out, i, mode_out, r, g, b, a); + if(error) break; + } + } + else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGBA) + { + error = getPixelColorsRGBA8(out, numpixels, 1, in, mode_in, fix_png); + } + else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGB) + { + error = getPixelColorsRGBA8(out, numpixels, 0, in, mode_in, fix_png); + } + else + { + unsigned char r = 0, g = 0, b = 0, a = 0; + for(i = 0; i < numpixels; i++) + { + error = getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in, fix_png); + if(error) break; + error = rgba8ToPixel(out, i, mode_out, &tree, r, g, b, a); + if(error) break; + } + } + + if(mode_out->colortype == LCT_PALETTE) + { + color_tree_cleanup(&tree); + } + + return error; +} + +#ifdef LODEPNG_COMPILE_ENCODER + +typedef struct ColorProfile +{ + unsigned char sixteenbit; /*needs more than 8 bits per channel*/ + unsigned char sixteenbit_done; + + + unsigned char colored; /*not greyscale*/ + unsigned char colored_done; + + unsigned char key; /*a color key is required, or more*/ + unsigned short key_r; /*these values are always in 16-bit bitdepth in the profile*/ + unsigned short key_g; + unsigned short key_b; + unsigned char alpha; /*alpha channel, or alpha palette, required*/ + unsigned char alpha_done; + + unsigned numcolors; + ColorTree tree; /*for listing the counted colors, up to 256*/ + unsigned char* palette; /*size 1024. Remember up to the first 256 RGBA colors*/ + unsigned maxnumcolors; /*if more than that amount counted*/ + unsigned char numcolors_done; + + unsigned greybits; /*amount of bits required for greyscale (1, 2, 4, 8). Does not take 16 bit into account.*/ + unsigned char greybits_done; + +} ColorProfile; + +static void color_profile_init(ColorProfile* profile, LodePNGColorMode* mode) +{ + profile->sixteenbit = 0; + profile->sixteenbit_done = mode->bitdepth == 16 ? 0 : 1; + + profile->colored = 0; + profile->colored_done = lodepng_is_greyscale_type(mode) ? 1 : 0; + + profile->key = 0; + profile->alpha = 0; + profile->alpha_done = lodepng_can_have_alpha(mode) ? 0 : 1; + + profile->numcolors = 0; + color_tree_init(&profile->tree); + profile->palette = (unsigned char*)lodepng_malloc(1024); + profile->maxnumcolors = 257; + if(lodepng_get_bpp(mode) <= 8) + { + int bpp = lodepng_get_bpp(mode); + profile->maxnumcolors = bpp == 1 ? 2 : (bpp == 2 ? 4 : (bpp == 4 ? 16 : 256)); + } + profile->numcolors_done = 0; + + profile->greybits = 1; + profile->greybits_done = lodepng_get_bpp(mode) == 1 ? 1 : 0; +} + +static void color_profile_cleanup(ColorProfile* profile) +{ + color_tree_cleanup(&profile->tree); + lodepng_free(profile->palette); +} + +/*function used for debug purposes with C++*/ +/*void printColorProfile(ColorProfile* p) +{ + std::cout << "sixteenbit: " << (int)p->sixteenbit << std::endl; + std::cout << "sixteenbit_done: " << (int)p->sixteenbit_done << std::endl; + std::cout << "colored: " << (int)p->colored << std::endl; + std::cout << "colored_done: " << (int)p->colored_done << std::endl; + std::cout << "key: " << (int)p->key << std::endl; + std::cout << "key_r: " << (int)p->key_r << std::endl; + std::cout << "key_g: " << (int)p->key_g << std::endl; + std::cout << "key_b: " << (int)p->key_b << std::endl; + std::cout << "alpha: " << (int)p->alpha << std::endl; + std::cout << "alpha_done: " << (int)p->alpha_done << std::endl; + std::cout << "numcolors: " << (int)p->numcolors << std::endl; + std::cout << "maxnumcolors: " << (int)p->maxnumcolors << std::endl; + std::cout << "numcolors_done: " << (int)p->numcolors_done << std::endl; + std::cout << "greybits: " << (int)p->greybits << std::endl; + std::cout << "greybits_done: " << (int)p->greybits_done << std::endl; +}*/ + +/*Returns how many bits needed to represent given value (max 8 bit)*/ +unsigned getValueRequiredBits(unsigned short value) +{ + if(value == 0 || value == 255) return 1; + /*The scaling of 2-bit and 4-bit values uses multiples of 85 and 17*/ + if(value % 17 == 0) return value % 85 == 0 ? 2 : 4; + return 8; +} + +/*profile must already have been inited with mode. +It's ok to set some parameters of profile to done already.*/ +static unsigned get_color_profile(ColorProfile* profile, + const unsigned char* in, + size_t numpixels /*must be full image size, for certain filesize based choices*/, + LodePNGColorMode* mode, + unsigned fix_png) +{ + unsigned error = 0; + size_t i; + + if(mode->bitdepth == 16) + { + for(i = 0; i < numpixels; i++) + { + unsigned short r, g, b, a; + error = getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode); + if(error) break; + + /*a color is considered good for 8-bit if the first byte and the second byte are equal, + (so if it's divisible through 257), NOT necessarily if the second byte is 0*/ + if(!profile->sixteenbit_done + && (((r & 255) != ((r >> 8) & 255)) + || ((g & 255) != ((g >> 8) & 255)) + || ((b & 255) != ((b >> 8) & 255)))) + { + profile->sixteenbit = 1; + profile->sixteenbit_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore at 16-bit*/ + profile->numcolors_done = 1; /*counting colors no longer useful, palette doesn't support 16-bit*/ + } + + if(!profile->colored_done && (r != g || r != b)) + { + profile->colored = 1; + profile->colored_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore*/ + } + + if(!profile->alpha_done && a != 65535) + { + /*only use color key if numpixels large enough to justify tRNS chunk size*/ + if(a == 0 && numpixels > 16 && !(profile->key && (r != profile->key_r || g != profile->key_g || b != profile->key_b))) + { + if(!profile->alpha && !profile->key) + { + profile->key = 1; + profile->key_r = r; + profile->key_g = g; + profile->key_b = b; + } + } + else + { + profile->alpha = 1; + profile->alpha_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore*/ + } + } + + /* Color key cannot be used if an opaque pixel also has that RGB color. */ + if(!profile->alpha_done && a == 65535 && profile->key + && r == profile->key_r && g == profile->key_g && b == profile->key_b) + { + profile->alpha = 1; + profile->alpha_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore*/ + } + + if(!profile->greybits_done) + { + /*assuming 8-bit r, this test does not care about 16-bit*/ + unsigned bits = getValueRequiredBits(r); + if(bits > profile->greybits) profile->greybits = bits; + if(profile->greybits >= 8) profile->greybits_done = 1; + } + + if(!profile->numcolors_done) + { + /*assuming 8-bit rgba, this test does not care about 16-bit*/ + if(!color_tree_has(&profile->tree, (unsigned char)r, (unsigned char)g, (unsigned char)b, (unsigned char)a)) + { + color_tree_add(&profile->tree, (unsigned char)r, (unsigned char)g, (unsigned char)b, (unsigned char)a, + profile->numcolors); + if(profile->numcolors < 256) + { + unsigned char* p = profile->palette; + unsigned i = profile->numcolors; + p[i * 4 + 0] = (unsigned char)r; + p[i * 4 + 1] = (unsigned char)g; + p[i * 4 + 2] = (unsigned char)b; + p[i * 4 + 3] = (unsigned char)a; + } + profile->numcolors++; + if(profile->numcolors >= profile->maxnumcolors) profile->numcolors_done = 1; + } + } + + if(profile->alpha_done && profile->numcolors_done + && profile->colored_done && profile->sixteenbit_done && profile->greybits_done) + { + break; + } + }; + } + else /* < 16-bit */ + { + for(i = 0; i < numpixels; i++) + { + unsigned char r = 0, g = 0, b = 0, a = 0; + error = getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode, fix_png); + if(error) break; + + if(!profile->colored_done && (r != g || r != b)) + { + profile->colored = 1; + profile->colored_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore*/ + } + + if(!profile->alpha_done && a != 255) + { + if(a == 0 && !(profile->key && (r != profile->key_r || g != profile->key_g || b != profile->key_b))) + { + if(!profile->key) + { + profile->key = 1; + profile->key_r = r; + profile->key_g = g; + profile->key_b = b; + } + } + else + { + profile->alpha = 1; + profile->alpha_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore*/ + } + } + + /* Color key cannot be used if an opaque pixel also has that RGB color. */ + if(!profile->alpha_done && a == 255 && profile->key + && r == profile->key_r && g == profile->key_g && b == profile->key_b) + { + profile->alpha = 1; + profile->alpha_done = 1; + profile->greybits_done = 1; /*greybits is not applicable anymore*/ + } + + if(!profile->greybits_done) + { + unsigned bits = getValueRequiredBits(r); + if(bits > profile->greybits) profile->greybits = bits; + if(profile->greybits >= 8) profile->greybits_done = 1; + } + + if(!profile->numcolors_done) + { + if(!color_tree_has(&profile->tree, r, g, b, a)) + { + + color_tree_add(&profile->tree, r, g, b, a, profile->numcolors); + if(profile->numcolors < 256) + { + unsigned char* p = profile->palette; + unsigned i = profile->numcolors; + p[i * 4 + 0] = r; + p[i * 4 + 1] = g; + p[i * 4 + 2] = b; + p[i * 4 + 3] = a; + } + profile->numcolors++; + if(profile->numcolors >= profile->maxnumcolors) profile->numcolors_done = 1; + } + } + + if(profile->alpha_done && profile->numcolors_done && profile->colored_done && profile->greybits_done) + { + break; + } + }; + } + + /*make the profile's key always 16-bit for consistency*/ + if(mode->bitdepth < 16) + { + /*repeat each byte twice*/ + profile->key_r *= 257; + profile->key_g *= 257; + profile->key_b *= 257; + } + + return error; +} + +static void setColorKeyFrom16bit(LodePNGColorMode* mode_out, unsigned r, unsigned g, unsigned b, unsigned bitdepth) +{ + unsigned mask = (1 << bitdepth) - 1; + mode_out->key_defined = 1; + mode_out->key_r = r & mask; + mode_out->key_g = g & mask; + mode_out->key_b = b & mask; +} + +/*updates values of mode with a potentially smaller color model. mode_out should +contain the user chosen color model, but will be overwritten with the new chosen one.*/ +static unsigned doAutoChooseColor(LodePNGColorMode* mode_out, + const unsigned char* image, unsigned w, unsigned h, LodePNGColorMode* mode_in, + LodePNGAutoConvert auto_convert) +{ + ColorProfile profile; + unsigned error = 0; + int no_nibbles = auto_convert == LAC_AUTO_NO_NIBBLES || auto_convert == LAC_AUTO_NO_NIBBLES_NO_PALETTE; + int no_palette = auto_convert == LAC_AUTO_NO_PALETTE || auto_convert == LAC_AUTO_NO_NIBBLES_NO_PALETTE; + + if(auto_convert == LAC_ALPHA) + { + if(mode_out->colortype != LCT_RGBA && mode_out->colortype != LCT_GREY_ALPHA) return 0; + } + + color_profile_init(&profile, mode_in); + if(auto_convert == LAC_ALPHA) + { + profile.colored_done = 1; + profile.greybits_done = 1; + profile.numcolors_done = 1; + profile.sixteenbit_done = 1; + } + error = get_color_profile(&profile, image, w * h, mode_in, 0 /*fix_png*/); + if(!error && auto_convert == LAC_ALPHA) + { + if(!profile.alpha) + { + mode_out->colortype = (mode_out->colortype == LCT_RGBA ? LCT_RGB : LCT_GREY); + if(profile.key) setColorKeyFrom16bit(mode_out, profile.key_r, profile.key_g, profile.key_b, mode_out->bitdepth); + } + } + else if(!error && auto_convert != LAC_ALPHA) + { + mode_out->key_defined = 0; + + if(profile.sixteenbit) + { + mode_out->bitdepth = 16; + if(profile.alpha) + { + mode_out->colortype = profile.colored ? LCT_RGBA : LCT_GREY_ALPHA; + } + else + { + mode_out->colortype = profile.colored ? LCT_RGB : LCT_GREY; + if(profile.key) setColorKeyFrom16bit(mode_out, profile.key_r, profile.key_g, profile.key_b, mode_out->bitdepth); + } + } + else /*less than 16 bits per channel*/ + { + /*don't add palette overhead if image hasn't got a lot of pixels*/ + unsigned n = profile.numcolors; + int palette_ok = !no_palette && n <= 256 && (n * 2 < w * h); + unsigned palettebits = n <= 2 ? 1 : (n <= 4 ? 2 : (n <= 16 ? 4 : 8)); + int grey_ok = !profile.colored && !profile.alpha; /*grey without alpha, with potentially low bits*/ + if(palette_ok || grey_ok) + { + if(!palette_ok || (grey_ok && profile.greybits <= palettebits)) + { + unsigned grey = profile.key_r; + mode_out->colortype = LCT_GREY; + mode_out->bitdepth = profile.greybits; + if(profile.key) setColorKeyFrom16bit(mode_out, grey, grey, grey, mode_out->bitdepth); + } + else + { + /*fill in the palette*/ + unsigned i; + unsigned char* p = profile.palette; + /*remove potential earlier palette*/ + lodepng_palette_clear(mode_out); + for(i = 0; i < profile.numcolors; i++) + { + error = lodepng_palette_add(mode_out, p[i * 4 + 0], p[i * 4 + 1], p[i * 4 + 2], p[i * 4 + 3]); + if(error) break; + } + + mode_out->colortype = LCT_PALETTE; + mode_out->bitdepth = palettebits; + } + } + else /*8-bit per channel*/ + { + mode_out->bitdepth = 8; + if(profile.alpha) + { + mode_out->colortype = profile.colored ? LCT_RGBA : LCT_GREY_ALPHA; + } + else + { + mode_out->colortype = profile.colored ? LCT_RGB : LCT_GREY /*LCT_GREY normally won't occur, already done earlier*/; + if(profile.key) setColorKeyFrom16bit(mode_out, profile.key_r, profile.key_g, profile.key_b, mode_out->bitdepth); + } + } + } + } + + color_profile_cleanup(&profile); + + if(mode_out->colortype == LCT_PALETTE && mode_in->palettesize == mode_out->palettesize) + { + /*In this case keep the palette order of the input, so that the user can choose an optimal one*/ + size_t i; + for(i = 0; i < mode_in->palettesize * 4; i++) + { + mode_out->palette[i] = mode_in->palette[i]; + } + } + + if(no_nibbles && mode_out->bitdepth < 8) + { + /*palette can keep its small amount of colors, as long as no indices use it*/ + mode_out->bitdepth = 8; + } + + return error; +} + +#endif /* #ifdef LODEPNG_COMPILE_ENCODER */ + +/* +Paeth predicter, used by PNG filter type 4 +The parameters are of type short, but should come from unsigned chars, the shorts +are only needed to make the paeth calculation correct. +*/ +static unsigned char paethPredictor(short a, short b, short c) +{ + short pa = abs(b - c); + short pb = abs(a - c); + short pc = abs(a + b - c - c); + + if(pc < pa && pc < pb) return (unsigned char)c; + else if(pb < pa) return (unsigned char)b; + else return (unsigned char)a; +} + +/*shared values used by multiple Adam7 related functions*/ + +static const unsigned ADAM7_IX[7] = { 0, 4, 0, 2, 0, 1, 0 }; /*x start values*/ +static const unsigned ADAM7_IY[7] = { 0, 0, 4, 0, 2, 0, 1 }; /*y start values*/ +static const unsigned ADAM7_DX[7] = { 8, 8, 4, 4, 2, 2, 1 }; /*x delta values*/ +static const unsigned ADAM7_DY[7] = { 8, 8, 8, 4, 4, 2, 2 }; /*y delta values*/ + +/* +Outputs various dimensions and positions in the image related to the Adam7 reduced images. +passw: output containing the width of the 7 passes +passh: output containing the height of the 7 passes +filter_passstart: output containing the index of the start and end of each + reduced image with filter bytes +padded_passstart output containing the index of the start and end of each + reduced image when without filter bytes but with padded scanlines +passstart: output containing the index of the start and end of each reduced + image without padding between scanlines, but still padding between the images +w, h: width and height of non-interlaced image +bpp: bits per pixel +"padded" is only relevant if bpp is less than 8 and a scanline or image does not + end at a full byte +*/ +static void Adam7_getpassvalues(unsigned passw[7], unsigned passh[7], size_t filter_passstart[8], + size_t padded_passstart[8], size_t passstart[8], unsigned w, unsigned h, unsigned bpp) +{ + /*the passstart values have 8 values: the 8th one indicates the byte after the end of the 7th (= last) pass*/ + unsigned i; + + /*calculate width and height in pixels of each pass*/ + for(i = 0; i < 7; i++) + { + passw[i] = (w + ADAM7_DX[i] - ADAM7_IX[i] - 1) / ADAM7_DX[i]; + passh[i] = (h + ADAM7_DY[i] - ADAM7_IY[i] - 1) / ADAM7_DY[i]; + if(passw[i] == 0) passh[i] = 0; + if(passh[i] == 0) passw[i] = 0; + } + + filter_passstart[0] = padded_passstart[0] = passstart[0] = 0; + for(i = 0; i < 7; i++) + { + /*if passw[i] is 0, it's 0 bytes, not 1 (no filtertype-byte)*/ + filter_passstart[i + 1] = filter_passstart[i] + + ((passw[i] && passh[i]) ? passh[i] * (1 + (passw[i] * bpp + 7) / 8) : 0); + /*bits padded if needed to fill full byte at end of each scanline*/ + padded_passstart[i + 1] = padded_passstart[i] + passh[i] * ((passw[i] * bpp + 7) / 8); + /*only padded at end of reduced image*/ + passstart[i + 1] = passstart[i] + (passh[i] * passw[i] * bpp + 7) / 8; + } +} + +#ifdef LODEPNG_COMPILE_DECODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / PNG Decoder / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*read the information from the header and store it in the LodePNGInfo. return value is error*/ +unsigned lodepng_inspect(unsigned* w, unsigned* h, LodePNGState* state, + const unsigned char* in, size_t insize) +{ + LodePNGInfo* info = &state->info_png; + if(insize == 0 || in == 0) + { + CERROR_RETURN_ERROR(state->error, 48); /*error: the given data is empty*/ + } + if(insize < 29) + { + CERROR_RETURN_ERROR(state->error, 27); /*error: the data length is smaller than the length of a PNG header*/ + } + + /*when decoding a new PNG image, make sure all parameters created after previous decoding are reset*/ + lodepng_info_cleanup(info); + lodepng_info_init(info); + + if(in[0] != 137 || in[1] != 80 || in[2] != 78 || in[3] != 71 + || in[4] != 13 || in[5] != 10 || in[6] != 26 || in[7] != 10) + { + CERROR_RETURN_ERROR(state->error, 28); /*error: the first 8 bytes are not the correct PNG signature*/ + } + if(in[12] != 'I' || in[13] != 'H' || in[14] != 'D' || in[15] != 'R') + { + CERROR_RETURN_ERROR(state->error, 29); /*error: it doesn't start with a IHDR chunk!*/ + } + + /*read the values given in the header*/ + *w = lodepng_read32bitInt(&in[16]); + *h = lodepng_read32bitInt(&in[20]); + info->color.bitdepth = in[24]; + info->color.colortype = (LodePNGColorType)in[25]; + info->compression_method = in[26]; + info->filter_method = in[27]; + info->interlace_method = in[28]; + + if(!state->decoder.ignore_crc) + { + unsigned CRC = lodepng_read32bitInt(&in[29]); + unsigned checksum = lodepng_crc32(&in[12], 17); + if(CRC != checksum) + { + CERROR_RETURN_ERROR(state->error, 57); /*invalid CRC*/ + } + } + + /*error: only compression method 0 is allowed in the specification*/ + if(info->compression_method != 0) CERROR_RETURN_ERROR(state->error, 32); + /*error: only filter method 0 is allowed in the specification*/ + if(info->filter_method != 0) CERROR_RETURN_ERROR(state->error, 33); + /*error: only interlace methods 0 and 1 exist in the specification*/ + if(info->interlace_method > 1) CERROR_RETURN_ERROR(state->error, 34); + + state->error = checkColorValidity(info->color.colortype, info->color.bitdepth); + return state->error; +} + +static unsigned unfilterScanline(unsigned char* recon, const unsigned char* scanline, const unsigned char* precon, + size_t bytewidth, unsigned char filterType, size_t length) +{ + /* + For PNG filter method 0 + unfilter a PNG image scanline by scanline. when the pixels are smaller than 1 byte, + the filter works byte per byte (bytewidth = 1) + precon is the previous unfiltered scanline, recon the result, scanline the current one + the incoming scanlines do NOT include the filtertype byte, that one is given in the parameter filterType instead + recon and scanline MAY be the same memory address! precon must be disjoint. + */ + + size_t i; + switch(filterType) + { + case 0: + for(i = 0; i < length; i++) recon[i] = scanline[i]; + break; + case 1: + for(i = 0; i < bytewidth; i++) recon[i] = scanline[i]; + for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + recon[i - bytewidth]; + break; + case 2: + if(precon) + { + for(i = 0; i < length; i++) recon[i] = scanline[i] + precon[i]; + } + else + { + for(i = 0; i < length; i++) recon[i] = scanline[i]; + } + break; + case 3: + if(precon) + { + for(i = 0; i < bytewidth; i++) recon[i] = scanline[i] + precon[i] / 2; + for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + ((recon[i - bytewidth] + precon[i]) / 2); + } + else + { + for(i = 0; i < bytewidth; i++) recon[i] = scanline[i]; + for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + recon[i - bytewidth] / 2; + } + break; + case 4: + if(precon) + { + for(i = 0; i < bytewidth; i++) + { + recon[i] = (scanline[i] + precon[i]); /*paethPredictor(0, precon[i], 0) is always precon[i]*/ + } + for(i = bytewidth; i < length; i++) + { + recon[i] = (scanline[i] + paethPredictor(recon[i - bytewidth], precon[i], precon[i - bytewidth])); + } + } + else + { + for(i = 0; i < bytewidth; i++) + { + recon[i] = scanline[i]; + } + for(i = bytewidth; i < length; i++) + { + /*paethPredictor(recon[i - bytewidth], 0, 0) is always recon[i - bytewidth]*/ + recon[i] = (scanline[i] + recon[i - bytewidth]); + } + } + break; + default: return 36; /*error: unexisting filter type given*/ + } + return 0; +} + +static unsigned unfilter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) +{ + /* + For PNG filter method 0 + this function unfilters a single image (e.g. without interlacing this is called once, with Adam7 seven times) + out must have enough bytes allocated already, in must have the scanlines + 1 filtertype byte per scanline + w and h are image dimensions or dimensions of reduced image, bpp is bits per pixel + in and out are allowed to be the same memory address (but aren't the same size since in has the extra filter bytes) + */ + + unsigned y; + unsigned char* prevline = 0; + + /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ + size_t bytewidth = (bpp + 7) / 8; + size_t linebytes = (w * bpp + 7) / 8; + + for(y = 0; y < h; y++) + { + size_t outindex = linebytes * y; + size_t inindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ + unsigned char filterType = in[inindex]; + + CERROR_TRY_RETURN(unfilterScanline(&out[outindex], &in[inindex + 1], prevline, bytewidth, filterType, linebytes)); + + prevline = &out[outindex]; + } + + return 0; +} + +/* +in: Adam7 interlaced image, with no padding bits between scanlines, but between + reduced images so that each reduced image starts at a byte. +out: the same pixels, but re-ordered so that they're now a non-interlaced image with size w*h +bpp: bits per pixel +out has the following size in bits: w * h * bpp. +in is possibly bigger due to padding bits between reduced images. +out must be big enough AND must be 0 everywhere if bpp < 8 in the current implementation +(because that's likely a little bit faster) +NOTE: comments about padding bits are only relevant if bpp < 8 +*/ +static void Adam7_deinterlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) +{ + unsigned passw[7], passh[7]; + size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned i; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + if(bpp >= 8) + { + for(i = 0; i < 7; i++) + { + unsigned x, y, b; + size_t bytewidth = bpp / 8; + for(y = 0; y < passh[i]; y++) + for(x = 0; x < passw[i]; x++) + { + size_t pixelinstart = passstart[i] + (y * passw[i] + x) * bytewidth; + size_t pixeloutstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; + for(b = 0; b < bytewidth; b++) + { + out[pixeloutstart + b] = in[pixelinstart + b]; + } + } + } + } + else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ + { + for(i = 0; i < 7; i++) + { + unsigned x, y, b; + unsigned ilinebits = bpp * passw[i]; + unsigned olinebits = bpp * w; + size_t obp, ibp; /*bit pointers (for out and in buffer)*/ + for(y = 0; y < passh[i]; y++) + for(x = 0; x < passw[i]; x++) + { + ibp = (8 * passstart[i]) + (y * ilinebits + x * bpp); + obp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; + for(b = 0; b < bpp; b++) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + /*note that this function assumes the out buffer is completely 0, use setBitOfReversedStream otherwise*/ + setBitOfReversedStream0(&obp, out, bit); + } + } + } + } +} + +static void removePaddingBits(unsigned char* out, const unsigned char* in, + size_t olinebits, size_t ilinebits, unsigned h) +{ + /* + After filtering there are still padding bits if scanlines have non multiple of 8 bit amounts. They need + to be removed (except at last scanline of (Adam7-reduced) image) before working with pure image buffers + for the Adam7 code, the color convert code and the output to the user. + in and out are allowed to be the same buffer, in may also be higher but still overlapping; in must + have >= ilinebits*h bits, out must have >= olinebits*h bits, olinebits must be <= ilinebits + also used to move bits after earlier such operations happened, e.g. in a sequence of reduced images from Adam7 + only useful if (ilinebits - olinebits) is a value in the range 1..7 + */ + unsigned y; + size_t diff = ilinebits - olinebits; + size_t ibp = 0, obp = 0; /*input and output bit pointers*/ + for(y = 0; y < h; y++) + { + size_t x; + for(x = 0; x < olinebits; x++) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + setBitOfReversedStream(&obp, out, bit); + } + ibp += diff; + } +} + +/*out must be buffer big enough to contain full image, and in must contain the full decompressed data from +the IDAT chunks (with filter index bytes and possible padding bits) +return value is error*/ +static unsigned postProcessScanlines(unsigned char* out, unsigned char* in, + unsigned w, unsigned h, const LodePNGInfo* info_png) +{ + /* + This function converts the filtered-padded-interlaced data into pure 2D image buffer with the PNG's colortype. + Steps: + *) if no Adam7: 1) unfilter 2) remove padding bits (= posible extra bits per scanline if bpp < 8) + *) if adam7: 1) 7x unfilter 2) 7x remove padding bits 3) Adam7_deinterlace + NOTE: the in buffer will be overwritten with intermediate data! + */ + unsigned bpp = lodepng_get_bpp(&info_png->color); + if(bpp == 0) return 31; /*error: invalid colortype*/ + + if(info_png->interlace_method == 0) + { + if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) + { + CERROR_TRY_RETURN(unfilter(in, in, w, h, bpp)); + removePaddingBits(out, in, w * bpp, ((w * bpp + 7) / 8) * 8, h); + } + /*we can immediatly filter into the out buffer, no other steps needed*/ + else CERROR_TRY_RETURN(unfilter(out, in, w, h, bpp)); + } + else /*interlace_method is 1 (Adam7)*/ + { + unsigned passw[7], passh[7]; size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned i; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + for(i = 0; i < 7; i++) + { + CERROR_TRY_RETURN(unfilter(&in[padded_passstart[i]], &in[filter_passstart[i]], passw[i], passh[i], bpp)); + /*TODO: possible efficiency improvement: if in this reduced image the bits fit nicely in 1 scanline, + move bytes instead of bits or move not at all*/ + if(bpp < 8) + { + /*remove padding bits in scanlines; after this there still may be padding + bits between the different reduced images: each reduced image still starts nicely at a byte*/ + removePaddingBits(&in[passstart[i]], &in[padded_passstart[i]], passw[i] * bpp, + ((passw[i] * bpp + 7) / 8) * 8, passh[i]); + } + } + + Adam7_deinterlace(out, in, w, h, bpp); + } + + return 0; +} + +static unsigned readChunk_PLTE(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) +{ + unsigned pos = 0, i; + if(color->palette) lodepng_free(color->palette); + color->palettesize = chunkLength / 3; + color->palette = (unsigned char*)lodepng_malloc(4 * color->palettesize); + if(!color->palette && color->palettesize) + { + color->palettesize = 0; + return 83; /*alloc fail*/ + } + if(color->palettesize > 256) return 38; /*error: palette too big*/ + + for(i = 0; i < color->palettesize; i++) + { + color->palette[4 * i + 0] = data[pos++]; /*R*/ + color->palette[4 * i + 1] = data[pos++]; /*G*/ + color->palette[4 * i + 2] = data[pos++]; /*B*/ + color->palette[4 * i + 3] = 255; /*alpha*/ + } + + return 0; /* OK */ +} + +static unsigned readChunk_tRNS(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) +{ + unsigned i; + if(color->colortype == LCT_PALETTE) + { + /*error: more alpha values given than there are palette entries*/ + if(chunkLength > color->palettesize) return 38; + + for(i = 0; i < chunkLength; i++) color->palette[4 * i + 3] = data[i]; + } + else if(color->colortype == LCT_GREY) + { + /*error: this chunk must be 2 bytes for greyscale image*/ + if(chunkLength != 2) return 30; + + color->key_defined = 1; + color->key_r = color->key_g = color->key_b = 256 * data[0] + data[1]; + } + else if(color->colortype == LCT_RGB) + { + /*error: this chunk must be 6 bytes for RGB image*/ + if(chunkLength != 6) return 41; + + color->key_defined = 1; + color->key_r = 256 * data[0] + data[1]; + color->key_g = 256 * data[2] + data[3]; + color->key_b = 256 * data[4] + data[5]; + } + else return 42; /*error: tRNS chunk not allowed for other color models*/ + + return 0; /* OK */ +} + + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +/*background color chunk (bKGD)*/ +static unsigned readChunk_bKGD(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + if(info->color.colortype == LCT_PALETTE) + { + /*error: this chunk must be 1 byte for indexed color image*/ + if(chunkLength != 1) return 43; + + info->background_defined = 1; + info->background_r = info->background_g = info->background_b = data[0]; + } + else if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) + { + /*error: this chunk must be 2 bytes for greyscale image*/ + if(chunkLength != 2) return 44; + + info->background_defined = 1; + info->background_r = info->background_g = info->background_b + = 256 * data[0] + data[1]; + } + else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) + { + /*error: this chunk must be 6 bytes for greyscale image*/ + if(chunkLength != 6) return 45; + + info->background_defined = 1; + info->background_r = 256 * data[0] + data[1]; + info->background_g = 256 * data[2] + data[3]; + info->background_b = 256 * data[4] + data[5]; + } + + return 0; /* OK */ +} + +/*text chunk (tEXt)*/ +static unsigned readChunk_tEXt(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + unsigned error = 0; + char *key = 0, *str = 0; + unsigned i; + + while(!error) /*not really a while loop, only used to break on error*/ + { + unsigned length, string2_begin; + + length = 0; + while(length < chunkLength && data[length] != 0) length++; + /*even though it's not allowed by the standard, no error is thrown if + there's no null termination char, if the text is empty*/ + if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ + + key = (char*)lodepng_malloc(length + 1); + if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ + + key[length] = 0; + for(i = 0; i < length; i++) key[i] = data[i]; + + string2_begin = length + 1; /*skip keyword null terminator*/ + + length = chunkLength < string2_begin ? 0 : chunkLength - string2_begin; + str = (char*)lodepng_malloc(length + 1); + if(!str) CERROR_BREAK(error, 83); /*alloc fail*/ + + str[length] = 0; + for(i = 0; i < length; i++) str[i] = data[string2_begin + i]; + + error = lodepng_add_text(info, key, str); + + break; + } + + lodepng_free(key); + lodepng_free(str); + + return error; +} + +/*compressed text chunk (zTXt)*/ +static unsigned readChunk_zTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, + const unsigned char* data, size_t chunkLength) +{ + unsigned error = 0; + unsigned i; + + unsigned length, string2_begin; + char *key = 0; + ucvector decoded; + + ucvector_init(&decoded); + + while(!error) /*not really a while loop, only used to break on error*/ + { + for(length = 0; length < chunkLength && data[length] != 0; length++) ; + if(length + 2 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ + if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ + + key = (char*)lodepng_malloc(length + 1); + if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ + + key[length] = 0; + for(i = 0; i < length; i++) key[i] = data[i]; + + if(data[length + 1] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ + + string2_begin = length + 2; + if(string2_begin > chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ + + length = chunkLength - string2_begin; + /*will fail if zlib error, e.g. if length is too small*/ + error = zlib_decompress(&decoded.data, &decoded.size, + (unsigned char*)(&data[string2_begin]), + length, zlibsettings); + if(error) break; + ucvector_push_back(&decoded, 0); + + error = lodepng_add_text(info, key, (char*)decoded.data); + + break; + } + + lodepng_free(key); + ucvector_cleanup(&decoded); + + return error; +} + +/*international text chunk (iTXt)*/ +static unsigned readChunk_iTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, + const unsigned char* data, size_t chunkLength) +{ + unsigned error = 0; + unsigned i; + + unsigned length, begin, compressed; + char *key = 0, *langtag = 0, *transkey = 0; + ucvector decoded; + ucvector_init(&decoded); + + while(!error) /*not really a while loop, only used to break on error*/ + { + /*Quick check if the chunk length isn't too small. Even without check + it'd still fail with other error checks below if it's too short. This just gives a different error code.*/ + if(chunkLength < 5) CERROR_BREAK(error, 30); /*iTXt chunk too short*/ + + /*read the key*/ + for(length = 0; length < chunkLength && data[length] != 0; length++) ; + if(length + 3 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination char, corrupt?*/ + if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ + + key = (char*)lodepng_malloc(length + 1); + if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ + + key[length] = 0; + for(i = 0; i < length; i++) key[i] = data[i]; + + /*read the compression method*/ + compressed = data[length + 1]; + if(data[length + 2] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ + + /*even though it's not allowed by the standard, no error is thrown if + there's no null termination char, if the text is empty for the next 3 texts*/ + + /*read the langtag*/ + begin = length + 3; + length = 0; + for(i = begin; i < chunkLength && data[i] != 0; i++) length++; + + langtag = (char*)lodepng_malloc(length + 1); + if(!langtag) CERROR_BREAK(error, 83); /*alloc fail*/ + + langtag[length] = 0; + for(i = 0; i < length; i++) langtag[i] = data[begin + i]; + + /*read the transkey*/ + begin += length + 1; + length = 0; + for(i = begin; i < chunkLength && data[i] != 0; i++) length++; + + transkey = (char*)lodepng_malloc(length + 1); + if(!transkey) CERROR_BREAK(error, 83); /*alloc fail*/ + + transkey[length] = 0; + for(i = 0; i < length; i++) transkey[i] = data[begin + i]; + + /*read the actual text*/ + begin += length + 1; + + length = chunkLength < begin ? 0 : chunkLength - begin; + + if(compressed) + { + /*will fail if zlib error, e.g. if length is too small*/ + error = zlib_decompress(&decoded.data, &decoded.size, + (unsigned char*)(&data[begin]), + length, zlibsettings); + if(error) break; + if(decoded.allocsize < decoded.size) decoded.allocsize = decoded.size; + ucvector_push_back(&decoded, 0); + } + else + { + if(!ucvector_resize(&decoded, length + 1)) CERROR_BREAK(error, 83 /*alloc fail*/); + + decoded.data[length] = 0; + for(i = 0; i < length; i++) decoded.data[i] = data[begin + i]; + } + + error = lodepng_add_itext(info, key, langtag, transkey, (char*)decoded.data); + + break; + } + + lodepng_free(key); + lodepng_free(langtag); + lodepng_free(transkey); + ucvector_cleanup(&decoded); + + return error; +} + +static unsigned readChunk_tIME(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + if(chunkLength != 7) return 73; /*invalid tIME chunk size*/ + + info->time_defined = 1; + info->time.year = 256 * data[0] + data[+ 1]; + info->time.month = data[2]; + info->time.day = data[3]; + info->time.hour = data[4]; + info->time.minute = data[5]; + info->time.second = data[6]; + + return 0; /* OK */ +} + +static unsigned readChunk_pHYs(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) +{ + if(chunkLength != 9) return 74; /*invalid pHYs chunk size*/ + + info->phys_defined = 1; + info->phys_x = 16777216 * data[0] + 65536 * data[1] + 256 * data[2] + data[3]; + info->phys_y = 16777216 * data[4] + 65536 * data[5] + 256 * data[6] + data[7]; + info->phys_unit = data[8]; + + return 0; /* OK */ +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +/*read a PNG, the result will be in the same color type as the PNG (hence "generic")*/ +static void decodeGeneric(unsigned char** out, unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize) +{ + unsigned char IEND = 0; + const unsigned char* chunk; + size_t i; + ucvector idat; /*the data from idat chunks*/ + + /*for unknown chunk order*/ + unsigned unknown = 0; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + unsigned critical_pos = 1; /*1 = after IHDR, 2 = after PLTE, 3 = after IDAT*/ +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + + /*provide some proper output values if error will happen*/ + *out = 0; + + state->error = lodepng_inspect(w, h, state, in, insize); /*reads header and resets other parameters in state->info_png*/ + if(state->error) return; + + ucvector_init(&idat); + chunk = &in[33]; /*first byte of the first chunk after the header*/ + + /*loop through the chunks, ignoring unknown chunks and stopping at IEND chunk. + IDAT data is put at the start of the in buffer*/ + while(!IEND && !state->error) + { + unsigned chunkLength; + const unsigned char* data; /*the data in the chunk*/ + + /*error: size of the in buffer too small to contain next chunk*/ + if((size_t)((chunk - in) + 12) > insize || chunk < in) CERROR_BREAK(state->error, 30); + + /*length of the data of the chunk, excluding the length bytes, chunk type and CRC bytes*/ + chunkLength = lodepng_chunk_length(chunk); + /*error: chunk length larger than the max PNG chunk size*/ + if(chunkLength > 2147483647) CERROR_BREAK(state->error, 63); + + if((size_t)((chunk - in) + chunkLength + 12) > insize || (chunk + chunkLength + 12) < in) + { + CERROR_BREAK(state->error, 64); /*error: size of the in buffer too small to contain next chunk*/ + } + + data = lodepng_chunk_data_const(chunk); + + /*IDAT chunk, containing compressed image data*/ + if(lodepng_chunk_type_equals(chunk, "IDAT")) + { + size_t oldsize = idat.size; + if(!ucvector_resize(&idat, oldsize + chunkLength)) CERROR_BREAK(state->error, 83 /*alloc fail*/); + for(i = 0; i < chunkLength; i++) idat.data[oldsize + i] = data[i]; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + critical_pos = 3; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + } + /*IEND chunk*/ + else if(lodepng_chunk_type_equals(chunk, "IEND")) + { + IEND = 1; + } + /*palette chunk (PLTE)*/ + else if(lodepng_chunk_type_equals(chunk, "PLTE")) + { + state->error = readChunk_PLTE(&state->info_png.color, data, chunkLength); + if(state->error) break; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + critical_pos = 2; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + } + /*palette transparency chunk (tRNS)*/ + else if(lodepng_chunk_type_equals(chunk, "tRNS")) + { + state->error = readChunk_tRNS(&state->info_png.color, data, chunkLength); + if(state->error) break; + } +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*background color chunk (bKGD)*/ + else if(lodepng_chunk_type_equals(chunk, "bKGD")) + { + state->error = readChunk_bKGD(&state->info_png, data, chunkLength); + if(state->error) break; + } + /*text chunk (tEXt)*/ + else if(lodepng_chunk_type_equals(chunk, "tEXt")) + { + if(state->decoder.read_text_chunks) + { + state->error = readChunk_tEXt(&state->info_png, data, chunkLength); + if(state->error) break; + } + } + /*compressed text chunk (zTXt)*/ + else if(lodepng_chunk_type_equals(chunk, "zTXt")) + { + if(state->decoder.read_text_chunks) + { + state->error = readChunk_zTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); + if(state->error) break; + } + } + /*international text chunk (iTXt)*/ + else if(lodepng_chunk_type_equals(chunk, "iTXt")) + { + if(state->decoder.read_text_chunks) + { + state->error = readChunk_iTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); + if(state->error) break; + } + } + else if(lodepng_chunk_type_equals(chunk, "tIME")) + { + state->error = readChunk_tIME(&state->info_png, data, chunkLength); + if(state->error) break; + } + else if(lodepng_chunk_type_equals(chunk, "pHYs")) + { + state->error = readChunk_pHYs(&state->info_png, data, chunkLength); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + else /*it's not an implemented chunk type, so ignore it: skip over the data*/ + { + /*error: unknown critical chunk (5th bit of first byte of chunk type is 0)*/ + if(!lodepng_chunk_ancillary(chunk)) CERROR_BREAK(state->error, 69); + + unknown = 1; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + if(state->decoder.remember_unknown_chunks) + { + state->error = lodepng_chunk_append(&state->info_png.unknown_chunks_data[critical_pos - 1], + &state->info_png.unknown_chunks_size[critical_pos - 1], chunk); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + } + + if(!state->decoder.ignore_crc && !unknown) /*check CRC if wanted, only on known chunk types*/ + { + if(lodepng_chunk_check_crc(chunk)) CERROR_BREAK(state->error, 57); /*invalid CRC*/ + } + + if(!IEND) chunk = lodepng_chunk_next_const(chunk); + } + + if(!state->error) + { + ucvector scanlines; + ucvector_init(&scanlines); + + /*maximum final image length is already reserved in the vector's length - this is not really necessary*/ + if(!ucvector_resize(&scanlines, lodepng_get_raw_size(*w, *h, &state->info_png.color) + *h)) + { + state->error = 83; /*alloc fail*/ + } + if(!state->error) + { + /*decompress with the Zlib decompressor*/ + state->error = zlib_decompress(&scanlines.data, &scanlines.size, idat.data, + idat.size, &state->decoder.zlibsettings); + } + + if(!state->error) + { + ucvector outv; + ucvector_init(&outv); + if(!ucvector_resizev(&outv, + lodepng_get_raw_size(*w, *h, &state->info_png.color), 0)) state->error = 83; /*alloc fail*/ + if(!state->error) state->error = postProcessScanlines(outv.data, scanlines.data, *w, *h, &state->info_png); + *out = outv.data; + } + ucvector_cleanup(&scanlines); + } + + ucvector_cleanup(&idat); +} + +unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize) +{ + *out = 0; + decodeGeneric(out, w, h, state, in, insize); + if(state->error) return state->error; + if(!state->decoder.color_convert || lodepng_color_mode_equal(&state->info_raw, &state->info_png.color)) + { + /*same color type, no copying or converting of data needed*/ + /*store the info_png color settings on the info_raw so that the info_raw still reflects what colortype + the raw image has to the end user*/ + if(!state->decoder.color_convert) + { + state->error = lodepng_color_mode_copy(&state->info_raw, &state->info_png.color); + if(state->error) return state->error; + } + } + else + { + /*color conversion needed; sort of copy of the data*/ + unsigned char* data = *out; + size_t outsize; + + /*TODO: check if this works according to the statement in the documentation: "The converter can convert + from greyscale input color type, to 8-bit greyscale or greyscale with alpha"*/ + if(!(state->info_raw.colortype == LCT_RGB || state->info_raw.colortype == LCT_RGBA) + && !(state->info_raw.bitdepth == 8)) + { + return 56; /*unsupported color mode conversion*/ + } + + outsize = lodepng_get_raw_size(*w, *h, &state->info_raw); + *out = (unsigned char*)lodepng_malloc(outsize); + if(!(*out)) + { + state->error = 83; /*alloc fail*/ + } + else state->error = lodepng_convert(*out, data, &state->info_raw, &state->info_png.color, *w, *h, state->decoder.fix_png); + lodepng_free(data); + } + return state->error; +} + +unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, + size_t insize, LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned error; + LodePNGState state; + lodepng_state_init(&state); + state.info_raw.colortype = colortype; + state.info_raw.bitdepth = bitdepth; + error = lodepng_decode(out, w, h, &state, in, insize); + lodepng_state_cleanup(&state); + return error; +} + +unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) +{ + return lodepng_decode_memory(out, w, h, in, insize, LCT_RGBA, 8); +} + +unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) +{ + return lodepng_decode_memory(out, w, h, in, insize, LCT_RGB, 8); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename, + LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error; + error = lodepng_load_file(&buffer, &buffersize, filename); + if(!error) error = lodepng_decode_memory(out, w, h, buffer, buffersize, colortype, bitdepth); + lodepng_free(buffer); + return error; +} + +unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) +{ + return lodepng_decode_file(out, w, h, filename, LCT_RGBA, 8); +} + +unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) +{ + return lodepng_decode_file(out, w, h, filename, LCT_RGB, 8); +} +#endif /*LODEPNG_COMPILE_DISK*/ + +void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings) +{ + settings->color_convert = 1; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + settings->read_text_chunks = 1; + settings->remember_unknown_chunks = 0; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + settings->ignore_crc = 0; + settings->fix_png = 0; + lodepng_decompress_settings_init(&settings->zlibsettings); +} + +#endif /*LODEPNG_COMPILE_DECODER*/ + +#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) + +void lodepng_state_init(LodePNGState* state) +{ +#ifdef LODEPNG_COMPILE_DECODER + lodepng_decoder_settings_init(&state->decoder); +#endif /*LODEPNG_COMPILE_DECODER*/ +#ifdef LODEPNG_COMPILE_ENCODER + lodepng_encoder_settings_init(&state->encoder); +#endif /*LODEPNG_COMPILE_ENCODER*/ + lodepng_color_mode_init(&state->info_raw); + lodepng_info_init(&state->info_png); + state->error = 1; +} + +void lodepng_state_cleanup(LodePNGState* state) +{ + lodepng_color_mode_cleanup(&state->info_raw); + lodepng_info_cleanup(&state->info_png); +} + +void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source) +{ + lodepng_state_cleanup(dest); + *dest = *source; + lodepng_color_mode_init(&dest->info_raw); + lodepng_info_init(&dest->info_png); + dest->error = lodepng_color_mode_copy(&dest->info_raw, &source->info_raw); if(dest->error) return; + dest->error = lodepng_info_copy(&dest->info_png, &source->info_png); if(dest->error) return; +} + +#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ + +#ifdef LODEPNG_COMPILE_ENCODER + +/* ////////////////////////////////////////////////////////////////////////// */ +/* / PNG Encoder / */ +/* ////////////////////////////////////////////////////////////////////////// */ + +/*chunkName must be string of 4 characters*/ +static unsigned addChunk(ucvector* out, const char* chunkName, const unsigned char* data, size_t length) +{ + CERROR_TRY_RETURN(lodepng_chunk_create(&out->data, &out->size, (unsigned)length, chunkName, data)); + out->allocsize = out->size; /*fix the allocsize again*/ + return 0; +} + +static void writeSignature(ucvector* out) +{ + /*8 bytes PNG signature, aka the magic bytes*/ + ucvector_push_back(out, 137); + ucvector_push_back(out, 80); + ucvector_push_back(out, 78); + ucvector_push_back(out, 71); + ucvector_push_back(out, 13); + ucvector_push_back(out, 10); + ucvector_push_back(out, 26); + ucvector_push_back(out, 10); +} + +static unsigned addChunk_IHDR(ucvector* out, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth, unsigned interlace_method) +{ + unsigned error = 0; + ucvector header; + ucvector_init(&header); + + lodepng_add32bitInt(&header, w); /*width*/ + lodepng_add32bitInt(&header, h); /*height*/ + ucvector_push_back(&header, (unsigned char)bitdepth); /*bit depth*/ + ucvector_push_back(&header, (unsigned char)colortype); /*color type*/ + ucvector_push_back(&header, 0); /*compression method*/ + ucvector_push_back(&header, 0); /*filter method*/ + ucvector_push_back(&header, interlace_method); /*interlace method*/ + + error = addChunk(out, "IHDR", header.data, header.size); + ucvector_cleanup(&header); + + return error; +} + +static unsigned addChunk_PLTE(ucvector* out, const LodePNGColorMode* info) +{ + unsigned error = 0; + size_t i; + ucvector PLTE; + ucvector_init(&PLTE); + for(i = 0; i < info->palettesize * 4; i++) + { + /*add all channels except alpha channel*/ + if(i % 4 != 3) ucvector_push_back(&PLTE, info->palette[i]); + } + error = addChunk(out, "PLTE", PLTE.data, PLTE.size); + ucvector_cleanup(&PLTE); + + return error; +} + +static unsigned addChunk_tRNS(ucvector* out, const LodePNGColorMode* info) +{ + unsigned error = 0; + size_t i; + ucvector tRNS; + ucvector_init(&tRNS); + if(info->colortype == LCT_PALETTE) + { + size_t amount = info->palettesize; + /*the tail of palette values that all have 255 as alpha, does not have to be encoded*/ + for(i = info->palettesize; i > 0; i--) + { + if(info->palette[4 * (i - 1) + 3] == 255) amount--; + else break; + } + /*add only alpha channel*/ + for(i = 0; i < amount; i++) ucvector_push_back(&tRNS, info->palette[4 * i + 3]); + } + else if(info->colortype == LCT_GREY) + { + if(info->key_defined) + { + ucvector_push_back(&tRNS, (unsigned char)(info->key_r / 256)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_r % 256)); + } + } + else if(info->colortype == LCT_RGB) + { + if(info->key_defined) + { + ucvector_push_back(&tRNS, (unsigned char)(info->key_r / 256)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_r % 256)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_g / 256)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_g % 256)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_b / 256)); + ucvector_push_back(&tRNS, (unsigned char)(info->key_b % 256)); + } + } + + error = addChunk(out, "tRNS", tRNS.data, tRNS.size); + ucvector_cleanup(&tRNS); + + return error; +} + +static unsigned addChunk_IDAT(ucvector* out, const unsigned char* data, size_t datasize, + LodePNGCompressSettings* zlibsettings) +{ + ucvector zlibdata; + unsigned error = 0; + + /*compress with the Zlib compressor*/ + ucvector_init(&zlibdata); + error = zlib_compress(&zlibdata.data, &zlibdata.size, data, datasize, zlibsettings); + if(!error) error = addChunk(out, "IDAT", zlibdata.data, zlibdata.size); + ucvector_cleanup(&zlibdata); + + return error; +} + +static unsigned addChunk_IEND(ucvector* out) +{ + unsigned error = 0; + error = addChunk(out, "IEND", 0, 0); + return error; +} + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + +static unsigned addChunk_tEXt(ucvector* out, const char* keyword, const char* textstring) +{ + unsigned error = 0; + size_t i; + ucvector text; + ucvector_init(&text); + for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&text, (unsigned char)keyword[i]); + if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ + ucvector_push_back(&text, 0); /*0 termination char*/ + for(i = 0; textstring[i] != 0; i++) ucvector_push_back(&text, (unsigned char)textstring[i]); + error = addChunk(out, "tEXt", text.data, text.size); + ucvector_cleanup(&text); + + return error; +} + +static unsigned addChunk_zTXt(ucvector* out, const char* keyword, const char* textstring, + LodePNGCompressSettings* zlibsettings) +{ + unsigned error = 0; + ucvector data, compressed; + size_t i, textsize = strlen(textstring); + + ucvector_init(&data); + ucvector_init(&compressed); + for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&data, (unsigned char)keyword[i]); + if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ + ucvector_push_back(&data, 0); /*0 termination char*/ + ucvector_push_back(&data, 0); /*compression method: 0*/ + + error = zlib_compress(&compressed.data, &compressed.size, + (unsigned char*)textstring, textsize, zlibsettings); + if(!error) + { + for(i = 0; i < compressed.size; i++) ucvector_push_back(&data, compressed.data[i]); + error = addChunk(out, "zTXt", data.data, data.size); + } + + ucvector_cleanup(&compressed); + ucvector_cleanup(&data); + return error; +} + +static unsigned addChunk_iTXt(ucvector* out, unsigned compressed, const char* keyword, const char* langtag, + const char* transkey, const char* textstring, LodePNGCompressSettings* zlibsettings) +{ + unsigned error = 0; + ucvector data; + size_t i, textsize = strlen(textstring); + + ucvector_init(&data); + + for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&data, (unsigned char)keyword[i]); + if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ + ucvector_push_back(&data, 0); /*null termination char*/ + ucvector_push_back(&data, compressed ? 1 : 0); /*compression flag*/ + ucvector_push_back(&data, 0); /*compression method*/ + for(i = 0; langtag[i] != 0; i++) ucvector_push_back(&data, (unsigned char)langtag[i]); + ucvector_push_back(&data, 0); /*null termination char*/ + for(i = 0; transkey[i] != 0; i++) ucvector_push_back(&data, (unsigned char)transkey[i]); + ucvector_push_back(&data, 0); /*null termination char*/ + + if(compressed) + { + ucvector compressed_data; + ucvector_init(&compressed_data); + error = zlib_compress(&compressed_data.data, &compressed_data.size, + (unsigned char*)textstring, textsize, zlibsettings); + if(!error) + { + for(i = 0; i < compressed_data.size; i++) ucvector_push_back(&data, compressed_data.data[i]); + } + ucvector_cleanup(&compressed_data); + } + else /*not compressed*/ + { + for(i = 0; textstring[i] != 0; i++) ucvector_push_back(&data, (unsigned char)textstring[i]); + } + + if(!error) error = addChunk(out, "iTXt", data.data, data.size); + ucvector_cleanup(&data); + return error; +} + +static unsigned addChunk_bKGD(ucvector* out, const LodePNGInfo* info) +{ + unsigned error = 0; + ucvector bKGD; + ucvector_init(&bKGD); + if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) + { + ucvector_push_back(&bKGD, (unsigned char)(info->background_r / 256)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); + } + else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) + { + ucvector_push_back(&bKGD, (unsigned char)(info->background_r / 256)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_g / 256)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_g % 256)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_b / 256)); + ucvector_push_back(&bKGD, (unsigned char)(info->background_b % 256)); + } + else if(info->color.colortype == LCT_PALETTE) + { + ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); /*palette index*/ + } + + error = addChunk(out, "bKGD", bKGD.data, bKGD.size); + ucvector_cleanup(&bKGD); + + return error; +} + +static unsigned addChunk_tIME(ucvector* out, const LodePNGTime* time) +{ + unsigned error = 0; + unsigned char* data = (unsigned char*)lodepng_malloc(7); + if(!data) return 83; /*alloc fail*/ + data[0] = (unsigned char)(time->year / 256); + data[1] = (unsigned char)(time->year % 256); + data[2] = time->month; + data[3] = time->day; + data[4] = time->hour; + data[5] = time->minute; + data[6] = time->second; + error = addChunk(out, "tIME", data, 7); + lodepng_free(data); + return error; +} + +static unsigned addChunk_pHYs(ucvector* out, const LodePNGInfo* info) +{ + unsigned error = 0; + ucvector data; + ucvector_init(&data); + + lodepng_add32bitInt(&data, info->phys_x); + lodepng_add32bitInt(&data, info->phys_y); + ucvector_push_back(&data, info->phys_unit); + + error = addChunk(out, "pHYs", data.data, data.size); + ucvector_cleanup(&data); + + return error; +} + +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +static void filterScanline(unsigned char* out, const unsigned char* scanline, const unsigned char* prevline, + size_t length, size_t bytewidth, unsigned char filterType) +{ + size_t i; + switch(filterType) + { + case 0: /*None*/ + for(i = 0; i < length; i++) out[i] = scanline[i]; + break; + case 1: /*Sub*/ + if(prevline) + { + for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; + for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth]; + } + else + { + for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; + for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth]; + } + break; + case 2: /*Up*/ + if(prevline) + { + for(i = 0; i < length; i++) out[i] = scanline[i] - prevline[i]; + } + else + { + for(i = 0; i < length; i++) out[i] = scanline[i]; + } + break; + case 3: /*Average*/ + if(prevline) + { + for(i = 0; i < bytewidth; i++) out[i] = scanline[i] - prevline[i] / 2; + for(i = bytewidth; i < length; i++) out[i] = scanline[i] - ((scanline[i - bytewidth] + prevline[i]) / 2); + } + else + { + for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; + for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth] / 2; + } + break; + case 4: /*Paeth*/ + if(prevline) + { + /*paethPredictor(0, prevline[i], 0) is always prevline[i]*/ + for(i = 0; i < bytewidth; i++) out[i] = (scanline[i] - prevline[i]); + for(i = bytewidth; i < length; i++) + { + out[i] = (scanline[i] - paethPredictor(scanline[i - bytewidth], prevline[i], prevline[i - bytewidth])); + } + } + else + { + for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; + /*paethPredictor(scanline[i - bytewidth], 0, 0) is always scanline[i - bytewidth]*/ + for(i = bytewidth; i < length; i++) out[i] = (scanline[i] - scanline[i - bytewidth]); + } + break; + default: return; /*unexisting filter type given*/ + } +} + +/* log2 approximation. A slight bit faster than std::log. */ +static float flog2(float f) +{ + float result = 0; + while(f > 32) { result += 4; f /= 16; } + while(f > 2) { result++; f /= 2; } + return result + 1.442695f * (f * f * f / 3 - 3 * f * f / 2 + 3 * f - 1.83333f); +} + +static unsigned filter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, + const LodePNGColorMode* info, const LodePNGEncoderSettings* settings) +{ + /* + For PNG filter method 0 + out must be a buffer with as size: h + (w * h * bpp + 7) / 8, because there are + the scanlines with 1 extra byte per scanline + */ + + unsigned bpp = lodepng_get_bpp(info); + /*the width of a scanline in bytes, not including the filter type*/ + size_t linebytes = (w * bpp + 7) / 8; + /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ + size_t bytewidth = (bpp + 7) / 8; + const unsigned char* prevline = 0; + unsigned x, y; + unsigned error = 0; + LodePNGFilterStrategy strategy = settings->filter_strategy; + + /* + There is a heuristic called the minimum sum of absolute differences heuristic, suggested by the PNG standard: + * If the image type is Palette, or the bit depth is smaller than 8, then do not filter the image (i.e. + use fixed filtering, with the filter None). + * (The other case) If the image type is Grayscale or RGB (with or without Alpha), and the bit depth is + not smaller than 8, then use adaptive filtering heuristic as follows: independently for each row, apply + all five filters and select the filter that produces the smallest sum of absolute values per row. + This heuristic is used if filter strategy is LFS_MINSUM and filter_palette_zero is true. + + If filter_palette_zero is true and filter_strategy is not LFS_MINSUM, the above heuristic is followed, + but for "the other case", whatever strategy filter_strategy is set to instead of the minimum sum + heuristic is used. + */ + if(settings->filter_palette_zero && + (info->colortype == LCT_PALETTE || info->bitdepth < 8)) strategy = LFS_ZERO; + + if(bpp == 0) return 31; /*error: invalid color type*/ + + if(strategy == LFS_ZERO) + { + for(y = 0; y < h; y++) + { + size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ + size_t inindex = linebytes * y; + out[outindex] = 0; /*filter type byte*/ + filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, 0); + prevline = &in[inindex]; + } + } + else if(strategy == LFS_MINSUM) + { + /*adaptive filtering*/ + size_t sum[5]; + ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ + size_t smallest = 0; + unsigned type, bestType = 0; + + for(type = 0; type < 5; type++) + { + ucvector_init(&attempt[type]); + if(!ucvector_resize(&attempt[type], linebytes)) return 83; /*alloc fail*/ + } + + if(!error) + { + for(y = 0; y < h; y++) + { + /*try the 5 filter types*/ + for(type = 0; type < 5; type++) + { + filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); + + /*calculate the sum of the result*/ + sum[type] = 0; + if(type == 0) + { + for(x = 0; x < linebytes; x++) sum[type] += (unsigned char)(attempt[type].data[x]); + } + else + { + for(x = 0; x < linebytes; x++) + { + /*For differences, each byte should be treated as signed, values above 127 are negative + (converted to signed char). Filtertype 0 isn't a difference though, so use unsigned there. + This means filtertype 0 is almost never chosen, but that is justified.*/ + signed char s = (signed char)(attempt[type].data[x]); + sum[type] += s < 0 ? -s : s; + } + } + + /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ + if(type == 0 || sum[type] < smallest) + { + bestType = type; + smallest = sum[type]; + } + } + + prevline = &in[y * linebytes]; + + /*now fill the out values*/ + out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ + for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; + } + } + + for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); + } + else if(strategy == LFS_ENTROPY) + { + float sum[5]; + ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ + float smallest = 0; + unsigned type, bestType = 0; + unsigned count[256]; + + for(type = 0; type < 5; type++) + { + ucvector_init(&attempt[type]); + if(!ucvector_resize(&attempt[type], linebytes)) return 83; /*alloc fail*/ + } + + for(y = 0; y < h; y++) + { + /*try the 5 filter types*/ + for(type = 0; type < 5; type++) + { + filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); + for(x = 0; x < 256; x++) count[x] = 0; + for(x = 0; x < linebytes; x++) count[attempt[type].data[x]]++; + count[type]++; /*the filter type itself is part of the scanline*/ + sum[type] = 0; + for(x = 0; x < 256; x++) + { + float p = count[x] / (float)(linebytes + 1); + sum[type] += count[x] == 0 ? 0 : flog2(1 / p) * p; + } + /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ + if(type == 0 || sum[type] < smallest) + { + bestType = type; + smallest = sum[type]; + } + } + + prevline = &in[y * linebytes]; + + /*now fill the out values*/ + out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ + for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; + } + + for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); + } + else if(strategy == LFS_PREDEFINED) + { + for(y = 0; y < h; y++) + { + size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ + size_t inindex = linebytes * y; + unsigned type = settings->predefined_filters[y]; + out[outindex] = type; /*filter type byte*/ + filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, type); + prevline = &in[inindex]; + } + } + else if(strategy == LFS_BRUTE_FORCE) + { + /*brute force filter chooser. + deflate the scanline after every filter attempt to see which one deflates best. + This is very slow and gives only slightly smaller, sometimes even larger, result*/ + size_t size[5]; + ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ + size_t smallest = 0; + unsigned type = 0, bestType = 0; + unsigned char* dummy; + LodePNGCompressSettings zlibsettings = settings->zlibsettings; + /*use fixed tree on the attempts so that the tree is not adapted to the filtertype on purpose, + to simulate the true case where the tree is the same for the whole image. Sometimes it gives + better result with dynamic tree anyway. Using the fixed tree sometimes gives worse, but in rare + cases better compression. It does make this a bit less slow, so it's worth doing this.*/ + zlibsettings.btype = 1; + /*a custom encoder likely doesn't read the btype setting and is optimized for complete PNG + images only, so disable it*/ + zlibsettings.custom_zlib = 0; + zlibsettings.custom_deflate = 0; + for(type = 0; type < 5; type++) + { + ucvector_init(&attempt[type]); + ucvector_resize(&attempt[type], linebytes); /*todo: give error if resize failed*/ + } + for(y = 0; y < h; y++) /*try the 5 filter types*/ + { + for(type = 0; type < 5; type++) + { + unsigned testsize = attempt[type].size; + /*if(testsize > 8) testsize /= 8;*/ /*it already works good enough by testing a part of the row*/ + + filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); + size[type] = 0; + dummy = 0; + zlib_compress(&dummy, &size[type], attempt[type].data, testsize, &zlibsettings); + lodepng_free(dummy); + /*check if this is smallest size (or if type == 0 it's the first case so always store the values)*/ + if(type == 0 || size[type] < smallest) + { + bestType = type; + smallest = size[type]; + } + } + prevline = &in[y * linebytes]; + out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ + for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; + } + for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); + } + else return 88; /* unknown filter strategy */ + + return error; +} + +static void addPaddingBits(unsigned char* out, const unsigned char* in, + size_t olinebits, size_t ilinebits, unsigned h) +{ + /*The opposite of the removePaddingBits function + olinebits must be >= ilinebits*/ + unsigned y; + size_t diff = olinebits - ilinebits; + size_t obp = 0, ibp = 0; /*bit pointers*/ + for(y = 0; y < h; y++) + { + size_t x; + for(x = 0; x < ilinebits; x++) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + setBitOfReversedStream(&obp, out, bit); + } + /*obp += diff; --> no, fill in some value in the padding bits too, to avoid + "Use of uninitialised value of size ###" warning from valgrind*/ + for(x = 0; x < diff; x++) setBitOfReversedStream(&obp, out, 0); + } +} + +/* +in: non-interlaced image with size w*h +out: the same pixels, but re-ordered according to PNG's Adam7 interlacing, with + no padding bits between scanlines, but between reduced images so that each + reduced image starts at a byte. +bpp: bits per pixel +there are no padding bits, not between scanlines, not between reduced images +in has the following size in bits: w * h * bpp. +out is possibly bigger due to padding bits between reduced images +NOTE: comments about padding bits are only relevant if bpp < 8 +*/ +static void Adam7_interlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) +{ + unsigned passw[7], passh[7]; + size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned i; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + if(bpp >= 8) + { + for(i = 0; i < 7; i++) + { + unsigned x, y, b; + size_t bytewidth = bpp / 8; + for(y = 0; y < passh[i]; y++) + for(x = 0; x < passw[i]; x++) + { + size_t pixelinstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; + size_t pixeloutstart = passstart[i] + (y * passw[i] + x) * bytewidth; + for(b = 0; b < bytewidth; b++) + { + out[pixeloutstart + b] = in[pixelinstart + b]; + } + } + } + } + else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ + { + for(i = 0; i < 7; i++) + { + unsigned x, y, b; + unsigned ilinebits = bpp * passw[i]; + unsigned olinebits = bpp * w; + size_t obp, ibp; /*bit pointers (for out and in buffer)*/ + for(y = 0; y < passh[i]; y++) + for(x = 0; x < passw[i]; x++) + { + ibp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; + obp = (8 * passstart[i]) + (y * ilinebits + x * bpp); + for(b = 0; b < bpp; b++) + { + unsigned char bit = readBitFromReversedStream(&ibp, in); + setBitOfReversedStream(&obp, out, bit); + } + } + } + } +} + +/*out must be buffer big enough to contain uncompressed IDAT chunk data, and in must contain the full image. +return value is error**/ +static unsigned preProcessScanlines(unsigned char** out, size_t* outsize, const unsigned char* in, + unsigned w, unsigned h, + const LodePNGInfo* info_png, const LodePNGEncoderSettings* settings) +{ + /* + This function converts the pure 2D image with the PNG's colortype, into filtered-padded-interlaced data. Steps: + *) if no Adam7: 1) add padding bits (= posible extra bits per scanline if bpp < 8) 2) filter + *) if adam7: 1) Adam7_interlace 2) 7x add padding bits 3) 7x filter + */ + unsigned bpp = lodepng_get_bpp(&info_png->color); + unsigned error = 0; + + if(info_png->interlace_method == 0) + { + *outsize = h + (h * ((w * bpp + 7) / 8)); /*image size plus an extra byte per scanline + possible padding bits*/ + *out = (unsigned char*)lodepng_malloc(*outsize); + if(!(*out) && (*outsize)) error = 83; /*alloc fail*/ + + if(!error) + { + /*non multiple of 8 bits per scanline, padding bits needed per scanline*/ + if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) + { + unsigned char* padded = (unsigned char*)lodepng_malloc(h * ((w * bpp + 7) / 8)); + if(!padded) error = 83; /*alloc fail*/ + if(!error) + { + addPaddingBits(padded, in, ((w * bpp + 7) / 8) * 8, w * bpp, h); + error = filter(*out, padded, w, h, &info_png->color, settings); + } + lodepng_free(padded); + } + else + { + /*we can immediatly filter into the out buffer, no other steps needed*/ + error = filter(*out, in, w, h, &info_png->color, settings); + } + } + } + else /*interlace_method is 1 (Adam7)*/ + { + unsigned passw[7], passh[7]; + size_t filter_passstart[8], padded_passstart[8], passstart[8]; + unsigned char* adam7; + + Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); + + *outsize = filter_passstart[7]; /*image size plus an extra byte per scanline + possible padding bits*/ + *out = (unsigned char*)lodepng_malloc(*outsize); + if(!(*out)) error = 83; /*alloc fail*/ + + adam7 = (unsigned char*)lodepng_malloc(passstart[7]); + if(!adam7 && passstart[7]) error = 83; /*alloc fail*/ + + if(!error) + { + unsigned i; + + Adam7_interlace(adam7, in, w, h, bpp); + for(i = 0; i < 7; i++) + { + if(bpp < 8) + { + unsigned char* padded = (unsigned char*)lodepng_malloc(padded_passstart[i + 1] - padded_passstart[i]); + if(!padded) ERROR_BREAK(83); /*alloc fail*/ + addPaddingBits(padded, &adam7[passstart[i]], + ((passw[i] * bpp + 7) / 8) * 8, passw[i] * bpp, passh[i]); + error = filter(&(*out)[filter_passstart[i]], padded, + passw[i], passh[i], &info_png->color, settings); + lodepng_free(padded); + } + else + { + error = filter(&(*out)[filter_passstart[i]], &adam7[padded_passstart[i]], + passw[i], passh[i], &info_png->color, settings); + } + + if(error) break; + } + } + + lodepng_free(adam7); + } + + return error; +} + +/* +palette must have 4 * palettesize bytes allocated, and given in format RGBARGBARGBARGBA... +returns 0 if the palette is opaque, +returns 1 if the palette has a single color with alpha 0 ==> color key +returns 2 if the palette is semi-translucent. +*/ +static unsigned getPaletteTranslucency(const unsigned char* palette, size_t palettesize) +{ + size_t i, key = 0; + unsigned r = 0, g = 0, b = 0; /*the value of the color with alpha 0, so long as color keying is possible*/ + for(i = 0; i < palettesize; i++) + { + if(!key && palette[4 * i + 3] == 0) + { + r = palette[4 * i + 0]; g = palette[4 * i + 1]; b = palette[4 * i + 2]; + key = 1; + i = (size_t)(-1); /*restart from beginning, to detect earlier opaque colors with key's value*/ + } + else if(palette[4 * i + 3] != 255) return 2; + /*when key, no opaque RGB may have key's RGB*/ + else if(key && r == palette[i * 4 + 0] && g == palette[i * 4 + 1] && b == palette[i * 4 + 2]) return 2; + } + return key; +} + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +static unsigned addUnknownChunks(ucvector* out, unsigned char* data, size_t datasize) +{ + unsigned char* inchunk = data; + while((size_t)(inchunk - data) < datasize) + { + CERROR_TRY_RETURN(lodepng_chunk_append(&out->data, &out->size, inchunk)); + out->allocsize = out->size; /*fix the allocsize again*/ + inchunk = lodepng_chunk_next(inchunk); + } + return 0; +} +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +unsigned lodepng_encode(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h, + LodePNGState* state) +{ + LodePNGInfo info; + ucvector outv; + unsigned char* data = 0; /*uncompressed version of the IDAT chunk data*/ + size_t datasize = 0; + + /*provide some proper output values if error will happen*/ + *out = 0; + *outsize = 0; + state->error = 0; + + lodepng_info_init(&info); + lodepng_info_copy(&info, &state->info_png); + + if((info.color.colortype == LCT_PALETTE || state->encoder.force_palette) + && (info.color.palettesize == 0 || info.color.palettesize > 256)) + { + state->error = 68; /*invalid palette size, it is only allowed to be 1-256*/ + return state->error; + } + + if(state->encoder.auto_convert != LAC_NO) + { + state->error = doAutoChooseColor(&info.color, image, w, h, &state->info_raw, + state->encoder.auto_convert); + } + if(state->error) return state->error; + + if(state->encoder.zlibsettings.windowsize > 32768) + { + CERROR_RETURN_ERROR(state->error, 60); /*error: windowsize larger than allowed*/ + } + if(state->encoder.zlibsettings.btype > 2) + { + CERROR_RETURN_ERROR(state->error, 61); /*error: unexisting btype*/ + } + if(state->info_png.interlace_method > 1) + { + CERROR_RETURN_ERROR(state->error, 71); /*error: unexisting interlace mode*/ + } + + state->error = checkColorValidity(info.color.colortype, info.color.bitdepth); + if(state->error) return state->error; /*error: unexisting color type given*/ + state->error = checkColorValidity(state->info_raw.colortype, state->info_raw.bitdepth); + if(state->error) return state->error; /*error: unexisting color type given*/ + + if(!lodepng_color_mode_equal(&state->info_raw, &info.color)) + { + unsigned char* converted; + size_t size = (w * h * lodepng_get_bpp(&info.color) + 7) / 8; + + converted = (unsigned char*)lodepng_malloc(size); + if(!converted && size) state->error = 83; /*alloc fail*/ + if(!state->error) + { + state->error = lodepng_convert(converted, image, &info.color, &state->info_raw, w, h, 0 /*fix_png*/); + } + if(!state->error) preProcessScanlines(&data, &datasize, converted, w, h, &info, &state->encoder); + lodepng_free(converted); + } + else preProcessScanlines(&data, &datasize, image, w, h, &info, &state->encoder); + + ucvector_init(&outv); + while(!state->error) /*while only executed once, to break on error*/ + { +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + size_t i; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*write signature and chunks*/ + writeSignature(&outv); + /*IHDR*/ + addChunk_IHDR(&outv, w, h, info.color.colortype, info.color.bitdepth, info.interlace_method); +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*unknown chunks between IHDR and PLTE*/ + if(info.unknown_chunks_data[0]) + { + state->error = addUnknownChunks(&outv, info.unknown_chunks_data[0], info.unknown_chunks_size[0]); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*PLTE*/ + if(info.color.colortype == LCT_PALETTE) + { + addChunk_PLTE(&outv, &info.color); + } + if(state->encoder.force_palette && (info.color.colortype == LCT_RGB || info.color.colortype == LCT_RGBA)) + { + addChunk_PLTE(&outv, &info.color); + } + /*tRNS*/ + if(info.color.colortype == LCT_PALETTE && getPaletteTranslucency(info.color.palette, info.color.palettesize) != 0) + { + addChunk_tRNS(&outv, &info.color); + } + if((info.color.colortype == LCT_GREY || info.color.colortype == LCT_RGB) && info.color.key_defined) + { + addChunk_tRNS(&outv, &info.color); + } +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*bKGD (must come between PLTE and the IDAt chunks*/ + if(info.background_defined) addChunk_bKGD(&outv, &info); + /*pHYs (must come before the IDAT chunks)*/ + if(info.phys_defined) addChunk_pHYs(&outv, &info); + + /*unknown chunks between PLTE and IDAT*/ + if(info.unknown_chunks_data[1]) + { + state->error = addUnknownChunks(&outv, info.unknown_chunks_data[1], info.unknown_chunks_size[1]); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*IDAT (multiple IDAT chunks must be consecutive)*/ + state->error = addChunk_IDAT(&outv, data, datasize, &state->encoder.zlibsettings); + if(state->error) break; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*tIME*/ + if(info.time_defined) addChunk_tIME(&outv, &info.time); + /*tEXt and/or zTXt*/ + for(i = 0; i < info.text_num; i++) + { + if(strlen(info.text_keys[i]) > 79) + { + state->error = 66; /*text chunk too large*/ + break; + } + if(strlen(info.text_keys[i]) < 1) + { + state->error = 67; /*text chunk too small*/ + break; + } + if(state->encoder.text_compression) + addChunk_zTXt(&outv, info.text_keys[i], info.text_strings[i], &state->encoder.zlibsettings); + else + addChunk_tEXt(&outv, info.text_keys[i], info.text_strings[i]); + } + /*LodePNG version id in text chunk*/ + if(state->encoder.add_id) + { + unsigned alread_added_id_text = 0; + for(i = 0; i < info.text_num; i++) + { + if(!strcmp(info.text_keys[i], "LodePNG")) + { + alread_added_id_text = 1; + break; + } + } + if(alread_added_id_text == 0) + addChunk_tEXt(&outv, "LodePNG", VERSION_STRING); /*it's shorter as tEXt than as zTXt chunk*/ + } + /*iTXt*/ + for(i = 0; i < info.itext_num; i++) + { + if(strlen(info.itext_keys[i]) > 79) + { + state->error = 66; /*text chunk too large*/ + break; + } + if(strlen(info.itext_keys[i]) < 1) + { + state->error = 67; /*text chunk too small*/ + break; + } + addChunk_iTXt(&outv, state->encoder.text_compression, + info.itext_keys[i], info.itext_langtags[i], info.itext_transkeys[i], info.itext_strings[i], + &state->encoder.zlibsettings); + } + + /*unknown chunks between IDAT and IEND*/ + if(info.unknown_chunks_data[2]) + { + state->error = addUnknownChunks(&outv, info.unknown_chunks_data[2], info.unknown_chunks_size[2]); + if(state->error) break; + } +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + /*IEND*/ + addChunk_IEND(&outv); + + break; /*this isn't really a while loop; no error happened so break out now!*/ + } + + lodepng_info_cleanup(&info); + lodepng_free(data); + /*instead of cleaning the vector up, give it to the output*/ + *out = outv.data; + *outsize = outv.size; + + return state->error; +} + +unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, const unsigned char* image, + unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned error; + LodePNGState state; + lodepng_state_init(&state); + state.info_raw.colortype = colortype; + state.info_raw.bitdepth = bitdepth; + state.info_png.color.colortype = colortype; + state.info_png.color.bitdepth = bitdepth; + lodepng_encode(out, outsize, image, w, h, &state); + error = state.error; + lodepng_state_cleanup(&state); + return error; +} + +unsigned lodepng_encode32(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGBA, 8); +} + +unsigned lodepng_encode24(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGB, 8); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned lodepng_encode_file(const char* filename, const unsigned char* image, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error = lodepng_encode_memory(&buffer, &buffersize, image, w, h, colortype, bitdepth); + if(!error) error = lodepng_save_file(buffer, buffersize, filename); + lodepng_free(buffer); + return error; +} + +unsigned lodepng_encode32_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_file(filename, image, w, h, LCT_RGBA, 8); +} + +unsigned lodepng_encode24_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) +{ + return lodepng_encode_file(filename, image, w, h, LCT_RGB, 8); +} +#endif /*LODEPNG_COMPILE_DISK*/ + +void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings) +{ + lodepng_compress_settings_init(&settings->zlibsettings); + settings->filter_palette_zero = 1; + settings->filter_strategy = LFS_MINSUM; + settings->auto_convert = LAC_AUTO; + settings->force_palette = 0; + settings->predefined_filters = 0; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + settings->add_id = 0; + settings->text_compression = 1; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} + +#endif /*LODEPNG_COMPILE_ENCODER*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ERROR_TEXT +/* +This returns the description of a numerical error code in English. This is also +the documentation of all the error codes. +*/ +const char* lodepng_error_text(unsigned code) +{ + switch(code) + { + case 0: return "no error, everything went ok"; + case 1: return "nothing done yet"; /*the Encoder/Decoder has done nothing yet, error checking makes no sense yet*/ + case 10: return "end of input memory reached without huffman end code"; /*while huffman decoding*/ + case 11: return "error in code tree made it jump outside of huffman tree"; /*while huffman decoding*/ + case 13: return "problem while processing dynamic deflate block"; + case 14: return "problem while processing dynamic deflate block"; + case 15: return "problem while processing dynamic deflate block"; + case 16: return "unexisting code while processing dynamic deflate block"; + case 17: return "end of out buffer memory reached while inflating"; + case 18: return "invalid distance code while inflating"; + case 19: return "end of out buffer memory reached while inflating"; + case 20: return "invalid deflate block BTYPE encountered while decoding"; + case 21: return "NLEN is not ones complement of LEN in a deflate block"; + /*end of out buffer memory reached while inflating: + This can happen if the inflated deflate data is longer than the amount of bytes required to fill up + all the pixels of the image, given the color depth and image dimensions. Something that doesn't + happen in a normal, well encoded, PNG image.*/ + case 22: return "end of out buffer memory reached while inflating"; + case 23: return "end of in buffer memory reached while inflating"; + case 24: return "invalid FCHECK in zlib header"; + case 25: return "invalid compression method in zlib header"; + case 26: return "FDICT encountered in zlib header while it's not used for PNG"; + case 27: return "PNG file is smaller than a PNG header"; + /*Checks the magic file header, the first 8 bytes of the PNG file*/ + case 28: return "incorrect PNG signature, it's no PNG or corrupted"; + case 29: return "first chunk is not the header chunk"; + case 30: return "chunk length too large, chunk broken off at end of file"; + case 31: return "illegal PNG color type or bpp"; + case 32: return "illegal PNG compression method"; + case 33: return "illegal PNG filter method"; + case 34: return "illegal PNG interlace method"; + case 35: return "chunk length of a chunk is too large or the chunk too small"; + case 36: return "illegal PNG filter type encountered"; + case 37: return "illegal bit depth for this color type given"; + case 38: return "the palette is too big"; /*more than 256 colors*/ + case 39: return "more palette alpha values given in tRNS chunk than there are colors in the palette"; + case 40: return "tRNS chunk has wrong size for greyscale image"; + case 41: return "tRNS chunk has wrong size for RGB image"; + case 42: return "tRNS chunk appeared while it was not allowed for this color type"; + case 43: return "bKGD chunk has wrong size for palette image"; + case 44: return "bKGD chunk has wrong size for greyscale image"; + case 45: return "bKGD chunk has wrong size for RGB image"; + /*Is the palette too small?*/ + case 46: return "a value in indexed image is larger than the palette size (bitdepth = 8)"; + /*Is the palette too small?*/ + case 47: return "a value in indexed image is larger than the palette size (bitdepth < 8)"; + /*the input data is empty, maybe a PNG file doesn't exist or is in the wrong path*/ + case 48: return "empty input or file doesn't exist"; + case 49: return "jumped past memory while generating dynamic huffman tree"; + case 50: return "jumped past memory while generating dynamic huffman tree"; + case 51: return "jumped past memory while inflating huffman block"; + case 52: return "jumped past memory while inflating"; + case 53: return "size of zlib data too small"; + case 54: return "repeat symbol in tree while there was no value symbol yet"; + /*jumped past tree while generating huffman tree, this could be when the + tree will have more leaves than symbols after generating it out of the + given lenghts. They call this an oversubscribed dynamic bit lengths tree in zlib.*/ + case 55: return "jumped past tree while generating huffman tree"; + case 56: return "given output image colortype or bitdepth not supported for color conversion"; + case 57: return "invalid CRC encountered (checking CRC can be disabled)"; + case 58: return "invalid ADLER32 encountered (checking ADLER32 can be disabled)"; + case 59: return "requested color conversion not supported"; + case 60: return "invalid window size given in the settings of the encoder (must be 0-32768)"; + case 61: return "invalid BTYPE given in the settings of the encoder (only 0, 1 and 2 are allowed)"; + /*LodePNG leaves the choice of RGB to greyscale conversion formula to the user.*/ + case 62: return "conversion from color to greyscale not supported"; + case 63: return "length of a chunk too long, max allowed for PNG is 2147483647 bytes per chunk"; /*(2^31-1)*/ + /*this would result in the inability of a deflated block to ever contain an end code. It must be at least 1.*/ + case 64: return "the length of the END symbol 256 in the Huffman tree is 0"; + case 66: return "the length of a text chunk keyword given to the encoder is longer than the maximum of 79 bytes"; + case 67: return "the length of a text chunk keyword given to the encoder is smaller than the minimum of 1 byte"; + case 68: return "tried to encode a PLTE chunk with a palette that has less than 1 or more than 256 colors"; + case 69: return "unknown chunk type with 'critical' flag encountered by the decoder"; + case 71: return "unexisting interlace mode given to encoder (must be 0 or 1)"; + case 72: return "while decoding, unexisting compression method encountering in zTXt or iTXt chunk (it must be 0)"; + case 73: return "invalid tIME chunk size"; + case 74: return "invalid pHYs chunk size"; + /*length could be wrong, or data chopped off*/ + case 75: return "no null termination char found while decoding text chunk"; + case 76: return "iTXt chunk too short to contain required bytes"; + case 77: return "integer overflow in buffer size"; + case 78: return "failed to open file for reading"; /*file doesn't exist or couldn't be opened for reading*/ + case 79: return "failed to open file for writing"; + case 80: return "tried creating a tree of 0 symbols"; + case 81: return "lazy matching at pos 0 is impossible"; + case 82: return "color conversion to palette requested while a color isn't in palette"; + case 83: return "memory allocation failed"; + case 84: return "given image too small to contain all pixels to be encoded"; + case 85: return "internal color conversion bug"; + case 86: return "impossible offset in lz77 encoding (internal bug)"; + case 87: return "must provide custom zlib function pointer if LODEPNG_COMPILE_ZLIB is not defined"; + case 88: return "invalid filter strategy given for LodePNGEncoderSettings.filter_strategy"; + case 89: return "text chunk keyword too short or long: must have size 1-79"; + } + return "unknown error code"; +} +#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ + +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* // C++ Wrapper // */ +/* ////////////////////////////////////////////////////////////////////////// */ +/* ////////////////////////////////////////////////////////////////////////// */ + + +#ifdef LODEPNG_COMPILE_CPP +namespace lodepng +{ + +#ifdef LODEPNG_COMPILE_DISK +void load_file(std::vector& buffer, const std::string& filename) +{ + std::ifstream file(filename.c_str(), std::ios::in|std::ios::binary|std::ios::ate); + + /*get filesize*/ + std::streamsize size = 0; + if(file.seekg(0, std::ios::end).good()) size = file.tellg(); + if(file.seekg(0, std::ios::beg).good()) size -= file.tellg(); + + /*read contents of the file into the vector*/ + buffer.resize(size_t(size)); + if(size > 0) file.read((char*)(&buffer[0]), size); +} + +/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ +void save_file(const std::vector& buffer, const std::string& filename) +{ + std::ofstream file(filename.c_str(), std::ios::out|std::ios::binary); + file.write(buffer.empty() ? 0 : (char*)&buffer[0], std::streamsize(buffer.size())); +} +#endif //LODEPNG_COMPILE_DISK + +#ifdef LODEPNG_COMPILE_ZLIB +#ifdef LODEPNG_COMPILE_DECODER +unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGDecompressSettings& settings) +{ + unsigned char* buffer = 0; + size_t buffersize = 0; + unsigned error = zlib_decompress(&buffer, &buffersize, in, insize, &settings); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned decompress(std::vector& out, const std::vector& in, + const LodePNGDecompressSettings& settings) +{ + return decompress(out, in.empty() ? 0 : &in[0], in.size(), settings); +} +#endif //LODEPNG_COMPILE_DECODER + +#ifdef LODEPNG_COMPILE_ENCODER +unsigned compress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGCompressSettings& settings) +{ + unsigned char* buffer = 0; + size_t buffersize = 0; + unsigned error = zlib_compress(&buffer, &buffersize, in, insize, &settings); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned compress(std::vector& out, const std::vector& in, + const LodePNGCompressSettings& settings) +{ + return compress(out, in.empty() ? 0 : &in[0], in.size(), settings); +} +#endif //LODEPNG_COMPILE_ENCODER +#endif //LODEPNG_COMPILE_ZLIB + + +#ifdef LODEPNG_COMPILE_PNG + +State::State() +{ + lodepng_state_init(this); +} + +State::State(const State& other) +{ + lodepng_state_init(this); + lodepng_state_copy(this, &other); +} + +State::~State() +{ + lodepng_state_cleanup(this); +} + +State& State::operator=(const State& other) +{ + lodepng_state_copy(this, &other); + return *this; +} + +#ifdef LODEPNG_COMPILE_DECODER + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, const unsigned char* in, + size_t insize, LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + unsigned error = lodepng_decode_memory(&buffer, &w, &h, in, insize, colortype, bitdepth); + if(buffer && !error) + { + State state; + state.info_raw.colortype = colortype; + state.info_raw.bitdepth = bitdepth; + size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const std::vector& in, LodePNGColorType colortype, unsigned bitdepth) +{ + return decode(out, w, h, in.empty() ? 0 : &in[0], (unsigned)in.size(), colortype, bitdepth); +} + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const unsigned char* in, size_t insize) +{ + unsigned char* buffer; + unsigned error = lodepng_decode(&buffer, &w, &h, &state, in, insize); + if(buffer && !error) + { + size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const std::vector& in) +{ + return decode(out, w, h, state, in.empty() ? 0 : &in[0], in.size()); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned decode(std::vector& out, unsigned& w, unsigned& h, const std::string& filename, + LodePNGColorType colortype, unsigned bitdepth) +{ + std::vector buffer; + load_file(buffer, filename); + return decode(out, w, h, buffer, colortype, bitdepth); +} +#endif //LODEPNG_COMPILE_DECODER +#endif //LODEPNG_COMPILE_DISK + +#ifdef LODEPNG_COMPILE_ENCODER +unsigned encode(std::vector& out, const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error = lodepng_encode_memory(&buffer, &buffersize, in, w, h, colortype, bitdepth); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; + return encode(out, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); +} + +unsigned encode(std::vector& out, + const unsigned char* in, unsigned w, unsigned h, + State& state) +{ + unsigned char* buffer; + size_t buffersize; + unsigned error = lodepng_encode(&buffer, &buffersize, in, w, h, &state); + if(buffer) + { + out.insert(out.end(), &buffer[0], &buffer[buffersize]); + lodepng_free(buffer); + } + return error; +} + +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + State& state) +{ + if(lodepng_get_raw_size(w, h, &state.info_raw) > in.size()) return 84; + return encode(out, in.empty() ? 0 : &in[0], w, h, state); +} + +#ifdef LODEPNG_COMPILE_DISK +unsigned encode(const std::string& filename, + const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + std::vector buffer; + unsigned error = encode(buffer, in, w, h, colortype, bitdepth); + if(!error) save_file(buffer, filename); + return error; +} + +unsigned encode(const std::string& filename, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth) +{ + if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; + return encode(filename, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); +} +#endif //LODEPNG_COMPILE_DISK +#endif //LODEPNG_COMPILE_ENCODER +#endif //LODEPNG_COMPILE_PNG +} //namespace lodepng +#endif /*LODEPNG_COMPILE_CPP*/ diff --git a/lodepng/lodepng.h b/lodepng/lodepng.h new file mode 100644 index 0000000000..3c1c392820 --- /dev/null +++ b/lodepng/lodepng.h @@ -0,0 +1,1705 @@ +/* +LodePNG version 20130831 + +Copyright (c) 2005-2013 Lode Vandevenne + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + + 1. The origin of this software must not be misrepresented; you must not + claim that you wrote the original software. If you use this software + in a product, an acknowledgment in the product documentation would be + appreciated but is not required. + + 2. Altered source versions must be plainly marked as such, and must not be + misrepresented as being the original software. + + 3. This notice may not be removed or altered from any source + distribution. +*/ + +#ifndef LODEPNG_H +#define LODEPNG_H + +#include /*for size_t*/ + +#ifdef __cplusplus +#include +#include +#endif /*__cplusplus*/ + +/* +The following #defines are used to create code sections. They can be disabled +to disable code sections, which can give faster compile time and smaller binary. +The "NO_COMPILE" defines are designed to be used to pass as defines to the +compiler command to disable them without modifying this header, e.g. +-DLODEPNG_NO_COMPILE_ZLIB for gcc. +*/ +/*deflate & zlib. If disabled, you must specify alternative zlib functions in +the custom_zlib field of the compress and decompress settings*/ +#ifndef LODEPNG_NO_COMPILE_ZLIB +#define LODEPNG_COMPILE_ZLIB +#endif +/*png encoder and png decoder*/ +#ifndef LODEPNG_NO_COMPILE_PNG +#define LODEPNG_COMPILE_PNG +#endif +/*deflate&zlib decoder and png decoder*/ +#ifndef LODEPNG_NO_COMPILE_DECODER +#define LODEPNG_COMPILE_DECODER +#endif +/*deflate&zlib encoder and png encoder*/ +#ifndef LODEPNG_NO_COMPILE_ENCODER +#define LODEPNG_COMPILE_ENCODER +#endif +/*the optional built in harddisk file loading and saving functions*/ +#ifndef LODEPNG_NO_COMPILE_DISK +#define LODEPNG_COMPILE_DISK +#endif +/*support for chunks other than IHDR, IDAT, PLTE, tRNS, IEND: ancillary and unknown chunks*/ +#ifndef LODEPNG_NO_COMPILE_ANCILLARY_CHUNKS +#define LODEPNG_COMPILE_ANCILLARY_CHUNKS +#endif +/*ability to convert error numerical codes to English text string*/ +#ifndef LODEPNG_NO_COMPILE_ERROR_TEXT +#define LODEPNG_COMPILE_ERROR_TEXT +#endif +/*Compile the default allocators (C's free, malloc and realloc). If you disable this, +you can define the functions lodepng_free, lodepng_malloc and lodepng_realloc in your +source files with custom allocators.*/ +#ifndef LODEPNG_NO_COMPILE_ALLOCATORS +#define LODEPNG_COMPILE_ALLOCATORS +#endif +/*compile the C++ version (you can disable the C++ wrapper here even when compiling for C++)*/ +#ifdef __cplusplus +#ifndef LODEPNG_NO_COMPILE_CPP +#define LODEPNG_COMPILE_CPP +#endif +#endif + +#ifdef LODEPNG_COMPILE_PNG +/*The PNG color types (also used for raw).*/ +typedef enum LodePNGColorType +{ + LCT_GREY = 0, /*greyscale: 1,2,4,8,16 bit*/ + LCT_RGB = 2, /*RGB: 8,16 bit*/ + LCT_PALETTE = 3, /*palette: 1,2,4,8 bit*/ + LCT_GREY_ALPHA = 4, /*greyscale with alpha: 8,16 bit*/ + LCT_RGBA = 6 /*RGB with alpha: 8,16 bit*/ +} LodePNGColorType; + +#ifdef LODEPNG_COMPILE_DECODER +/* +Converts PNG data in memory to raw pixel data. +out: Output parameter. Pointer to buffer that will contain the raw pixel data. + After decoding, its size is w * h * (bytes per pixel) bytes larger than + initially. Bytes per pixel depends on colortype and bitdepth. + Must be freed after usage with free(*out). + Note: for 16-bit per channel colors, uses big endian format like PNG does. +w: Output parameter. Pointer to width of pixel data. +h: Output parameter. Pointer to height of pixel data. +in: Memory buffer with the PNG file. +insize: size of the in buffer. +colortype: the desired color type for the raw output image. See explanation on PNG color types. +bitdepth: the desired bit depth for the raw output image. See explanation on PNG color types. +Return value: LodePNG error code (0 means no error). +*/ +unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, + const unsigned char* in, size_t insize, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_decode_memory, but always decodes to 32-bit RGBA raw image*/ +unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, + const unsigned char* in, size_t insize); + +/*Same as lodepng_decode_memory, but always decodes to 24-bit RGB raw image*/ +unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, + const unsigned char* in, size_t insize); + +#ifdef LODEPNG_COMPILE_DISK +/* +Load PNG from disk, from file with given name. +Same as the other decode functions, but instead takes a filename as input. +*/ +unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, + const char* filename, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_decode_file, but always decodes to 32-bit RGBA raw image.*/ +unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, + const char* filename); + +/*Same as lodepng_decode_file, but always decodes to 24-bit RGB raw image.*/ +unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, + const char* filename); +#endif /*LODEPNG_COMPILE_DISK*/ +#endif /*LODEPNG_COMPILE_DECODER*/ + + +#ifdef LODEPNG_COMPILE_ENCODER +/* +Converts raw pixel data into a PNG image in memory. The colortype and bitdepth + of the output PNG image cannot be chosen, they are automatically determined + by the colortype, bitdepth and content of the input pixel data. + Note: for 16-bit per channel colors, needs big endian format like PNG does. +out: Output parameter. Pointer to buffer that will contain the PNG image data. + Must be freed after usage with free(*out). +outsize: Output parameter. Pointer to the size in bytes of the out buffer. +image: The raw pixel data to encode. The size of this buffer should be + w * h * (bytes per pixel), bytes per pixel depends on colortype and bitdepth. +w: width of the raw pixel data in pixels. +h: height of the raw pixel data in pixels. +colortype: the color type of the raw input image. See explanation on PNG color types. +bitdepth: the bit depth of the raw input image. See explanation on PNG color types. +Return value: LodePNG error code (0 means no error). +*/ +unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_encode_memory, but always encodes from 32-bit RGBA raw image.*/ +unsigned lodepng_encode32(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h); + +/*Same as lodepng_encode_memory, but always encodes from 24-bit RGB raw image.*/ +unsigned lodepng_encode24(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h); + +#ifdef LODEPNG_COMPILE_DISK +/* +Converts raw pixel data into a PNG file on disk. +Same as the other encode functions, but instead takes a filename as output. +NOTE: This overwrites existing files without warning! +*/ +unsigned lodepng_encode_file(const char* filename, + const unsigned char* image, unsigned w, unsigned h, + LodePNGColorType colortype, unsigned bitdepth); + +/*Same as lodepng_encode_file, but always encodes from 32-bit RGBA raw image.*/ +unsigned lodepng_encode32_file(const char* filename, + const unsigned char* image, unsigned w, unsigned h); + +/*Same as lodepng_encode_file, but always encodes from 24-bit RGB raw image.*/ +unsigned lodepng_encode24_file(const char* filename, + const unsigned char* image, unsigned w, unsigned h); +#endif /*LODEPNG_COMPILE_DISK*/ +#endif /*LODEPNG_COMPILE_ENCODER*/ + + +#ifdef LODEPNG_COMPILE_CPP +namespace lodepng +{ +#ifdef LODEPNG_COMPILE_DECODER +/*Same as lodepng_decode_memory, but decodes to an std::vector.*/ +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const unsigned char* in, size_t insize, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const std::vector& in, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#ifdef LODEPNG_COMPILE_DISK +/* +Converts PNG file from disk to raw pixel data in memory. +Same as the other decode functions, but instead takes a filename as input. +*/ +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + const std::string& filename, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#endif //LODEPNG_COMPILE_DISK +#endif //LODEPNG_COMPILE_DECODER + +#ifdef LODEPNG_COMPILE_ENCODER +/*Same as lodepng_encode_memory, but encodes to an std::vector.*/ +unsigned encode(std::vector& out, + const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#ifdef LODEPNG_COMPILE_DISK +/* +Converts 32-bit RGBA raw pixel data into a PNG file on disk. +Same as the other encode functions, but instead takes a filename as output. +NOTE: This overwrites existing files without warning! +*/ +unsigned encode(const std::string& filename, + const unsigned char* in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +unsigned encode(const std::string& filename, + const std::vector& in, unsigned w, unsigned h, + LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); +#endif //LODEPNG_COMPILE_DISK +#endif //LODEPNG_COMPILE_ENCODER +} //namespace lodepng +#endif /*LODEPNG_COMPILE_CPP*/ +#endif /*LODEPNG_COMPILE_PNG*/ + +#ifdef LODEPNG_COMPILE_ERROR_TEXT +/*Returns an English description of the numerical error code.*/ +const char* lodepng_error_text(unsigned code); +#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ + +#ifdef LODEPNG_COMPILE_DECODER +/*Settings for zlib decompression*/ +typedef struct LodePNGDecompressSettings LodePNGDecompressSettings; +struct LodePNGDecompressSettings +{ + unsigned ignore_adler32; /*if 1, continue and don't give an error message if the Adler32 checksum is corrupted*/ + + /*use custom zlib decoder instead of built in one (default: null)*/ + unsigned (*custom_zlib)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGDecompressSettings*); + /*use custom deflate decoder instead of built in one (default: null) + if custom_zlib is used, custom_deflate is ignored since only the built in + zlib function will call custom_deflate*/ + unsigned (*custom_inflate)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGDecompressSettings*); + + void* custom_context; /*optional custom settings for custom functions*/ +}; + +extern const LodePNGDecompressSettings lodepng_default_decompress_settings; +void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/* +Settings for zlib compression. Tweaking these settings tweaks the balance +between speed and compression ratio. +*/ +typedef struct LodePNGCompressSettings LodePNGCompressSettings; +struct LodePNGCompressSettings /*deflate = compress*/ +{ + /*LZ77 related settings*/ + unsigned btype; /*the block type for LZ (0, 1, 2 or 3, see zlib standard). Should be 2 for proper compression.*/ + unsigned use_lz77; /*whether or not to use LZ77. Should be 1 for proper compression.*/ + unsigned windowsize; /*the maximum is 32768, higher gives more compression but is slower. Typical value: 2048.*/ + unsigned minmatch; /*mininum lz77 length. 3 is normally best, 6 can be better for some PNGs. Default: 0*/ + unsigned nicematch; /*stop searching if >= this length found. Set to 258 for best compression. Default: 128*/ + unsigned lazymatching; /*use lazy matching: better compression but a bit slower. Default: true*/ + + /*use custom zlib encoder instead of built in one (default: null)*/ + unsigned (*custom_zlib)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGCompressSettings*); + /*use custom deflate encoder instead of built in one (default: null) + if custom_zlib is used, custom_deflate is ignored since only the built in + zlib function will call custom_deflate*/ + unsigned (*custom_deflate)(unsigned char**, size_t*, + const unsigned char*, size_t, + const LodePNGCompressSettings*); + + void* custom_context; /*optional custom settings for custom functions*/ +}; + +extern const LodePNGCompressSettings lodepng_default_compress_settings; +void lodepng_compress_settings_init(LodePNGCompressSettings* settings); +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_PNG +/* +Color mode of an image. Contains all information required to decode the pixel +bits to RGBA colors. This information is the same as used in the PNG file +format, and is used both for PNG and raw image data in LodePNG. +*/ +typedef struct LodePNGColorMode +{ + /*header (IHDR)*/ + LodePNGColorType colortype; /*color type, see PNG standard or documentation further in this header file*/ + unsigned bitdepth; /*bits per sample, see PNG standard or documentation further in this header file*/ + + /* + palette (PLTE and tRNS) + + Dynamically allocated with the colors of the palette, including alpha. + When encoding a PNG, to store your colors in the palette of the LodePNGColorMode, first use + lodepng_palette_clear, then for each color use lodepng_palette_add. + If you encode an image without alpha with palette, don't forget to put value 255 in each A byte of the palette. + + When decoding, by default you can ignore this palette, since LodePNG already + fills the palette colors in the pixels of the raw RGBA output. + + The palette is only supported for color type 3. + */ + unsigned char* palette; /*palette in RGBARGBA... order. When allocated, must be either 0, or have size 1024*/ + size_t palettesize; /*palette size in number of colors (amount of bytes is 4 * palettesize)*/ + + /* + transparent color key (tRNS) + + This color uses the same bit depth as the bitdepth value in this struct, which can be 1-bit to 16-bit. + For greyscale PNGs, r, g and b will all 3 be set to the same. + + When decoding, by default you can ignore this information, since LodePNG sets + pixels with this key to transparent already in the raw RGBA output. + + The color key is only supported for color types 0 and 2. + */ + unsigned key_defined; /*is a transparent color key given? 0 = false, 1 = true*/ + unsigned key_r; /*red/greyscale component of color key*/ + unsigned key_g; /*green component of color key*/ + unsigned key_b; /*blue component of color key*/ +} LodePNGColorMode; + +/*init, cleanup and copy functions to use with this struct*/ +void lodepng_color_mode_init(LodePNGColorMode* info); +void lodepng_color_mode_cleanup(LodePNGColorMode* info); +/*return value is error code (0 means no error)*/ +unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source); + +void lodepng_palette_clear(LodePNGColorMode* info); +/*add 1 color to the palette*/ +unsigned lodepng_palette_add(LodePNGColorMode* info, + unsigned char r, unsigned char g, unsigned char b, unsigned char a); + +/*get the total amount of bits per pixel, based on colortype and bitdepth in the struct*/ +unsigned lodepng_get_bpp(const LodePNGColorMode* info); +/*get the amount of color channels used, based on colortype in the struct. +If a palette is used, it counts as 1 channel.*/ +unsigned lodepng_get_channels(const LodePNGColorMode* info); +/*is it a greyscale type? (only colortype 0 or 4)*/ +unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info); +/*has it got an alpha channel? (only colortype 2 or 6)*/ +unsigned lodepng_is_alpha_type(const LodePNGColorMode* info); +/*has it got a palette? (only colortype 3)*/ +unsigned lodepng_is_palette_type(const LodePNGColorMode* info); +/*only returns true if there is a palette and there is a value in the palette with alpha < 255. +Loops through the palette to check this.*/ +unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info); +/* +Check if the given color info indicates the possibility of having non-opaque pixels in the PNG image. +Returns true if the image can have translucent or invisible pixels (it still be opaque if it doesn't use such pixels). +Returns false if the image can only have opaque pixels. +In detail, it returns true only if it's a color type with alpha, or has a palette with non-opaque values, +or if "key_defined" is true. +*/ +unsigned lodepng_can_have_alpha(const LodePNGColorMode* info); +/*Returns the byte size of a raw image buffer with given width, height and color mode*/ +size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color); + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +/*The information of a Time chunk in PNG.*/ +typedef struct LodePNGTime +{ + unsigned year; /*2 bytes used (0-65535)*/ + unsigned month; /*1-12*/ + unsigned day; /*1-31*/ + unsigned hour; /*0-23*/ + unsigned minute; /*0-59*/ + unsigned second; /*0-60 (to allow for leap seconds)*/ +} LodePNGTime; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +/*Information about the PNG image, except pixels, width and height.*/ +typedef struct LodePNGInfo +{ + /*header (IHDR), palette (PLTE) and transparency (tRNS) chunks*/ + unsigned compression_method;/*compression method of the original file. Always 0.*/ + unsigned filter_method; /*filter method of the original file*/ + unsigned interlace_method; /*interlace method of the original file*/ + LodePNGColorMode color; /*color type and bits, palette and transparency of the PNG file*/ + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /* + suggested background color chunk (bKGD) + This color uses the same color mode as the PNG (except alpha channel), which can be 1-bit to 16-bit. + + For greyscale PNGs, r, g and b will all 3 be set to the same. When encoding + the encoder writes the red one. For palette PNGs: When decoding, the RGB value + will be stored, not a palette index. But when encoding, specify the index of + the palette in background_r, the other two are then ignored. + + The decoder does not use this background color to edit the color of pixels. + */ + unsigned background_defined; /*is a suggested background color given?*/ + unsigned background_r; /*red component of suggested background color*/ + unsigned background_g; /*green component of suggested background color*/ + unsigned background_b; /*blue component of suggested background color*/ + + /* + non-international text chunks (tEXt and zTXt) + + The char** arrays each contain num strings. The actual messages are in + text_strings, while text_keys are keywords that give a short description what + the actual text represents, e.g. Title, Author, Description, or anything else. + + A keyword is minimum 1 character and maximum 79 characters long. It's + discouraged to use a single line length longer than 79 characters for texts. + + Don't allocate these text buffers yourself. Use the init/cleanup functions + correctly and use lodepng_add_text and lodepng_clear_text. + */ + size_t text_num; /*the amount of texts in these char** buffers (there may be more texts in itext)*/ + char** text_keys; /*the keyword of a text chunk (e.g. "Comment")*/ + char** text_strings; /*the actual text*/ + + /* + international text chunks (iTXt) + Similar to the non-international text chunks, but with additional strings + "langtags" and "transkeys". + */ + size_t itext_num; /*the amount of international texts in this PNG*/ + char** itext_keys; /*the English keyword of the text chunk (e.g. "Comment")*/ + char** itext_langtags; /*language tag for this text's language, ISO/IEC 646 string, e.g. ISO 639 language tag*/ + char** itext_transkeys; /*keyword translated to the international language - UTF-8 string*/ + char** itext_strings; /*the actual international text - UTF-8 string*/ + + /*time chunk (tIME)*/ + unsigned time_defined; /*set to 1 to make the encoder generate a tIME chunk*/ + LodePNGTime time; + + /*phys chunk (pHYs)*/ + unsigned phys_defined; /*if 0, there is no pHYs chunk and the values below are undefined, if 1 else there is one*/ + unsigned phys_x; /*pixels per unit in x direction*/ + unsigned phys_y; /*pixels per unit in y direction*/ + unsigned phys_unit; /*may be 0 (unknown unit) or 1 (metre)*/ + + /* + unknown chunks + There are 3 buffers, one for each position in the PNG where unknown chunks can appear + each buffer contains all unknown chunks for that position consecutively + The 3 buffers are the unknown chunks between certain critical chunks: + 0: IHDR-PLTE, 1: PLTE-IDAT, 2: IDAT-IEND + Do not allocate or traverse this data yourself. Use the chunk traversing functions declared + later, such as lodepng_chunk_next and lodepng_chunk_append, to read/write this struct. + */ + unsigned char* unknown_chunks_data[3]; + size_t unknown_chunks_size[3]; /*size in bytes of the unknown chunks, given for protection*/ +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} LodePNGInfo; + +/*init, cleanup and copy functions to use with this struct*/ +void lodepng_info_init(LodePNGInfo* info); +void lodepng_info_cleanup(LodePNGInfo* info); +/*return value is error code (0 means no error)*/ +unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source); + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS +void lodepng_clear_text(LodePNGInfo* info); /*use this to clear the texts again after you filled them in*/ +unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str); /*push back both texts at once*/ + +void lodepng_clear_itext(LodePNGInfo* info); /*use this to clear the itexts again after you filled them in*/ +unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, + const char* transkey, const char* str); /*push back the 4 texts of 1 chunk at once*/ +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ + +/* +Converts raw buffer from one color type to another color type, based on +LodePNGColorMode structs to describe the input and output color type. +See the reference manual at the end of this header file to see which color conversions are supported. +return value = LodePNG error code (0 if all went ok, an error if the conversion isn't supported) +The out buffer must have size (w * h * bpp + 7) / 8, where bpp is the bits per pixel +of the output color type (lodepng_get_bpp) +The fix_png value works as described in struct LodePNGDecoderSettings. +Note: for 16-bit per channel colors, uses big endian format like PNG does. +*/ +unsigned lodepng_convert(unsigned char* out, const unsigned char* in, + LodePNGColorMode* mode_out, LodePNGColorMode* mode_in, + unsigned w, unsigned h, unsigned fix_png); + + +#ifdef LODEPNG_COMPILE_DECODER +/* +Settings for the decoder. This contains settings for the PNG and the Zlib +decoder, but not the Info settings from the Info structs. +*/ +typedef struct LodePNGDecoderSettings +{ + LodePNGDecompressSettings zlibsettings; /*in here is the setting to ignore Adler32 checksums*/ + + unsigned ignore_crc; /*ignore CRC checksums*/ + /* + The fix_png setting, if 1, makes the decoder tolerant towards some PNG images + that do not correctly follow the PNG specification. This only supports errors + that are fixable, were found in images that are actually used on the web, and + are silently tolerated by other decoders as well. Currently only one such fix + is implemented: if a palette index is out of bounds given the palette size, + interpret it as opaque black. + By default this value is 0, which makes it stop with an error on such images. + */ + unsigned fix_png; + unsigned color_convert; /*whether to convert the PNG to the color type you want. Default: yes*/ + +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + unsigned read_text_chunks; /*if false but remember_unknown_chunks is true, they're stored in the unknown chunks*/ + /*store all bytes from unknown chunks in the LodePNGInfo (off by default, useful for a png editor)*/ + unsigned remember_unknown_chunks; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} LodePNGDecoderSettings; + +void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/*automatically use color type with less bits per pixel if losslessly possible. Default: AUTO*/ +typedef enum LodePNGFilterStrategy +{ + /*every filter at zero*/ + LFS_ZERO, + /*Use filter that gives minumum sum, as described in the official PNG filter heuristic.*/ + LFS_MINSUM, + /*Use the filter type that gives smallest Shannon entropy for this scanline. Depending + on the image, this is better or worse than minsum.*/ + LFS_ENTROPY, + /* + Brute-force-search PNG filters by compressing each filter for each scanline. + Experimental, very slow, and only rarely gives better compression than MINSUM. + */ + LFS_BRUTE_FORCE, + /*use predefined_filters buffer: you specify the filter type for each scanline*/ + LFS_PREDEFINED +} LodePNGFilterStrategy; + +/*automatically use color type with less bits per pixel if losslessly possible. Default: LAC_AUTO*/ +typedef enum LodePNGAutoConvert +{ + LAC_NO, /*use color type user requested*/ + LAC_ALPHA, /*use color type user requested, but if only opaque pixels and RGBA or grey+alpha, use RGB or grey*/ + LAC_AUTO, /*use PNG color type that can losslessly represent the uncompressed image the smallest possible*/ + /* + like AUTO, but do not choose 1, 2 or 4 bit per pixel types. + sometimes a PNG image compresses worse if less than 8 bits per pixels. + */ + LAC_AUTO_NO_NIBBLES, + /* + like AUTO, but never choose palette color type. For small images, encoding + the palette may take more bytes than what is gained. Note that AUTO also + already prevents encoding the palette for extremely small images, but that may + not be sufficient because due to the compression it cannot predict when to + switch. + */ + LAC_AUTO_NO_PALETTE, + LAC_AUTO_NO_NIBBLES_NO_PALETTE +} LodePNGAutoConvert; + + +/*Settings for the encoder.*/ +typedef struct LodePNGEncoderSettings +{ + LodePNGCompressSettings zlibsettings; /*settings for the zlib encoder, such as window size, ...*/ + + LodePNGAutoConvert auto_convert; /*how to automatically choose output PNG color type, if at all*/ + + /*If true, follows the official PNG heuristic: if the PNG uses a palette or lower than + 8 bit depth, set all filters to zero. Otherwise use the filter_strategy. Note that to + completely follow the official PNG heuristic, filter_palette_zero must be true and + filter_strategy must be LFS_MINSUM*/ + unsigned filter_palette_zero; + /*Which filter strategy to use when not using zeroes due to filter_palette_zero. + Set filter_palette_zero to 0 to ensure always using your chosen strategy. Default: LFS_MINSUM*/ + LodePNGFilterStrategy filter_strategy; + /*used if filter_strategy is LFS_PREDEFINED. In that case, this must point to a buffer with + the same length as the amount of scanlines in the image, and each value must <= 5. You + have to cleanup this buffer, LodePNG will never free it. Don't forget that filter_palette_zero + must be set to 0 to ensure this is also used on palette or low bitdepth images.*/ + unsigned char* predefined_filters; + + /*force creating a PLTE chunk if colortype is 2 or 6 (= a suggested palette). + If colortype is 3, PLTE is _always_ created.*/ + unsigned force_palette; +#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS + /*add LodePNG identifier and version as a text chunk, for debugging*/ + unsigned add_id; + /*encode text chunks as zTXt chunks instead of tEXt chunks, and use compression in iTXt chunks*/ + unsigned text_compression; +#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ +} LodePNGEncoderSettings; + +void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings); +#endif /*LODEPNG_COMPILE_ENCODER*/ + + +#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) +/*The settings, state and information for extended encoding and decoding.*/ +typedef struct LodePNGState +{ +#ifdef LODEPNG_COMPILE_DECODER + LodePNGDecoderSettings decoder; /*the decoding settings*/ +#endif /*LODEPNG_COMPILE_DECODER*/ +#ifdef LODEPNG_COMPILE_ENCODER + LodePNGEncoderSettings encoder; /*the encoding settings*/ +#endif /*LODEPNG_COMPILE_ENCODER*/ + LodePNGColorMode info_raw; /*specifies the format in which you would like to get the raw pixel buffer*/ + LodePNGInfo info_png; /*info of the PNG image obtained after decoding*/ + unsigned error; +#ifdef LODEPNG_COMPILE_CPP + //For the lodepng::State subclass. + virtual ~LodePNGState(){} +#endif +} LodePNGState; + +/*init, cleanup and copy functions to use with this struct*/ +void lodepng_state_init(LodePNGState* state); +void lodepng_state_cleanup(LodePNGState* state); +void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source); +#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ + +#ifdef LODEPNG_COMPILE_DECODER +/* +Same as lodepng_decode_memory, but uses a LodePNGState to allow custom settings and +getting much more information about the PNG image and color mode. +*/ +unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize); + +/* +Read the PNG header, but not the actual data. This returns only the information +that is in the header chunk of the PNG, such as width, height and color type. The +information is placed in the info_png field of the LodePNGState. +*/ +unsigned lodepng_inspect(unsigned* w, unsigned* h, + LodePNGState* state, + const unsigned char* in, size_t insize); +#endif /*LODEPNG_COMPILE_DECODER*/ + + +#ifdef LODEPNG_COMPILE_ENCODER +/*This function allocates the out buffer with standard malloc and stores the size in *outsize.*/ +unsigned lodepng_encode(unsigned char** out, size_t* outsize, + const unsigned char* image, unsigned w, unsigned h, + LodePNGState* state); +#endif /*LODEPNG_COMPILE_ENCODER*/ + +/* +The lodepng_chunk functions are normally not needed, except to traverse the +unknown chunks stored in the LodePNGInfo struct, or add new ones to it. +It also allows traversing the chunks of an encoded PNG file yourself. + +PNG standard chunk naming conventions: +First byte: uppercase = critical, lowercase = ancillary +Second byte: uppercase = public, lowercase = private +Third byte: must be uppercase +Fourth byte: uppercase = unsafe to copy, lowercase = safe to copy +*/ + +/*get the length of the data of the chunk. Total chunk length has 12 bytes more.*/ +unsigned lodepng_chunk_length(const unsigned char* chunk); + +/*puts the 4-byte type in null terminated string*/ +void lodepng_chunk_type(char type[5], const unsigned char* chunk); + +/*check if the type is the given type*/ +unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type); + +/*0: it's one of the critical chunk types, 1: it's an ancillary chunk (see PNG standard)*/ +unsigned char lodepng_chunk_ancillary(const unsigned char* chunk); + +/*0: public, 1: private (see PNG standard)*/ +unsigned char lodepng_chunk_private(const unsigned char* chunk); + +/*0: the chunk is unsafe to copy, 1: the chunk is safe to copy (see PNG standard)*/ +unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk); + +/*get pointer to the data of the chunk, where the input points to the header of the chunk*/ +unsigned char* lodepng_chunk_data(unsigned char* chunk); +const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk); + +/*returns 0 if the crc is correct, 1 if it's incorrect (0 for OK as usual!)*/ +unsigned lodepng_chunk_check_crc(const unsigned char* chunk); + +/*generates the correct CRC from the data and puts it in the last 4 bytes of the chunk*/ +void lodepng_chunk_generate_crc(unsigned char* chunk); + +/*iterate to next chunks. don't use on IEND chunk, as there is no next chunk then*/ +unsigned char* lodepng_chunk_next(unsigned char* chunk); +const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk); + +/* +Appends chunk to the data in out. The given chunk should already have its chunk header. +The out variable and outlength are updated to reflect the new reallocated buffer. +Returns error code (0 if it went ok) +*/ +unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk); + +/* +Appends new chunk to out. The chunk to append is given by giving its length, type +and data separately. The type is a 4-letter string. +The out variable and outlength are updated to reflect the new reallocated buffer. +Returne error code (0 if it went ok) +*/ +unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, + const char* type, const unsigned char* data); + + +/*Calculate CRC32 of buffer*/ +unsigned lodepng_crc32(const unsigned char* buf, size_t len); +#endif /*LODEPNG_COMPILE_PNG*/ + + +#ifdef LODEPNG_COMPILE_ZLIB +/* +This zlib part can be used independently to zlib compress and decompress a +buffer. It cannot be used to create gzip files however, and it only supports the +part of zlib that is required for PNG, it does not support dictionaries. +*/ + +#ifdef LODEPNG_COMPILE_DECODER +/*Inflate a buffer. Inflate is the decompression step of deflate. Out buffer must be freed after use.*/ +unsigned lodepng_inflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings); + +/* +Decompresses Zlib data. Reallocates the out buffer and appends the data. The +data must be according to the zlib specification. +Either, *out must be NULL and *outsize must be 0, or, *out must be a valid +buffer and *outsize its size in bytes. out must be freed by user after usage. +*/ +unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGDecompressSettings* settings); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +/* +Compresses data with Zlib. Reallocates the out buffer and appends the data. +Zlib adds a small header and trailer around the deflate data. +The data is output in the format of the zlib specification. +Either, *out must be NULL and *outsize must be 0, or, *out must be a valid +buffer and *outsize its size in bytes. out must be freed by user after usage. +*/ +unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings); + +/* +Find length-limited Huffman code for given frequencies. This function is in the +public interface only for tests, it's used internally by lodepng_deflate. +*/ +unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, + size_t numcodes, unsigned maxbitlen); + +/*Compress a buffer with deflate. See RFC 1951. Out buffer must be freed after use.*/ +unsigned lodepng_deflate(unsigned char** out, size_t* outsize, + const unsigned char* in, size_t insize, + const LodePNGCompressSettings* settings); + +#endif /*LODEPNG_COMPILE_ENCODER*/ +#endif /*LODEPNG_COMPILE_ZLIB*/ + +#ifdef LODEPNG_COMPILE_DISK +/* +Load a file from disk into buffer. The function allocates the out buffer, and +after usage you should free it. +out: output parameter, contains pointer to loaded buffer. +outsize: output parameter, size of the allocated out buffer +filename: the path to the file to load +return value: error code (0 means ok) +*/ +unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename); + +/* +Save a file from buffer to disk. Warning, if it exists, this function overwrites +the file without warning! +buffer: the buffer to write +buffersize: size of the buffer to write +filename: the path to the file to save to +return value: error code (0 means ok) +*/ +unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename); +#endif /*LODEPNG_COMPILE_DISK*/ + +#ifdef LODEPNG_COMPILE_CPP +//The LodePNG C++ wrapper uses std::vectors instead of manually allocated memory buffers. +namespace lodepng +{ +#ifdef LODEPNG_COMPILE_PNG +class State : public LodePNGState +{ + public: + State(); + State(const State& other); + virtual ~State(); + State& operator=(const State& other); +}; + +#ifdef LODEPNG_COMPILE_DECODER +//Same as other lodepng::decode, but using a State for more settings and information. +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const unsigned char* in, size_t insize); +unsigned decode(std::vector& out, unsigned& w, unsigned& h, + State& state, + const std::vector& in); +#endif /*LODEPNG_COMPILE_DECODER*/ + +#ifdef LODEPNG_COMPILE_ENCODER +//Same as other lodepng::encode, but using a State for more settings and information. +unsigned encode(std::vector& out, + const unsigned char* in, unsigned w, unsigned h, + State& state); +unsigned encode(std::vector& out, + const std::vector& in, unsigned w, unsigned h, + State& state); +#endif /*LODEPNG_COMPILE_ENCODER*/ + +#ifdef LODEPNG_COMPILE_DISK +/* +Load a file from disk into an std::vector. If the vector is empty, then either +the file doesn't exist or is an empty file. +*/ +void load_file(std::vector& buffer, const std::string& filename); + +/* +Save the binary data in an std::vector to a file on disk. The file is overwritten +without warning. +*/ +void save_file(const std::vector& buffer, const std::string& filename); +#endif //LODEPNG_COMPILE_DISK +#endif //LODEPNG_COMPILE_PNG + +#ifdef LODEPNG_COMPILE_ZLIB +#ifdef LODEPNG_COMPILE_DECODER +//Zlib-decompress an unsigned char buffer +unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); + +//Zlib-decompress an std::vector +unsigned decompress(std::vector& out, const std::vector& in, + const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); +#endif //LODEPNG_COMPILE_DECODER + +#ifdef LODEPNG_COMPILE_ENCODER +//Zlib-compress an unsigned char buffer +unsigned compress(std::vector& out, const unsigned char* in, size_t insize, + const LodePNGCompressSettings& settings = lodepng_default_compress_settings); + +//Zlib-compress an std::vector +unsigned compress(std::vector& out, const std::vector& in, + const LodePNGCompressSettings& settings = lodepng_default_compress_settings); +#endif //LODEPNG_COMPILE_ENCODER +#endif //LODEPNG_COMPILE_ZLIB +} //namespace lodepng +#endif /*LODEPNG_COMPILE_CPP*/ + +/* +TODO: +[.] test if there are no memory leaks or security exploits - done a lot but needs to be checked often +[.] check compatibility with vareous compilers - done but needs to be redone for every newer version +[X] converting color to 16-bit per channel types +[ ] read all public PNG chunk types (but never let the color profile and gamma ones touch RGB values) +[ ] make sure encoder generates no chunks with size > (2^31)-1 +[ ] partial decoding (stream processing) +[X] let the "isFullyOpaque" function check color keys and transparent palettes too +[X] better name for the variables "codes", "codesD", "codelengthcodes", "clcl" and "lldl" +[ ] don't stop decoding on errors like 69, 57, 58 (make warnings) +[ ] make option to choose if the raw image with non multiple of 8 bits per scanline should have padding bits or not +[ ] let the C++ wrapper catch exceptions coming from the standard library and return LodePNG error codes +*/ + +#endif /*LODEPNG_H inclusion guard*/ + +/* +LodePNG Documentation +--------------------- + +0. table of contents +-------------------- + + 1. about + 1.1. supported features + 1.2. features not supported + 2. C and C++ version + 3. security + 4. decoding + 5. encoding + 6. color conversions + 6.1. PNG color types + 6.2. color conversions + 6.3. padding bits + 6.4. A note about 16-bits per channel and endianness + 7. error values + 8. chunks and PNG editing + 9. compiler support + 10. examples + 10.1. decoder C++ example + 10.2. decoder C example + 11. changes + 12. contact information + + +1. about +-------- + +PNG is a file format to store raster images losslessly with good compression, +supporting different color types and alpha channel. + +LodePNG is a PNG codec according to the Portable Network Graphics (PNG) +Specification (Second Edition) - W3C Recommendation 10 November 2003. + +The specifications used are: + +*) Portable Network Graphics (PNG) Specification (Second Edition): + http://www.w3.org/TR/2003/REC-PNG-20031110 +*) RFC 1950 ZLIB Compressed Data Format version 3.3: + http://www.gzip.org/zlib/rfc-zlib.html +*) RFC 1951 DEFLATE Compressed Data Format Specification ver 1.3: + http://www.gzip.org/zlib/rfc-deflate.html + +The most recent version of LodePNG can currently be found at +http://lodev.org/lodepng/ + +LodePNG works both in C (ISO C90) and C++, with a C++ wrapper that adds +extra functionality. + +LodePNG exists out of two files: +-lodepng.h: the header file for both C and C++ +-lodepng.c(pp): give it the name lodepng.c or lodepng.cpp (or .cc) depending on your usage + +If you want to start using LodePNG right away without reading this doc, get the +examples from the LodePNG website to see how to use it in code, or check the +smaller examples in chapter 13 here. + +LodePNG is simple but only supports the basic requirements. To achieve +simplicity, the following design choices were made: There are no dependencies +on any external library. There are functions to decode and encode a PNG with +a single function call, and extended versions of these functions taking a +LodePNGState struct allowing to specify or get more information. By default +the colors of the raw image are always RGB or RGBA, no matter what color type +the PNG file uses. To read and write files, there are simple functions to +convert the files to/from buffers in memory. + +This all makes LodePNG suitable for loading textures in games, demos and small +programs, ... It's less suitable for full fledged image editors, loading PNGs +over network (it requires all the image data to be available before decoding can +begin), life-critical systems, ... + +1.1. supported features +----------------------- + +The following features are supported by the decoder: + +*) decoding of PNGs with any color type, bit depth and interlace mode, to a 24- or 32-bit color raw image, + or the same color type as the PNG +*) encoding of PNGs, from any raw image to 24- or 32-bit color, or the same color type as the raw image +*) Adam7 interlace and deinterlace for any color type +*) loading the image from harddisk or decoding it from a buffer from other sources than harddisk +*) support for alpha channels, including RGBA color model, translucent palettes and color keying +*) zlib decompression (inflate) +*) zlib compression (deflate) +*) CRC32 and ADLER32 checksums +*) handling of unknown chunks, allowing making a PNG editor that stores custom and unknown chunks. +*) the following chunks are supported (generated/interpreted) by both encoder and decoder: + IHDR: header information + PLTE: color palette + IDAT: pixel data + IEND: the final chunk + tRNS: transparency for palettized images + tEXt: textual information + zTXt: compressed textual information + iTXt: international textual information + bKGD: suggested background color + pHYs: physical dimensions + tIME: modification time + +1.2. features not supported +--------------------------- + +The following features are _not_ supported: + +*) some features needed to make a conformant PNG-Editor might be still missing. +*) partial loading/stream processing. All data must be available and is processed in one call. +*) The following public chunks are not supported but treated as unknown chunks by LodePNG + cHRM, gAMA, iCCP, sRGB, sBIT, hIST, sPLT + Some of these are not supported on purpose: LodePNG wants to provide the RGB values + stored in the pixels, not values modified by system dependent gamma or color models. + + +2. C and C++ version +-------------------- + +The C version uses buffers allocated with alloc that you need to free() +yourself. You need to use init and cleanup functions for each struct whenever +using a struct from the C version to avoid exploits and memory leaks. + +The C++ version has extra functions with std::vectors in the interface and the +lodepng::State class which is a LodePNGState with constructor and destructor. + +These files work without modification for both C and C++ compilers because all +the additional C++ code is in "#ifdef __cplusplus" blocks that make C-compilers +ignore it, and the C code is made to compile both with strict ISO C90 and C++. + +To use the C++ version, you need to rename the source file to lodepng.cpp +(instead of lodepng.c), and compile it with a C++ compiler. + +To use the C version, you need to rename the source file to lodepng.c (instead +of lodepng.cpp), and compile it with a C compiler. + + +3. Security +----------- + +Even if carefully designed, it's always possible that LodePNG contains possible +exploits. If you discover one, please let me know, and it will be fixed. + +When using LodePNG, care has to be taken with the C version of LodePNG, as well +as the C-style structs when working with C++. The following conventions are used +for all C-style structs: + +-if a struct has a corresponding init function, always call the init function when making a new one +-if a struct has a corresponding cleanup function, call it before the struct disappears to avoid memory leaks +-if a struct has a corresponding copy function, use the copy function instead of "=". + The destination must also be inited already. + + +4. Decoding +----------- + +Decoding converts a PNG compressed image to a raw pixel buffer. + +Most documentation on using the decoder is at its declarations in the header +above. For C, simple decoding can be done with functions such as +lodepng_decode32, and more advanced decoding can be done with the struct +LodePNGState and lodepng_decode. For C++, all decoding can be done with the +various lodepng::decode functions, and lodepng::State can be used for advanced +features. + +When using the LodePNGState, it uses the following fields for decoding: +*) LodePNGInfo info_png: it stores extra information about the PNG (the input) in here +*) LodePNGColorMode info_raw: here you can say what color mode of the raw image (the output) you want to get +*) LodePNGDecoderSettings decoder: you can specify a few extra settings for the decoder to use + +LodePNGInfo info_png +-------------------- + +After decoding, this contains extra information of the PNG image, except the actual +pixels, width and height because these are already gotten directly from the decoder +functions. + +It contains for example the original color type of the PNG image, text comments, +suggested background color, etc... More details about the LodePNGInfo struct are +at its declaration documentation. + +LodePNGColorMode info_raw +------------------------- + +When decoding, here you can specify which color type you want +the resulting raw image to be. If this is different from the colortype of the +PNG, then the decoder will automatically convert the result. This conversion +always works, except if you want it to convert a color PNG to greyscale or to +a palette with missing colors. + +By default, 32-bit color is used for the result. + +LodePNGDecoderSettings decoder +------------------------------ + +The settings can be used to ignore the errors created by invalid CRC and Adler32 +chunks, and to disable the decoding of tEXt chunks. + +There's also a setting color_convert, true by default. If false, no conversion +is done, the resulting data will be as it was in the PNG (after decompression) +and you'll have to puzzle the colors of the pixels together yourself using the +color type information in the LodePNGInfo. + + +5. Encoding +----------- + +Encoding converts a raw pixel buffer to a PNG compressed image. + +Most documentation on using the encoder is at its declarations in the header +above. For C, simple encoding can be done with functions such as +lodepng_encode32, and more advanced decoding can be done with the struct +LodePNGState and lodepng_encode. For C++, all encoding can be done with the +various lodepng::encode functions, and lodepng::State can be used for advanced +features. + +Like the decoder, the encoder can also give errors. However it gives less errors +since the encoder input is trusted, the decoder input (a PNG image that could +be forged by anyone) is not trusted. + +When using the LodePNGState, it uses the following fields for encoding: +*) LodePNGInfo info_png: here you specify how you want the PNG (the output) to be. +*) LodePNGColorMode info_raw: here you say what color type of the raw image (the input) has +*) LodePNGEncoderSettings encoder: you can specify a few settings for the encoder to use + +LodePNGInfo info_png +-------------------- + +When encoding, you use this the opposite way as when decoding: for encoding, +you fill in the values you want the PNG to have before encoding. By default it's +not needed to specify a color type for the PNG since it's automatically chosen, +but it's possible to choose it yourself given the right settings. + +The encoder will not always exactly match the LodePNGInfo struct you give, +it tries as close as possible. Some things are ignored by the encoder. The +encoder uses, for example, the following settings from it when applicable: +colortype and bitdepth, text chunks, time chunk, the color key, the palette, the +background color, the interlace method, unknown chunks, ... + +When encoding to a PNG with colortype 3, the encoder will generate a PLTE chunk. +If the palette contains any colors for which the alpha channel is not 255 (so +there are translucent colors in the palette), it'll add a tRNS chunk. + +LodePNGColorMode info_raw +------------------------- + +You specify the color type of the raw image that you give to the input here, +including a possible transparent color key and palette you happen to be using in +your raw image data. + +By default, 32-bit color is assumed, meaning your input has to be in RGBA +format with 4 bytes (unsigned chars) per pixel. + +LodePNGEncoderSettings encoder +------------------------------ + +The following settings are supported (some are in sub-structs): +*) auto_convert: when this option is enabled, the encoder will +automatically choose the smallest possible color mode (including color key) that +can encode the colors of all pixels without information loss. +*) btype: the block type for LZ77. 0 = uncompressed, 1 = fixed huffman tree, + 2 = dynamic huffman tree (best compression). Should be 2 for proper + compression. +*) use_lz77: whether or not to use LZ77 for compressed block types. Should be + true for proper compression. +*) windowsize: the window size used by the LZ77 encoder (1 - 32768). Has value + 2048 by default, but can be set to 32768 for better, but slow, compression. +*) force_palette: if colortype is 2 or 6, you can make the encoder write a PLTE + chunk if force_palette is true. This can used as suggested palette to convert + to by viewers that don't support more than 256 colors (if those still exist) +*) add_id: add text chunk "Encoder: LodePNG " to the image. +*) text_compression: default 1. If 1, it'll store texts as zTXt instead of tEXt chunks. + zTXt chunks use zlib compression on the text. This gives a smaller result on + large texts but a larger result on small texts (such as a single program name). + It's all tEXt or all zTXt though, there's no separate setting per text yet. + + +6. color conversions +-------------------- + +An important thing to note about LodePNG, is that the color type of the PNG, and +the color type of the raw image, are completely independent. By default, when +you decode a PNG, you get the result as a raw image in the color type you want, +no matter whether the PNG was encoded with a palette, greyscale or RGBA color. +And if you encode an image, by default LodePNG will automatically choose the PNG +color type that gives good compression based on the values of colors and amount +of colors in the image. It can be configured to let you control it instead as +well, though. + +To be able to do this, LodePNG does conversions from one color mode to another. +It can convert from almost any color type to any other color type, except the +following conversions: RGB to greyscale is not supported, and converting to a +palette when the palette doesn't have a required color is not supported. This is +not supported on purpose: this is information loss which requires a color +reduction algorithm that is beyong the scope of a PNG encoder (yes, RGB to grey +is easy, but there are multiple ways if you want to give some channels more +weight). + +By default, when decoding, you get the raw image in 32-bit RGBA or 24-bit RGB +color, no matter what color type the PNG has. And by default when encoding, +LodePNG automatically picks the best color model for the output PNG, and expects +the input image to be 32-bit RGBA or 24-bit RGB. So, unless you want to control +the color format of the images yourself, you can skip this chapter. + +6.1. PNG color types +-------------------- + +A PNG image can have many color types, ranging from 1-bit color to 64-bit color, +as well as palettized color modes. After the zlib decompression and unfiltering +in the PNG image is done, the raw pixel data will have that color type and thus +a certain amount of bits per pixel. If you want the output raw image after +decoding to have another color type, a conversion is done by LodePNG. + +The PNG specification gives the following color types: + +0: greyscale, bit depths 1, 2, 4, 8, 16 +2: RGB, bit depths 8 and 16 +3: palette, bit depths 1, 2, 4 and 8 +4: greyscale with alpha, bit depths 8 and 16 +6: RGBA, bit depths 8 and 16 + +Bit depth is the amount of bits per pixel per color channel. So the total amount +of bits per pixel is: amount of channels * bitdepth. + +6.2. color conversions +---------------------- + +As explained in the sections about the encoder and decoder, you can specify +color types and bit depths in info_png and info_raw to change the default +behaviour. + +If, when decoding, you want the raw image to be something else than the default, +you need to set the color type and bit depth you want in the LodePNGColorMode, +or the parameters of the simple function of LodePNG you're using. + +If, when encoding, you use another color type than the default in the input +image, you need to specify its color type and bit depth in the LodePNGColorMode +of the raw image, or use the parameters of the simplefunction of LodePNG you're +using. + +If, when encoding, you don't want LodePNG to choose the output PNG color type +but control it yourself, you need to set auto_convert in the encoder settings +to LAC_NONE, and specify the color type you want in the LodePNGInfo of the +encoder. + +If you do any of the above, LodePNG may need to do a color conversion, which +follows the rules below, and may sometimes not be allowed. + +To avoid some confusion: +-the decoder converts from PNG to raw image +-the encoder converts from raw image to PNG +-the colortype and bitdepth in LodePNGColorMode info_raw, are those of the raw image +-the colortype and bitdepth in the color field of LodePNGInfo info_png, are those of the PNG +-when encoding, the color type in LodePNGInfo is ignored if auto_convert + is enabled, it is automatically generated instead +-when decoding, the color type in LodePNGInfo is set by the decoder to that of the original + PNG image, but it can be ignored since the raw image has the color type you requested instead +-if the color type of the LodePNGColorMode and PNG image aren't the same, a conversion + between the color types is done if the color types are supported. If it is not + supported, an error is returned. If the types are the same, no conversion is done. +-even though some conversions aren't supported, LodePNG supports loading PNGs from any + colortype and saving PNGs to any colortype, sometimes it just requires preparing + the raw image correctly before encoding. +-both encoder and decoder use the same color converter. + +Non supported color conversions: +-color to greyscale: no error is thrown, but the result will look ugly because +only the red channel is taken +-anything, to palette when that palette does not have that color in it: in this +case an error is thrown + +Supported color conversions: +-anything to 8-bit RGB, 8-bit RGBA, 16-bit RGB, 16-bit RGBA +-any grey or grey+alpha, to grey or grey+alpha +-anything to a palette, as long as the palette has the requested colors in it +-removing alpha channel +-higher to smaller bitdepth, and vice versa + +If you want no color conversion to be done: +-In the encoder, you can make it save a PNG with any color type by giving the +raw color mode and LodePNGInfo the same color mode, and setting auto_convert to +LAC_NO. +-In the decoder, you can make it store the pixel data in the same color type +as the PNG has, by setting the color_convert setting to false. Settings in +info_raw are then ignored. + +The function lodepng_convert does the color conversion. It is available in the +interface but normally isn't needed since the encoder and decoder already call +it. + +6.3. padding bits +----------------- + +In the PNG file format, if a less than 8-bit per pixel color type is used and the scanlines +have a bit amount that isn't a multiple of 8, then padding bits are used so that each +scanline starts at a fresh byte. But that is NOT true for the LodePNG raw input and output. +The raw input image you give to the encoder, and the raw output image you get from the decoder +will NOT have these padding bits, e.g. in the case of a 1-bit image with a width +of 7 pixels, the first pixel of the second scanline will the the 8th bit of the first byte, +not the first bit of a new byte. + +6.4. A note about 16-bits per channel and endianness +---------------------------------------------------- + +LodePNG uses unsigned char arrays for 16-bit per channel colors too, just like +for any other color format. The 16-bit values are stored in big endian (most +significant byte first) in these arrays. This is the opposite order of the +little endian used by x86 CPU's. + +LodePNG always uses big endian because the PNG file format does so internally. +Conversions to other formats than PNG uses internally are not supported by +LodePNG on purpose, there are myriads of formats, including endianness of 16-bit +colors, the order in which you store R, G, B and A, and so on. Supporting and +converting to/from all that is outside the scope of LodePNG. + +This may mean that, depending on your use case, you may want to convert the big +endian output of LodePNG to little endian with a for loop. This is certainly not +always needed, many applications and libraries support big endian 16-bit colors +anyway, but it means you cannot simply cast the unsigned char* buffer to an +unsigned short* buffer on x86 CPUs. + + +7. error values +--------------- + +All functions in LodePNG that return an error code, return 0 if everything went +OK, or a non-zero code if there was an error. + +The meaning of the LodePNG error values can be retrieved with the function +lodepng_error_text: given the numerical error code, it returns a description +of the error in English as a string. + +Check the implementation of lodepng_error_text to see the meaning of each code. + + +8. chunks and PNG editing +------------------------- + +If you want to add extra chunks to a PNG you encode, or use LodePNG for a PNG +editor that should follow the rules about handling of unknown chunks, or if your +program is able to read other types of chunks than the ones handled by LodePNG, +then that's possible with the chunk functions of LodePNG. + +A PNG chunk has the following layout: + +4 bytes length +4 bytes type name +length bytes data +4 bytes CRC + +8.1. iterating through chunks +----------------------------- + +If you have a buffer containing the PNG image data, then the first chunk (the +IHDR chunk) starts at byte number 8 of that buffer. The first 8 bytes are the +signature of the PNG and are not part of a chunk. But if you start at byte 8 +then you have a chunk, and can check the following things of it. + +NOTE: none of these functions check for memory buffer boundaries. To avoid +exploits, always make sure the buffer contains all the data of the chunks. +When using lodepng_chunk_next, make sure the returned value is within the +allocated memory. + +unsigned lodepng_chunk_length(const unsigned char* chunk): + +Get the length of the chunk's data. The total chunk length is this length + 12. + +void lodepng_chunk_type(char type[5], const unsigned char* chunk): +unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type): + +Get the type of the chunk or compare if it's a certain type + +unsigned char lodepng_chunk_critical(const unsigned char* chunk): +unsigned char lodepng_chunk_private(const unsigned char* chunk): +unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk): + +Check if the chunk is critical in the PNG standard (only IHDR, PLTE, IDAT and IEND are). +Check if the chunk is private (public chunks are part of the standard, private ones not). +Check if the chunk is safe to copy. If it's not, then, when modifying data in a critical +chunk, unsafe to copy chunks of the old image may NOT be saved in the new one if your +program doesn't handle that type of unknown chunk. + +unsigned char* lodepng_chunk_data(unsigned char* chunk): +const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk): + +Get a pointer to the start of the data of the chunk. + +unsigned lodepng_chunk_check_crc(const unsigned char* chunk): +void lodepng_chunk_generate_crc(unsigned char* chunk): + +Check if the crc is correct or generate a correct one. + +unsigned char* lodepng_chunk_next(unsigned char* chunk): +const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk): + +Iterate to the next chunk. This works if you have a buffer with consecutive chunks. Note that these +functions do no boundary checking of the allocated data whatsoever, so make sure there is enough +data available in the buffer to be able to go to the next chunk. + +unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk): +unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, + const char* type, const unsigned char* data): + +These functions are used to create new chunks that are appended to the data in *out that has +length *outlength. The append function appends an existing chunk to the new data. The create +function creates a new chunk with the given parameters and appends it. Type is the 4-letter +name of the chunk. + +8.2. chunks in info_png +----------------------- + +The LodePNGInfo struct contains fields with the unknown chunk in it. It has 3 +buffers (each with size) to contain 3 types of unknown chunks: +the ones that come before the PLTE chunk, the ones that come between the PLTE +and the IDAT chunks, and the ones that come after the IDAT chunks. +It's necessary to make the distionction between these 3 cases because the PNG +standard forces to keep the ordering of unknown chunks compared to the critical +chunks, but does not force any other ordering rules. + +info_png.unknown_chunks_data[0] is the chunks before PLTE +info_png.unknown_chunks_data[1] is the chunks after PLTE, before IDAT +info_png.unknown_chunks_data[2] is the chunks after IDAT + +The chunks in these 3 buffers can be iterated through and read by using the same +way described in the previous subchapter. + +When using the decoder to decode a PNG, you can make it store all unknown chunks +if you set the option settings.remember_unknown_chunks to 1. By default, this +option is off (0). + +The encoder will always encode unknown chunks that are stored in the info_png. +If you need it to add a particular chunk that isn't known by LodePNG, you can +use lodepng_chunk_append or lodepng_chunk_create to the chunk data in +info_png.unknown_chunks_data[x]. + +Chunks that are known by LodePNG should not be added in that way. E.g. to make +LodePNG add a bKGD chunk, set background_defined to true and add the correct +parameters there instead. + + +9. compiler support +------------------- + +No libraries other than the current standard C library are needed to compile +LodePNG. For the C++ version, only the standard C++ library is needed on top. +Add the files lodepng.c(pp) and lodepng.h to your project, include +lodepng.h where needed, and your program can read/write PNG files. + +If performance is important, use optimization when compiling! For both the +encoder and decoder, this makes a large difference. + +Make sure that LodePNG is compiled with the same compiler of the same version +and with the same settings as the rest of the program, or the interfaces with +std::vectors and std::strings in C++ can be incompatible. + +CHAR_BITS must be 8 or higher, because LodePNG uses unsigned chars for octets. + +*) gcc and g++ + +LodePNG is developed in gcc so this compiler is natively supported. It gives no +warnings with compiler options "-Wall -Wextra -pedantic -ansi", with gcc and g++ +version 4.7.1 on Linux, 32-bit and 64-bit. + +*) Mingw + +The Mingw compiler (a port of gcc) for Windows is fully supported by LodePNG. + +*) Visual Studio 2005 and up, Visual C++ Express Edition 2005 and up + +Visual Studio may give warnings about 'fopen' being deprecated. A multiplatform library +can't support the proposed Visual Studio alternative however, so LodePNG keeps using +fopen. If you don't want to see the deprecated warnings, put this on top of lodepng.h +before the inclusions: +#define _CRT_SECURE_NO_DEPRECATE + +Other than the above warnings, LodePNG should be warning-free with warning +level 3 (W3). Warning level 4 (W4) will give warnings about integer conversions. +I'm not planning to resolve these warnings. To get rid of them, let Visual +Studio use warning level W3 for lodepng.cpp only: right click lodepng.cpp, +Properties, C/C++, General, Warning Level: Level 3 (/W3). + +Visual Studio may want "stdafx.h" files to be included in each source file and +give an error "unexpected end of file while looking for precompiled header". +That is not standard C++ and will not be added to the stock LodePNG. You can +disable it for lodepng.cpp only by right clicking it, Properties, C/C++, +Precompiled Headers, and set it to Not Using Precompiled Headers there. + +*) Visual Studio 6.0 + +LodePNG support for Visual Studio 6.0 is not guaranteed because VS6 doesn't +follow the C++ standard correctly. + +*) Comeau C/C++ + +Vesion 20070107 compiles without problems on the Comeau C/C++ Online Test Drive +at http://www.comeaucomputing.com/tryitout in both C90 and C++ mode. + +*) Compilers on Macintosh + +LodePNG has been reported to work both with the gcc and LLVM for Macintosh, both +for C and C++. + +*) Other Compilers + +If you encounter problems on other compilers, feel free to let me know and I may +try to fix it if the compiler is modern standards complient. + + +10. examples +------------ + +This decoder example shows the most basic usage of LodePNG. More complex +examples can be found on the LodePNG website. + +10.1. decoder C++ example +------------------------- + +#include "lodepng.h" +#include + +int main(int argc, char *argv[]) +{ + const char* filename = argc > 1 ? argv[1] : "test.png"; + + //load and decode + std::vector image; + unsigned width, height; + unsigned error = lodepng::decode(image, width, height, filename); + + //if there's an error, display it + if(error) std::cout << "decoder error " << error << ": " << lodepng_error_text(error) << std::endl; + + //the pixels are now in the vector "image", 4 bytes per pixel, ordered RGBARGBA..., use it as texture, draw it, ... +} + +10.2. decoder C example +----------------------- + +#include "lodepng.h" + +int main(int argc, char *argv[]) +{ + unsigned error; + unsigned char* image; + size_t width, height; + const char* filename = argc > 1 ? argv[1] : "test.png"; + + error = lodepng_decode32_file(&image, &width, &height, filename); + + if(error) printf("decoder error %u: %s\n", error, lodepng_error_text(error)); + + / * use image here * / + + free(image); + return 0; +} + + +11. changes +----------- + +The version number of LodePNG is the date of the change given in the format +yyyymmdd. + +Some changes aren't backwards compatible. Those are indicated with a (!) +symbol. + +*) 15 apr 2013: Fixed bug with LAC_ALPHA and color key. +*) 25 mar 2013: Added an optional feature to ignore some PNG errors (fix_png). +*) 11 mar 2013 (!): Bugfix with custom free. Changed from "my" to "lodepng_" + prefix for the custom allocators and made it possible with a new #define to + use custom ones in your project without needing to change lodepng's code. +*) 28 jan 2013: Bugfix with color key. +*) 27 okt 2012: Tweaks in text chunk keyword length error handling. +*) 8 okt 2012 (!): Added new filter strategy (entropy) and new auto color mode. + (no palette). Better deflate tree encoding. New compression tweak settings. + Faster color conversions while decoding. Some internal cleanups. +*) 23 sep 2012: Reduced warnings in Visual Studio a little bit. +*) 1 sep 2012 (!): Removed #define's for giving custom (de)compression functions + and made it work with function pointers instead. +*) 23 jun 2012: Added more filter strategies. Made it easier to use custom alloc + and free functions and toggle #defines from compiler flags. Small fixes. +*) 6 may 2012 (!): Made plugging in custom zlib/deflate functions more flexible. +*) 22 apr 2012 (!): Made interface more consistent, renaming a lot. Removed + redundant C++ codec classes. Reduced amount of structs. Everything changed, + but it is cleaner now imho and functionality remains the same. Also fixed + several bugs and shrinked the implementation code. Made new samples. +*) 6 nov 2011 (!): By default, the encoder now automatically chooses the best + PNG color model and bit depth, based on the amount and type of colors of the + raw image. For this, autoLeaveOutAlphaChannel replaced by auto_choose_color. +*) 9 okt 2011: simpler hash chain implementation for the encoder. +*) 8 sep 2011: lz77 encoder lazy matching instead of greedy matching. +*) 23 aug 2011: tweaked the zlib compression parameters after benchmarking. + A bug with the PNG filtertype heuristic was fixed, so that it chooses much + better ones (it's quite significant). A setting to do an experimental, slow, + brute force search for PNG filter types is added. +*) 17 aug 2011 (!): changed some C zlib related function names. +*) 16 aug 2011: made the code less wide (max 120 characters per line). +*) 17 apr 2011: code cleanup. Bugfixes. Convert low to 16-bit per sample colors. +*) 21 feb 2011: fixed compiling for C90. Fixed compiling with sections disabled. +*) 11 dec 2010: encoding is made faster, based on suggestion by Peter Eastman + to optimize long sequences of zeros. +*) 13 nov 2010: added LodePNG_InfoColor_hasPaletteAlpha and + LodePNG_InfoColor_canHaveAlpha functions for convenience. +*) 7 nov 2010: added LodePNG_error_text function to get error code description. +*) 30 okt 2010: made decoding slightly faster +*) 26 okt 2010: (!) changed some C function and struct names (more consistent). + Reorganized the documentation and the declaration order in the header. +*) 08 aug 2010: only changed some comments and external samples. +*) 05 jul 2010: fixed bug thanks to warnings in the new gcc version. +*) 14 mar 2010: fixed bug where too much memory was allocated for char buffers. +*) 02 sep 2008: fixed bug where it could create empty tree that linux apps could + read by ignoring the problem but windows apps couldn't. +*) 06 jun 2008: added more error checks for out of memory cases. +*) 26 apr 2008: added a few more checks here and there to ensure more safety. +*) 06 mar 2008: crash with encoding of strings fixed +*) 02 feb 2008: support for international text chunks added (iTXt) +*) 23 jan 2008: small cleanups, and #defines to divide code in sections +*) 20 jan 2008: support for unknown chunks allowing using LodePNG for an editor. +*) 18 jan 2008: support for tIME and pHYs chunks added to encoder and decoder. +*) 17 jan 2008: ability to encode and decode compressed zTXt chunks added + Also vareous fixes, such as in the deflate and the padding bits code. +*) 13 jan 2008: Added ability to encode Adam7-interlaced images. Improved + filtering code of encoder. +*) 07 jan 2008: (!) changed LodePNG to use ISO C90 instead of C++. A + C++ wrapper around this provides an interface almost identical to before. + Having LodePNG be pure ISO C90 makes it more portable. The C and C++ code + are together in these files but it works both for C and C++ compilers. +*) 29 dec 2007: (!) changed most integer types to unsigned int + other tweaks +*) 30 aug 2007: bug fixed which makes this Borland C++ compatible +*) 09 aug 2007: some VS2005 warnings removed again +*) 21 jul 2007: deflate code placed in new namespace separate from zlib code +*) 08 jun 2007: fixed bug with 2- and 4-bit color, and small interlaced images +*) 04 jun 2007: improved support for Visual Studio 2005: crash with accessing + invalid std::vector element [0] fixed, and level 3 and 4 warnings removed +*) 02 jun 2007: made the encoder add a tag with version by default +*) 27 may 2007: zlib and png code separated (but still in the same file), + simple encoder/decoder functions added for more simple usage cases +*) 19 may 2007: minor fixes, some code cleaning, new error added (error 69), + moved some examples from here to lodepng_examples.cpp +*) 12 may 2007: palette decoding bug fixed +*) 24 apr 2007: changed the license from BSD to the zlib license +*) 11 mar 2007: very simple addition: ability to encode bKGD chunks. +*) 04 mar 2007: (!) tEXt chunk related fixes, and support for encoding + palettized PNG images. Plus little interface change with palette and texts. +*) 03 mar 2007: Made it encode dynamic Huffman shorter with repeat codes. + Fixed a bug where the end code of a block had length 0 in the Huffman tree. +*) 26 feb 2007: Huffman compression with dynamic trees (BTYPE 2) now implemented + and supported by the encoder, resulting in smaller PNGs at the output. +*) 27 jan 2007: Made the Adler-32 test faster so that a timewaste is gone. +*) 24 jan 2007: gave encoder an error interface. Added color conversion from any + greyscale type to 8-bit greyscale with or without alpha. +*) 21 jan 2007: (!) Totally changed the interface. It allows more color types + to convert to and is more uniform. See the manual for how it works now. +*) 07 jan 2007: Some cleanup & fixes, and a few changes over the last days: + encode/decode custom tEXt chunks, separate classes for zlib & deflate, and + at last made the decoder give errors for incorrect Adler32 or Crc. +*) 01 jan 2007: Fixed bug with encoding PNGs with less than 8 bits per channel. +*) 29 dec 2006: Added support for encoding images without alpha channel, and + cleaned out code as well as making certain parts faster. +*) 28 dec 2006: Added "Settings" to the encoder. +*) 26 dec 2006: The encoder now does LZ77 encoding and produces much smaller files now. + Removed some code duplication in the decoder. Fixed little bug in an example. +*) 09 dec 2006: (!) Placed output parameters of public functions as first parameter. + Fixed a bug of the decoder with 16-bit per color. +*) 15 okt 2006: Changed documentation structure +*) 09 okt 2006: Encoder class added. It encodes a valid PNG image from the + given image buffer, however for now it's not compressed. +*) 08 sep 2006: (!) Changed to interface with a Decoder class +*) 30 jul 2006: (!) LodePNG_InfoPng , width and height are now retrieved in different + way. Renamed decodePNG to decodePNGGeneric. +*) 29 jul 2006: (!) Changed the interface: image info is now returned as a + struct of type LodePNG::LodePNG_Info, instead of a vector, which was a bit clumsy. +*) 28 jul 2006: Cleaned the code and added new error checks. + Corrected terminology "deflate" into "inflate". +*) 23 jun 2006: Added SDL example in the documentation in the header, this + example allows easy debugging by displaying the PNG and its transparency. +*) 22 jun 2006: (!) Changed way to obtain error value. Added + loadFile function for convenience. Made decodePNG32 faster. +*) 21 jun 2006: (!) Changed type of info vector to unsigned. + Changed position of palette in info vector. Fixed an important bug that + happened on PNGs with an uncompressed block. +*) 16 jun 2006: Internally changed unsigned into unsigned where + needed, and performed some optimizations. +*) 07 jun 2006: (!) Renamed functions to decodePNG and placed them + in LodePNG namespace. Changed the order of the parameters. Rewrote the + documentation in the header. Renamed files to lodepng.cpp and lodepng.h +*) 22 apr 2006: Optimized and improved some code +*) 07 sep 2005: (!) Changed to std::vector interface +*) 12 aug 2005: Initial release (C++, decoder only) + + +12. contact information +----------------------- + +Feel free to contact me with suggestions, problems, comments, ... concerning +LodePNG. If you encounter a PNG image that doesn't work properly with this +decoder, feel free to send it and I'll use it to find and fix the problem. + +My email address is (puzzle the account and domain together with an @ symbol): +Domain: gmail dot com. +Account: lode dot vandevenne. + + +Copyright (c) 2005-2013 Lode Vandevenne +*/ diff --git a/rpcs3.sln b/rpcs3.sln index 5d0afea291..532495113f 100644 --- a/rpcs3.sln +++ b/rpcs3.sln @@ -1,5 +1,5 @@ -Microsoft Visual Studio Solution File, Format Version 11.00 -# Visual Studio 2010 +Microsoft Visual Studio Solution File, Format Version 12.00 +# Visual Studio 2012 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "rpcs3", "rpcs3\rpcs3.vcxproj", "{70CD65B0-91D6-4FAE-9A7B-4AF55D0D1B12}" ProjectSection(ProjectDependencies) = postProject {3111D679-7796-23C4-BA0C-271F1145DA24} = {3111D679-7796-23C4-BA0C-271F1145DA24} @@ -57,6 +57,12 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "xml", "wxWidgets\build\msw\ EndProject Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "xrc", "wxWidgets\build\msw\wx_xrc.vcxproj", "{8BECCA95-C7D7-CFF8-FDB1-4950E9F8E8E6}" EndProject +Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "lodepng", "lodepng", "{48675C2D-D949-400D-8758-E22FB6DF3D58}" + ProjectSection(SolutionItems) = preProject + lodepng\lodepng.cpp = lodepng\lodepng.cpp + lodepng\lodepng.h = lodepng\lodepng.h + EndProjectSection +EndProject Global GlobalSection(SolutionConfigurationPlatforms) = preSolution Debug|Win32 = Debug|Win32 diff --git a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp new file mode 100644 index 0000000000..d8f78a9b86 --- /dev/null +++ b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp @@ -0,0 +1,233 @@ +#include "stdafx.h" +#include "Emu/SysCalls/SysCalls.h" +#include "Emu/SysCalls/SC_FUNC.h" + +#include "lodepng/lodepng.h" +#include "lodepng/lodepng.cpp" + +void cellPngDec_init(); +Module cellPngDec(0x0018, cellPngDec_init); + +//Return Codes +enum +{ + CELL_PNGDEC_ERROR_HEADER = 0x80611201, + CELL_PNGDEC_ERROR_STREAM_FORMAT = 0x80611202, + CELL_PNGDEC_ERROR_ARG = 0x80611203, + CELL_PNGDEC_ERROR_SEQ = 0x80611204, + CELL_PNGDEC_ERROR_BUSY = 0x80611205, + CELL_PNGDEC_ERROR_FATAL = 0x80611206, + CELL_PNGDEC_ERROR_OPEN_FILE = 0x80611207, + CELL_PNGDEC_ERROR_SPU_UNSUPPORT = 0x80611208, + CELL_PNGDEC_ERROR_SPU_ERROR = 0x80611209, + CELL_PNGDEC_ERROR_CB_PARAM = 0x8061120a, +}; + +struct CellPngDecInfo +{ + u32 imageWidth; + u32 imageHeight; + u32 numComponents; + u32 colorSpace; // CellPngDecColorSpace + u32 bitDepth; + u32 interlaceMethod; // CellPngDecInterlaceMode + u32 chunkInformation; +}; + +struct CellPngDecSrc +{ + u32 srcSelect; // CellPngDecStreamSrcSel + u32 fileName_addr; // const char* + u64 fileOffset; // int64_t + u32 fileSize; + u32 streamPtr; + u32 streamSize; + u32 spuThreadEnable; // CellPngDecSpuThreadEna +}; + +CellPngDecInfo current_info; +CellPngDecSrc current_src; + + +int cellPngDecCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam) +{ + UNIMPLEMENTED_FUNC(cellPngDec); + return CELL_OK; +} + +int cellPngDecDestroy(u32 mainHandle) +{ + UNIMPLEMENTED_FUNC(cellPngDec); + return CELL_OK; +} + +int cellPngDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInfo) +{ + //current_src.srcSelect = Memory.Read32(src_addr); + current_src.fileName_addr = Memory.Read32(src_addr+4); + //current_src.fileOffset = Memory.Read32(src_addr+8); + //current_src.fileSize = Memory.Read32(src_addr+12); + //current_src.streamPtr = Memory.Read32(src_addr+16); + //current_src.streamSize = Memory.Read32(src_addr+20); + //current_src.spuThreadEnable = Memory.Read32(src_addr+24); + + u32& fd_addr = subHandle_addr; // Set file descriptor as sub handler of the decoder + int ret = cellFsOpen(current_src.fileName_addr, 0, fd_addr, NULL, 0); + if(ret != 0) return CELL_PNGDEC_ERROR_OPEN_FILE; + + return CELL_OK; +} + +int cellPngDecClose(u32 mainHandle, u32 subHandle) +{ + u32& fd = subHandle; + cellFsClose(fd); + + return CELL_OK; +} + +int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) +{ + u32& fd = subHandle; + + //Check size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(fd, sb_addr); + u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + if(fileSize < 29) return CELL_PNGDEC_ERROR_HEADER; // Error: The file is smaller than the length of a PNG header + + + //Write the header to buffer + u32 buffer = Memory.Alloc(34,1); // Alloc buffer for PNG header + u32 pos_addr = Memory.Alloc(8,1); + cellFsLseek(fd, 0, 0, pos_addr); + cellFsRead(fd, buffer, 34, NULL); + + if (Memory.Read32(buffer) != 0x89504E47 || Memory.Read32(buffer+4) != 0x0D0A1A0A) + return CELL_PNGDEC_ERROR_HEADER; // Error: The first 8 bytes are not a valid PNG signature + + if (Memory.Read32(buffer+12) != 0x49484452) + return CELL_PNGDEC_ERROR_HEADER; // Error: The PNG file does not start with an IHDR chunk + + current_info.imageWidth = Memory.Read32(buffer+16); + current_info.imageHeight = Memory.Read32(buffer+20); + current_info.numComponents = 0; // Unimplemented + current_info.colorSpace = Memory.Read8(buffer+25); + current_info.bitDepth = Memory.Read8(buffer+24); + current_info.interlaceMethod = Memory.Read8(buffer+28); + current_info.chunkInformation = 0; // Unimplemented + + mem_class_t info(info_addr); + info += current_info.imageWidth; + info += current_info.imageHeight; + info += current_info.numComponents; + info += current_info.colorSpace; + info += current_info.bitDepth; + info += current_info.interlaceMethod; + info += current_info.chunkInformation; + + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + + return CELL_OK; +} + +int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataCtrlParam_addr, u32 dataOutInfo_addr) +{ + u32& fd = subHandle; + + //Check size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(fd, sb_addr); + u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + if(fileSize < 29) return CELL_PNGDEC_ERROR_HEADER; // Error: The file is smaller than the length of a PNG header + + //Write the header to buffer + u32 buffer = Memory.Alloc(34,1); // Alloc buffer for PNG header + u32 pos_addr = Memory.Alloc(8,1); + cellFsLseek(fd, 0, 0, pos_addr); + cellFsRead(fd, buffer, 34, NULL); + + if (Memory.Read32(buffer) != 0x89504E47 || Memory.Read32(buffer+4) != 0x0D0A1A0A) + return CELL_PNGDEC_ERROR_HEADER; // Error: The first 8 bytes are not a valid PNG signature + + if (Memory.Read32(buffer+12) != 0x49484452) + return CELL_PNGDEC_ERROR_HEADER; // Error: The PNG file does not start with an IHDR chunk + + current_info.imageWidth = Memory.Read32(buffer+16); + current_info.imageHeight = Memory.Read32(buffer+20); + Memory.Free(buffer); + + //Decode PNG file. (TODO: Is there any faster alternative? Can we do it without external libraries?) + buffer = Memory.Alloc(fileSize,1); + cellFsLseek(fd,0,0,pos_addr); + cellFsRead(fd, buffer, fileSize, NULL); + + std::vector png; // PNG buffer + std::vector image; // Raw buffer + + //Load contents in png buffer + png.resize(size_t(fileSize)); + for(u32 i = 0; i < fileSize; i++){ + png[i] = Memory.Read8(buffer+i); + } + + //Decode + lodepng::decode(image, current_info.imageWidth, current_info.imageHeight, png); + + u32 image_size = image.size(); + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+3]); + Memory.Write8(data_addr+i+1, image[i+0]); + Memory.Write8(data_addr+i+2, image[i+1]); + Memory.Write8(data_addr+i+3, image[i+2]); + } + + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + + return CELL_OK; +} + +int cellPngDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outParam) +{ + UNIMPLEMENTED_FUNC(cellPngDec); + return CELL_OK; +} + + +void cellPngDec_init() +{ + cellPngDec.AddFunc(0x157d30c5, cellPngDecCreate); + cellPngDec.AddFunc(0x820dae1a, cellPngDecDestroy); + cellPngDec.AddFunc(0xd2bc5bfd, cellPngDecOpen); + cellPngDec.AddFunc(0x5b3d1ff1, cellPngDecClose); + cellPngDec.AddFunc(0x9ccdcc95, cellPngDecReadHeader); + cellPngDec.AddFunc(0x2310f155, cellPngDecDecodeData); + cellPngDec.AddFunc(0xe97c9bd4, cellPngDecSetParameter); + + /*cellPngDec.AddFunc(0x48436b2d, cellPngDecExtCreate); + cellPngDec.AddFunc(0x0c515302, cellPngDecExtOpen); + cellPngDec.AddFunc(0x8b33f863, cellPngDecExtReadHeader); + cellPngDec.AddFunc(0x726fc1d0, cellPngDecExtDecodeData); + cellPngDec.AddFunc(0x9e9d7d42, cellPngDecExtSetParameter); + cellPngDec.AddFunc(0x7585a275, cellPngDecGetbKGD); + cellPngDec.AddFunc(0x7a062d26, cellPngDecGetcHRM); + cellPngDec.AddFunc(0xb153629c, cellPngDecGetgAMA); + cellPngDec.AddFunc(0xb905ebb7, cellPngDecGethIST); + cellPngDec.AddFunc(0xf44b6c30, cellPngDecGetiCCP); + cellPngDec.AddFunc(0x27c921b5, cellPngDecGetoFFs); + cellPngDec.AddFunc(0xb4fe75e1, cellPngDecGetpCAL); + cellPngDec.AddFunc(0x3d50016a, cellPngDecGetpHYs); + cellPngDec.AddFunc(0x30cb334a, cellPngDecGetsBIT); + cellPngDec.AddFunc(0xc41e1198, cellPngDecGetsCAL); + cellPngDec.AddFunc(0xa5cdf57e, cellPngDecGetsPLT); + cellPngDec.AddFunc(0xe4416e82, cellPngDecGetsRGB); + cellPngDec.AddFunc(0x35a6846c, cellPngDecGettIME); + cellPngDec.AddFunc(0xb96fb26e, cellPngDecGettRNS); + cellPngDec.AddFunc(0xe163977f, cellPngDecGetPLTE); + cellPngDec.AddFunc(0x609ec7d5, cellPngDecUnknownChunks); + cellPngDec.AddFunc(0xb40ca175, cellPngDecGetTextChunk);*/ +} \ No newline at end of file diff --git a/rpcs3/rpcs3.vcxproj b/rpcs3/rpcs3.vcxproj index fb39fdb24f..5277d1d0d1 100644 --- a/rpcs3/rpcs3.vcxproj +++ b/rpcs3/rpcs3.vcxproj @@ -245,6 +245,7 @@ + diff --git a/rpcs3/rpcs3.vcxproj.filters b/rpcs3/rpcs3.vcxproj.filters index d9e380c208..ae85f01d96 100644 --- a/rpcs3/rpcs3.vcxproj.filters +++ b/rpcs3/rpcs3.vcxproj.filters @@ -286,6 +286,9 @@ Emu\SysCalls\Modules + + Emu\SysCalls\Modules + From 0275b430cf46b0f2ed9aaf23650fd9d016659a66 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Mon, 9 Sep 2013 18:27:31 +0200 Subject: [PATCH 2/7] cellJpgDec Module Added *Replaced "sys_jpgdec" with "cellJpgDec" in "rpcs3\Emu\SysCalls\Modules.cpp". *Implemented 'cellJpgDec', using 'sys_fs' to access the files and the 'jpg-compressor' (jpgd) library to decode JPG buffers to Raw-pixel buffers that can be used as textures. *Fixed memory leak issue in cellPngDec and minor changes. --- jpeg-compressor/jpgd.cpp | 3172 +++++++++++++++++++++ jpeg-compressor/jpgd.h | 319 +++ rpcs3.sln | 6 + rpcs3/Emu/SysCalls/Modules.cpp | 2 +- rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp | 203 ++ rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp | 58 +- rpcs3/rpcs3.vcxproj | 1 + rpcs3/rpcs3.vcxproj.filters | 3 + 8 files changed, 3731 insertions(+), 33 deletions(-) create mode 100644 jpeg-compressor/jpgd.cpp create mode 100644 jpeg-compressor/jpgd.h create mode 100644 rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp diff --git a/jpeg-compressor/jpgd.cpp b/jpeg-compressor/jpgd.cpp new file mode 100644 index 0000000000..fad9a37a9a --- /dev/null +++ b/jpeg-compressor/jpgd.cpp @@ -0,0 +1,3172 @@ +// jpgd.cpp - C++ class for JPEG decompression. +// Public domain, Rich Geldreich +// Alex Evans: Linear memory allocator (taken from jpge.h). +// v1.04, May. 19, 2012: Code tweaks to fix VS2008 static code analysis warnings (all looked harmless) +// +// Supports progressive and baseline sequential JPEG image files, and the most common chroma subsampling factors: Y, H1V1, H2V1, H1V2, and H2V2. +// +// Chroma upsampling quality: H2V2 is upsampled in the frequency domain, H2V1 and H1V2 are upsampled using point sampling. +// Chroma upsampling reference: "Fast Scheme for Image Size Change in the Compressed Domain" +// http://vision.ai.uiuc.edu/~dugad/research/dct/index.html + +#include "jpgd.h" +#include + +#include +#define JPGD_ASSERT(x) assert(x) + +#ifdef _MSC_VER +#pragma warning (disable : 4611) // warning C4611: interaction between '_setjmp' and C++ object destruction is non-portable +#endif + +// Set to 1 to enable freq. domain chroma upsampling on images using H2V2 subsampling (0=faster nearest neighbor sampling). +// This is slower, but results in higher quality on images with highly saturated colors. +#define JPGD_SUPPORT_FREQ_DOMAIN_UPSAMPLING 1 + +#define JPGD_TRUE (1) +#define JPGD_FALSE (0) + +#define JPGD_MAX(a,b) (((a)>(b)) ? (a) : (b)) +#define JPGD_MIN(a,b) (((a)<(b)) ? (a) : (b)) + +namespace jpgd { + +static inline void *jpgd_malloc(size_t nSize) { return malloc(nSize); } +static inline void jpgd_free(void *p) { free(p); } + +// DCT coefficients are stored in this sequence. +static int g_ZAG[64] = { 0,1,8,16,9,2,3,10,17,24,32,25,18,11,4,5,12,19,26,33,40,48,41,34,27,20,13,6,7,14,21,28,35,42,49,56,57,50,43,36,29,22,15,23,30,37,44,51,58,59,52,45,38,31,39,46,53,60,61,54,47,55,62,63 }; + +enum JPEG_MARKER +{ + M_SOF0 = 0xC0, M_SOF1 = 0xC1, M_SOF2 = 0xC2, M_SOF3 = 0xC3, M_SOF5 = 0xC5, M_SOF6 = 0xC6, M_SOF7 = 0xC7, M_JPG = 0xC8, + M_SOF9 = 0xC9, M_SOF10 = 0xCA, M_SOF11 = 0xCB, M_SOF13 = 0xCD, M_SOF14 = 0xCE, M_SOF15 = 0xCF, M_DHT = 0xC4, M_DAC = 0xCC, + M_RST0 = 0xD0, M_RST1 = 0xD1, M_RST2 = 0xD2, M_RST3 = 0xD3, M_RST4 = 0xD4, M_RST5 = 0xD5, M_RST6 = 0xD6, M_RST7 = 0xD7, + M_SOI = 0xD8, M_EOI = 0xD9, M_SOS = 0xDA, M_DQT = 0xDB, M_DNL = 0xDC, M_DRI = 0xDD, M_DHP = 0xDE, M_EXP = 0xDF, + M_APP0 = 0xE0, M_APP15 = 0xEF, M_JPG0 = 0xF0, M_JPG13 = 0xFD, M_COM = 0xFE, M_TEM = 0x01, M_ERROR = 0x100, RST0 = 0xD0 +}; + +enum JPEG_SUBSAMPLING { JPGD_GRAYSCALE = 0, JPGD_YH1V1, JPGD_YH2V1, JPGD_YH1V2, JPGD_YH2V2 }; + +#define CONST_BITS 13 +#define PASS1_BITS 2 +#define SCALEDONE ((int32)1) + +#define FIX_0_298631336 ((int32)2446) /* FIX(0.298631336) */ +#define FIX_0_390180644 ((int32)3196) /* FIX(0.390180644) */ +#define FIX_0_541196100 ((int32)4433) /* FIX(0.541196100) */ +#define FIX_0_765366865 ((int32)6270) /* FIX(0.765366865) */ +#define FIX_0_899976223 ((int32)7373) /* FIX(0.899976223) */ +#define FIX_1_175875602 ((int32)9633) /* FIX(1.175875602) */ +#define FIX_1_501321110 ((int32)12299) /* FIX(1.501321110) */ +#define FIX_1_847759065 ((int32)15137) /* FIX(1.847759065) */ +#define FIX_1_961570560 ((int32)16069) /* FIX(1.961570560) */ +#define FIX_2_053119869 ((int32)16819) /* FIX(2.053119869) */ +#define FIX_2_562915447 ((int32)20995) /* FIX(2.562915447) */ +#define FIX_3_072711026 ((int32)25172) /* FIX(3.072711026) */ + +#define DESCALE(x,n) (((x) + (SCALEDONE << ((n)-1))) >> (n)) +#define DESCALE_ZEROSHIFT(x,n) (((x) + (128 << (n)) + (SCALEDONE << ((n)-1))) >> (n)) + +#define MULTIPLY(var, cnst) ((var) * (cnst)) + +#define CLAMP(i) ((static_cast(i) > 255) ? (((~i) >> 31) & 0xFF) : (i)) + +// Compiler creates a fast path 1D IDCT for X non-zero columns +template +struct Row +{ + static void idct(int* pTemp, const jpgd_block_t* pSrc) + { + // ACCESS_COL() will be optimized at compile time to either an array access, or 0. + #define ACCESS_COL(x) (((x) < NONZERO_COLS) ? (int)pSrc[x] : 0) + + const int z2 = ACCESS_COL(2), z3 = ACCESS_COL(6); + + const int z1 = MULTIPLY(z2 + z3, FIX_0_541196100); + const int tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); + const int tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); + + const int tmp0 = (ACCESS_COL(0) + ACCESS_COL(4)) << CONST_BITS; + const int tmp1 = (ACCESS_COL(0) - ACCESS_COL(4)) << CONST_BITS; + + const int tmp10 = tmp0 + tmp3, tmp13 = tmp0 - tmp3, tmp11 = tmp1 + tmp2, tmp12 = tmp1 - tmp2; + + const int atmp0 = ACCESS_COL(7), atmp1 = ACCESS_COL(5), atmp2 = ACCESS_COL(3), atmp3 = ACCESS_COL(1); + + const int bz1 = atmp0 + atmp3, bz2 = atmp1 + atmp2, bz3 = atmp0 + atmp2, bz4 = atmp1 + atmp3; + const int bz5 = MULTIPLY(bz3 + bz4, FIX_1_175875602); + + const int az1 = MULTIPLY(bz1, - FIX_0_899976223); + const int az2 = MULTIPLY(bz2, - FIX_2_562915447); + const int az3 = MULTIPLY(bz3, - FIX_1_961570560) + bz5; + const int az4 = MULTIPLY(bz4, - FIX_0_390180644) + bz5; + + const int btmp0 = MULTIPLY(atmp0, FIX_0_298631336) + az1 + az3; + const int btmp1 = MULTIPLY(atmp1, FIX_2_053119869) + az2 + az4; + const int btmp2 = MULTIPLY(atmp2, FIX_3_072711026) + az2 + az3; + const int btmp3 = MULTIPLY(atmp3, FIX_1_501321110) + az1 + az4; + + pTemp[0] = DESCALE(tmp10 + btmp3, CONST_BITS-PASS1_BITS); + pTemp[7] = DESCALE(tmp10 - btmp3, CONST_BITS-PASS1_BITS); + pTemp[1] = DESCALE(tmp11 + btmp2, CONST_BITS-PASS1_BITS); + pTemp[6] = DESCALE(tmp11 - btmp2, CONST_BITS-PASS1_BITS); + pTemp[2] = DESCALE(tmp12 + btmp1, CONST_BITS-PASS1_BITS); + pTemp[5] = DESCALE(tmp12 - btmp1, CONST_BITS-PASS1_BITS); + pTemp[3] = DESCALE(tmp13 + btmp0, CONST_BITS-PASS1_BITS); + pTemp[4] = DESCALE(tmp13 - btmp0, CONST_BITS-PASS1_BITS); + } +}; + +template <> +struct Row<0> +{ + static void idct(int* pTemp, const jpgd_block_t* pSrc) + { +#ifdef _MSC_VER + pTemp; pSrc; +#endif + } +}; + +template <> +struct Row<1> +{ + static void idct(int* pTemp, const jpgd_block_t* pSrc) + { + const int dcval = (pSrc[0] << PASS1_BITS); + + pTemp[0] = dcval; + pTemp[1] = dcval; + pTemp[2] = dcval; + pTemp[3] = dcval; + pTemp[4] = dcval; + pTemp[5] = dcval; + pTemp[6] = dcval; + pTemp[7] = dcval; + } +}; + +// Compiler creates a fast path 1D IDCT for X non-zero rows +template +struct Col +{ + static void idct(uint8* pDst_ptr, const int* pTemp) + { + // ACCESS_ROW() will be optimized at compile time to either an array access, or 0. + #define ACCESS_ROW(x) (((x) < NONZERO_ROWS) ? pTemp[x * 8] : 0) + + const int z2 = ACCESS_ROW(2); + const int z3 = ACCESS_ROW(6); + + const int z1 = MULTIPLY(z2 + z3, FIX_0_541196100); + const int tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); + const int tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); + + const int tmp0 = (ACCESS_ROW(0) + ACCESS_ROW(4)) << CONST_BITS; + const int tmp1 = (ACCESS_ROW(0) - ACCESS_ROW(4)) << CONST_BITS; + + const int tmp10 = tmp0 + tmp3, tmp13 = tmp0 - tmp3, tmp11 = tmp1 + tmp2, tmp12 = tmp1 - tmp2; + + const int atmp0 = ACCESS_ROW(7), atmp1 = ACCESS_ROW(5), atmp2 = ACCESS_ROW(3), atmp3 = ACCESS_ROW(1); + + const int bz1 = atmp0 + atmp3, bz2 = atmp1 + atmp2, bz3 = atmp0 + atmp2, bz4 = atmp1 + atmp3; + const int bz5 = MULTIPLY(bz3 + bz4, FIX_1_175875602); + + const int az1 = MULTIPLY(bz1, - FIX_0_899976223); + const int az2 = MULTIPLY(bz2, - FIX_2_562915447); + const int az3 = MULTIPLY(bz3, - FIX_1_961570560) + bz5; + const int az4 = MULTIPLY(bz4, - FIX_0_390180644) + bz5; + + const int btmp0 = MULTIPLY(atmp0, FIX_0_298631336) + az1 + az3; + const int btmp1 = MULTIPLY(atmp1, FIX_2_053119869) + az2 + az4; + const int btmp2 = MULTIPLY(atmp2, FIX_3_072711026) + az2 + az3; + const int btmp3 = MULTIPLY(atmp3, FIX_1_501321110) + az1 + az4; + + int i = DESCALE_ZEROSHIFT(tmp10 + btmp3, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*0] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp10 - btmp3, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*7] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp11 + btmp2, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*1] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp11 - btmp2, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*6] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp12 + btmp1, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*2] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp12 - btmp1, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*5] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp13 + btmp0, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*3] = (uint8)CLAMP(i); + + i = DESCALE_ZEROSHIFT(tmp13 - btmp0, CONST_BITS+PASS1_BITS+3); + pDst_ptr[8*4] = (uint8)CLAMP(i); + } +}; + +template <> +struct Col<1> +{ + static void idct(uint8* pDst_ptr, const int* pTemp) + { + int dcval = DESCALE_ZEROSHIFT(pTemp[0], PASS1_BITS+3); + const uint8 dcval_clamped = (uint8)CLAMP(dcval); + pDst_ptr[0*8] = dcval_clamped; + pDst_ptr[1*8] = dcval_clamped; + pDst_ptr[2*8] = dcval_clamped; + pDst_ptr[3*8] = dcval_clamped; + pDst_ptr[4*8] = dcval_clamped; + pDst_ptr[5*8] = dcval_clamped; + pDst_ptr[6*8] = dcval_clamped; + pDst_ptr[7*8] = dcval_clamped; + } +}; + +static const uint8 s_idct_row_table[] = +{ + 1,0,0,0,0,0,0,0, 2,0,0,0,0,0,0,0, 2,1,0,0,0,0,0,0, 2,1,1,0,0,0,0,0, 2,2,1,0,0,0,0,0, 3,2,1,0,0,0,0,0, 4,2,1,0,0,0,0,0, 4,3,1,0,0,0,0,0, + 4,3,2,0,0,0,0,0, 4,3,2,1,0,0,0,0, 4,3,2,1,1,0,0,0, 4,3,2,2,1,0,0,0, 4,3,3,2,1,0,0,0, 4,4,3,2,1,0,0,0, 5,4,3,2,1,0,0,0, 6,4,3,2,1,0,0,0, + 6,5,3,2,1,0,0,0, 6,5,4,2,1,0,0,0, 6,5,4,3,1,0,0,0, 6,5,4,3,2,0,0,0, 6,5,4,3,2,1,0,0, 6,5,4,3,2,1,1,0, 6,5,4,3,2,2,1,0, 6,5,4,3,3,2,1,0, + 6,5,4,4,3,2,1,0, 6,5,5,4,3,2,1,0, 6,6,5,4,3,2,1,0, 7,6,5,4,3,2,1,0, 8,6,5,4,3,2,1,0, 8,7,5,4,3,2,1,0, 8,7,6,4,3,2,1,0, 8,7,6,5,3,2,1,0, + 8,7,6,5,4,2,1,0, 8,7,6,5,4,3,1,0, 8,7,6,5,4,3,2,0, 8,7,6,5,4,3,2,1, 8,7,6,5,4,3,2,2, 8,7,6,5,4,3,3,2, 8,7,6,5,4,4,3,2, 8,7,6,5,5,4,3,2, + 8,7,6,6,5,4,3,2, 8,7,7,6,5,4,3,2, 8,8,7,6,5,4,3,2, 8,8,8,6,5,4,3,2, 8,8,8,7,5,4,3,2, 8,8,8,7,6,4,3,2, 8,8,8,7,6,5,3,2, 8,8,8,7,6,5,4,2, + 8,8,8,7,6,5,4,3, 8,8,8,7,6,5,4,4, 8,8,8,7,6,5,5,4, 8,8,8,7,6,6,5,4, 8,8,8,7,7,6,5,4, 8,8,8,8,7,6,5,4, 8,8,8,8,8,6,5,4, 8,8,8,8,8,7,5,4, + 8,8,8,8,8,7,6,4, 8,8,8,8,8,7,6,5, 8,8,8,8,8,7,6,6, 8,8,8,8,8,7,7,6, 8,8,8,8,8,8,7,6, 8,8,8,8,8,8,8,6, 8,8,8,8,8,8,8,7, 8,8,8,8,8,8,8,8, +}; + +static const uint8 s_idct_col_table[] = { 1, 1, 2, 3, 3, 3, 3, 3, 3, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8 }; + +void idct(const jpgd_block_t* pSrc_ptr, uint8* pDst_ptr, int block_max_zag) +{ + JPGD_ASSERT(block_max_zag >= 1); + JPGD_ASSERT(block_max_zag <= 64); + + if (block_max_zag <= 1) + { + int k = ((pSrc_ptr[0] + 4) >> 3) + 128; + k = CLAMP(k); + k = k | (k<<8); + k = k | (k<<16); + + for (int i = 8; i > 0; i--) + { + *(int*)&pDst_ptr[0] = k; + *(int*)&pDst_ptr[4] = k; + pDst_ptr += 8; + } + return; + } + + int temp[64]; + + const jpgd_block_t* pSrc = pSrc_ptr; + int* pTemp = temp; + + const uint8* pRow_tab = &s_idct_row_table[(block_max_zag - 1) * 8]; + int i; + for (i = 8; i > 0; i--, pRow_tab++) + { + switch (*pRow_tab) + { + case 0: Row<0>::idct(pTemp, pSrc); break; + case 1: Row<1>::idct(pTemp, pSrc); break; + case 2: Row<2>::idct(pTemp, pSrc); break; + case 3: Row<3>::idct(pTemp, pSrc); break; + case 4: Row<4>::idct(pTemp, pSrc); break; + case 5: Row<5>::idct(pTemp, pSrc); break; + case 6: Row<6>::idct(pTemp, pSrc); break; + case 7: Row<7>::idct(pTemp, pSrc); break; + case 8: Row<8>::idct(pTemp, pSrc); break; + } + + pSrc += 8; + pTemp += 8; + } + + pTemp = temp; + + const int nonzero_rows = s_idct_col_table[block_max_zag - 1]; + for (i = 8; i > 0; i--) + { + switch (nonzero_rows) + { + case 1: Col<1>::idct(pDst_ptr, pTemp); break; + case 2: Col<2>::idct(pDst_ptr, pTemp); break; + case 3: Col<3>::idct(pDst_ptr, pTemp); break; + case 4: Col<4>::idct(pDst_ptr, pTemp); break; + case 5: Col<5>::idct(pDst_ptr, pTemp); break; + case 6: Col<6>::idct(pDst_ptr, pTemp); break; + case 7: Col<7>::idct(pDst_ptr, pTemp); break; + case 8: Col<8>::idct(pDst_ptr, pTemp); break; + } + + pTemp++; + pDst_ptr++; + } +} + +void idct_4x4(const jpgd_block_t* pSrc_ptr, uint8* pDst_ptr) +{ + int temp[64]; + int* pTemp = temp; + const jpgd_block_t* pSrc = pSrc_ptr; + + for (int i = 4; i > 0; i--) + { + Row<4>::idct(pTemp, pSrc); + pSrc += 8; + pTemp += 8; + } + + pTemp = temp; + for (int i = 8; i > 0; i--) + { + Col<4>::idct(pDst_ptr, pTemp); + pTemp++; + pDst_ptr++; + } +} + +// Retrieve one character from the input stream. +inline uint jpeg_decoder::get_char() +{ + // Any bytes remaining in buffer? + if (!m_in_buf_left) + { + // Try to get more bytes. + prep_in_buffer(); + // Still nothing to get? + if (!m_in_buf_left) + { + // Pad the end of the stream with 0xFF 0xD9 (EOI marker) + int t = m_tem_flag; + m_tem_flag ^= 1; + if (t) + return 0xD9; + else + return 0xFF; + } + } + + uint c = *m_pIn_buf_ofs++; + m_in_buf_left--; + + return c; +} + +// Same as previous method, except can indicate if the character is a pad character or not. +inline uint jpeg_decoder::get_char(bool *pPadding_flag) +{ + if (!m_in_buf_left) + { + prep_in_buffer(); + if (!m_in_buf_left) + { + *pPadding_flag = true; + int t = m_tem_flag; + m_tem_flag ^= 1; + if (t) + return 0xD9; + else + return 0xFF; + } + } + + *pPadding_flag = false; + + uint c = *m_pIn_buf_ofs++; + m_in_buf_left--; + + return c; +} + +// Inserts a previously retrieved character back into the input buffer. +inline void jpeg_decoder::stuff_char(uint8 q) +{ + *(--m_pIn_buf_ofs) = q; + m_in_buf_left++; +} + +// Retrieves one character from the input stream, but does not read past markers. Will continue to return 0xFF when a marker is encountered. +inline uint8 jpeg_decoder::get_octet() +{ + bool padding_flag; + int c = get_char(&padding_flag); + + if (c == 0xFF) + { + if (padding_flag) + return 0xFF; + + c = get_char(&padding_flag); + if (padding_flag) + { + stuff_char(0xFF); + return 0xFF; + } + + if (c == 0x00) + return 0xFF; + else + { + stuff_char(static_cast(c)); + stuff_char(0xFF); + return 0xFF; + } + } + + return static_cast(c); +} + +// Retrieves a variable number of bits from the input stream. Does not recognize markers. +inline uint jpeg_decoder::get_bits(int num_bits) +{ + if (!num_bits) + return 0; + + uint i = m_bit_buf >> (32 - num_bits); + + if ((m_bits_left -= num_bits) <= 0) + { + m_bit_buf <<= (num_bits += m_bits_left); + + uint c1 = get_char(); + uint c2 = get_char(); + m_bit_buf = (m_bit_buf & 0xFFFF0000) | (c1 << 8) | c2; + + m_bit_buf <<= -m_bits_left; + + m_bits_left += 16; + + JPGD_ASSERT(m_bits_left >= 0); + } + else + m_bit_buf <<= num_bits; + + return i; +} + +// Retrieves a variable number of bits from the input stream. Markers will not be read into the input bit buffer. Instead, an infinite number of all 1's will be returned when a marker is encountered. +inline uint jpeg_decoder::get_bits_no_markers(int num_bits) +{ + if (!num_bits) + return 0; + + uint i = m_bit_buf >> (32 - num_bits); + + if ((m_bits_left -= num_bits) <= 0) + { + m_bit_buf <<= (num_bits += m_bits_left); + + if ((m_in_buf_left < 2) || (m_pIn_buf_ofs[0] == 0xFF) || (m_pIn_buf_ofs[1] == 0xFF)) + { + uint c1 = get_octet(); + uint c2 = get_octet(); + m_bit_buf |= (c1 << 8) | c2; + } + else + { + m_bit_buf |= ((uint)m_pIn_buf_ofs[0] << 8) | m_pIn_buf_ofs[1]; + m_in_buf_left -= 2; + m_pIn_buf_ofs += 2; + } + + m_bit_buf <<= -m_bits_left; + + m_bits_left += 16; + + JPGD_ASSERT(m_bits_left >= 0); + } + else + m_bit_buf <<= num_bits; + + return i; +} + +// Decodes a Huffman encoded symbol. +inline int jpeg_decoder::huff_decode(huff_tables *pH) +{ + int symbol; + + // Check first 8-bits: do we have a complete symbol? + if ((symbol = pH->look_up[m_bit_buf >> 24]) < 0) + { + // Decode more bits, use a tree traversal to find symbol. + int ofs = 23; + do + { + symbol = pH->tree[-(int)(symbol + ((m_bit_buf >> ofs) & 1))]; + ofs--; + } while (symbol < 0); + + get_bits_no_markers(8 + (23 - ofs)); + } + else + get_bits_no_markers(pH->code_size[symbol]); + + return symbol; +} + +// Decodes a Huffman encoded symbol. +inline int jpeg_decoder::huff_decode(huff_tables *pH, int& extra_bits) +{ + int symbol; + + // Check first 8-bits: do we have a complete symbol? + if ((symbol = pH->look_up2[m_bit_buf >> 24]) < 0) + { + // Use a tree traversal to find symbol. + int ofs = 23; + do + { + symbol = pH->tree[-(int)(symbol + ((m_bit_buf >> ofs) & 1))]; + ofs--; + } while (symbol < 0); + + get_bits_no_markers(8 + (23 - ofs)); + + extra_bits = get_bits_no_markers(symbol & 0xF); + } + else + { + JPGD_ASSERT(((symbol >> 8) & 31) == pH->code_size[symbol & 255] + ((symbol & 0x8000) ? (symbol & 15) : 0)); + + if (symbol & 0x8000) + { + get_bits_no_markers((symbol >> 8) & 31); + extra_bits = symbol >> 16; + } + else + { + int code_size = (symbol >> 8) & 31; + int num_extra_bits = symbol & 0xF; + int bits = code_size + num_extra_bits; + if (bits <= (m_bits_left + 16)) + extra_bits = get_bits_no_markers(bits) & ((1 << num_extra_bits) - 1); + else + { + get_bits_no_markers(code_size); + extra_bits = get_bits_no_markers(num_extra_bits); + } + } + + symbol &= 0xFF; + } + + return symbol; +} + +// Tables and macro used to fully decode the DPCM differences. +static const int s_extend_test[16] = { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080, 0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 }; +static const int s_extend_offset[16] = { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1, ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1, ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1, ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 }; +static const int s_extend_mask[] = { 0, (1<<0), (1<<1), (1<<2), (1<<3), (1<<4), (1<<5), (1<<6), (1<<7), (1<<8), (1<<9), (1<<10), (1<<11), (1<<12), (1<<13), (1<<14), (1<<15), (1<<16) }; +// The logical AND's in this macro are to shut up static code analysis (aren't really necessary - couldn't find another way to do this) +#define JPGD_HUFF_EXTEND(x, s) (((x) < s_extend_test[s & 15]) ? ((x) + s_extend_offset[s & 15]) : (x)) + +// Clamps a value between 0-255. +inline uint8 jpeg_decoder::clamp(int i) +{ + if (static_cast(i) > 255) + i = (((~i) >> 31) & 0xFF); + + return static_cast(i); +} + +namespace DCT_Upsample +{ + struct Matrix44 + { + typedef int Element_Type; + enum { NUM_ROWS = 4, NUM_COLS = 4 }; + + Element_Type v[NUM_ROWS][NUM_COLS]; + + inline int rows() const { return NUM_ROWS; } + inline int cols() const { return NUM_COLS; } + + inline const Element_Type & at(int r, int c) const { return v[r][c]; } + inline Element_Type & at(int r, int c) { return v[r][c]; } + + inline Matrix44() { } + + inline Matrix44& operator += (const Matrix44& a) + { + for (int r = 0; r < NUM_ROWS; r++) + { + at(r, 0) += a.at(r, 0); + at(r, 1) += a.at(r, 1); + at(r, 2) += a.at(r, 2); + at(r, 3) += a.at(r, 3); + } + return *this; + } + + inline Matrix44& operator -= (const Matrix44& a) + { + for (int r = 0; r < NUM_ROWS; r++) + { + at(r, 0) -= a.at(r, 0); + at(r, 1) -= a.at(r, 1); + at(r, 2) -= a.at(r, 2); + at(r, 3) -= a.at(r, 3); + } + return *this; + } + + friend inline Matrix44 operator + (const Matrix44& a, const Matrix44& b) + { + Matrix44 ret; + for (int r = 0; r < NUM_ROWS; r++) + { + ret.at(r, 0) = a.at(r, 0) + b.at(r, 0); + ret.at(r, 1) = a.at(r, 1) + b.at(r, 1); + ret.at(r, 2) = a.at(r, 2) + b.at(r, 2); + ret.at(r, 3) = a.at(r, 3) + b.at(r, 3); + } + return ret; + } + + friend inline Matrix44 operator - (const Matrix44& a, const Matrix44& b) + { + Matrix44 ret; + for (int r = 0; r < NUM_ROWS; r++) + { + ret.at(r, 0) = a.at(r, 0) - b.at(r, 0); + ret.at(r, 1) = a.at(r, 1) - b.at(r, 1); + ret.at(r, 2) = a.at(r, 2) - b.at(r, 2); + ret.at(r, 3) = a.at(r, 3) - b.at(r, 3); + } + return ret; + } + + static inline void add_and_store(jpgd_block_t* pDst, const Matrix44& a, const Matrix44& b) + { + for (int r = 0; r < 4; r++) + { + pDst[0*8 + r] = static_cast(a.at(r, 0) + b.at(r, 0)); + pDst[1*8 + r] = static_cast(a.at(r, 1) + b.at(r, 1)); + pDst[2*8 + r] = static_cast(a.at(r, 2) + b.at(r, 2)); + pDst[3*8 + r] = static_cast(a.at(r, 3) + b.at(r, 3)); + } + } + + static inline void sub_and_store(jpgd_block_t* pDst, const Matrix44& a, const Matrix44& b) + { + for (int r = 0; r < 4; r++) + { + pDst[0*8 + r] = static_cast(a.at(r, 0) - b.at(r, 0)); + pDst[1*8 + r] = static_cast(a.at(r, 1) - b.at(r, 1)); + pDst[2*8 + r] = static_cast(a.at(r, 2) - b.at(r, 2)); + pDst[3*8 + r] = static_cast(a.at(r, 3) - b.at(r, 3)); + } + } + }; + + const int FRACT_BITS = 10; + const int SCALE = 1 << FRACT_BITS; + + typedef int Temp_Type; + #define D(i) (((i) + (SCALE >> 1)) >> FRACT_BITS) + #define F(i) ((int)((i) * SCALE + .5f)) + + // Any decent C++ compiler will optimize this at compile time to a 0, or an array access. + #define AT(c, r) ((((c)>=NUM_COLS)||((r)>=NUM_ROWS)) ? 0 : pSrc[(c)+(r)*8]) + + // NUM_ROWS/NUM_COLS = # of non-zero rows/cols in input matrix + template + struct P_Q + { + static void calc(Matrix44& P, Matrix44& Q, const jpgd_block_t* pSrc) + { + // 4x8 = 4x8 times 8x8, matrix 0 is constant + const Temp_Type X000 = AT(0, 0); + const Temp_Type X001 = AT(0, 1); + const Temp_Type X002 = AT(0, 2); + const Temp_Type X003 = AT(0, 3); + const Temp_Type X004 = AT(0, 4); + const Temp_Type X005 = AT(0, 5); + const Temp_Type X006 = AT(0, 6); + const Temp_Type X007 = AT(0, 7); + const Temp_Type X010 = D(F(0.415735f) * AT(1, 0) + F(0.791065f) * AT(3, 0) + F(-0.352443f) * AT(5, 0) + F(0.277785f) * AT(7, 0)); + const Temp_Type X011 = D(F(0.415735f) * AT(1, 1) + F(0.791065f) * AT(3, 1) + F(-0.352443f) * AT(5, 1) + F(0.277785f) * AT(7, 1)); + const Temp_Type X012 = D(F(0.415735f) * AT(1, 2) + F(0.791065f) * AT(3, 2) + F(-0.352443f) * AT(5, 2) + F(0.277785f) * AT(7, 2)); + const Temp_Type X013 = D(F(0.415735f) * AT(1, 3) + F(0.791065f) * AT(3, 3) + F(-0.352443f) * AT(5, 3) + F(0.277785f) * AT(7, 3)); + const Temp_Type X014 = D(F(0.415735f) * AT(1, 4) + F(0.791065f) * AT(3, 4) + F(-0.352443f) * AT(5, 4) + F(0.277785f) * AT(7, 4)); + const Temp_Type X015 = D(F(0.415735f) * AT(1, 5) + F(0.791065f) * AT(3, 5) + F(-0.352443f) * AT(5, 5) + F(0.277785f) * AT(7, 5)); + const Temp_Type X016 = D(F(0.415735f) * AT(1, 6) + F(0.791065f) * AT(3, 6) + F(-0.352443f) * AT(5, 6) + F(0.277785f) * AT(7, 6)); + const Temp_Type X017 = D(F(0.415735f) * AT(1, 7) + F(0.791065f) * AT(3, 7) + F(-0.352443f) * AT(5, 7) + F(0.277785f) * AT(7, 7)); + const Temp_Type X020 = AT(4, 0); + const Temp_Type X021 = AT(4, 1); + const Temp_Type X022 = AT(4, 2); + const Temp_Type X023 = AT(4, 3); + const Temp_Type X024 = AT(4, 4); + const Temp_Type X025 = AT(4, 5); + const Temp_Type X026 = AT(4, 6); + const Temp_Type X027 = AT(4, 7); + const Temp_Type X030 = D(F(0.022887f) * AT(1, 0) + F(-0.097545f) * AT(3, 0) + F(0.490393f) * AT(5, 0) + F(0.865723f) * AT(7, 0)); + const Temp_Type X031 = D(F(0.022887f) * AT(1, 1) + F(-0.097545f) * AT(3, 1) + F(0.490393f) * AT(5, 1) + F(0.865723f) * AT(7, 1)); + const Temp_Type X032 = D(F(0.022887f) * AT(1, 2) + F(-0.097545f) * AT(3, 2) + F(0.490393f) * AT(5, 2) + F(0.865723f) * AT(7, 2)); + const Temp_Type X033 = D(F(0.022887f) * AT(1, 3) + F(-0.097545f) * AT(3, 3) + F(0.490393f) * AT(5, 3) + F(0.865723f) * AT(7, 3)); + const Temp_Type X034 = D(F(0.022887f) * AT(1, 4) + F(-0.097545f) * AT(3, 4) + F(0.490393f) * AT(5, 4) + F(0.865723f) * AT(7, 4)); + const Temp_Type X035 = D(F(0.022887f) * AT(1, 5) + F(-0.097545f) * AT(3, 5) + F(0.490393f) * AT(5, 5) + F(0.865723f) * AT(7, 5)); + const Temp_Type X036 = D(F(0.022887f) * AT(1, 6) + F(-0.097545f) * AT(3, 6) + F(0.490393f) * AT(5, 6) + F(0.865723f) * AT(7, 6)); + const Temp_Type X037 = D(F(0.022887f) * AT(1, 7) + F(-0.097545f) * AT(3, 7) + F(0.490393f) * AT(5, 7) + F(0.865723f) * AT(7, 7)); + + // 4x4 = 4x8 times 8x4, matrix 1 is constant + P.at(0, 0) = X000; + P.at(0, 1) = D(X001 * F(0.415735f) + X003 * F(0.791065f) + X005 * F(-0.352443f) + X007 * F(0.277785f)); + P.at(0, 2) = X004; + P.at(0, 3) = D(X001 * F(0.022887f) + X003 * F(-0.097545f) + X005 * F(0.490393f) + X007 * F(0.865723f)); + P.at(1, 0) = X010; + P.at(1, 1) = D(X011 * F(0.415735f) + X013 * F(0.791065f) + X015 * F(-0.352443f) + X017 * F(0.277785f)); + P.at(1, 2) = X014; + P.at(1, 3) = D(X011 * F(0.022887f) + X013 * F(-0.097545f) + X015 * F(0.490393f) + X017 * F(0.865723f)); + P.at(2, 0) = X020; + P.at(2, 1) = D(X021 * F(0.415735f) + X023 * F(0.791065f) + X025 * F(-0.352443f) + X027 * F(0.277785f)); + P.at(2, 2) = X024; + P.at(2, 3) = D(X021 * F(0.022887f) + X023 * F(-0.097545f) + X025 * F(0.490393f) + X027 * F(0.865723f)); + P.at(3, 0) = X030; + P.at(3, 1) = D(X031 * F(0.415735f) + X033 * F(0.791065f) + X035 * F(-0.352443f) + X037 * F(0.277785f)); + P.at(3, 2) = X034; + P.at(3, 3) = D(X031 * F(0.022887f) + X033 * F(-0.097545f) + X035 * F(0.490393f) + X037 * F(0.865723f)); + // 40 muls 24 adds + + // 4x4 = 4x8 times 8x4, matrix 1 is constant + Q.at(0, 0) = D(X001 * F(0.906127f) + X003 * F(-0.318190f) + X005 * F(0.212608f) + X007 * F(-0.180240f)); + Q.at(0, 1) = X002; + Q.at(0, 2) = D(X001 * F(-0.074658f) + X003 * F(0.513280f) + X005 * F(0.768178f) + X007 * F(-0.375330f)); + Q.at(0, 3) = X006; + Q.at(1, 0) = D(X011 * F(0.906127f) + X013 * F(-0.318190f) + X015 * F(0.212608f) + X017 * F(-0.180240f)); + Q.at(1, 1) = X012; + Q.at(1, 2) = D(X011 * F(-0.074658f) + X013 * F(0.513280f) + X015 * F(0.768178f) + X017 * F(-0.375330f)); + Q.at(1, 3) = X016; + Q.at(2, 0) = D(X021 * F(0.906127f) + X023 * F(-0.318190f) + X025 * F(0.212608f) + X027 * F(-0.180240f)); + Q.at(2, 1) = X022; + Q.at(2, 2) = D(X021 * F(-0.074658f) + X023 * F(0.513280f) + X025 * F(0.768178f) + X027 * F(-0.375330f)); + Q.at(2, 3) = X026; + Q.at(3, 0) = D(X031 * F(0.906127f) + X033 * F(-0.318190f) + X035 * F(0.212608f) + X037 * F(-0.180240f)); + Q.at(3, 1) = X032; + Q.at(3, 2) = D(X031 * F(-0.074658f) + X033 * F(0.513280f) + X035 * F(0.768178f) + X037 * F(-0.375330f)); + Q.at(3, 3) = X036; + // 40 muls 24 adds + } + }; + + template + struct R_S + { + static void calc(Matrix44& R, Matrix44& S, const jpgd_block_t* pSrc) + { + // 4x8 = 4x8 times 8x8, matrix 0 is constant + const Temp_Type X100 = D(F(0.906127f) * AT(1, 0) + F(-0.318190f) * AT(3, 0) + F(0.212608f) * AT(5, 0) + F(-0.180240f) * AT(7, 0)); + const Temp_Type X101 = D(F(0.906127f) * AT(1, 1) + F(-0.318190f) * AT(3, 1) + F(0.212608f) * AT(5, 1) + F(-0.180240f) * AT(7, 1)); + const Temp_Type X102 = D(F(0.906127f) * AT(1, 2) + F(-0.318190f) * AT(3, 2) + F(0.212608f) * AT(5, 2) + F(-0.180240f) * AT(7, 2)); + const Temp_Type X103 = D(F(0.906127f) * AT(1, 3) + F(-0.318190f) * AT(3, 3) + F(0.212608f) * AT(5, 3) + F(-0.180240f) * AT(7, 3)); + const Temp_Type X104 = D(F(0.906127f) * AT(1, 4) + F(-0.318190f) * AT(3, 4) + F(0.212608f) * AT(5, 4) + F(-0.180240f) * AT(7, 4)); + const Temp_Type X105 = D(F(0.906127f) * AT(1, 5) + F(-0.318190f) * AT(3, 5) + F(0.212608f) * AT(5, 5) + F(-0.180240f) * AT(7, 5)); + const Temp_Type X106 = D(F(0.906127f) * AT(1, 6) + F(-0.318190f) * AT(3, 6) + F(0.212608f) * AT(5, 6) + F(-0.180240f) * AT(7, 6)); + const Temp_Type X107 = D(F(0.906127f) * AT(1, 7) + F(-0.318190f) * AT(3, 7) + F(0.212608f) * AT(5, 7) + F(-0.180240f) * AT(7, 7)); + const Temp_Type X110 = AT(2, 0); + const Temp_Type X111 = AT(2, 1); + const Temp_Type X112 = AT(2, 2); + const Temp_Type X113 = AT(2, 3); + const Temp_Type X114 = AT(2, 4); + const Temp_Type X115 = AT(2, 5); + const Temp_Type X116 = AT(2, 6); + const Temp_Type X117 = AT(2, 7); + const Temp_Type X120 = D(F(-0.074658f) * AT(1, 0) + F(0.513280f) * AT(3, 0) + F(0.768178f) * AT(5, 0) + F(-0.375330f) * AT(7, 0)); + const Temp_Type X121 = D(F(-0.074658f) * AT(1, 1) + F(0.513280f) * AT(3, 1) + F(0.768178f) * AT(5, 1) + F(-0.375330f) * AT(7, 1)); + const Temp_Type X122 = D(F(-0.074658f) * AT(1, 2) + F(0.513280f) * AT(3, 2) + F(0.768178f) * AT(5, 2) + F(-0.375330f) * AT(7, 2)); + const Temp_Type X123 = D(F(-0.074658f) * AT(1, 3) + F(0.513280f) * AT(3, 3) + F(0.768178f) * AT(5, 3) + F(-0.375330f) * AT(7, 3)); + const Temp_Type X124 = D(F(-0.074658f) * AT(1, 4) + F(0.513280f) * AT(3, 4) + F(0.768178f) * AT(5, 4) + F(-0.375330f) * AT(7, 4)); + const Temp_Type X125 = D(F(-0.074658f) * AT(1, 5) + F(0.513280f) * AT(3, 5) + F(0.768178f) * AT(5, 5) + F(-0.375330f) * AT(7, 5)); + const Temp_Type X126 = D(F(-0.074658f) * AT(1, 6) + F(0.513280f) * AT(3, 6) + F(0.768178f) * AT(5, 6) + F(-0.375330f) * AT(7, 6)); + const Temp_Type X127 = D(F(-0.074658f) * AT(1, 7) + F(0.513280f) * AT(3, 7) + F(0.768178f) * AT(5, 7) + F(-0.375330f) * AT(7, 7)); + const Temp_Type X130 = AT(6, 0); + const Temp_Type X131 = AT(6, 1); + const Temp_Type X132 = AT(6, 2); + const Temp_Type X133 = AT(6, 3); + const Temp_Type X134 = AT(6, 4); + const Temp_Type X135 = AT(6, 5); + const Temp_Type X136 = AT(6, 6); + const Temp_Type X137 = AT(6, 7); + // 80 muls 48 adds + + // 4x4 = 4x8 times 8x4, matrix 1 is constant + R.at(0, 0) = X100; + R.at(0, 1) = D(X101 * F(0.415735f) + X103 * F(0.791065f) + X105 * F(-0.352443f) + X107 * F(0.277785f)); + R.at(0, 2) = X104; + R.at(0, 3) = D(X101 * F(0.022887f) + X103 * F(-0.097545f) + X105 * F(0.490393f) + X107 * F(0.865723f)); + R.at(1, 0) = X110; + R.at(1, 1) = D(X111 * F(0.415735f) + X113 * F(0.791065f) + X115 * F(-0.352443f) + X117 * F(0.277785f)); + R.at(1, 2) = X114; + R.at(1, 3) = D(X111 * F(0.022887f) + X113 * F(-0.097545f) + X115 * F(0.490393f) + X117 * F(0.865723f)); + R.at(2, 0) = X120; + R.at(2, 1) = D(X121 * F(0.415735f) + X123 * F(0.791065f) + X125 * F(-0.352443f) + X127 * F(0.277785f)); + R.at(2, 2) = X124; + R.at(2, 3) = D(X121 * F(0.022887f) + X123 * F(-0.097545f) + X125 * F(0.490393f) + X127 * F(0.865723f)); + R.at(3, 0) = X130; + R.at(3, 1) = D(X131 * F(0.415735f) + X133 * F(0.791065f) + X135 * F(-0.352443f) + X137 * F(0.277785f)); + R.at(3, 2) = X134; + R.at(3, 3) = D(X131 * F(0.022887f) + X133 * F(-0.097545f) + X135 * F(0.490393f) + X137 * F(0.865723f)); + // 40 muls 24 adds + // 4x4 = 4x8 times 8x4, matrix 1 is constant + S.at(0, 0) = D(X101 * F(0.906127f) + X103 * F(-0.318190f) + X105 * F(0.212608f) + X107 * F(-0.180240f)); + S.at(0, 1) = X102; + S.at(0, 2) = D(X101 * F(-0.074658f) + X103 * F(0.513280f) + X105 * F(0.768178f) + X107 * F(-0.375330f)); + S.at(0, 3) = X106; + S.at(1, 0) = D(X111 * F(0.906127f) + X113 * F(-0.318190f) + X115 * F(0.212608f) + X117 * F(-0.180240f)); + S.at(1, 1) = X112; + S.at(1, 2) = D(X111 * F(-0.074658f) + X113 * F(0.513280f) + X115 * F(0.768178f) + X117 * F(-0.375330f)); + S.at(1, 3) = X116; + S.at(2, 0) = D(X121 * F(0.906127f) + X123 * F(-0.318190f) + X125 * F(0.212608f) + X127 * F(-0.180240f)); + S.at(2, 1) = X122; + S.at(2, 2) = D(X121 * F(-0.074658f) + X123 * F(0.513280f) + X125 * F(0.768178f) + X127 * F(-0.375330f)); + S.at(2, 3) = X126; + S.at(3, 0) = D(X131 * F(0.906127f) + X133 * F(-0.318190f) + X135 * F(0.212608f) + X137 * F(-0.180240f)); + S.at(3, 1) = X132; + S.at(3, 2) = D(X131 * F(-0.074658f) + X133 * F(0.513280f) + X135 * F(0.768178f) + X137 * F(-0.375330f)); + S.at(3, 3) = X136; + // 40 muls 24 adds + } + }; +} // end namespace DCT_Upsample + +// Unconditionally frees all allocated m_blocks. +void jpeg_decoder::free_all_blocks() +{ + m_pStream = NULL; + for (mem_block *b = m_pMem_blocks; b; ) + { + mem_block *n = b->m_pNext; + jpgd_free(b); + b = n; + } + m_pMem_blocks = NULL; +} + +// This method handles all errors. It will never return. +// It could easily be changed to use C++ exceptions. +JPGD_NORETURN void jpeg_decoder::stop_decoding(jpgd_status status) +{ + m_error_code = status; + free_all_blocks(); + longjmp(m_jmp_state, status); +} + +void *jpeg_decoder::alloc(size_t nSize, bool zero) +{ + nSize = (JPGD_MAX(nSize, 1) + 3) & ~3; + char *rv = NULL; + for (mem_block *b = m_pMem_blocks; b; b = b->m_pNext) + { + if ((b->m_used_count + nSize) <= b->m_size) + { + rv = b->m_data + b->m_used_count; + b->m_used_count += nSize; + break; + } + } + if (!rv) + { + int capacity = JPGD_MAX(32768 - 256, (nSize + 2047) & ~2047); + mem_block *b = (mem_block*)jpgd_malloc(sizeof(mem_block) + capacity); + if (!b) { stop_decoding(JPGD_NOTENOUGHMEM); } + b->m_pNext = m_pMem_blocks; m_pMem_blocks = b; + b->m_used_count = nSize; + b->m_size = capacity; + rv = b->m_data; + } + if (zero) memset(rv, 0, nSize); + return rv; +} + +void jpeg_decoder::word_clear(void *p, uint16 c, uint n) +{ + uint8 *pD = (uint8*)p; + const uint8 l = c & 0xFF, h = (c >> 8) & 0xFF; + while (n) + { + pD[0] = l; pD[1] = h; pD += 2; + n--; + } +} + +// Refill the input buffer. +// This method will sit in a loop until (A) the buffer is full or (B) +// the stream's read() method reports and end of file condition. +void jpeg_decoder::prep_in_buffer() +{ + m_in_buf_left = 0; + m_pIn_buf_ofs = m_in_buf; + + if (m_eof_flag) + return; + + do + { + int bytes_read = m_pStream->read(m_in_buf + m_in_buf_left, JPGD_IN_BUF_SIZE - m_in_buf_left, &m_eof_flag); + if (bytes_read == -1) + stop_decoding(JPGD_STREAM_READ); + + m_in_buf_left += bytes_read; + } while ((m_in_buf_left < JPGD_IN_BUF_SIZE) && (!m_eof_flag)); + + m_total_bytes_read += m_in_buf_left; + + // Pad the end of the block with M_EOI (prevents the decompressor from going off the rails if the stream is invalid). + // (This dates way back to when this decompressor was written in C/asm, and the all-asm Huffman decoder did some fancy things to increase perf.) + word_clear(m_pIn_buf_ofs + m_in_buf_left, 0xD9FF, 64); +} + +// Read a Huffman code table. +void jpeg_decoder::read_dht_marker() +{ + int i, index, count; + uint8 huff_num[17]; + uint8 huff_val[256]; + + uint num_left = get_bits(16); + + if (num_left < 2) + stop_decoding(JPGD_BAD_DHT_MARKER); + + num_left -= 2; + + while (num_left) + { + index = get_bits(8); + + huff_num[0] = 0; + + count = 0; + + for (i = 1; i <= 16; i++) + { + huff_num[i] = static_cast(get_bits(8)); + count += huff_num[i]; + } + + if (count > 255) + stop_decoding(JPGD_BAD_DHT_COUNTS); + + for (i = 0; i < count; i++) + huff_val[i] = static_cast(get_bits(8)); + + i = 1 + 16 + count; + + if (num_left < (uint)i) + stop_decoding(JPGD_BAD_DHT_MARKER); + + num_left -= i; + + if ((index & 0x10) > 0x10) + stop_decoding(JPGD_BAD_DHT_INDEX); + + index = (index & 0x0F) + ((index & 0x10) >> 4) * (JPGD_MAX_HUFF_TABLES >> 1); + + if (index >= JPGD_MAX_HUFF_TABLES) + stop_decoding(JPGD_BAD_DHT_INDEX); + + if (!m_huff_num[index]) + m_huff_num[index] = (uint8 *)alloc(17); + + if (!m_huff_val[index]) + m_huff_val[index] = (uint8 *)alloc(256); + + m_huff_ac[index] = (index & 0x10) != 0; + memcpy(m_huff_num[index], huff_num, 17); + memcpy(m_huff_val[index], huff_val, 256); + } +} + +// Read a quantization table. +void jpeg_decoder::read_dqt_marker() +{ + int n, i, prec; + uint num_left; + uint temp; + + num_left = get_bits(16); + + if (num_left < 2) + stop_decoding(JPGD_BAD_DQT_MARKER); + + num_left -= 2; + + while (num_left) + { + n = get_bits(8); + prec = n >> 4; + n &= 0x0F; + + if (n >= JPGD_MAX_QUANT_TABLES) + stop_decoding(JPGD_BAD_DQT_TABLE); + + if (!m_quant[n]) + m_quant[n] = (jpgd_quant_t *)alloc(64 * sizeof(jpgd_quant_t)); + + // read quantization entries, in zag order + for (i = 0; i < 64; i++) + { + temp = get_bits(8); + + if (prec) + temp = (temp << 8) + get_bits(8); + + m_quant[n][i] = static_cast(temp); + } + + i = 64 + 1; + + if (prec) + i += 64; + + if (num_left < (uint)i) + stop_decoding(JPGD_BAD_DQT_LENGTH); + + num_left -= i; + } +} + +// Read the start of frame (SOF) marker. +void jpeg_decoder::read_sof_marker() +{ + int i; + uint num_left; + + num_left = get_bits(16); + + if (get_bits(8) != 8) /* precision: sorry, only 8-bit precision is supported right now */ + stop_decoding(JPGD_BAD_PRECISION); + + m_image_y_size = get_bits(16); + + if ((m_image_y_size < 1) || (m_image_y_size > JPGD_MAX_HEIGHT)) + stop_decoding(JPGD_BAD_HEIGHT); + + m_image_x_size = get_bits(16); + + if ((m_image_x_size < 1) || (m_image_x_size > JPGD_MAX_WIDTH)) + stop_decoding(JPGD_BAD_WIDTH); + + m_comps_in_frame = get_bits(8); + + if (m_comps_in_frame > JPGD_MAX_COMPONENTS) + stop_decoding(JPGD_TOO_MANY_COMPONENTS); + + if (num_left != (uint)(m_comps_in_frame * 3 + 8)) + stop_decoding(JPGD_BAD_SOF_LENGTH); + + for (i = 0; i < m_comps_in_frame; i++) + { + m_comp_ident[i] = get_bits(8); + m_comp_h_samp[i] = get_bits(4); + m_comp_v_samp[i] = get_bits(4); + m_comp_quant[i] = get_bits(8); + } +} + +// Used to skip unrecognized markers. +void jpeg_decoder::skip_variable_marker() +{ + uint num_left; + + num_left = get_bits(16); + + if (num_left < 2) + stop_decoding(JPGD_BAD_VARIABLE_MARKER); + + num_left -= 2; + + while (num_left) + { + get_bits(8); + num_left--; + } +} + +// Read a define restart interval (DRI) marker. +void jpeg_decoder::read_dri_marker() +{ + if (get_bits(16) != 4) + stop_decoding(JPGD_BAD_DRI_LENGTH); + + m_restart_interval = get_bits(16); +} + +// Read a start of scan (SOS) marker. +void jpeg_decoder::read_sos_marker() +{ + uint num_left; + int i, ci, n, c, cc; + + num_left = get_bits(16); + + n = get_bits(8); + + m_comps_in_scan = n; + + num_left -= 3; + + if ( (num_left != (uint)(n * 2 + 3)) || (n < 1) || (n > JPGD_MAX_COMPS_IN_SCAN) ) + stop_decoding(JPGD_BAD_SOS_LENGTH); + + for (i = 0; i < n; i++) + { + cc = get_bits(8); + c = get_bits(8); + num_left -= 2; + + for (ci = 0; ci < m_comps_in_frame; ci++) + if (cc == m_comp_ident[ci]) + break; + + if (ci >= m_comps_in_frame) + stop_decoding(JPGD_BAD_SOS_COMP_ID); + + m_comp_list[i] = ci; + m_comp_dc_tab[ci] = (c >> 4) & 15; + m_comp_ac_tab[ci] = (c & 15) + (JPGD_MAX_HUFF_TABLES >> 1); + } + + m_spectral_start = get_bits(8); + m_spectral_end = get_bits(8); + m_successive_high = get_bits(4); + m_successive_low = get_bits(4); + + if (!m_progressive_flag) + { + m_spectral_start = 0; + m_spectral_end = 63; + } + + num_left -= 3; + + while (num_left) /* read past whatever is num_left */ + { + get_bits(8); + num_left--; + } +} + +// Finds the next marker. +int jpeg_decoder::next_marker() +{ + uint c, bytes; + + bytes = 0; + + do + { + do + { + bytes++; + c = get_bits(8); + } while (c != 0xFF); + + do + { + c = get_bits(8); + } while (c == 0xFF); + + } while (c == 0); + + // If bytes > 0 here, there where extra bytes before the marker (not good). + + return c; +} + +// Process markers. Returns when an SOFx, SOI, EOI, or SOS marker is +// encountered. +int jpeg_decoder::process_markers() +{ + int c; + + for ( ; ; ) + { + c = next_marker(); + + switch (c) + { + case M_SOF0: + case M_SOF1: + case M_SOF2: + case M_SOF3: + case M_SOF5: + case M_SOF6: + case M_SOF7: +// case M_JPG: + case M_SOF9: + case M_SOF10: + case M_SOF11: + case M_SOF13: + case M_SOF14: + case M_SOF15: + case M_SOI: + case M_EOI: + case M_SOS: + { + return c; + } + case M_DHT: + { + read_dht_marker(); + break; + } + // No arithmitic support - dumb patents! + case M_DAC: + { + stop_decoding(JPGD_NO_ARITHMITIC_SUPPORT); + break; + } + case M_DQT: + { + read_dqt_marker(); + break; + } + case M_DRI: + { + read_dri_marker(); + break; + } + //case M_APP0: /* no need to read the JFIF marker */ + + case M_JPG: + case M_RST0: /* no parameters */ + case M_RST1: + case M_RST2: + case M_RST3: + case M_RST4: + case M_RST5: + case M_RST6: + case M_RST7: + case M_TEM: + { + stop_decoding(JPGD_UNEXPECTED_MARKER); + break; + } + default: /* must be DNL, DHP, EXP, APPn, JPGn, COM, or RESn or APP0 */ + { + skip_variable_marker(); + break; + } + } + } +} + +// Finds the start of image (SOI) marker. +// This code is rather defensive: it only checks the first 512 bytes to avoid +// false positives. +void jpeg_decoder::locate_soi_marker() +{ + uint lastchar, thischar; + uint bytesleft; + + lastchar = get_bits(8); + + thischar = get_bits(8); + + /* ok if it's a normal JPEG file without a special header */ + + if ((lastchar == 0xFF) && (thischar == M_SOI)) + return; + + bytesleft = 4096; //512; + + for ( ; ; ) + { + if (--bytesleft == 0) + stop_decoding(JPGD_NOT_JPEG); + + lastchar = thischar; + + thischar = get_bits(8); + + if (lastchar == 0xFF) + { + if (thischar == M_SOI) + break; + else if (thischar == M_EOI) // get_bits will keep returning M_EOI if we read past the end + stop_decoding(JPGD_NOT_JPEG); + } + } + + // Check the next character after marker: if it's not 0xFF, it can't be the start of the next marker, so the file is bad. + thischar = (m_bit_buf >> 24) & 0xFF; + + if (thischar != 0xFF) + stop_decoding(JPGD_NOT_JPEG); +} + +// Find a start of frame (SOF) marker. +void jpeg_decoder::locate_sof_marker() +{ + locate_soi_marker(); + + int c = process_markers(); + + switch (c) + { + case M_SOF2: + m_progressive_flag = JPGD_TRUE; + case M_SOF0: /* baseline DCT */ + case M_SOF1: /* extended sequential DCT */ + { + read_sof_marker(); + break; + } + case M_SOF9: /* Arithmitic coding */ + { + stop_decoding(JPGD_NO_ARITHMITIC_SUPPORT); + break; + } + default: + { + stop_decoding(JPGD_UNSUPPORTED_MARKER); + break; + } + } +} + +// Find a start of scan (SOS) marker. +int jpeg_decoder::locate_sos_marker() +{ + int c; + + c = process_markers(); + + if (c == M_EOI) + return JPGD_FALSE; + else if (c != M_SOS) + stop_decoding(JPGD_UNEXPECTED_MARKER); + + read_sos_marker(); + + return JPGD_TRUE; +} + +// Reset everything to default/uninitialized state. +void jpeg_decoder::init(jpeg_decoder_stream *pStream) +{ + m_pMem_blocks = NULL; + m_error_code = JPGD_SUCCESS; + m_ready_flag = false; + m_image_x_size = m_image_y_size = 0; + m_pStream = pStream; + m_progressive_flag = JPGD_FALSE; + + memset(m_huff_ac, 0, sizeof(m_huff_ac)); + memset(m_huff_num, 0, sizeof(m_huff_num)); + memset(m_huff_val, 0, sizeof(m_huff_val)); + memset(m_quant, 0, sizeof(m_quant)); + + m_scan_type = 0; + m_comps_in_frame = 0; + + memset(m_comp_h_samp, 0, sizeof(m_comp_h_samp)); + memset(m_comp_v_samp, 0, sizeof(m_comp_v_samp)); + memset(m_comp_quant, 0, sizeof(m_comp_quant)); + memset(m_comp_ident, 0, sizeof(m_comp_ident)); + memset(m_comp_h_blocks, 0, sizeof(m_comp_h_blocks)); + memset(m_comp_v_blocks, 0, sizeof(m_comp_v_blocks)); + + m_comps_in_scan = 0; + memset(m_comp_list, 0, sizeof(m_comp_list)); + memset(m_comp_dc_tab, 0, sizeof(m_comp_dc_tab)); + memset(m_comp_ac_tab, 0, sizeof(m_comp_ac_tab)); + + m_spectral_start = 0; + m_spectral_end = 0; + m_successive_low = 0; + m_successive_high = 0; + m_max_mcu_x_size = 0; + m_max_mcu_y_size = 0; + m_blocks_per_mcu = 0; + m_max_blocks_per_row = 0; + m_mcus_per_row = 0; + m_mcus_per_col = 0; + m_expanded_blocks_per_component = 0; + m_expanded_blocks_per_mcu = 0; + m_expanded_blocks_per_row = 0; + m_freq_domain_chroma_upsample = false; + + memset(m_mcu_org, 0, sizeof(m_mcu_org)); + + m_total_lines_left = 0; + m_mcu_lines_left = 0; + m_real_dest_bytes_per_scan_line = 0; + m_dest_bytes_per_scan_line = 0; + m_dest_bytes_per_pixel = 0; + + memset(m_pHuff_tabs, 0, sizeof(m_pHuff_tabs)); + + memset(m_dc_coeffs, 0, sizeof(m_dc_coeffs)); + memset(m_ac_coeffs, 0, sizeof(m_ac_coeffs)); + memset(m_block_y_mcu, 0, sizeof(m_block_y_mcu)); + + m_eob_run = 0; + + memset(m_block_y_mcu, 0, sizeof(m_block_y_mcu)); + + m_pIn_buf_ofs = m_in_buf; + m_in_buf_left = 0; + m_eof_flag = false; + m_tem_flag = 0; + + memset(m_in_buf_pad_start, 0, sizeof(m_in_buf_pad_start)); + memset(m_in_buf, 0, sizeof(m_in_buf)); + memset(m_in_buf_pad_end, 0, sizeof(m_in_buf_pad_end)); + + m_restart_interval = 0; + m_restarts_left = 0; + m_next_restart_num = 0; + + m_max_mcus_per_row = 0; + m_max_blocks_per_mcu = 0; + m_max_mcus_per_col = 0; + + memset(m_last_dc_val, 0, sizeof(m_last_dc_val)); + m_pMCU_coefficients = NULL; + m_pSample_buf = NULL; + + m_total_bytes_read = 0; + + m_pScan_line_0 = NULL; + m_pScan_line_1 = NULL; + + // Ready the input buffer. + prep_in_buffer(); + + // Prime the bit buffer. + m_bits_left = 16; + m_bit_buf = 0; + + get_bits(16); + get_bits(16); + + for (int i = 0; i < JPGD_MAX_BLOCKS_PER_MCU; i++) + m_mcu_block_max_zag[i] = 64; +} + +#define SCALEBITS 16 +#define ONE_HALF ((int) 1 << (SCALEBITS-1)) +#define FIX(x) ((int) ((x) * (1L<> SCALEBITS; + m_cbb[i] = ( FIX(1.77200f) * k + ONE_HALF) >> SCALEBITS; + m_crg[i] = (-FIX(0.71414f)) * k; + m_cbg[i] = (-FIX(0.34414f)) * k + ONE_HALF; + } +} + +// This method throws back into the stream any bytes that where read +// into the bit buffer during initial marker scanning. +void jpeg_decoder::fix_in_buffer() +{ + // In case any 0xFF's where pulled into the buffer during marker scanning. + JPGD_ASSERT((m_bits_left & 7) == 0); + + if (m_bits_left == 16) + stuff_char( (uint8)(m_bit_buf & 0xFF)); + + if (m_bits_left >= 8) + stuff_char( (uint8)((m_bit_buf >> 8) & 0xFF)); + + stuff_char((uint8)((m_bit_buf >> 16) & 0xFF)); + stuff_char((uint8)((m_bit_buf >> 24) & 0xFF)); + + m_bits_left = 16; + get_bits_no_markers(16); + get_bits_no_markers(16); +} + +void jpeg_decoder::transform_mcu(int mcu_row) +{ + jpgd_block_t* pSrc_ptr = m_pMCU_coefficients; + uint8* pDst_ptr = m_pSample_buf + mcu_row * m_blocks_per_mcu * 64; + + for (int mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) + { + idct(pSrc_ptr, pDst_ptr, m_mcu_block_max_zag[mcu_block]); + pSrc_ptr += 64; + pDst_ptr += 64; + } +} + +static const uint8 s_max_rc[64] = +{ + 17, 18, 34, 50, 50, 51, 52, 52, 52, 68, 84, 84, 84, 84, 85, 86, 86, 86, 86, 86, + 102, 118, 118, 118, 118, 118, 118, 119, 120, 120, 120, 120, 120, 120, 120, 136, + 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, + 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136 +}; + +void jpeg_decoder::transform_mcu_expand(int mcu_row) +{ + jpgd_block_t* pSrc_ptr = m_pMCU_coefficients; + uint8* pDst_ptr = m_pSample_buf + mcu_row * m_expanded_blocks_per_mcu * 64; + + // Y IDCT + int mcu_block; + for (mcu_block = 0; mcu_block < m_expanded_blocks_per_component; mcu_block++) + { + idct(pSrc_ptr, pDst_ptr, m_mcu_block_max_zag[mcu_block]); + pSrc_ptr += 64; + pDst_ptr += 64; + } + + // Chroma IDCT, with upsampling + jpgd_block_t temp_block[64]; + + for (int i = 0; i < 2; i++) + { + DCT_Upsample::Matrix44 P, Q, R, S; + + JPGD_ASSERT(m_mcu_block_max_zag[mcu_block] >= 1); + JPGD_ASSERT(m_mcu_block_max_zag[mcu_block] <= 64); + + int max_zag = m_mcu_block_max_zag[mcu_block++] - 1; + if (max_zag <= 0) max_zag = 0; // should never happen, only here to shut up static analysis + switch (s_max_rc[max_zag]) + { + case 1*16+1: + DCT_Upsample::P_Q<1, 1>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<1, 1>::calc(R, S, pSrc_ptr); + break; + case 1*16+2: + DCT_Upsample::P_Q<1, 2>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<1, 2>::calc(R, S, pSrc_ptr); + break; + case 2*16+2: + DCT_Upsample::P_Q<2, 2>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<2, 2>::calc(R, S, pSrc_ptr); + break; + case 3*16+2: + DCT_Upsample::P_Q<3, 2>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<3, 2>::calc(R, S, pSrc_ptr); + break; + case 3*16+3: + DCT_Upsample::P_Q<3, 3>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<3, 3>::calc(R, S, pSrc_ptr); + break; + case 3*16+4: + DCT_Upsample::P_Q<3, 4>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<3, 4>::calc(R, S, pSrc_ptr); + break; + case 4*16+4: + DCT_Upsample::P_Q<4, 4>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<4, 4>::calc(R, S, pSrc_ptr); + break; + case 5*16+4: + DCT_Upsample::P_Q<5, 4>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<5, 4>::calc(R, S, pSrc_ptr); + break; + case 5*16+5: + DCT_Upsample::P_Q<5, 5>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<5, 5>::calc(R, S, pSrc_ptr); + break; + case 5*16+6: + DCT_Upsample::P_Q<5, 6>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<5, 6>::calc(R, S, pSrc_ptr); + break; + case 6*16+6: + DCT_Upsample::P_Q<6, 6>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<6, 6>::calc(R, S, pSrc_ptr); + break; + case 7*16+6: + DCT_Upsample::P_Q<7, 6>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<7, 6>::calc(R, S, pSrc_ptr); + break; + case 7*16+7: + DCT_Upsample::P_Q<7, 7>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<7, 7>::calc(R, S, pSrc_ptr); + break; + case 7*16+8: + DCT_Upsample::P_Q<7, 8>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<7, 8>::calc(R, S, pSrc_ptr); + break; + case 8*16+8: + DCT_Upsample::P_Q<8, 8>::calc(P, Q, pSrc_ptr); + DCT_Upsample::R_S<8, 8>::calc(R, S, pSrc_ptr); + break; + default: + JPGD_ASSERT(false); + } + + DCT_Upsample::Matrix44 a(P + Q); P -= Q; + DCT_Upsample::Matrix44& b = P; + DCT_Upsample::Matrix44 c(R + S); R -= S; + DCT_Upsample::Matrix44& d = R; + + DCT_Upsample::Matrix44::add_and_store(temp_block, a, c); + idct_4x4(temp_block, pDst_ptr); + pDst_ptr += 64; + + DCT_Upsample::Matrix44::sub_and_store(temp_block, a, c); + idct_4x4(temp_block, pDst_ptr); + pDst_ptr += 64; + + DCT_Upsample::Matrix44::add_and_store(temp_block, b, d); + idct_4x4(temp_block, pDst_ptr); + pDst_ptr += 64; + + DCT_Upsample::Matrix44::sub_and_store(temp_block, b, d); + idct_4x4(temp_block, pDst_ptr); + pDst_ptr += 64; + + pSrc_ptr += 64; + } +} + +// Loads and dequantizes the next row of (already decoded) coefficients. +// Progressive images only. +void jpeg_decoder::load_next_row() +{ + int i; + jpgd_block_t *p; + jpgd_quant_t *q; + int mcu_row, mcu_block, row_block = 0; + int component_num, component_id; + int block_x_mcu[JPGD_MAX_COMPONENTS]; + + memset(block_x_mcu, 0, JPGD_MAX_COMPONENTS * sizeof(int)); + + for (mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) + { + int block_x_mcu_ofs = 0, block_y_mcu_ofs = 0; + + for (mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) + { + component_id = m_mcu_org[mcu_block]; + q = m_quant[m_comp_quant[component_id]]; + + p = m_pMCU_coefficients + 64 * mcu_block; + + jpgd_block_t* pAC = coeff_buf_getp(m_ac_coeffs[component_id], block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs); + jpgd_block_t* pDC = coeff_buf_getp(m_dc_coeffs[component_id], block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs); + p[0] = pDC[0]; + memcpy(&p[1], &pAC[1], 63 * sizeof(jpgd_block_t)); + + for (i = 63; i > 0; i--) + if (p[g_ZAG[i]]) + break; + + m_mcu_block_max_zag[mcu_block] = i + 1; + + for ( ; i >= 0; i--) + if (p[g_ZAG[i]]) + p[g_ZAG[i]] = static_cast(p[g_ZAG[i]] * q[i]); + + row_block++; + + if (m_comps_in_scan == 1) + block_x_mcu[component_id]++; + else + { + if (++block_x_mcu_ofs == m_comp_h_samp[component_id]) + { + block_x_mcu_ofs = 0; + + if (++block_y_mcu_ofs == m_comp_v_samp[component_id]) + { + block_y_mcu_ofs = 0; + + block_x_mcu[component_id] += m_comp_h_samp[component_id]; + } + } + } + } + + if (m_freq_domain_chroma_upsample) + transform_mcu_expand(mcu_row); + else + transform_mcu(mcu_row); + } + + if (m_comps_in_scan == 1) + m_block_y_mcu[m_comp_list[0]]++; + else + { + for (component_num = 0; component_num < m_comps_in_scan; component_num++) + { + component_id = m_comp_list[component_num]; + + m_block_y_mcu[component_id] += m_comp_v_samp[component_id]; + } + } +} + +// Restart interval processing. +void jpeg_decoder::process_restart() +{ + int i; + int c = 0; + + // Align to a byte boundry + // FIXME: Is this really necessary? get_bits_no_markers() never reads in markers! + //get_bits_no_markers(m_bits_left & 7); + + // Let's scan a little bit to find the marker, but not _too_ far. + // 1536 is a "fudge factor" that determines how much to scan. + for (i = 1536; i > 0; i--) + if (get_char() == 0xFF) + break; + + if (i == 0) + stop_decoding(JPGD_BAD_RESTART_MARKER); + + for ( ; i > 0; i--) + if ((c = get_char()) != 0xFF) + break; + + if (i == 0) + stop_decoding(JPGD_BAD_RESTART_MARKER); + + // Is it the expected marker? If not, something bad happened. + if (c != (m_next_restart_num + M_RST0)) + stop_decoding(JPGD_BAD_RESTART_MARKER); + + // Reset each component's DC prediction values. + memset(&m_last_dc_val, 0, m_comps_in_frame * sizeof(uint)); + + m_eob_run = 0; + + m_restarts_left = m_restart_interval; + + m_next_restart_num = (m_next_restart_num + 1) & 7; + + // Get the bit buffer going again... + + m_bits_left = 16; + get_bits_no_markers(16); + get_bits_no_markers(16); +} + +static inline int dequantize_ac(int c, int q) { c *= q; return c; } + +// Decodes and dequantizes the next row of coefficients. +void jpeg_decoder::decode_next_row() +{ + int row_block = 0; + + for (int mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) + { + if ((m_restart_interval) && (m_restarts_left == 0)) + process_restart(); + + jpgd_block_t* p = m_pMCU_coefficients; + for (int mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++, p += 64) + { + int component_id = m_mcu_org[mcu_block]; + jpgd_quant_t* q = m_quant[m_comp_quant[component_id]]; + + int r, s; + s = huff_decode(m_pHuff_tabs[m_comp_dc_tab[component_id]], r); + s = JPGD_HUFF_EXTEND(r, s); + + m_last_dc_val[component_id] = (s += m_last_dc_val[component_id]); + + p[0] = static_cast(s * q[0]); + + int prev_num_set = m_mcu_block_max_zag[mcu_block]; + + huff_tables *pH = m_pHuff_tabs[m_comp_ac_tab[component_id]]; + + int k; + for (k = 1; k < 64; k++) + { + int extra_bits; + s = huff_decode(pH, extra_bits); + + r = s >> 4; + s &= 15; + + if (s) + { + if (r) + { + if ((k + r) > 63) + stop_decoding(JPGD_DECODE_ERROR); + + if (k < prev_num_set) + { + int n = JPGD_MIN(r, prev_num_set - k); + int kt = k; + while (n--) + p[g_ZAG[kt++]] = 0; + } + + k += r; + } + + s = JPGD_HUFF_EXTEND(extra_bits, s); + + JPGD_ASSERT(k < 64); + + p[g_ZAG[k]] = static_cast(dequantize_ac(s, q[k])); //s * q[k]; + } + else + { + if (r == 15) + { + if ((k + 16) > 64) + stop_decoding(JPGD_DECODE_ERROR); + + if (k < prev_num_set) + { + int n = JPGD_MIN(16, prev_num_set - k); + int kt = k; + while (n--) + { + JPGD_ASSERT(kt <= 63); + p[g_ZAG[kt++]] = 0; + } + } + + k += 16 - 1; // - 1 because the loop counter is k + JPGD_ASSERT(p[g_ZAG[k]] == 0); + } + else + break; + } + } + + if (k < prev_num_set) + { + int kt = k; + while (kt < prev_num_set) + p[g_ZAG[kt++]] = 0; + } + + m_mcu_block_max_zag[mcu_block] = k; + + row_block++; + } + + if (m_freq_domain_chroma_upsample) + transform_mcu_expand(mcu_row); + else + transform_mcu(mcu_row); + + m_restarts_left--; + } +} + +// YCbCr H1V1 (1x1:1:1, 3 m_blocks per MCU) to RGB +void jpeg_decoder::H1V1Convert() +{ + int row = m_max_mcu_y_size - m_mcu_lines_left; + uint8 *d = m_pScan_line_0; + uint8 *s = m_pSample_buf + row * 8; + + for (int i = m_max_mcus_per_row; i > 0; i--) + { + for (int j = 0; j < 8; j++) + { + int y = s[j]; + int cb = s[64+j]; + int cr = s[128+j]; + + d[0] = clamp(y + m_crr[cr]); + d[1] = clamp(y + ((m_crg[cr] + m_cbg[cb]) >> 16)); + d[2] = clamp(y + m_cbb[cb]); + d[3] = 255; + + d += 4; + } + + s += 64*3; + } +} + +// YCbCr H2V1 (2x1:1:1, 4 m_blocks per MCU) to RGB +void jpeg_decoder::H2V1Convert() +{ + int row = m_max_mcu_y_size - m_mcu_lines_left; + uint8 *d0 = m_pScan_line_0; + uint8 *y = m_pSample_buf + row * 8; + uint8 *c = m_pSample_buf + 2*64 + row * 8; + + for (int i = m_max_mcus_per_row; i > 0; i--) + { + for (int l = 0; l < 2; l++) + { + for (int j = 0; j < 4; j++) + { + int cb = c[0]; + int cr = c[64]; + + int rc = m_crr[cr]; + int gc = ((m_crg[cr] + m_cbg[cb]) >> 16); + int bc = m_cbb[cb]; + + int yy = y[j<<1]; + d0[0] = clamp(yy+rc); + d0[1] = clamp(yy+gc); + d0[2] = clamp(yy+bc); + d0[3] = 255; + + yy = y[(j<<1)+1]; + d0[4] = clamp(yy+rc); + d0[5] = clamp(yy+gc); + d0[6] = clamp(yy+bc); + d0[7] = 255; + + d0 += 8; + + c++; + } + y += 64; + } + + y += 64*4 - 64*2; + c += 64*4 - 8; + } +} + +// YCbCr H2V1 (1x2:1:1, 4 m_blocks per MCU) to RGB +void jpeg_decoder::H1V2Convert() +{ + int row = m_max_mcu_y_size - m_mcu_lines_left; + uint8 *d0 = m_pScan_line_0; + uint8 *d1 = m_pScan_line_1; + uint8 *y; + uint8 *c; + + if (row < 8) + y = m_pSample_buf + row * 8; + else + y = m_pSample_buf + 64*1 + (row & 7) * 8; + + c = m_pSample_buf + 64*2 + (row >> 1) * 8; + + for (int i = m_max_mcus_per_row; i > 0; i--) + { + for (int j = 0; j < 8; j++) + { + int cb = c[0+j]; + int cr = c[64+j]; + + int rc = m_crr[cr]; + int gc = ((m_crg[cr] + m_cbg[cb]) >> 16); + int bc = m_cbb[cb]; + + int yy = y[j]; + d0[0] = clamp(yy+rc); + d0[1] = clamp(yy+gc); + d0[2] = clamp(yy+bc); + d0[3] = 255; + + yy = y[8+j]; + d1[0] = clamp(yy+rc); + d1[1] = clamp(yy+gc); + d1[2] = clamp(yy+bc); + d1[3] = 255; + + d0 += 4; + d1 += 4; + } + + y += 64*4; + c += 64*4; + } +} + +// YCbCr H2V2 (2x2:1:1, 6 m_blocks per MCU) to RGB +void jpeg_decoder::H2V2Convert() +{ + int row = m_max_mcu_y_size - m_mcu_lines_left; + uint8 *d0 = m_pScan_line_0; + uint8 *d1 = m_pScan_line_1; + uint8 *y; + uint8 *c; + + if (row < 8) + y = m_pSample_buf + row * 8; + else + y = m_pSample_buf + 64*2 + (row & 7) * 8; + + c = m_pSample_buf + 64*4 + (row >> 1) * 8; + + for (int i = m_max_mcus_per_row; i > 0; i--) + { + for (int l = 0; l < 2; l++) + { + for (int j = 0; j < 8; j += 2) + { + int cb = c[0]; + int cr = c[64]; + + int rc = m_crr[cr]; + int gc = ((m_crg[cr] + m_cbg[cb]) >> 16); + int bc = m_cbb[cb]; + + int yy = y[j]; + d0[0] = clamp(yy+rc); + d0[1] = clamp(yy+gc); + d0[2] = clamp(yy+bc); + d0[3] = 255; + + yy = y[j+1]; + d0[4] = clamp(yy+rc); + d0[5] = clamp(yy+gc); + d0[6] = clamp(yy+bc); + d0[7] = 255; + + yy = y[j+8]; + d1[0] = clamp(yy+rc); + d1[1] = clamp(yy+gc); + d1[2] = clamp(yy+bc); + d1[3] = 255; + + yy = y[j+8+1]; + d1[4] = clamp(yy+rc); + d1[5] = clamp(yy+gc); + d1[6] = clamp(yy+bc); + d1[7] = 255; + + d0 += 8; + d1 += 8; + + c++; + } + y += 64; + } + + y += 64*6 - 64*2; + c += 64*6 - 8; + } +} + +// Y (1 block per MCU) to 8-bit grayscale +void jpeg_decoder::gray_convert() +{ + int row = m_max_mcu_y_size - m_mcu_lines_left; + uint8 *d = m_pScan_line_0; + uint8 *s = m_pSample_buf + row * 8; + + for (int i = m_max_mcus_per_row; i > 0; i--) + { + *(uint *)d = *(uint *)s; + *(uint *)(&d[4]) = *(uint *)(&s[4]); + + s += 64; + d += 8; + } +} + +void jpeg_decoder::expanded_convert() +{ + int row = m_max_mcu_y_size - m_mcu_lines_left; + + uint8* Py = m_pSample_buf + (row / 8) * 64 * m_comp_h_samp[0] + (row & 7) * 8; + + uint8* d = m_pScan_line_0; + + for (int i = m_max_mcus_per_row; i > 0; i--) + { + for (int k = 0; k < m_max_mcu_x_size; k += 8) + { + const int Y_ofs = k * 8; + const int Cb_ofs = Y_ofs + 64 * m_expanded_blocks_per_component; + const int Cr_ofs = Y_ofs + 64 * m_expanded_blocks_per_component * 2; + for (int j = 0; j < 8; j++) + { + int y = Py[Y_ofs + j]; + int cb = Py[Cb_ofs + j]; + int cr = Py[Cr_ofs + j]; + + d[0] = clamp(y + m_crr[cr]); + d[1] = clamp(y + ((m_crg[cr] + m_cbg[cb]) >> 16)); + d[2] = clamp(y + m_cbb[cb]); + d[3] = 255; + + d += 4; + } + } + + Py += 64 * m_expanded_blocks_per_mcu; + } +} + +// Find end of image (EOI) marker, so we can return to the user the exact size of the input stream. +void jpeg_decoder::find_eoi() +{ + if (!m_progressive_flag) + { + // Attempt to read the EOI marker. + //get_bits_no_markers(m_bits_left & 7); + + // Prime the bit buffer + m_bits_left = 16; + get_bits(16); + get_bits(16); + + // The next marker _should_ be EOI + process_markers(); + } + + m_total_bytes_read -= m_in_buf_left; +} + +int jpeg_decoder::decode(const void** pScan_line, uint* pScan_line_len) +{ + if ((m_error_code) || (!m_ready_flag)) + return JPGD_FAILED; + + if (m_total_lines_left == 0) + return JPGD_DONE; + + if (m_mcu_lines_left == 0) + { + if (setjmp(m_jmp_state)) + return JPGD_FAILED; + + if (m_progressive_flag) + load_next_row(); + else + decode_next_row(); + + // Find the EOI marker if that was the last row. + if (m_total_lines_left <= m_max_mcu_y_size) + find_eoi(); + + m_mcu_lines_left = m_max_mcu_y_size; + } + + if (m_freq_domain_chroma_upsample) + { + expanded_convert(); + *pScan_line = m_pScan_line_0; + } + else + { + switch (m_scan_type) + { + case JPGD_YH2V2: + { + if ((m_mcu_lines_left & 1) == 0) + { + H2V2Convert(); + *pScan_line = m_pScan_line_0; + } + else + *pScan_line = m_pScan_line_1; + + break; + } + case JPGD_YH2V1: + { + H2V1Convert(); + *pScan_line = m_pScan_line_0; + break; + } + case JPGD_YH1V2: + { + if ((m_mcu_lines_left & 1) == 0) + { + H1V2Convert(); + *pScan_line = m_pScan_line_0; + } + else + *pScan_line = m_pScan_line_1; + + break; + } + case JPGD_YH1V1: + { + H1V1Convert(); + *pScan_line = m_pScan_line_0; + break; + } + case JPGD_GRAYSCALE: + { + gray_convert(); + *pScan_line = m_pScan_line_0; + + break; + } + } + } + + *pScan_line_len = m_real_dest_bytes_per_scan_line; + + m_mcu_lines_left--; + m_total_lines_left--; + + return JPGD_SUCCESS; +} + +// Creates the tables needed for efficient Huffman decoding. +void jpeg_decoder::make_huff_table(int index, huff_tables *pH) +{ + int p, i, l, si; + uint8 huffsize[257]; + uint huffcode[257]; + uint code; + uint subtree; + int code_size; + int lastp; + int nextfreeentry; + int currententry; + + pH->ac_table = m_huff_ac[index] != 0; + + p = 0; + + for (l = 1; l <= 16; l++) + { + for (i = 1; i <= m_huff_num[index][l]; i++) + huffsize[p++] = static_cast(l); + } + + huffsize[p] = 0; + + lastp = p; + + code = 0; + si = huffsize[0]; + p = 0; + + while (huffsize[p]) + { + while (huffsize[p] == si) + { + huffcode[p++] = code; + code++; + } + + code <<= 1; + si++; + } + + memset(pH->look_up, 0, sizeof(pH->look_up)); + memset(pH->look_up2, 0, sizeof(pH->look_up2)); + memset(pH->tree, 0, sizeof(pH->tree)); + memset(pH->code_size, 0, sizeof(pH->code_size)); + + nextfreeentry = -1; + + p = 0; + + while (p < lastp) + { + i = m_huff_val[index][p]; + code = huffcode[p]; + code_size = huffsize[p]; + + pH->code_size[i] = static_cast(code_size); + + if (code_size <= 8) + { + code <<= (8 - code_size); + + for (l = 1 << (8 - code_size); l > 0; l--) + { + JPGD_ASSERT(i < 256); + + pH->look_up[code] = i; + + bool has_extrabits = false; + int extra_bits = 0; + int num_extra_bits = i & 15; + + int bits_to_fetch = code_size; + if (num_extra_bits) + { + int total_codesize = code_size + num_extra_bits; + if (total_codesize <= 8) + { + has_extrabits = true; + extra_bits = ((1 << num_extra_bits) - 1) & (code >> (8 - total_codesize)); + JPGD_ASSERT(extra_bits <= 0x7FFF); + bits_to_fetch += num_extra_bits; + } + } + + if (!has_extrabits) + pH->look_up2[code] = i | (bits_to_fetch << 8); + else + pH->look_up2[code] = i | 0x8000 | (extra_bits << 16) | (bits_to_fetch << 8); + + code++; + } + } + else + { + subtree = (code >> (code_size - 8)) & 0xFF; + + currententry = pH->look_up[subtree]; + + if (currententry == 0) + { + pH->look_up[subtree] = currententry = nextfreeentry; + pH->look_up2[subtree] = currententry = nextfreeentry; + + nextfreeentry -= 2; + } + + code <<= (16 - (code_size - 8)); + + for (l = code_size; l > 9; l--) + { + if ((code & 0x8000) == 0) + currententry--; + + if (pH->tree[-currententry - 1] == 0) + { + pH->tree[-currententry - 1] = nextfreeentry; + + currententry = nextfreeentry; + + nextfreeentry -= 2; + } + else + currententry = pH->tree[-currententry - 1]; + + code <<= 1; + } + + if ((code & 0x8000) == 0) + currententry--; + + pH->tree[-currententry - 1] = i; + } + + p++; + } +} + +// Verifies the quantization tables needed for this scan are available. +void jpeg_decoder::check_quant_tables() +{ + for (int i = 0; i < m_comps_in_scan; i++) + if (m_quant[m_comp_quant[m_comp_list[i]]] == NULL) + stop_decoding(JPGD_UNDEFINED_QUANT_TABLE); +} + +// Verifies that all the Huffman tables needed for this scan are available. +void jpeg_decoder::check_huff_tables() +{ + for (int i = 0; i < m_comps_in_scan; i++) + { + if ((m_spectral_start == 0) && (m_huff_num[m_comp_dc_tab[m_comp_list[i]]] == NULL)) + stop_decoding(JPGD_UNDEFINED_HUFF_TABLE); + + if ((m_spectral_end > 0) && (m_huff_num[m_comp_ac_tab[m_comp_list[i]]] == NULL)) + stop_decoding(JPGD_UNDEFINED_HUFF_TABLE); + } + + for (int i = 0; i < JPGD_MAX_HUFF_TABLES; i++) + if (m_huff_num[i]) + { + if (!m_pHuff_tabs[i]) + m_pHuff_tabs[i] = (huff_tables *)alloc(sizeof(huff_tables)); + + make_huff_table(i, m_pHuff_tabs[i]); + } +} + +// Determines the component order inside each MCU. +// Also calcs how many MCU's are on each row, etc. +void jpeg_decoder::calc_mcu_block_order() +{ + int component_num, component_id; + int max_h_samp = 0, max_v_samp = 0; + + for (component_id = 0; component_id < m_comps_in_frame; component_id++) + { + if (m_comp_h_samp[component_id] > max_h_samp) + max_h_samp = m_comp_h_samp[component_id]; + + if (m_comp_v_samp[component_id] > max_v_samp) + max_v_samp = m_comp_v_samp[component_id]; + } + + for (component_id = 0; component_id < m_comps_in_frame; component_id++) + { + m_comp_h_blocks[component_id] = ((((m_image_x_size * m_comp_h_samp[component_id]) + (max_h_samp - 1)) / max_h_samp) + 7) / 8; + m_comp_v_blocks[component_id] = ((((m_image_y_size * m_comp_v_samp[component_id]) + (max_v_samp - 1)) / max_v_samp) + 7) / 8; + } + + if (m_comps_in_scan == 1) + { + m_mcus_per_row = m_comp_h_blocks[m_comp_list[0]]; + m_mcus_per_col = m_comp_v_blocks[m_comp_list[0]]; + } + else + { + m_mcus_per_row = (((m_image_x_size + 7) / 8) + (max_h_samp - 1)) / max_h_samp; + m_mcus_per_col = (((m_image_y_size + 7) / 8) + (max_v_samp - 1)) / max_v_samp; + } + + if (m_comps_in_scan == 1) + { + m_mcu_org[0] = m_comp_list[0]; + + m_blocks_per_mcu = 1; + } + else + { + m_blocks_per_mcu = 0; + + for (component_num = 0; component_num < m_comps_in_scan; component_num++) + { + int num_blocks; + + component_id = m_comp_list[component_num]; + + num_blocks = m_comp_h_samp[component_id] * m_comp_v_samp[component_id]; + + while (num_blocks--) + m_mcu_org[m_blocks_per_mcu++] = component_id; + } + } +} + +// Starts a new scan. +int jpeg_decoder::init_scan() +{ + if (!locate_sos_marker()) + return JPGD_FALSE; + + calc_mcu_block_order(); + + check_huff_tables(); + + check_quant_tables(); + + memset(m_last_dc_val, 0, m_comps_in_frame * sizeof(uint)); + + m_eob_run = 0; + + if (m_restart_interval) + { + m_restarts_left = m_restart_interval; + m_next_restart_num = 0; + } + + fix_in_buffer(); + + return JPGD_TRUE; +} + +// Starts a frame. Determines if the number of components or sampling factors +// are supported. +void jpeg_decoder::init_frame() +{ + int i; + + if (m_comps_in_frame == 1) + { + if ((m_comp_h_samp[0] != 1) || (m_comp_v_samp[0] != 1)) + stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS); + + m_scan_type = JPGD_GRAYSCALE; + m_max_blocks_per_mcu = 1; + m_max_mcu_x_size = 8; + m_max_mcu_y_size = 8; + } + else if (m_comps_in_frame == 3) + { + if ( ((m_comp_h_samp[1] != 1) || (m_comp_v_samp[1] != 1)) || + ((m_comp_h_samp[2] != 1) || (m_comp_v_samp[2] != 1)) ) + stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS); + + if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 1)) + { + m_scan_type = JPGD_YH1V1; + + m_max_blocks_per_mcu = 3; + m_max_mcu_x_size = 8; + m_max_mcu_y_size = 8; + } + else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 1)) + { + m_scan_type = JPGD_YH2V1; + m_max_blocks_per_mcu = 4; + m_max_mcu_x_size = 16; + m_max_mcu_y_size = 8; + } + else if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 2)) + { + m_scan_type = JPGD_YH1V2; + m_max_blocks_per_mcu = 4; + m_max_mcu_x_size = 8; + m_max_mcu_y_size = 16; + } + else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 2)) + { + m_scan_type = JPGD_YH2V2; + m_max_blocks_per_mcu = 6; + m_max_mcu_x_size = 16; + m_max_mcu_y_size = 16; + } + else + stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS); + } + else + stop_decoding(JPGD_UNSUPPORTED_COLORSPACE); + + m_max_mcus_per_row = (m_image_x_size + (m_max_mcu_x_size - 1)) / m_max_mcu_x_size; + m_max_mcus_per_col = (m_image_y_size + (m_max_mcu_y_size - 1)) / m_max_mcu_y_size; + + // These values are for the *destination* pixels: after conversion. + if (m_scan_type == JPGD_GRAYSCALE) + m_dest_bytes_per_pixel = 1; + else + m_dest_bytes_per_pixel = 4; + + m_dest_bytes_per_scan_line = ((m_image_x_size + 15) & 0xFFF0) * m_dest_bytes_per_pixel; + + m_real_dest_bytes_per_scan_line = (m_image_x_size * m_dest_bytes_per_pixel); + + // Initialize two scan line buffers. + m_pScan_line_0 = (uint8 *)alloc(m_dest_bytes_per_scan_line, true); + if ((m_scan_type == JPGD_YH1V2) || (m_scan_type == JPGD_YH2V2)) + m_pScan_line_1 = (uint8 *)alloc(m_dest_bytes_per_scan_line, true); + + m_max_blocks_per_row = m_max_mcus_per_row * m_max_blocks_per_mcu; + + // Should never happen + if (m_max_blocks_per_row > JPGD_MAX_BLOCKS_PER_ROW) + stop_decoding(JPGD_ASSERTION_ERROR); + + // Allocate the coefficient buffer, enough for one MCU + m_pMCU_coefficients = (jpgd_block_t*)alloc(m_max_blocks_per_mcu * 64 * sizeof(jpgd_block_t)); + + for (i = 0; i < m_max_blocks_per_mcu; i++) + m_mcu_block_max_zag[i] = 64; + + m_expanded_blocks_per_component = m_comp_h_samp[0] * m_comp_v_samp[0]; + m_expanded_blocks_per_mcu = m_expanded_blocks_per_component * m_comps_in_frame; + m_expanded_blocks_per_row = m_max_mcus_per_row * m_expanded_blocks_per_mcu; + // Freq. domain chroma upsampling is only supported for H2V2 subsampling factor (the most common one I've seen). + m_freq_domain_chroma_upsample = false; +#if JPGD_SUPPORT_FREQ_DOMAIN_UPSAMPLING + m_freq_domain_chroma_upsample = (m_expanded_blocks_per_mcu == 4*3); +#endif + + if (m_freq_domain_chroma_upsample) + m_pSample_buf = (uint8 *)alloc(m_expanded_blocks_per_row * 64); + else + m_pSample_buf = (uint8 *)alloc(m_max_blocks_per_row * 64); + + m_total_lines_left = m_image_y_size; + + m_mcu_lines_left = 0; + + create_look_ups(); +} + +// The coeff_buf series of methods originally stored the coefficients +// into a "virtual" file which was located in EMS, XMS, or a disk file. A cache +// was used to make this process more efficient. Now, we can store the entire +// thing in RAM. +jpeg_decoder::coeff_buf* jpeg_decoder::coeff_buf_open(int block_num_x, int block_num_y, int block_len_x, int block_len_y) +{ + coeff_buf* cb = (coeff_buf*)alloc(sizeof(coeff_buf)); + + cb->block_num_x = block_num_x; + cb->block_num_y = block_num_y; + cb->block_len_x = block_len_x; + cb->block_len_y = block_len_y; + cb->block_size = (block_len_x * block_len_y) * sizeof(jpgd_block_t); + cb->pData = (uint8 *)alloc(cb->block_size * block_num_x * block_num_y, true); + return cb; +} + +inline jpgd_block_t *jpeg_decoder::coeff_buf_getp(coeff_buf *cb, int block_x, int block_y) +{ + JPGD_ASSERT((block_x < cb->block_num_x) && (block_y < cb->block_num_y)); + return (jpgd_block_t *)(cb->pData + block_x * cb->block_size + block_y * (cb->block_size * cb->block_num_x)); +} + +// The following methods decode the various types of m_blocks encountered +// in progressively encoded images. +void jpeg_decoder::decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y) +{ + int s, r; + jpgd_block_t *p = pD->coeff_buf_getp(pD->m_dc_coeffs[component_id], block_x, block_y); + + if ((s = pD->huff_decode(pD->m_pHuff_tabs[pD->m_comp_dc_tab[component_id]])) != 0) + { + r = pD->get_bits_no_markers(s); + s = JPGD_HUFF_EXTEND(r, s); + } + + pD->m_last_dc_val[component_id] = (s += pD->m_last_dc_val[component_id]); + + p[0] = static_cast(s << pD->m_successive_low); +} + +void jpeg_decoder::decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y) +{ + if (pD->get_bits_no_markers(1)) + { + jpgd_block_t *p = pD->coeff_buf_getp(pD->m_dc_coeffs[component_id], block_x, block_y); + + p[0] |= (1 << pD->m_successive_low); + } +} + +void jpeg_decoder::decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y) +{ + int k, s, r; + + if (pD->m_eob_run) + { + pD->m_eob_run--; + return; + } + + jpgd_block_t *p = pD->coeff_buf_getp(pD->m_ac_coeffs[component_id], block_x, block_y); + + for (k = pD->m_spectral_start; k <= pD->m_spectral_end; k++) + { + s = pD->huff_decode(pD->m_pHuff_tabs[pD->m_comp_ac_tab[component_id]]); + + r = s >> 4; + s &= 15; + + if (s) + { + if ((k += r) > 63) + pD->stop_decoding(JPGD_DECODE_ERROR); + + r = pD->get_bits_no_markers(s); + s = JPGD_HUFF_EXTEND(r, s); + + p[g_ZAG[k]] = static_cast(s << pD->m_successive_low); + } + else + { + if (r == 15) + { + if ((k += 15) > 63) + pD->stop_decoding(JPGD_DECODE_ERROR); + } + else + { + pD->m_eob_run = 1 << r; + + if (r) + pD->m_eob_run += pD->get_bits_no_markers(r); + + pD->m_eob_run--; + + break; + } + } + } +} + +void jpeg_decoder::decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y) +{ + int s, k, r; + int p1 = 1 << pD->m_successive_low; + int m1 = (-1) << pD->m_successive_low; + jpgd_block_t *p = pD->coeff_buf_getp(pD->m_ac_coeffs[component_id], block_x, block_y); + + JPGD_ASSERT(pD->m_spectral_end <= 63); + + k = pD->m_spectral_start; + + if (pD->m_eob_run == 0) + { + for ( ; k <= pD->m_spectral_end; k++) + { + s = pD->huff_decode(pD->m_pHuff_tabs[pD->m_comp_ac_tab[component_id]]); + + r = s >> 4; + s &= 15; + + if (s) + { + if (s != 1) + pD->stop_decoding(JPGD_DECODE_ERROR); + + if (pD->get_bits_no_markers(1)) + s = p1; + else + s = m1; + } + else + { + if (r != 15) + { + pD->m_eob_run = 1 << r; + + if (r) + pD->m_eob_run += pD->get_bits_no_markers(r); + + break; + } + } + + do + { + jpgd_block_t *this_coef = p + g_ZAG[k & 63]; + + if (*this_coef != 0) + { + if (pD->get_bits_no_markers(1)) + { + if ((*this_coef & p1) == 0) + { + if (*this_coef >= 0) + *this_coef = static_cast(*this_coef + p1); + else + *this_coef = static_cast(*this_coef + m1); + } + } + } + else + { + if (--r < 0) + break; + } + + k++; + + } while (k <= pD->m_spectral_end); + + if ((s) && (k < 64)) + { + p[g_ZAG[k]] = static_cast(s); + } + } + } + + if (pD->m_eob_run > 0) + { + for ( ; k <= pD->m_spectral_end; k++) + { + jpgd_block_t *this_coef = p + g_ZAG[k & 63]; // logical AND to shut up static code analysis + + if (*this_coef != 0) + { + if (pD->get_bits_no_markers(1)) + { + if ((*this_coef & p1) == 0) + { + if (*this_coef >= 0) + *this_coef = static_cast(*this_coef + p1); + else + *this_coef = static_cast(*this_coef + m1); + } + } + } + } + + pD->m_eob_run--; + } +} + +// Decode a scan in a progressively encoded image. +void jpeg_decoder::decode_scan(pDecode_block_func decode_block_func) +{ + int mcu_row, mcu_col, mcu_block; + int block_x_mcu[JPGD_MAX_COMPONENTS], m_block_y_mcu[JPGD_MAX_COMPONENTS]; + + memset(m_block_y_mcu, 0, sizeof(m_block_y_mcu)); + + for (mcu_col = 0; mcu_col < m_mcus_per_col; mcu_col++) + { + int component_num, component_id; + + memset(block_x_mcu, 0, sizeof(block_x_mcu)); + + for (mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) + { + int block_x_mcu_ofs = 0, block_y_mcu_ofs = 0; + + if ((m_restart_interval) && (m_restarts_left == 0)) + process_restart(); + + for (mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) + { + component_id = m_mcu_org[mcu_block]; + + decode_block_func(this, component_id, block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs); + + if (m_comps_in_scan == 1) + block_x_mcu[component_id]++; + else + { + if (++block_x_mcu_ofs == m_comp_h_samp[component_id]) + { + block_x_mcu_ofs = 0; + + if (++block_y_mcu_ofs == m_comp_v_samp[component_id]) + { + block_y_mcu_ofs = 0; + block_x_mcu[component_id] += m_comp_h_samp[component_id]; + } + } + } + } + + m_restarts_left--; + } + + if (m_comps_in_scan == 1) + m_block_y_mcu[m_comp_list[0]]++; + else + { + for (component_num = 0; component_num < m_comps_in_scan; component_num++) + { + component_id = m_comp_list[component_num]; + m_block_y_mcu[component_id] += m_comp_v_samp[component_id]; + } + } + } +} + +// Decode a progressively encoded image. +void jpeg_decoder::init_progressive() +{ + int i; + + if (m_comps_in_frame == 4) + stop_decoding(JPGD_UNSUPPORTED_COLORSPACE); + + // Allocate the coefficient buffers. + for (i = 0; i < m_comps_in_frame; i++) + { + m_dc_coeffs[i] = coeff_buf_open(m_max_mcus_per_row * m_comp_h_samp[i], m_max_mcus_per_col * m_comp_v_samp[i], 1, 1); + m_ac_coeffs[i] = coeff_buf_open(m_max_mcus_per_row * m_comp_h_samp[i], m_max_mcus_per_col * m_comp_v_samp[i], 8, 8); + } + + for ( ; ; ) + { + int dc_only_scan, refinement_scan; + pDecode_block_func decode_block_func; + + if (!init_scan()) + break; + + dc_only_scan = (m_spectral_start == 0); + refinement_scan = (m_successive_high != 0); + + if ((m_spectral_start > m_spectral_end) || (m_spectral_end > 63)) + stop_decoding(JPGD_BAD_SOS_SPECTRAL); + + if (dc_only_scan) + { + if (m_spectral_end) + stop_decoding(JPGD_BAD_SOS_SPECTRAL); + } + else if (m_comps_in_scan != 1) /* AC scans can only contain one component */ + stop_decoding(JPGD_BAD_SOS_SPECTRAL); + + if ((refinement_scan) && (m_successive_low != m_successive_high - 1)) + stop_decoding(JPGD_BAD_SOS_SUCCESSIVE); + + if (dc_only_scan) + { + if (refinement_scan) + decode_block_func = decode_block_dc_refine; + else + decode_block_func = decode_block_dc_first; + } + else + { + if (refinement_scan) + decode_block_func = decode_block_ac_refine; + else + decode_block_func = decode_block_ac_first; + } + + decode_scan(decode_block_func); + + m_bits_left = 16; + get_bits(16); + get_bits(16); + } + + m_comps_in_scan = m_comps_in_frame; + + for (i = 0; i < m_comps_in_frame; i++) + m_comp_list[i] = i; + + calc_mcu_block_order(); +} + +void jpeg_decoder::init_sequential() +{ + if (!init_scan()) + stop_decoding(JPGD_UNEXPECTED_MARKER); +} + +void jpeg_decoder::decode_start() +{ + init_frame(); + + if (m_progressive_flag) + init_progressive(); + else + init_sequential(); +} + +void jpeg_decoder::decode_init(jpeg_decoder_stream *pStream) +{ + init(pStream); + locate_sof_marker(); +} + +jpeg_decoder::jpeg_decoder(jpeg_decoder_stream *pStream) +{ + if (setjmp(m_jmp_state)) + return; + decode_init(pStream); +} + +int jpeg_decoder::begin_decoding() +{ + if (m_ready_flag) + return JPGD_SUCCESS; + + if (m_error_code) + return JPGD_FAILED; + + if (setjmp(m_jmp_state)) + return JPGD_FAILED; + + decode_start(); + + m_ready_flag = true; + + return JPGD_SUCCESS; +} + +jpeg_decoder::~jpeg_decoder() +{ + free_all_blocks(); +} + +jpeg_decoder_file_stream::jpeg_decoder_file_stream() +{ + m_pFile = NULL; + m_eof_flag = false; + m_error_flag = false; +} + +void jpeg_decoder_file_stream::close() +{ + if (m_pFile) + { + fclose(m_pFile); + m_pFile = NULL; + } + + m_eof_flag = false; + m_error_flag = false; +} + +jpeg_decoder_file_stream::~jpeg_decoder_file_stream() +{ + close(); +} + +bool jpeg_decoder_file_stream::open(const char *Pfilename) +{ + close(); + + m_eof_flag = false; + m_error_flag = false; + +#if defined(_MSC_VER) + m_pFile = NULL; + fopen_s(&m_pFile, Pfilename, "rb"); +#else + m_pFile = fopen(Pfilename, "rb"); +#endif + return m_pFile != NULL; +} + +int jpeg_decoder_file_stream::read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) +{ + if (!m_pFile) + return -1; + + if (m_eof_flag) + { + *pEOF_flag = true; + return 0; + } + + if (m_error_flag) + return -1; + + int bytes_read = static_cast(fread(pBuf, 1, max_bytes_to_read, m_pFile)); + if (bytes_read < max_bytes_to_read) + { + if (ferror(m_pFile)) + { + m_error_flag = true; + return -1; + } + + m_eof_flag = true; + *pEOF_flag = true; + } + + return bytes_read; +} + +bool jpeg_decoder_mem_stream::open(const uint8 *pSrc_data, uint size) +{ + close(); + m_pSrc_data = pSrc_data; + m_ofs = 0; + m_size = size; + return true; +} + +int jpeg_decoder_mem_stream::read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) +{ + *pEOF_flag = false; + + if (!m_pSrc_data) + return -1; + + uint bytes_remaining = m_size - m_ofs; + if ((uint)max_bytes_to_read > bytes_remaining) + { + max_bytes_to_read = bytes_remaining; + *pEOF_flag = true; + } + + memcpy(pBuf, m_pSrc_data + m_ofs, max_bytes_to_read); + m_ofs += max_bytes_to_read; + + return max_bytes_to_read; +} + +unsigned char *decompress_jpeg_image_from_stream(jpeg_decoder_stream *pStream, int *width, int *height, int *actual_comps, int req_comps) +{ + if (!actual_comps) + return NULL; + *actual_comps = 0; + + if ((!pStream) || (!width) || (!height) || (!req_comps)) + return NULL; + + if ((req_comps != 1) && (req_comps != 3) && (req_comps != 4)) + return NULL; + + jpeg_decoder decoder(pStream); + if (decoder.get_error_code() != JPGD_SUCCESS) + return NULL; + + const int image_width = decoder.get_width(), image_height = decoder.get_height(); + *width = image_width; + *height = image_height; + *actual_comps = decoder.get_num_components(); + + if (decoder.begin_decoding() != JPGD_SUCCESS) + return NULL; + + const int dst_bpl = image_width * req_comps; + + uint8 *pImage_data = (uint8*)jpgd_malloc(dst_bpl * image_height); + if (!pImage_data) + return NULL; + + for (int y = 0; y < image_height; y++) + { + const uint8* pScan_line; + uint scan_line_len; + if (decoder.decode((const void**)&pScan_line, &scan_line_len) != JPGD_SUCCESS) + { + jpgd_free(pImage_data); + return NULL; + } + + uint8 *pDst = pImage_data + y * dst_bpl; + + if (((req_comps == 1) && (decoder.get_num_components() == 1)) || ((req_comps == 4) && (decoder.get_num_components() == 3))) + memcpy(pDst, pScan_line, dst_bpl); + else if (decoder.get_num_components() == 1) + { + if (req_comps == 3) + { + for (int x = 0; x < image_width; x++) + { + uint8 luma = pScan_line[x]; + pDst[0] = luma; + pDst[1] = luma; + pDst[2] = luma; + pDst += 3; + } + } + else + { + for (int x = 0; x < image_width; x++) + { + uint8 luma = pScan_line[x]; + pDst[0] = luma; + pDst[1] = luma; + pDst[2] = luma; + pDst[3] = 255; + pDst += 4; + } + } + } + else if (decoder.get_num_components() == 3) + { + if (req_comps == 1) + { + const int YR = 19595, YG = 38470, YB = 7471; + for (int x = 0; x < image_width; x++) + { + int r = pScan_line[x*4+0]; + int g = pScan_line[x*4+1]; + int b = pScan_line[x*4+2]; + *pDst++ = static_cast((r * YR + g * YG + b * YB + 32768) >> 16); + } + } + else + { + for (int x = 0; x < image_width; x++) + { + pDst[0] = pScan_line[x*4+0]; + pDst[1] = pScan_line[x*4+1]; + pDst[2] = pScan_line[x*4+2]; + pDst += 3; + } + } + } + } + + return pImage_data; +} + +unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, int *width, int *height, int *actual_comps, int req_comps) +{ + jpgd::jpeg_decoder_mem_stream mem_stream(pSrc_data, src_data_size); + return decompress_jpeg_image_from_stream(&mem_stream, width, height, actual_comps, req_comps); +} + +unsigned char *decompress_jpeg_image_from_file(const char *pSrc_filename, int *width, int *height, int *actual_comps, int req_comps) +{ + jpgd::jpeg_decoder_file_stream file_stream; + if (!file_stream.open(pSrc_filename)) + return NULL; + return decompress_jpeg_image_from_stream(&file_stream, width, height, actual_comps, req_comps); +} + +} // namespace jpgd \ No newline at end of file diff --git a/jpeg-compressor/jpgd.h b/jpeg-compressor/jpgd.h new file mode 100644 index 0000000000..150b9a0b26 --- /dev/null +++ b/jpeg-compressor/jpgd.h @@ -0,0 +1,319 @@ +// jpgd.h - C++ class for JPEG decompression. +// Public domain, Rich Geldreich +#ifndef JPEG_DECODER_H +#define JPEG_DECODER_H + +#include +#include +#include + +#ifdef _MSC_VER + #define JPGD_NORETURN __declspec(noreturn) +#elif defined(__GNUC__) + #define JPGD_NORETURN __attribute__ ((noreturn)) +#else + #define JPGD_NORETURN +#endif + +namespace jpgd +{ + typedef unsigned char uint8; + typedef signed short int16; + typedef unsigned short uint16; + typedef unsigned int uint; + typedef signed int int32; + + // Loads a JPEG image from a memory buffer or a file. + // req_comps can be 1 (grayscale), 3 (RGB), or 4 (RGBA). + // On return, width/height will be set to the image's dimensions, and actual_comps will be set to the either 1 (grayscale) or 3 (RGB). + // Notes: For more control over where and how the source data is read, see the decompress_jpeg_image_from_stream() function below, or call the jpeg_decoder class directly. + // Requesting a 8 or 32bpp image is currently a little faster than 24bpp because the jpeg_decoder class itself currently always unpacks to either 8 or 32bpp. + unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, int *width, int *height, int *actual_comps, int req_comps); + unsigned char *decompress_jpeg_image_from_file(const char *pSrc_filename, int *width, int *height, int *actual_comps, int req_comps); + + // Success/failure error codes. + enum jpgd_status + { + JPGD_SUCCESS = 0, JPGD_FAILED = -1, JPGD_DONE = 1, + JPGD_BAD_DHT_COUNTS = -256, JPGD_BAD_DHT_INDEX, JPGD_BAD_DHT_MARKER, JPGD_BAD_DQT_MARKER, JPGD_BAD_DQT_TABLE, + JPGD_BAD_PRECISION, JPGD_BAD_HEIGHT, JPGD_BAD_WIDTH, JPGD_TOO_MANY_COMPONENTS, + JPGD_BAD_SOF_LENGTH, JPGD_BAD_VARIABLE_MARKER, JPGD_BAD_DRI_LENGTH, JPGD_BAD_SOS_LENGTH, + JPGD_BAD_SOS_COMP_ID, JPGD_W_EXTRA_BYTES_BEFORE_MARKER, JPGD_NO_ARITHMITIC_SUPPORT, JPGD_UNEXPECTED_MARKER, + JPGD_NOT_JPEG, JPGD_UNSUPPORTED_MARKER, JPGD_BAD_DQT_LENGTH, JPGD_TOO_MANY_BLOCKS, + JPGD_UNDEFINED_QUANT_TABLE, JPGD_UNDEFINED_HUFF_TABLE, JPGD_NOT_SINGLE_SCAN, JPGD_UNSUPPORTED_COLORSPACE, + JPGD_UNSUPPORTED_SAMP_FACTORS, JPGD_DECODE_ERROR, JPGD_BAD_RESTART_MARKER, JPGD_ASSERTION_ERROR, + JPGD_BAD_SOS_SPECTRAL, JPGD_BAD_SOS_SUCCESSIVE, JPGD_STREAM_READ, JPGD_NOTENOUGHMEM + }; + + // Input stream interface. + // Derive from this class to read input data from sources other than files or memory. Set m_eof_flag to true when no more data is available. + // The decoder is rather greedy: it will keep on calling this method until its internal input buffer is full, or until the EOF flag is set. + // It the input stream contains data after the JPEG stream's EOI (end of image) marker it will probably be pulled into the internal buffer. + // Call the get_total_bytes_read() method to determine the actual size of the JPEG stream after successful decoding. + class jpeg_decoder_stream + { + public: + jpeg_decoder_stream() { } + virtual ~jpeg_decoder_stream() { } + + // The read() method is called when the internal input buffer is empty. + // Parameters: + // pBuf - input buffer + // max_bytes_to_read - maximum bytes that can be written to pBuf + // pEOF_flag - set this to true if at end of stream (no more bytes remaining) + // Returns -1 on error, otherwise return the number of bytes actually written to the buffer (which may be 0). + // Notes: This method will be called in a loop until you set *pEOF_flag to true or the internal buffer is full. + virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) = 0; + }; + + // stdio FILE stream class. + class jpeg_decoder_file_stream : public jpeg_decoder_stream + { + jpeg_decoder_file_stream(const jpeg_decoder_file_stream &); + jpeg_decoder_file_stream &operator =(const jpeg_decoder_file_stream &); + + FILE *m_pFile; + bool m_eof_flag, m_error_flag; + + public: + jpeg_decoder_file_stream(); + virtual ~jpeg_decoder_file_stream(); + + bool open(const char *Pfilename); + void close(); + + virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag); + }; + + // Memory stream class. + class jpeg_decoder_mem_stream : public jpeg_decoder_stream + { + const uint8 *m_pSrc_data; + uint m_ofs, m_size; + + public: + jpeg_decoder_mem_stream() : m_pSrc_data(NULL), m_ofs(0), m_size(0) { } + jpeg_decoder_mem_stream(const uint8 *pSrc_data, uint size) : m_pSrc_data(pSrc_data), m_ofs(0), m_size(size) { } + + virtual ~jpeg_decoder_mem_stream() { } + + bool open(const uint8 *pSrc_data, uint size); + void close() { m_pSrc_data = NULL; m_ofs = 0; m_size = 0; } + + virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag); + }; + + // Loads JPEG file from a jpeg_decoder_stream. + unsigned char *decompress_jpeg_image_from_stream(jpeg_decoder_stream *pStream, int *width, int *height, int *actual_comps, int req_comps); + + enum + { + JPGD_IN_BUF_SIZE = 8192, JPGD_MAX_BLOCKS_PER_MCU = 10, JPGD_MAX_HUFF_TABLES = 8, JPGD_MAX_QUANT_TABLES = 4, + JPGD_MAX_COMPONENTS = 4, JPGD_MAX_COMPS_IN_SCAN = 4, JPGD_MAX_BLOCKS_PER_ROW = 8192, JPGD_MAX_HEIGHT = 16384, JPGD_MAX_WIDTH = 16384 + }; + + typedef int16 jpgd_quant_t; + typedef int16 jpgd_block_t; + + class jpeg_decoder + { + public: + // Call get_error_code() after constructing to determine if the stream is valid or not. You may call the get_width(), get_height(), etc. + // methods after the constructor is called. You may then either destruct the object, or begin decoding the image by calling begin_decoding(), then decode() on each scanline. + jpeg_decoder(jpeg_decoder_stream *pStream); + + ~jpeg_decoder(); + + // Call this method after constructing the object to begin decompression. + // If JPGD_SUCCESS is returned you may then call decode() on each scanline. + int begin_decoding(); + + // Returns the next scan line. + // For grayscale images, pScan_line will point to a buffer containing 8-bit pixels (get_bytes_per_pixel() will return 1). + // Otherwise, it will always point to a buffer containing 32-bit RGBA pixels (A will always be 255, and get_bytes_per_pixel() will return 4). + // Returns JPGD_SUCCESS if a scan line has been returned. + // Returns JPGD_DONE if all scan lines have been returned. + // Returns JPGD_FAILED if an error occurred. Call get_error_code() for a more info. + int decode(const void** pScan_line, uint* pScan_line_len); + + inline jpgd_status get_error_code() const { return m_error_code; } + + inline int get_width() const { return m_image_x_size; } + inline int get_height() const { return m_image_y_size; } + + inline int get_num_components() const { return m_comps_in_frame; } + + inline int get_bytes_per_pixel() const { return m_dest_bytes_per_pixel; } + inline int get_bytes_per_scan_line() const { return m_image_x_size * get_bytes_per_pixel(); } + + // Returns the total number of bytes actually consumed by the decoder (which should equal the actual size of the JPEG file). + inline int get_total_bytes_read() const { return m_total_bytes_read; } + + private: + jpeg_decoder(const jpeg_decoder &); + jpeg_decoder &operator =(const jpeg_decoder &); + + typedef void (*pDecode_block_func)(jpeg_decoder *, int, int, int); + + struct huff_tables + { + bool ac_table; + uint look_up[256]; + uint look_up2[256]; + uint8 code_size[256]; + uint tree[512]; + }; + + struct coeff_buf + { + uint8 *pData; + int block_num_x, block_num_y; + int block_len_x, block_len_y; + int block_size; + }; + + struct mem_block + { + mem_block *m_pNext; + size_t m_used_count; + size_t m_size; + char m_data[1]; + }; + + jmp_buf m_jmp_state; + mem_block *m_pMem_blocks; + int m_image_x_size; + int m_image_y_size; + jpeg_decoder_stream *m_pStream; + int m_progressive_flag; + uint8 m_huff_ac[JPGD_MAX_HUFF_TABLES]; + uint8* m_huff_num[JPGD_MAX_HUFF_TABLES]; // pointer to number of Huffman codes per bit size + uint8* m_huff_val[JPGD_MAX_HUFF_TABLES]; // pointer to Huffman codes per bit size + jpgd_quant_t* m_quant[JPGD_MAX_QUANT_TABLES]; // pointer to quantization tables + int m_scan_type; // Gray, Yh1v1, Yh1v2, Yh2v1, Yh2v2 (CMYK111, CMYK4114 no longer supported) + int m_comps_in_frame; // # of components in frame + int m_comp_h_samp[JPGD_MAX_COMPONENTS]; // component's horizontal sampling factor + int m_comp_v_samp[JPGD_MAX_COMPONENTS]; // component's vertical sampling factor + int m_comp_quant[JPGD_MAX_COMPONENTS]; // component's quantization table selector + int m_comp_ident[JPGD_MAX_COMPONENTS]; // component's ID + int m_comp_h_blocks[JPGD_MAX_COMPONENTS]; + int m_comp_v_blocks[JPGD_MAX_COMPONENTS]; + int m_comps_in_scan; // # of components in scan + int m_comp_list[JPGD_MAX_COMPS_IN_SCAN]; // components in this scan + int m_comp_dc_tab[JPGD_MAX_COMPONENTS]; // component's DC Huffman coding table selector + int m_comp_ac_tab[JPGD_MAX_COMPONENTS]; // component's AC Huffman coding table selector + int m_spectral_start; // spectral selection start + int m_spectral_end; // spectral selection end + int m_successive_low; // successive approximation low + int m_successive_high; // successive approximation high + int m_max_mcu_x_size; // MCU's max. X size in pixels + int m_max_mcu_y_size; // MCU's max. Y size in pixels + int m_blocks_per_mcu; + int m_max_blocks_per_row; + int m_mcus_per_row, m_mcus_per_col; + int m_mcu_org[JPGD_MAX_BLOCKS_PER_MCU]; + int m_total_lines_left; // total # lines left in image + int m_mcu_lines_left; // total # lines left in this MCU + int m_real_dest_bytes_per_scan_line; + int m_dest_bytes_per_scan_line; // rounded up + int m_dest_bytes_per_pixel; // 4 (RGB) or 1 (Y) + huff_tables* m_pHuff_tabs[JPGD_MAX_HUFF_TABLES]; + coeff_buf* m_dc_coeffs[JPGD_MAX_COMPONENTS]; + coeff_buf* m_ac_coeffs[JPGD_MAX_COMPONENTS]; + int m_eob_run; + int m_block_y_mcu[JPGD_MAX_COMPONENTS]; + uint8* m_pIn_buf_ofs; + int m_in_buf_left; + int m_tem_flag; + bool m_eof_flag; + uint8 m_in_buf_pad_start[128]; + uint8 m_in_buf[JPGD_IN_BUF_SIZE + 128]; + uint8 m_in_buf_pad_end[128]; + int m_bits_left; + uint m_bit_buf; + int m_restart_interval; + int m_restarts_left; + int m_next_restart_num; + int m_max_mcus_per_row; + int m_max_blocks_per_mcu; + int m_expanded_blocks_per_mcu; + int m_expanded_blocks_per_row; + int m_expanded_blocks_per_component; + bool m_freq_domain_chroma_upsample; + int m_max_mcus_per_col; + uint m_last_dc_val[JPGD_MAX_COMPONENTS]; + jpgd_block_t* m_pMCU_coefficients; + int m_mcu_block_max_zag[JPGD_MAX_BLOCKS_PER_MCU]; + uint8* m_pSample_buf; + int m_crr[256]; + int m_cbb[256]; + int m_crg[256]; + int m_cbg[256]; + uint8* m_pScan_line_0; + uint8* m_pScan_line_1; + jpgd_status m_error_code; + bool m_ready_flag; + int m_total_bytes_read; + + void free_all_blocks(); + JPGD_NORETURN void stop_decoding(jpgd_status status); + void *alloc(size_t n, bool zero = false); + void word_clear(void *p, uint16 c, uint n); + void prep_in_buffer(); + void read_dht_marker(); + void read_dqt_marker(); + void read_sof_marker(); + void skip_variable_marker(); + void read_dri_marker(); + void read_sos_marker(); + int next_marker(); + int process_markers(); + void locate_soi_marker(); + void locate_sof_marker(); + int locate_sos_marker(); + void init(jpeg_decoder_stream * pStream); + void create_look_ups(); + void fix_in_buffer(); + void transform_mcu(int mcu_row); + void transform_mcu_expand(int mcu_row); + coeff_buf* coeff_buf_open(int block_num_x, int block_num_y, int block_len_x, int block_len_y); + inline jpgd_block_t *coeff_buf_getp(coeff_buf *cb, int block_x, int block_y); + void load_next_row(); + void decode_next_row(); + void make_huff_table(int index, huff_tables *pH); + void check_quant_tables(); + void check_huff_tables(); + void calc_mcu_block_order(); + int init_scan(); + void init_frame(); + void process_restart(); + void decode_scan(pDecode_block_func decode_block_func); + void init_progressive(); + void init_sequential(); + void decode_start(); + void decode_init(jpeg_decoder_stream * pStream); + void H2V2Convert(); + void H2V1Convert(); + void H1V2Convert(); + void H1V1Convert(); + void gray_convert(); + void expanded_convert(); + void find_eoi(); + inline uint get_char(); + inline uint get_char(bool *pPadding_flag); + inline void stuff_char(uint8 q); + inline uint8 get_octet(); + inline uint get_bits(int num_bits); + inline uint get_bits_no_markers(int numbits); + inline int huff_decode(huff_tables *pH); + inline int huff_decode(huff_tables *pH, int& extrabits); + static inline uint8 clamp(int i); + static void decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y); + static void decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y); + static void decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y); + static void decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y); + }; + +} // namespace jpgd + +#endif // JPEG_DECODER_H diff --git a/rpcs3.sln b/rpcs3.sln index 532495113f..2d6dce268f 100644 --- a/rpcs3.sln +++ b/rpcs3.sln @@ -63,6 +63,12 @@ Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "lodepng", "lodepng", "{4867 lodepng\lodepng.h = lodepng\lodepng.h EndProjectSection EndProject +Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "jpeg-compressor", "jpeg-compressor", "{9D9C8EDB-4469-42B3-A3C7-712CDABC43B0}" + ProjectSection(SolutionItems) = preProject + jpeg-compressor\jpgd.cpp = jpeg-compressor\jpgd.cpp + jpeg-compressor\jpgd.h = jpeg-compressor\jpgd.h + EndProjectSection +EndProject Global GlobalSection(SolutionConfigurationPlatforms) = preSolution Debug|Win32 = Debug|Win32 diff --git a/rpcs3/Emu/SysCalls/Modules.cpp b/rpcs3/Emu/SysCalls/Modules.cpp index e8a5021980..cbac0e1e82 100644 --- a/rpcs3/Emu/SysCalls/Modules.cpp +++ b/rpcs3/Emu/SysCalls/Modules.cpp @@ -29,7 +29,7 @@ static const g_module_list[] = {0x000c, "cellSheap"}, {0x000d, "sys_sync"}, {0x000e, "sys_fs"}, - {0x000f, "sys_jpgdec"}, + {0x000f, "cellJpgDec"}, {0x0010, "cellGcmSys"}, {0x0011, "cellAudio"}, {0x0012, "cellPamf"}, diff --git a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp new file mode 100644 index 0000000000..5aefb12eba --- /dev/null +++ b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp @@ -0,0 +1,203 @@ +#include "stdafx.h" +#include "Emu/SysCalls/SysCalls.h" +#include "Emu/SysCalls/SC_FUNC.h" + +#include "jpeg-compressor/jpgd.cpp" + +void cellJpgDec_init(); +Module cellJpgDec(0x000f, cellJpgDec_init); + +//Return Codes +enum +{ + CELL_JPGDEC_ERROR_HEADER = 0x80611101, + CELL_JPGDEC_ERROR_STREAM_FORMAT = 0x80611102, + CELL_JPGDEC_ERROR_ARG = 0x80611103, + CELL_JPGDEC_ERROR_SEQ = 0x80611104, + CELL_JPGDEC_ERROR_BUSY = 0x80611105, + CELL_JPGDEC_ERROR_FATAL = 0x80611106, + CELL_JPGDEC_ERROR_OPEN_FILE = 0x80611107, + CELL_JPGDEC_ERROR_SPU_UNSUPPORT = 0x80611108, + CELL_JPGDEC_ERROR_CB_PARAM = 0x80611109, +}; + +struct CellJpgDecInfo +{ + u32 imageWidth; + u32 imageHeight; + u32 numComponents; + u32 colorSpace; // CellJpgDecColorSpace +}; + +struct CellJpgDecSrc +{ + u32 srcSelect; // CellJpgDecStreamSrcSel + u32 fileName; // const char* + u64 fileOffset; // int64_t + u32 fileSize; + u32 streamPtr; + u32 streamSize; + u32 spuThreadEnable; // CellJpgDecSpuThreadEna +}; + +CellJpgDecInfo current_info; +CellJpgDecSrc current_src; + + +int cellJpgDecCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam) +{ + UNIMPLEMENTED_FUNC(cellJpgDec); + return CELL_OK; +} + +int cellJpgDecExtCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam) +{ + UNIMPLEMENTED_FUNC(cellJpgDec); + return CELL_OK; +} + +int cellJpgDecDestroy(u32 mainHandle) +{ + UNIMPLEMENTED_FUNC(cellJpgDec); + return CELL_OK; +} + +int cellJpgDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInfo) +{ + //current_src.srcSelect = Memory.Read32(src_addr); + current_src.fileName = Memory.Read32(src_addr+4); + //current_src.fileOffset = Memory.Read32(src_addr+8); + //current_src.fileSize = Memory.Read32(src_addr+12); + //current_src.streamPtr = Memory.Read32(src_addr+16); + //current_src.streamSize = Memory.Read32(src_addr+20); + //current_src.spuThreadEnable = Memory.Read32(src_addr+24); + + u32& fd_addr = subHandle_addr; // Set file descriptor as sub handler of the decoder + int ret = cellFsOpen(current_src.fileName, 0, fd_addr, NULL, 0); + if(ret != 0) return CELL_JPGDEC_ERROR_OPEN_FILE; + + return CELL_OK; +} + +int cellJpgDecClose(u32 mainHandle, u32 subHandle) +{ + u32& fd = subHandle; + cellFsClose(fd); + + return CELL_OK; +} + +int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) +{ + u32& fd = subHandle; + + //Get size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(fd, sb_addr); + u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + + //Copy the JPG file to a buffer + u32 buffer = Memory.Alloc(fileSize,1); + u32 pos_addr = Memory.Alloc(8,1); + cellFsLseek(fd, 0, 0, pos_addr); + cellFsRead(fd, buffer, fileSize, NULL); + + if (Memory.Read32(buffer) != 0xFFD8FFE0 || // Error: Not a valid SOI header + Memory.Read32(buffer+6) != 0x4A464946) // Error: Not a valid JFIF string + { + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + return CELL_JPGDEC_ERROR_HEADER; + } + + u32 i = 4; + u16 block_length = Memory.Read8(buffer+i)*0xFF + Memory.Read8(buffer+i+1); + while(i < fileSize) + { + i += block_length; // Increase the file index to get to the next block + if (i >= fileSize) return CELL_JPGDEC_ERROR_HEADER; // Check to protect against segmentation faults + if(Memory.Read8(buffer+i) != 0xFF) return CELL_JPGDEC_ERROR_HEADER; // Check that we are truly at the start of another block + if(Memory.Read8(buffer+i+1) == 0xC0) break; // 0xFFC0 is the "Start of frame" marker which contains the file size + i += 2; // Skip the block marker + block_length = Memory.Read8(buffer+i)*0xFF + Memory.Read8(buffer+i+1); // Go to the next block + } + + current_info.imageWidth = Memory.Read8(buffer+i+7)*256 + Memory.Read8(buffer+i+8); + current_info.imageHeight = Memory.Read8(buffer+i+5)*256 + Memory.Read8(buffer+i+6); + current_info.numComponents = 0; // Unimplemented + current_info.colorSpace = 0; // Unimplemented + + mem_class_t info(info_addr); + info += current_info.imageWidth; + info += current_info.imageHeight; + info += current_info.numComponents; + info += current_info.colorSpace; + + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + + return CELL_OK; +} + +int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataCtrlParam_addr, u32 dataOutInfo_addr) +{ + u32& fd = subHandle; + + //Get size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(fd, sb_addr); + u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + + //Copy the JPG file to a buffer + u32 buffer = Memory.Alloc(fileSize,1); + u32 pos_addr = Memory.Alloc(8,1); + cellFsLseek(fd, 0, 0, pos_addr); + cellFsRead(fd, buffer, fileSize, NULL); + + int width, height, actual_components; + unsigned char *jpg = new unsigned char [fileSize]; + for(u32 i = 0; i < fileSize; i++){ + jpg[i] = Memory.Read8(buffer+i); + } + + unsigned char *image = jpgd::decompress_jpeg_image_from_memory((const unsigned char*)jpg, fileSize, &width, &height, &actual_components, 4); + u32 image_size = width * height * 4; + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+3]); + Memory.Write8(data_addr+i+1, image[i+0]); + Memory.Write8(data_addr+i+2, image[i+1]); + Memory.Write8(data_addr+i+3, image[i+2]); + } + + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + + return CELL_OK; +} + +int cellJpgDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outParam) +{ + UNIMPLEMENTED_FUNC(cellJpgDec); + return CELL_OK; +} + + +void cellJpgDec_init() +{ + cellJpgDec.AddFunc(0xa7978f59, cellJpgDecCreate); + cellJpgDec.AddFunc(0x8b300f66, cellJpgDecExtCreate); + cellJpgDec.AddFunc(0x976ca5c2, cellJpgDecOpen); + cellJpgDec.AddFunc(0x6d9ebccf, cellJpgDecReadHeader); + cellJpgDec.AddFunc(0xe08f3910, cellJpgDecSetParameter); + cellJpgDec.AddFunc(0xaf8bb012, cellJpgDecDecodeData); + cellJpgDec.AddFunc(0x9338a07a, cellJpgDecClose); + cellJpgDec.AddFunc(0xd8ea91f8, cellJpgDecDestroy); + + /*cellJpgDec.AddFunc(0xa9f703e3, cellJpgDecExtOpen); + cellJpgDec.AddFunc(0xb91eb3d2, cellJpgDecExtReadHeader); + cellJpgDec.AddFunc(0x65cbbb16, cellJpgDecExtSetParameter); + cellJpgDec.AddFunc(0x716f8792, cellJpgDecExtDecodeData);*/ +} \ No newline at end of file diff --git a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp index d8f78a9b86..c570e5230b 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp @@ -2,7 +2,6 @@ #include "Emu/SysCalls/SysCalls.h" #include "Emu/SysCalls/SC_FUNC.h" -#include "lodepng/lodepng.h" #include "lodepng/lodepng.cpp" void cellPngDec_init(); @@ -37,7 +36,7 @@ struct CellPngDecInfo struct CellPngDecSrc { u32 srcSelect; // CellPngDecStreamSrcSel - u32 fileName_addr; // const char* + u32 fileName; // const char* u64 fileOffset; // int64_t u32 fileSize; u32 streamPtr; @@ -64,7 +63,7 @@ int cellPngDecDestroy(u32 mainHandle) int cellPngDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInfo) { //current_src.srcSelect = Memory.Read32(src_addr); - current_src.fileName_addr = Memory.Read32(src_addr+4); + current_src.fileName = Memory.Read32(src_addr+4); //current_src.fileOffset = Memory.Read32(src_addr+8); //current_src.fileSize = Memory.Read32(src_addr+12); //current_src.streamPtr = Memory.Read32(src_addr+16); @@ -72,7 +71,7 @@ int cellPngDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInf //current_src.spuThreadEnable = Memory.Read32(src_addr+24); u32& fd_addr = subHandle_addr; // Set file descriptor as sub handler of the decoder - int ret = cellFsOpen(current_src.fileName_addr, 0, fd_addr, NULL, 0); + int ret = cellFsOpen(current_src.fileName, 0, fd_addr, NULL, 0); if(ret != 0) return CELL_PNGDEC_ERROR_OPEN_FILE; return CELL_OK; @@ -96,18 +95,21 @@ int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size if(fileSize < 29) return CELL_PNGDEC_ERROR_HEADER; // Error: The file is smaller than the length of a PNG header - //Write the header to buffer u32 buffer = Memory.Alloc(34,1); // Alloc buffer for PNG header u32 pos_addr = Memory.Alloc(8,1); cellFsLseek(fd, 0, 0, pos_addr); cellFsRead(fd, buffer, 34, NULL); - if (Memory.Read32(buffer) != 0x89504E47 || Memory.Read32(buffer+4) != 0x0D0A1A0A) - return CELL_PNGDEC_ERROR_HEADER; // Error: The first 8 bytes are not a valid PNG signature - - if (Memory.Read32(buffer+12) != 0x49484452) - return CELL_PNGDEC_ERROR_HEADER; // Error: The PNG file does not start with an IHDR chunk + if (Memory.Read32(buffer) != 0x89504E47 || + Memory.Read32(buffer+4) != 0x0D0A1A0A || // Error: The first 8 bytes are not a valid PNG signature + Memory.Read32(buffer+12) != 0x49484452) // Error: The PNG file does not start with an IHDR chunk + { + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + return CELL_PNGDEC_ERROR_HEADER; + } current_info.imageWidth = Memory.Read32(buffer+16); current_info.imageHeight = Memory.Read32(buffer+20); @@ -137,33 +139,18 @@ int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC { u32& fd = subHandle; - //Check size of file + //Get size of file u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct cellFsFstat(fd, sb_addr); u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - if(fileSize < 29) return CELL_PNGDEC_ERROR_HEADER; // Error: The file is smaller than the length of a PNG header - //Write the header to buffer - u32 buffer = Memory.Alloc(34,1); // Alloc buffer for PNG header + //Copy the PNG file to a buffer + u32 buffer = Memory.Alloc(fileSize,1); // Alloc buffer for PNG header u32 pos_addr = Memory.Alloc(8,1); cellFsLseek(fd, 0, 0, pos_addr); - cellFsRead(fd, buffer, 34, NULL); - - if (Memory.Read32(buffer) != 0x89504E47 || Memory.Read32(buffer+4) != 0x0D0A1A0A) - return CELL_PNGDEC_ERROR_HEADER; // Error: The first 8 bytes are not a valid PNG signature - - if (Memory.Read32(buffer+12) != 0x49484452) - return CELL_PNGDEC_ERROR_HEADER; // Error: The PNG file does not start with an IHDR chunk - - current_info.imageWidth = Memory.Read32(buffer+16); - current_info.imageHeight = Memory.Read32(buffer+20); - Memory.Free(buffer); - - //Decode PNG file. (TODO: Is there any faster alternative? Can we do it without external libraries?) - buffer = Memory.Alloc(fileSize,1); - cellFsLseek(fd,0,0,pos_addr); cellFsRead(fd, buffer, fileSize, NULL); + //Decode PNG file. (TODO: Is there any faster alternative? Can we do it without external libraries?) std::vector png; // PNG buffer std::vector image; // Raw buffer @@ -174,8 +161,16 @@ int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC } //Decode - lodepng::decode(image, current_info.imageWidth, current_info.imageHeight, png); - + unsigned width, height; + unsigned error = lodepng::decode(image, width, height, png); + if (error) + { + Memory.Free(sb_addr); + Memory.Free(pos_addr); + Memory.Free(buffer); + return CELL_PNGDEC_ERROR_STREAM_FORMAT; + } + u32 image_size = image.size(); for(u32 i = 0; i < image_size; i+=4){ Memory.Write8(data_addr+i+0, image[i+3]); @@ -197,7 +192,6 @@ int cellPngDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outPa return CELL_OK; } - void cellPngDec_init() { cellPngDec.AddFunc(0x157d30c5, cellPngDecCreate); diff --git a/rpcs3/rpcs3.vcxproj b/rpcs3/rpcs3.vcxproj index 5277d1d0d1..e55b47a184 100644 --- a/rpcs3/rpcs3.vcxproj +++ b/rpcs3/rpcs3.vcxproj @@ -245,6 +245,7 @@ + diff --git a/rpcs3/rpcs3.vcxproj.filters b/rpcs3/rpcs3.vcxproj.filters index ae85f01d96..53aacbdd8e 100644 --- a/rpcs3/rpcs3.vcxproj.filters +++ b/rpcs3/rpcs3.vcxproj.filters @@ -289,6 +289,9 @@ Emu\SysCalls\Modules + + Emu\SysCalls\Modules + From aea7f4aa771f2592f7ed0c007e50ecee1b3cb2c7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Tue, 10 Sep 2013 15:17:02 +0200 Subject: [PATCH 3/7] cellGifDec Module Added *Implemented 'cellGifDec', using 'sys_fs' to access the files and the 'stb_image' (stblib) library to decode GIF buffers to Raw-pixel buffers that can be used as textures. *Replace 'lodepng' and 'jpeg-compressor' libraries with 'stb_image' in cellPngDec and cellJpgDec, respectively. *Fixed minor issues in cellPngDec and cellJpgDec. --- jpeg-compressor/jpgd.cpp | 3172 ----------- jpeg-compressor/jpgd.h | 319 -- lodepng/lodepng.cpp | 6267 --------------------- lodepng/lodepng.h | 1705 ------ rpcs3.sln | 12 +- rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp | 206 + rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp | 38 +- rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp | 36 +- rpcs3/rpcs3.vcxproj | 1 + rpcs3/rpcs3.vcxproj.filters | 3 + stblib/stb_image.c | 4401 +++++++++++++++ stblib/stb_image.h | 329 ++ 12 files changed, 4979 insertions(+), 11510 deletions(-) delete mode 100644 jpeg-compressor/jpgd.cpp delete mode 100644 jpeg-compressor/jpgd.h delete mode 100644 lodepng/lodepng.cpp delete mode 100644 lodepng/lodepng.h create mode 100644 rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp create mode 100644 stblib/stb_image.c create mode 100644 stblib/stb_image.h diff --git a/jpeg-compressor/jpgd.cpp b/jpeg-compressor/jpgd.cpp deleted file mode 100644 index fad9a37a9a..0000000000 --- a/jpeg-compressor/jpgd.cpp +++ /dev/null @@ -1,3172 +0,0 @@ -// jpgd.cpp - C++ class for JPEG decompression. -// Public domain, Rich Geldreich -// Alex Evans: Linear memory allocator (taken from jpge.h). -// v1.04, May. 19, 2012: Code tweaks to fix VS2008 static code analysis warnings (all looked harmless) -// -// Supports progressive and baseline sequential JPEG image files, and the most common chroma subsampling factors: Y, H1V1, H2V1, H1V2, and H2V2. -// -// Chroma upsampling quality: H2V2 is upsampled in the frequency domain, H2V1 and H1V2 are upsampled using point sampling. -// Chroma upsampling reference: "Fast Scheme for Image Size Change in the Compressed Domain" -// http://vision.ai.uiuc.edu/~dugad/research/dct/index.html - -#include "jpgd.h" -#include - -#include -#define JPGD_ASSERT(x) assert(x) - -#ifdef _MSC_VER -#pragma warning (disable : 4611) // warning C4611: interaction between '_setjmp' and C++ object destruction is non-portable -#endif - -// Set to 1 to enable freq. domain chroma upsampling on images using H2V2 subsampling (0=faster nearest neighbor sampling). -// This is slower, but results in higher quality on images with highly saturated colors. -#define JPGD_SUPPORT_FREQ_DOMAIN_UPSAMPLING 1 - -#define JPGD_TRUE (1) -#define JPGD_FALSE (0) - -#define JPGD_MAX(a,b) (((a)>(b)) ? (a) : (b)) -#define JPGD_MIN(a,b) (((a)<(b)) ? (a) : (b)) - -namespace jpgd { - -static inline void *jpgd_malloc(size_t nSize) { return malloc(nSize); } -static inline void jpgd_free(void *p) { free(p); } - -// DCT coefficients are stored in this sequence. -static int g_ZAG[64] = { 0,1,8,16,9,2,3,10,17,24,32,25,18,11,4,5,12,19,26,33,40,48,41,34,27,20,13,6,7,14,21,28,35,42,49,56,57,50,43,36,29,22,15,23,30,37,44,51,58,59,52,45,38,31,39,46,53,60,61,54,47,55,62,63 }; - -enum JPEG_MARKER -{ - M_SOF0 = 0xC0, M_SOF1 = 0xC1, M_SOF2 = 0xC2, M_SOF3 = 0xC3, M_SOF5 = 0xC5, M_SOF6 = 0xC6, M_SOF7 = 0xC7, M_JPG = 0xC8, - M_SOF9 = 0xC9, M_SOF10 = 0xCA, M_SOF11 = 0xCB, M_SOF13 = 0xCD, M_SOF14 = 0xCE, M_SOF15 = 0xCF, M_DHT = 0xC4, M_DAC = 0xCC, - M_RST0 = 0xD0, M_RST1 = 0xD1, M_RST2 = 0xD2, M_RST3 = 0xD3, M_RST4 = 0xD4, M_RST5 = 0xD5, M_RST6 = 0xD6, M_RST7 = 0xD7, - M_SOI = 0xD8, M_EOI = 0xD9, M_SOS = 0xDA, M_DQT = 0xDB, M_DNL = 0xDC, M_DRI = 0xDD, M_DHP = 0xDE, M_EXP = 0xDF, - M_APP0 = 0xE0, M_APP15 = 0xEF, M_JPG0 = 0xF0, M_JPG13 = 0xFD, M_COM = 0xFE, M_TEM = 0x01, M_ERROR = 0x100, RST0 = 0xD0 -}; - -enum JPEG_SUBSAMPLING { JPGD_GRAYSCALE = 0, JPGD_YH1V1, JPGD_YH2V1, JPGD_YH1V2, JPGD_YH2V2 }; - -#define CONST_BITS 13 -#define PASS1_BITS 2 -#define SCALEDONE ((int32)1) - -#define FIX_0_298631336 ((int32)2446) /* FIX(0.298631336) */ -#define FIX_0_390180644 ((int32)3196) /* FIX(0.390180644) */ -#define FIX_0_541196100 ((int32)4433) /* FIX(0.541196100) */ -#define FIX_0_765366865 ((int32)6270) /* FIX(0.765366865) */ -#define FIX_0_899976223 ((int32)7373) /* FIX(0.899976223) */ -#define FIX_1_175875602 ((int32)9633) /* FIX(1.175875602) */ -#define FIX_1_501321110 ((int32)12299) /* FIX(1.501321110) */ -#define FIX_1_847759065 ((int32)15137) /* FIX(1.847759065) */ -#define FIX_1_961570560 ((int32)16069) /* FIX(1.961570560) */ -#define FIX_2_053119869 ((int32)16819) /* FIX(2.053119869) */ -#define FIX_2_562915447 ((int32)20995) /* FIX(2.562915447) */ -#define FIX_3_072711026 ((int32)25172) /* FIX(3.072711026) */ - -#define DESCALE(x,n) (((x) + (SCALEDONE << ((n)-1))) >> (n)) -#define DESCALE_ZEROSHIFT(x,n) (((x) + (128 << (n)) + (SCALEDONE << ((n)-1))) >> (n)) - -#define MULTIPLY(var, cnst) ((var) * (cnst)) - -#define CLAMP(i) ((static_cast(i) > 255) ? (((~i) >> 31) & 0xFF) : (i)) - -// Compiler creates a fast path 1D IDCT for X non-zero columns -template -struct Row -{ - static void idct(int* pTemp, const jpgd_block_t* pSrc) - { - // ACCESS_COL() will be optimized at compile time to either an array access, or 0. - #define ACCESS_COL(x) (((x) < NONZERO_COLS) ? (int)pSrc[x] : 0) - - const int z2 = ACCESS_COL(2), z3 = ACCESS_COL(6); - - const int z1 = MULTIPLY(z2 + z3, FIX_0_541196100); - const int tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); - const int tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); - - const int tmp0 = (ACCESS_COL(0) + ACCESS_COL(4)) << CONST_BITS; - const int tmp1 = (ACCESS_COL(0) - ACCESS_COL(4)) << CONST_BITS; - - const int tmp10 = tmp0 + tmp3, tmp13 = tmp0 - tmp3, tmp11 = tmp1 + tmp2, tmp12 = tmp1 - tmp2; - - const int atmp0 = ACCESS_COL(7), atmp1 = ACCESS_COL(5), atmp2 = ACCESS_COL(3), atmp3 = ACCESS_COL(1); - - const int bz1 = atmp0 + atmp3, bz2 = atmp1 + atmp2, bz3 = atmp0 + atmp2, bz4 = atmp1 + atmp3; - const int bz5 = MULTIPLY(bz3 + bz4, FIX_1_175875602); - - const int az1 = MULTIPLY(bz1, - FIX_0_899976223); - const int az2 = MULTIPLY(bz2, - FIX_2_562915447); - const int az3 = MULTIPLY(bz3, - FIX_1_961570560) + bz5; - const int az4 = MULTIPLY(bz4, - FIX_0_390180644) + bz5; - - const int btmp0 = MULTIPLY(atmp0, FIX_0_298631336) + az1 + az3; - const int btmp1 = MULTIPLY(atmp1, FIX_2_053119869) + az2 + az4; - const int btmp2 = MULTIPLY(atmp2, FIX_3_072711026) + az2 + az3; - const int btmp3 = MULTIPLY(atmp3, FIX_1_501321110) + az1 + az4; - - pTemp[0] = DESCALE(tmp10 + btmp3, CONST_BITS-PASS1_BITS); - pTemp[7] = DESCALE(tmp10 - btmp3, CONST_BITS-PASS1_BITS); - pTemp[1] = DESCALE(tmp11 + btmp2, CONST_BITS-PASS1_BITS); - pTemp[6] = DESCALE(tmp11 - btmp2, CONST_BITS-PASS1_BITS); - pTemp[2] = DESCALE(tmp12 + btmp1, CONST_BITS-PASS1_BITS); - pTemp[5] = DESCALE(tmp12 - btmp1, CONST_BITS-PASS1_BITS); - pTemp[3] = DESCALE(tmp13 + btmp0, CONST_BITS-PASS1_BITS); - pTemp[4] = DESCALE(tmp13 - btmp0, CONST_BITS-PASS1_BITS); - } -}; - -template <> -struct Row<0> -{ - static void idct(int* pTemp, const jpgd_block_t* pSrc) - { -#ifdef _MSC_VER - pTemp; pSrc; -#endif - } -}; - -template <> -struct Row<1> -{ - static void idct(int* pTemp, const jpgd_block_t* pSrc) - { - const int dcval = (pSrc[0] << PASS1_BITS); - - pTemp[0] = dcval; - pTemp[1] = dcval; - pTemp[2] = dcval; - pTemp[3] = dcval; - pTemp[4] = dcval; - pTemp[5] = dcval; - pTemp[6] = dcval; - pTemp[7] = dcval; - } -}; - -// Compiler creates a fast path 1D IDCT for X non-zero rows -template -struct Col -{ - static void idct(uint8* pDst_ptr, const int* pTemp) - { - // ACCESS_ROW() will be optimized at compile time to either an array access, or 0. - #define ACCESS_ROW(x) (((x) < NONZERO_ROWS) ? pTemp[x * 8] : 0) - - const int z2 = ACCESS_ROW(2); - const int z3 = ACCESS_ROW(6); - - const int z1 = MULTIPLY(z2 + z3, FIX_0_541196100); - const int tmp2 = z1 + MULTIPLY(z3, - FIX_1_847759065); - const int tmp3 = z1 + MULTIPLY(z2, FIX_0_765366865); - - const int tmp0 = (ACCESS_ROW(0) + ACCESS_ROW(4)) << CONST_BITS; - const int tmp1 = (ACCESS_ROW(0) - ACCESS_ROW(4)) << CONST_BITS; - - const int tmp10 = tmp0 + tmp3, tmp13 = tmp0 - tmp3, tmp11 = tmp1 + tmp2, tmp12 = tmp1 - tmp2; - - const int atmp0 = ACCESS_ROW(7), atmp1 = ACCESS_ROW(5), atmp2 = ACCESS_ROW(3), atmp3 = ACCESS_ROW(1); - - const int bz1 = atmp0 + atmp3, bz2 = atmp1 + atmp2, bz3 = atmp0 + atmp2, bz4 = atmp1 + atmp3; - const int bz5 = MULTIPLY(bz3 + bz4, FIX_1_175875602); - - const int az1 = MULTIPLY(bz1, - FIX_0_899976223); - const int az2 = MULTIPLY(bz2, - FIX_2_562915447); - const int az3 = MULTIPLY(bz3, - FIX_1_961570560) + bz5; - const int az4 = MULTIPLY(bz4, - FIX_0_390180644) + bz5; - - const int btmp0 = MULTIPLY(atmp0, FIX_0_298631336) + az1 + az3; - const int btmp1 = MULTIPLY(atmp1, FIX_2_053119869) + az2 + az4; - const int btmp2 = MULTIPLY(atmp2, FIX_3_072711026) + az2 + az3; - const int btmp3 = MULTIPLY(atmp3, FIX_1_501321110) + az1 + az4; - - int i = DESCALE_ZEROSHIFT(tmp10 + btmp3, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*0] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp10 - btmp3, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*7] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp11 + btmp2, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*1] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp11 - btmp2, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*6] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp12 + btmp1, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*2] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp12 - btmp1, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*5] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp13 + btmp0, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*3] = (uint8)CLAMP(i); - - i = DESCALE_ZEROSHIFT(tmp13 - btmp0, CONST_BITS+PASS1_BITS+3); - pDst_ptr[8*4] = (uint8)CLAMP(i); - } -}; - -template <> -struct Col<1> -{ - static void idct(uint8* pDst_ptr, const int* pTemp) - { - int dcval = DESCALE_ZEROSHIFT(pTemp[0], PASS1_BITS+3); - const uint8 dcval_clamped = (uint8)CLAMP(dcval); - pDst_ptr[0*8] = dcval_clamped; - pDst_ptr[1*8] = dcval_clamped; - pDst_ptr[2*8] = dcval_clamped; - pDst_ptr[3*8] = dcval_clamped; - pDst_ptr[4*8] = dcval_clamped; - pDst_ptr[5*8] = dcval_clamped; - pDst_ptr[6*8] = dcval_clamped; - pDst_ptr[7*8] = dcval_clamped; - } -}; - -static const uint8 s_idct_row_table[] = -{ - 1,0,0,0,0,0,0,0, 2,0,0,0,0,0,0,0, 2,1,0,0,0,0,0,0, 2,1,1,0,0,0,0,0, 2,2,1,0,0,0,0,0, 3,2,1,0,0,0,0,0, 4,2,1,0,0,0,0,0, 4,3,1,0,0,0,0,0, - 4,3,2,0,0,0,0,0, 4,3,2,1,0,0,0,0, 4,3,2,1,1,0,0,0, 4,3,2,2,1,0,0,0, 4,3,3,2,1,0,0,0, 4,4,3,2,1,0,0,0, 5,4,3,2,1,0,0,0, 6,4,3,2,1,0,0,0, - 6,5,3,2,1,0,0,0, 6,5,4,2,1,0,0,0, 6,5,4,3,1,0,0,0, 6,5,4,3,2,0,0,0, 6,5,4,3,2,1,0,0, 6,5,4,3,2,1,1,0, 6,5,4,3,2,2,1,0, 6,5,4,3,3,2,1,0, - 6,5,4,4,3,2,1,0, 6,5,5,4,3,2,1,0, 6,6,5,4,3,2,1,0, 7,6,5,4,3,2,1,0, 8,6,5,4,3,2,1,0, 8,7,5,4,3,2,1,0, 8,7,6,4,3,2,1,0, 8,7,6,5,3,2,1,0, - 8,7,6,5,4,2,1,0, 8,7,6,5,4,3,1,0, 8,7,6,5,4,3,2,0, 8,7,6,5,4,3,2,1, 8,7,6,5,4,3,2,2, 8,7,6,5,4,3,3,2, 8,7,6,5,4,4,3,2, 8,7,6,5,5,4,3,2, - 8,7,6,6,5,4,3,2, 8,7,7,6,5,4,3,2, 8,8,7,6,5,4,3,2, 8,8,8,6,5,4,3,2, 8,8,8,7,5,4,3,2, 8,8,8,7,6,4,3,2, 8,8,8,7,6,5,3,2, 8,8,8,7,6,5,4,2, - 8,8,8,7,6,5,4,3, 8,8,8,7,6,5,4,4, 8,8,8,7,6,5,5,4, 8,8,8,7,6,6,5,4, 8,8,8,7,7,6,5,4, 8,8,8,8,7,6,5,4, 8,8,8,8,8,6,5,4, 8,8,8,8,8,7,5,4, - 8,8,8,8,8,7,6,4, 8,8,8,8,8,7,6,5, 8,8,8,8,8,7,6,6, 8,8,8,8,8,7,7,6, 8,8,8,8,8,8,7,6, 8,8,8,8,8,8,8,6, 8,8,8,8,8,8,8,7, 8,8,8,8,8,8,8,8, -}; - -static const uint8 s_idct_col_table[] = { 1, 1, 2, 3, 3, 3, 3, 3, 3, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 6, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8 }; - -void idct(const jpgd_block_t* pSrc_ptr, uint8* pDst_ptr, int block_max_zag) -{ - JPGD_ASSERT(block_max_zag >= 1); - JPGD_ASSERT(block_max_zag <= 64); - - if (block_max_zag <= 1) - { - int k = ((pSrc_ptr[0] + 4) >> 3) + 128; - k = CLAMP(k); - k = k | (k<<8); - k = k | (k<<16); - - for (int i = 8; i > 0; i--) - { - *(int*)&pDst_ptr[0] = k; - *(int*)&pDst_ptr[4] = k; - pDst_ptr += 8; - } - return; - } - - int temp[64]; - - const jpgd_block_t* pSrc = pSrc_ptr; - int* pTemp = temp; - - const uint8* pRow_tab = &s_idct_row_table[(block_max_zag - 1) * 8]; - int i; - for (i = 8; i > 0; i--, pRow_tab++) - { - switch (*pRow_tab) - { - case 0: Row<0>::idct(pTemp, pSrc); break; - case 1: Row<1>::idct(pTemp, pSrc); break; - case 2: Row<2>::idct(pTemp, pSrc); break; - case 3: Row<3>::idct(pTemp, pSrc); break; - case 4: Row<4>::idct(pTemp, pSrc); break; - case 5: Row<5>::idct(pTemp, pSrc); break; - case 6: Row<6>::idct(pTemp, pSrc); break; - case 7: Row<7>::idct(pTemp, pSrc); break; - case 8: Row<8>::idct(pTemp, pSrc); break; - } - - pSrc += 8; - pTemp += 8; - } - - pTemp = temp; - - const int nonzero_rows = s_idct_col_table[block_max_zag - 1]; - for (i = 8; i > 0; i--) - { - switch (nonzero_rows) - { - case 1: Col<1>::idct(pDst_ptr, pTemp); break; - case 2: Col<2>::idct(pDst_ptr, pTemp); break; - case 3: Col<3>::idct(pDst_ptr, pTemp); break; - case 4: Col<4>::idct(pDst_ptr, pTemp); break; - case 5: Col<5>::idct(pDst_ptr, pTemp); break; - case 6: Col<6>::idct(pDst_ptr, pTemp); break; - case 7: Col<7>::idct(pDst_ptr, pTemp); break; - case 8: Col<8>::idct(pDst_ptr, pTemp); break; - } - - pTemp++; - pDst_ptr++; - } -} - -void idct_4x4(const jpgd_block_t* pSrc_ptr, uint8* pDst_ptr) -{ - int temp[64]; - int* pTemp = temp; - const jpgd_block_t* pSrc = pSrc_ptr; - - for (int i = 4; i > 0; i--) - { - Row<4>::idct(pTemp, pSrc); - pSrc += 8; - pTemp += 8; - } - - pTemp = temp; - for (int i = 8; i > 0; i--) - { - Col<4>::idct(pDst_ptr, pTemp); - pTemp++; - pDst_ptr++; - } -} - -// Retrieve one character from the input stream. -inline uint jpeg_decoder::get_char() -{ - // Any bytes remaining in buffer? - if (!m_in_buf_left) - { - // Try to get more bytes. - prep_in_buffer(); - // Still nothing to get? - if (!m_in_buf_left) - { - // Pad the end of the stream with 0xFF 0xD9 (EOI marker) - int t = m_tem_flag; - m_tem_flag ^= 1; - if (t) - return 0xD9; - else - return 0xFF; - } - } - - uint c = *m_pIn_buf_ofs++; - m_in_buf_left--; - - return c; -} - -// Same as previous method, except can indicate if the character is a pad character or not. -inline uint jpeg_decoder::get_char(bool *pPadding_flag) -{ - if (!m_in_buf_left) - { - prep_in_buffer(); - if (!m_in_buf_left) - { - *pPadding_flag = true; - int t = m_tem_flag; - m_tem_flag ^= 1; - if (t) - return 0xD9; - else - return 0xFF; - } - } - - *pPadding_flag = false; - - uint c = *m_pIn_buf_ofs++; - m_in_buf_left--; - - return c; -} - -// Inserts a previously retrieved character back into the input buffer. -inline void jpeg_decoder::stuff_char(uint8 q) -{ - *(--m_pIn_buf_ofs) = q; - m_in_buf_left++; -} - -// Retrieves one character from the input stream, but does not read past markers. Will continue to return 0xFF when a marker is encountered. -inline uint8 jpeg_decoder::get_octet() -{ - bool padding_flag; - int c = get_char(&padding_flag); - - if (c == 0xFF) - { - if (padding_flag) - return 0xFF; - - c = get_char(&padding_flag); - if (padding_flag) - { - stuff_char(0xFF); - return 0xFF; - } - - if (c == 0x00) - return 0xFF; - else - { - stuff_char(static_cast(c)); - stuff_char(0xFF); - return 0xFF; - } - } - - return static_cast(c); -} - -// Retrieves a variable number of bits from the input stream. Does not recognize markers. -inline uint jpeg_decoder::get_bits(int num_bits) -{ - if (!num_bits) - return 0; - - uint i = m_bit_buf >> (32 - num_bits); - - if ((m_bits_left -= num_bits) <= 0) - { - m_bit_buf <<= (num_bits += m_bits_left); - - uint c1 = get_char(); - uint c2 = get_char(); - m_bit_buf = (m_bit_buf & 0xFFFF0000) | (c1 << 8) | c2; - - m_bit_buf <<= -m_bits_left; - - m_bits_left += 16; - - JPGD_ASSERT(m_bits_left >= 0); - } - else - m_bit_buf <<= num_bits; - - return i; -} - -// Retrieves a variable number of bits from the input stream. Markers will not be read into the input bit buffer. Instead, an infinite number of all 1's will be returned when a marker is encountered. -inline uint jpeg_decoder::get_bits_no_markers(int num_bits) -{ - if (!num_bits) - return 0; - - uint i = m_bit_buf >> (32 - num_bits); - - if ((m_bits_left -= num_bits) <= 0) - { - m_bit_buf <<= (num_bits += m_bits_left); - - if ((m_in_buf_left < 2) || (m_pIn_buf_ofs[0] == 0xFF) || (m_pIn_buf_ofs[1] == 0xFF)) - { - uint c1 = get_octet(); - uint c2 = get_octet(); - m_bit_buf |= (c1 << 8) | c2; - } - else - { - m_bit_buf |= ((uint)m_pIn_buf_ofs[0] << 8) | m_pIn_buf_ofs[1]; - m_in_buf_left -= 2; - m_pIn_buf_ofs += 2; - } - - m_bit_buf <<= -m_bits_left; - - m_bits_left += 16; - - JPGD_ASSERT(m_bits_left >= 0); - } - else - m_bit_buf <<= num_bits; - - return i; -} - -// Decodes a Huffman encoded symbol. -inline int jpeg_decoder::huff_decode(huff_tables *pH) -{ - int symbol; - - // Check first 8-bits: do we have a complete symbol? - if ((symbol = pH->look_up[m_bit_buf >> 24]) < 0) - { - // Decode more bits, use a tree traversal to find symbol. - int ofs = 23; - do - { - symbol = pH->tree[-(int)(symbol + ((m_bit_buf >> ofs) & 1))]; - ofs--; - } while (symbol < 0); - - get_bits_no_markers(8 + (23 - ofs)); - } - else - get_bits_no_markers(pH->code_size[symbol]); - - return symbol; -} - -// Decodes a Huffman encoded symbol. -inline int jpeg_decoder::huff_decode(huff_tables *pH, int& extra_bits) -{ - int symbol; - - // Check first 8-bits: do we have a complete symbol? - if ((symbol = pH->look_up2[m_bit_buf >> 24]) < 0) - { - // Use a tree traversal to find symbol. - int ofs = 23; - do - { - symbol = pH->tree[-(int)(symbol + ((m_bit_buf >> ofs) & 1))]; - ofs--; - } while (symbol < 0); - - get_bits_no_markers(8 + (23 - ofs)); - - extra_bits = get_bits_no_markers(symbol & 0xF); - } - else - { - JPGD_ASSERT(((symbol >> 8) & 31) == pH->code_size[symbol & 255] + ((symbol & 0x8000) ? (symbol & 15) : 0)); - - if (symbol & 0x8000) - { - get_bits_no_markers((symbol >> 8) & 31); - extra_bits = symbol >> 16; - } - else - { - int code_size = (symbol >> 8) & 31; - int num_extra_bits = symbol & 0xF; - int bits = code_size + num_extra_bits; - if (bits <= (m_bits_left + 16)) - extra_bits = get_bits_no_markers(bits) & ((1 << num_extra_bits) - 1); - else - { - get_bits_no_markers(code_size); - extra_bits = get_bits_no_markers(num_extra_bits); - } - } - - symbol &= 0xFF; - } - - return symbol; -} - -// Tables and macro used to fully decode the DPCM differences. -static const int s_extend_test[16] = { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080, 0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 }; -static const int s_extend_offset[16] = { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1, ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1, ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1, ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 }; -static const int s_extend_mask[] = { 0, (1<<0), (1<<1), (1<<2), (1<<3), (1<<4), (1<<5), (1<<6), (1<<7), (1<<8), (1<<9), (1<<10), (1<<11), (1<<12), (1<<13), (1<<14), (1<<15), (1<<16) }; -// The logical AND's in this macro are to shut up static code analysis (aren't really necessary - couldn't find another way to do this) -#define JPGD_HUFF_EXTEND(x, s) (((x) < s_extend_test[s & 15]) ? ((x) + s_extend_offset[s & 15]) : (x)) - -// Clamps a value between 0-255. -inline uint8 jpeg_decoder::clamp(int i) -{ - if (static_cast(i) > 255) - i = (((~i) >> 31) & 0xFF); - - return static_cast(i); -} - -namespace DCT_Upsample -{ - struct Matrix44 - { - typedef int Element_Type; - enum { NUM_ROWS = 4, NUM_COLS = 4 }; - - Element_Type v[NUM_ROWS][NUM_COLS]; - - inline int rows() const { return NUM_ROWS; } - inline int cols() const { return NUM_COLS; } - - inline const Element_Type & at(int r, int c) const { return v[r][c]; } - inline Element_Type & at(int r, int c) { return v[r][c]; } - - inline Matrix44() { } - - inline Matrix44& operator += (const Matrix44& a) - { - for (int r = 0; r < NUM_ROWS; r++) - { - at(r, 0) += a.at(r, 0); - at(r, 1) += a.at(r, 1); - at(r, 2) += a.at(r, 2); - at(r, 3) += a.at(r, 3); - } - return *this; - } - - inline Matrix44& operator -= (const Matrix44& a) - { - for (int r = 0; r < NUM_ROWS; r++) - { - at(r, 0) -= a.at(r, 0); - at(r, 1) -= a.at(r, 1); - at(r, 2) -= a.at(r, 2); - at(r, 3) -= a.at(r, 3); - } - return *this; - } - - friend inline Matrix44 operator + (const Matrix44& a, const Matrix44& b) - { - Matrix44 ret; - for (int r = 0; r < NUM_ROWS; r++) - { - ret.at(r, 0) = a.at(r, 0) + b.at(r, 0); - ret.at(r, 1) = a.at(r, 1) + b.at(r, 1); - ret.at(r, 2) = a.at(r, 2) + b.at(r, 2); - ret.at(r, 3) = a.at(r, 3) + b.at(r, 3); - } - return ret; - } - - friend inline Matrix44 operator - (const Matrix44& a, const Matrix44& b) - { - Matrix44 ret; - for (int r = 0; r < NUM_ROWS; r++) - { - ret.at(r, 0) = a.at(r, 0) - b.at(r, 0); - ret.at(r, 1) = a.at(r, 1) - b.at(r, 1); - ret.at(r, 2) = a.at(r, 2) - b.at(r, 2); - ret.at(r, 3) = a.at(r, 3) - b.at(r, 3); - } - return ret; - } - - static inline void add_and_store(jpgd_block_t* pDst, const Matrix44& a, const Matrix44& b) - { - for (int r = 0; r < 4; r++) - { - pDst[0*8 + r] = static_cast(a.at(r, 0) + b.at(r, 0)); - pDst[1*8 + r] = static_cast(a.at(r, 1) + b.at(r, 1)); - pDst[2*8 + r] = static_cast(a.at(r, 2) + b.at(r, 2)); - pDst[3*8 + r] = static_cast(a.at(r, 3) + b.at(r, 3)); - } - } - - static inline void sub_and_store(jpgd_block_t* pDst, const Matrix44& a, const Matrix44& b) - { - for (int r = 0; r < 4; r++) - { - pDst[0*8 + r] = static_cast(a.at(r, 0) - b.at(r, 0)); - pDst[1*8 + r] = static_cast(a.at(r, 1) - b.at(r, 1)); - pDst[2*8 + r] = static_cast(a.at(r, 2) - b.at(r, 2)); - pDst[3*8 + r] = static_cast(a.at(r, 3) - b.at(r, 3)); - } - } - }; - - const int FRACT_BITS = 10; - const int SCALE = 1 << FRACT_BITS; - - typedef int Temp_Type; - #define D(i) (((i) + (SCALE >> 1)) >> FRACT_BITS) - #define F(i) ((int)((i) * SCALE + .5f)) - - // Any decent C++ compiler will optimize this at compile time to a 0, or an array access. - #define AT(c, r) ((((c)>=NUM_COLS)||((r)>=NUM_ROWS)) ? 0 : pSrc[(c)+(r)*8]) - - // NUM_ROWS/NUM_COLS = # of non-zero rows/cols in input matrix - template - struct P_Q - { - static void calc(Matrix44& P, Matrix44& Q, const jpgd_block_t* pSrc) - { - // 4x8 = 4x8 times 8x8, matrix 0 is constant - const Temp_Type X000 = AT(0, 0); - const Temp_Type X001 = AT(0, 1); - const Temp_Type X002 = AT(0, 2); - const Temp_Type X003 = AT(0, 3); - const Temp_Type X004 = AT(0, 4); - const Temp_Type X005 = AT(0, 5); - const Temp_Type X006 = AT(0, 6); - const Temp_Type X007 = AT(0, 7); - const Temp_Type X010 = D(F(0.415735f) * AT(1, 0) + F(0.791065f) * AT(3, 0) + F(-0.352443f) * AT(5, 0) + F(0.277785f) * AT(7, 0)); - const Temp_Type X011 = D(F(0.415735f) * AT(1, 1) + F(0.791065f) * AT(3, 1) + F(-0.352443f) * AT(5, 1) + F(0.277785f) * AT(7, 1)); - const Temp_Type X012 = D(F(0.415735f) * AT(1, 2) + F(0.791065f) * AT(3, 2) + F(-0.352443f) * AT(5, 2) + F(0.277785f) * AT(7, 2)); - const Temp_Type X013 = D(F(0.415735f) * AT(1, 3) + F(0.791065f) * AT(3, 3) + F(-0.352443f) * AT(5, 3) + F(0.277785f) * AT(7, 3)); - const Temp_Type X014 = D(F(0.415735f) * AT(1, 4) + F(0.791065f) * AT(3, 4) + F(-0.352443f) * AT(5, 4) + F(0.277785f) * AT(7, 4)); - const Temp_Type X015 = D(F(0.415735f) * AT(1, 5) + F(0.791065f) * AT(3, 5) + F(-0.352443f) * AT(5, 5) + F(0.277785f) * AT(7, 5)); - const Temp_Type X016 = D(F(0.415735f) * AT(1, 6) + F(0.791065f) * AT(3, 6) + F(-0.352443f) * AT(5, 6) + F(0.277785f) * AT(7, 6)); - const Temp_Type X017 = D(F(0.415735f) * AT(1, 7) + F(0.791065f) * AT(3, 7) + F(-0.352443f) * AT(5, 7) + F(0.277785f) * AT(7, 7)); - const Temp_Type X020 = AT(4, 0); - const Temp_Type X021 = AT(4, 1); - const Temp_Type X022 = AT(4, 2); - const Temp_Type X023 = AT(4, 3); - const Temp_Type X024 = AT(4, 4); - const Temp_Type X025 = AT(4, 5); - const Temp_Type X026 = AT(4, 6); - const Temp_Type X027 = AT(4, 7); - const Temp_Type X030 = D(F(0.022887f) * AT(1, 0) + F(-0.097545f) * AT(3, 0) + F(0.490393f) * AT(5, 0) + F(0.865723f) * AT(7, 0)); - const Temp_Type X031 = D(F(0.022887f) * AT(1, 1) + F(-0.097545f) * AT(3, 1) + F(0.490393f) * AT(5, 1) + F(0.865723f) * AT(7, 1)); - const Temp_Type X032 = D(F(0.022887f) * AT(1, 2) + F(-0.097545f) * AT(3, 2) + F(0.490393f) * AT(5, 2) + F(0.865723f) * AT(7, 2)); - const Temp_Type X033 = D(F(0.022887f) * AT(1, 3) + F(-0.097545f) * AT(3, 3) + F(0.490393f) * AT(5, 3) + F(0.865723f) * AT(7, 3)); - const Temp_Type X034 = D(F(0.022887f) * AT(1, 4) + F(-0.097545f) * AT(3, 4) + F(0.490393f) * AT(5, 4) + F(0.865723f) * AT(7, 4)); - const Temp_Type X035 = D(F(0.022887f) * AT(1, 5) + F(-0.097545f) * AT(3, 5) + F(0.490393f) * AT(5, 5) + F(0.865723f) * AT(7, 5)); - const Temp_Type X036 = D(F(0.022887f) * AT(1, 6) + F(-0.097545f) * AT(3, 6) + F(0.490393f) * AT(5, 6) + F(0.865723f) * AT(7, 6)); - const Temp_Type X037 = D(F(0.022887f) * AT(1, 7) + F(-0.097545f) * AT(3, 7) + F(0.490393f) * AT(5, 7) + F(0.865723f) * AT(7, 7)); - - // 4x4 = 4x8 times 8x4, matrix 1 is constant - P.at(0, 0) = X000; - P.at(0, 1) = D(X001 * F(0.415735f) + X003 * F(0.791065f) + X005 * F(-0.352443f) + X007 * F(0.277785f)); - P.at(0, 2) = X004; - P.at(0, 3) = D(X001 * F(0.022887f) + X003 * F(-0.097545f) + X005 * F(0.490393f) + X007 * F(0.865723f)); - P.at(1, 0) = X010; - P.at(1, 1) = D(X011 * F(0.415735f) + X013 * F(0.791065f) + X015 * F(-0.352443f) + X017 * F(0.277785f)); - P.at(1, 2) = X014; - P.at(1, 3) = D(X011 * F(0.022887f) + X013 * F(-0.097545f) + X015 * F(0.490393f) + X017 * F(0.865723f)); - P.at(2, 0) = X020; - P.at(2, 1) = D(X021 * F(0.415735f) + X023 * F(0.791065f) + X025 * F(-0.352443f) + X027 * F(0.277785f)); - P.at(2, 2) = X024; - P.at(2, 3) = D(X021 * F(0.022887f) + X023 * F(-0.097545f) + X025 * F(0.490393f) + X027 * F(0.865723f)); - P.at(3, 0) = X030; - P.at(3, 1) = D(X031 * F(0.415735f) + X033 * F(0.791065f) + X035 * F(-0.352443f) + X037 * F(0.277785f)); - P.at(3, 2) = X034; - P.at(3, 3) = D(X031 * F(0.022887f) + X033 * F(-0.097545f) + X035 * F(0.490393f) + X037 * F(0.865723f)); - // 40 muls 24 adds - - // 4x4 = 4x8 times 8x4, matrix 1 is constant - Q.at(0, 0) = D(X001 * F(0.906127f) + X003 * F(-0.318190f) + X005 * F(0.212608f) + X007 * F(-0.180240f)); - Q.at(0, 1) = X002; - Q.at(0, 2) = D(X001 * F(-0.074658f) + X003 * F(0.513280f) + X005 * F(0.768178f) + X007 * F(-0.375330f)); - Q.at(0, 3) = X006; - Q.at(1, 0) = D(X011 * F(0.906127f) + X013 * F(-0.318190f) + X015 * F(0.212608f) + X017 * F(-0.180240f)); - Q.at(1, 1) = X012; - Q.at(1, 2) = D(X011 * F(-0.074658f) + X013 * F(0.513280f) + X015 * F(0.768178f) + X017 * F(-0.375330f)); - Q.at(1, 3) = X016; - Q.at(2, 0) = D(X021 * F(0.906127f) + X023 * F(-0.318190f) + X025 * F(0.212608f) + X027 * F(-0.180240f)); - Q.at(2, 1) = X022; - Q.at(2, 2) = D(X021 * F(-0.074658f) + X023 * F(0.513280f) + X025 * F(0.768178f) + X027 * F(-0.375330f)); - Q.at(2, 3) = X026; - Q.at(3, 0) = D(X031 * F(0.906127f) + X033 * F(-0.318190f) + X035 * F(0.212608f) + X037 * F(-0.180240f)); - Q.at(3, 1) = X032; - Q.at(3, 2) = D(X031 * F(-0.074658f) + X033 * F(0.513280f) + X035 * F(0.768178f) + X037 * F(-0.375330f)); - Q.at(3, 3) = X036; - // 40 muls 24 adds - } - }; - - template - struct R_S - { - static void calc(Matrix44& R, Matrix44& S, const jpgd_block_t* pSrc) - { - // 4x8 = 4x8 times 8x8, matrix 0 is constant - const Temp_Type X100 = D(F(0.906127f) * AT(1, 0) + F(-0.318190f) * AT(3, 0) + F(0.212608f) * AT(5, 0) + F(-0.180240f) * AT(7, 0)); - const Temp_Type X101 = D(F(0.906127f) * AT(1, 1) + F(-0.318190f) * AT(3, 1) + F(0.212608f) * AT(5, 1) + F(-0.180240f) * AT(7, 1)); - const Temp_Type X102 = D(F(0.906127f) * AT(1, 2) + F(-0.318190f) * AT(3, 2) + F(0.212608f) * AT(5, 2) + F(-0.180240f) * AT(7, 2)); - const Temp_Type X103 = D(F(0.906127f) * AT(1, 3) + F(-0.318190f) * AT(3, 3) + F(0.212608f) * AT(5, 3) + F(-0.180240f) * AT(7, 3)); - const Temp_Type X104 = D(F(0.906127f) * AT(1, 4) + F(-0.318190f) * AT(3, 4) + F(0.212608f) * AT(5, 4) + F(-0.180240f) * AT(7, 4)); - const Temp_Type X105 = D(F(0.906127f) * AT(1, 5) + F(-0.318190f) * AT(3, 5) + F(0.212608f) * AT(5, 5) + F(-0.180240f) * AT(7, 5)); - const Temp_Type X106 = D(F(0.906127f) * AT(1, 6) + F(-0.318190f) * AT(3, 6) + F(0.212608f) * AT(5, 6) + F(-0.180240f) * AT(7, 6)); - const Temp_Type X107 = D(F(0.906127f) * AT(1, 7) + F(-0.318190f) * AT(3, 7) + F(0.212608f) * AT(5, 7) + F(-0.180240f) * AT(7, 7)); - const Temp_Type X110 = AT(2, 0); - const Temp_Type X111 = AT(2, 1); - const Temp_Type X112 = AT(2, 2); - const Temp_Type X113 = AT(2, 3); - const Temp_Type X114 = AT(2, 4); - const Temp_Type X115 = AT(2, 5); - const Temp_Type X116 = AT(2, 6); - const Temp_Type X117 = AT(2, 7); - const Temp_Type X120 = D(F(-0.074658f) * AT(1, 0) + F(0.513280f) * AT(3, 0) + F(0.768178f) * AT(5, 0) + F(-0.375330f) * AT(7, 0)); - const Temp_Type X121 = D(F(-0.074658f) * AT(1, 1) + F(0.513280f) * AT(3, 1) + F(0.768178f) * AT(5, 1) + F(-0.375330f) * AT(7, 1)); - const Temp_Type X122 = D(F(-0.074658f) * AT(1, 2) + F(0.513280f) * AT(3, 2) + F(0.768178f) * AT(5, 2) + F(-0.375330f) * AT(7, 2)); - const Temp_Type X123 = D(F(-0.074658f) * AT(1, 3) + F(0.513280f) * AT(3, 3) + F(0.768178f) * AT(5, 3) + F(-0.375330f) * AT(7, 3)); - const Temp_Type X124 = D(F(-0.074658f) * AT(1, 4) + F(0.513280f) * AT(3, 4) + F(0.768178f) * AT(5, 4) + F(-0.375330f) * AT(7, 4)); - const Temp_Type X125 = D(F(-0.074658f) * AT(1, 5) + F(0.513280f) * AT(3, 5) + F(0.768178f) * AT(5, 5) + F(-0.375330f) * AT(7, 5)); - const Temp_Type X126 = D(F(-0.074658f) * AT(1, 6) + F(0.513280f) * AT(3, 6) + F(0.768178f) * AT(5, 6) + F(-0.375330f) * AT(7, 6)); - const Temp_Type X127 = D(F(-0.074658f) * AT(1, 7) + F(0.513280f) * AT(3, 7) + F(0.768178f) * AT(5, 7) + F(-0.375330f) * AT(7, 7)); - const Temp_Type X130 = AT(6, 0); - const Temp_Type X131 = AT(6, 1); - const Temp_Type X132 = AT(6, 2); - const Temp_Type X133 = AT(6, 3); - const Temp_Type X134 = AT(6, 4); - const Temp_Type X135 = AT(6, 5); - const Temp_Type X136 = AT(6, 6); - const Temp_Type X137 = AT(6, 7); - // 80 muls 48 adds - - // 4x4 = 4x8 times 8x4, matrix 1 is constant - R.at(0, 0) = X100; - R.at(0, 1) = D(X101 * F(0.415735f) + X103 * F(0.791065f) + X105 * F(-0.352443f) + X107 * F(0.277785f)); - R.at(0, 2) = X104; - R.at(0, 3) = D(X101 * F(0.022887f) + X103 * F(-0.097545f) + X105 * F(0.490393f) + X107 * F(0.865723f)); - R.at(1, 0) = X110; - R.at(1, 1) = D(X111 * F(0.415735f) + X113 * F(0.791065f) + X115 * F(-0.352443f) + X117 * F(0.277785f)); - R.at(1, 2) = X114; - R.at(1, 3) = D(X111 * F(0.022887f) + X113 * F(-0.097545f) + X115 * F(0.490393f) + X117 * F(0.865723f)); - R.at(2, 0) = X120; - R.at(2, 1) = D(X121 * F(0.415735f) + X123 * F(0.791065f) + X125 * F(-0.352443f) + X127 * F(0.277785f)); - R.at(2, 2) = X124; - R.at(2, 3) = D(X121 * F(0.022887f) + X123 * F(-0.097545f) + X125 * F(0.490393f) + X127 * F(0.865723f)); - R.at(3, 0) = X130; - R.at(3, 1) = D(X131 * F(0.415735f) + X133 * F(0.791065f) + X135 * F(-0.352443f) + X137 * F(0.277785f)); - R.at(3, 2) = X134; - R.at(3, 3) = D(X131 * F(0.022887f) + X133 * F(-0.097545f) + X135 * F(0.490393f) + X137 * F(0.865723f)); - // 40 muls 24 adds - // 4x4 = 4x8 times 8x4, matrix 1 is constant - S.at(0, 0) = D(X101 * F(0.906127f) + X103 * F(-0.318190f) + X105 * F(0.212608f) + X107 * F(-0.180240f)); - S.at(0, 1) = X102; - S.at(0, 2) = D(X101 * F(-0.074658f) + X103 * F(0.513280f) + X105 * F(0.768178f) + X107 * F(-0.375330f)); - S.at(0, 3) = X106; - S.at(1, 0) = D(X111 * F(0.906127f) + X113 * F(-0.318190f) + X115 * F(0.212608f) + X117 * F(-0.180240f)); - S.at(1, 1) = X112; - S.at(1, 2) = D(X111 * F(-0.074658f) + X113 * F(0.513280f) + X115 * F(0.768178f) + X117 * F(-0.375330f)); - S.at(1, 3) = X116; - S.at(2, 0) = D(X121 * F(0.906127f) + X123 * F(-0.318190f) + X125 * F(0.212608f) + X127 * F(-0.180240f)); - S.at(2, 1) = X122; - S.at(2, 2) = D(X121 * F(-0.074658f) + X123 * F(0.513280f) + X125 * F(0.768178f) + X127 * F(-0.375330f)); - S.at(2, 3) = X126; - S.at(3, 0) = D(X131 * F(0.906127f) + X133 * F(-0.318190f) + X135 * F(0.212608f) + X137 * F(-0.180240f)); - S.at(3, 1) = X132; - S.at(3, 2) = D(X131 * F(-0.074658f) + X133 * F(0.513280f) + X135 * F(0.768178f) + X137 * F(-0.375330f)); - S.at(3, 3) = X136; - // 40 muls 24 adds - } - }; -} // end namespace DCT_Upsample - -// Unconditionally frees all allocated m_blocks. -void jpeg_decoder::free_all_blocks() -{ - m_pStream = NULL; - for (mem_block *b = m_pMem_blocks; b; ) - { - mem_block *n = b->m_pNext; - jpgd_free(b); - b = n; - } - m_pMem_blocks = NULL; -} - -// This method handles all errors. It will never return. -// It could easily be changed to use C++ exceptions. -JPGD_NORETURN void jpeg_decoder::stop_decoding(jpgd_status status) -{ - m_error_code = status; - free_all_blocks(); - longjmp(m_jmp_state, status); -} - -void *jpeg_decoder::alloc(size_t nSize, bool zero) -{ - nSize = (JPGD_MAX(nSize, 1) + 3) & ~3; - char *rv = NULL; - for (mem_block *b = m_pMem_blocks; b; b = b->m_pNext) - { - if ((b->m_used_count + nSize) <= b->m_size) - { - rv = b->m_data + b->m_used_count; - b->m_used_count += nSize; - break; - } - } - if (!rv) - { - int capacity = JPGD_MAX(32768 - 256, (nSize + 2047) & ~2047); - mem_block *b = (mem_block*)jpgd_malloc(sizeof(mem_block) + capacity); - if (!b) { stop_decoding(JPGD_NOTENOUGHMEM); } - b->m_pNext = m_pMem_blocks; m_pMem_blocks = b; - b->m_used_count = nSize; - b->m_size = capacity; - rv = b->m_data; - } - if (zero) memset(rv, 0, nSize); - return rv; -} - -void jpeg_decoder::word_clear(void *p, uint16 c, uint n) -{ - uint8 *pD = (uint8*)p; - const uint8 l = c & 0xFF, h = (c >> 8) & 0xFF; - while (n) - { - pD[0] = l; pD[1] = h; pD += 2; - n--; - } -} - -// Refill the input buffer. -// This method will sit in a loop until (A) the buffer is full or (B) -// the stream's read() method reports and end of file condition. -void jpeg_decoder::prep_in_buffer() -{ - m_in_buf_left = 0; - m_pIn_buf_ofs = m_in_buf; - - if (m_eof_flag) - return; - - do - { - int bytes_read = m_pStream->read(m_in_buf + m_in_buf_left, JPGD_IN_BUF_SIZE - m_in_buf_left, &m_eof_flag); - if (bytes_read == -1) - stop_decoding(JPGD_STREAM_READ); - - m_in_buf_left += bytes_read; - } while ((m_in_buf_left < JPGD_IN_BUF_SIZE) && (!m_eof_flag)); - - m_total_bytes_read += m_in_buf_left; - - // Pad the end of the block with M_EOI (prevents the decompressor from going off the rails if the stream is invalid). - // (This dates way back to when this decompressor was written in C/asm, and the all-asm Huffman decoder did some fancy things to increase perf.) - word_clear(m_pIn_buf_ofs + m_in_buf_left, 0xD9FF, 64); -} - -// Read a Huffman code table. -void jpeg_decoder::read_dht_marker() -{ - int i, index, count; - uint8 huff_num[17]; - uint8 huff_val[256]; - - uint num_left = get_bits(16); - - if (num_left < 2) - stop_decoding(JPGD_BAD_DHT_MARKER); - - num_left -= 2; - - while (num_left) - { - index = get_bits(8); - - huff_num[0] = 0; - - count = 0; - - for (i = 1; i <= 16; i++) - { - huff_num[i] = static_cast(get_bits(8)); - count += huff_num[i]; - } - - if (count > 255) - stop_decoding(JPGD_BAD_DHT_COUNTS); - - for (i = 0; i < count; i++) - huff_val[i] = static_cast(get_bits(8)); - - i = 1 + 16 + count; - - if (num_left < (uint)i) - stop_decoding(JPGD_BAD_DHT_MARKER); - - num_left -= i; - - if ((index & 0x10) > 0x10) - stop_decoding(JPGD_BAD_DHT_INDEX); - - index = (index & 0x0F) + ((index & 0x10) >> 4) * (JPGD_MAX_HUFF_TABLES >> 1); - - if (index >= JPGD_MAX_HUFF_TABLES) - stop_decoding(JPGD_BAD_DHT_INDEX); - - if (!m_huff_num[index]) - m_huff_num[index] = (uint8 *)alloc(17); - - if (!m_huff_val[index]) - m_huff_val[index] = (uint8 *)alloc(256); - - m_huff_ac[index] = (index & 0x10) != 0; - memcpy(m_huff_num[index], huff_num, 17); - memcpy(m_huff_val[index], huff_val, 256); - } -} - -// Read a quantization table. -void jpeg_decoder::read_dqt_marker() -{ - int n, i, prec; - uint num_left; - uint temp; - - num_left = get_bits(16); - - if (num_left < 2) - stop_decoding(JPGD_BAD_DQT_MARKER); - - num_left -= 2; - - while (num_left) - { - n = get_bits(8); - prec = n >> 4; - n &= 0x0F; - - if (n >= JPGD_MAX_QUANT_TABLES) - stop_decoding(JPGD_BAD_DQT_TABLE); - - if (!m_quant[n]) - m_quant[n] = (jpgd_quant_t *)alloc(64 * sizeof(jpgd_quant_t)); - - // read quantization entries, in zag order - for (i = 0; i < 64; i++) - { - temp = get_bits(8); - - if (prec) - temp = (temp << 8) + get_bits(8); - - m_quant[n][i] = static_cast(temp); - } - - i = 64 + 1; - - if (prec) - i += 64; - - if (num_left < (uint)i) - stop_decoding(JPGD_BAD_DQT_LENGTH); - - num_left -= i; - } -} - -// Read the start of frame (SOF) marker. -void jpeg_decoder::read_sof_marker() -{ - int i; - uint num_left; - - num_left = get_bits(16); - - if (get_bits(8) != 8) /* precision: sorry, only 8-bit precision is supported right now */ - stop_decoding(JPGD_BAD_PRECISION); - - m_image_y_size = get_bits(16); - - if ((m_image_y_size < 1) || (m_image_y_size > JPGD_MAX_HEIGHT)) - stop_decoding(JPGD_BAD_HEIGHT); - - m_image_x_size = get_bits(16); - - if ((m_image_x_size < 1) || (m_image_x_size > JPGD_MAX_WIDTH)) - stop_decoding(JPGD_BAD_WIDTH); - - m_comps_in_frame = get_bits(8); - - if (m_comps_in_frame > JPGD_MAX_COMPONENTS) - stop_decoding(JPGD_TOO_MANY_COMPONENTS); - - if (num_left != (uint)(m_comps_in_frame * 3 + 8)) - stop_decoding(JPGD_BAD_SOF_LENGTH); - - for (i = 0; i < m_comps_in_frame; i++) - { - m_comp_ident[i] = get_bits(8); - m_comp_h_samp[i] = get_bits(4); - m_comp_v_samp[i] = get_bits(4); - m_comp_quant[i] = get_bits(8); - } -} - -// Used to skip unrecognized markers. -void jpeg_decoder::skip_variable_marker() -{ - uint num_left; - - num_left = get_bits(16); - - if (num_left < 2) - stop_decoding(JPGD_BAD_VARIABLE_MARKER); - - num_left -= 2; - - while (num_left) - { - get_bits(8); - num_left--; - } -} - -// Read a define restart interval (DRI) marker. -void jpeg_decoder::read_dri_marker() -{ - if (get_bits(16) != 4) - stop_decoding(JPGD_BAD_DRI_LENGTH); - - m_restart_interval = get_bits(16); -} - -// Read a start of scan (SOS) marker. -void jpeg_decoder::read_sos_marker() -{ - uint num_left; - int i, ci, n, c, cc; - - num_left = get_bits(16); - - n = get_bits(8); - - m_comps_in_scan = n; - - num_left -= 3; - - if ( (num_left != (uint)(n * 2 + 3)) || (n < 1) || (n > JPGD_MAX_COMPS_IN_SCAN) ) - stop_decoding(JPGD_BAD_SOS_LENGTH); - - for (i = 0; i < n; i++) - { - cc = get_bits(8); - c = get_bits(8); - num_left -= 2; - - for (ci = 0; ci < m_comps_in_frame; ci++) - if (cc == m_comp_ident[ci]) - break; - - if (ci >= m_comps_in_frame) - stop_decoding(JPGD_BAD_SOS_COMP_ID); - - m_comp_list[i] = ci; - m_comp_dc_tab[ci] = (c >> 4) & 15; - m_comp_ac_tab[ci] = (c & 15) + (JPGD_MAX_HUFF_TABLES >> 1); - } - - m_spectral_start = get_bits(8); - m_spectral_end = get_bits(8); - m_successive_high = get_bits(4); - m_successive_low = get_bits(4); - - if (!m_progressive_flag) - { - m_spectral_start = 0; - m_spectral_end = 63; - } - - num_left -= 3; - - while (num_left) /* read past whatever is num_left */ - { - get_bits(8); - num_left--; - } -} - -// Finds the next marker. -int jpeg_decoder::next_marker() -{ - uint c, bytes; - - bytes = 0; - - do - { - do - { - bytes++; - c = get_bits(8); - } while (c != 0xFF); - - do - { - c = get_bits(8); - } while (c == 0xFF); - - } while (c == 0); - - // If bytes > 0 here, there where extra bytes before the marker (not good). - - return c; -} - -// Process markers. Returns when an SOFx, SOI, EOI, or SOS marker is -// encountered. -int jpeg_decoder::process_markers() -{ - int c; - - for ( ; ; ) - { - c = next_marker(); - - switch (c) - { - case M_SOF0: - case M_SOF1: - case M_SOF2: - case M_SOF3: - case M_SOF5: - case M_SOF6: - case M_SOF7: -// case M_JPG: - case M_SOF9: - case M_SOF10: - case M_SOF11: - case M_SOF13: - case M_SOF14: - case M_SOF15: - case M_SOI: - case M_EOI: - case M_SOS: - { - return c; - } - case M_DHT: - { - read_dht_marker(); - break; - } - // No arithmitic support - dumb patents! - case M_DAC: - { - stop_decoding(JPGD_NO_ARITHMITIC_SUPPORT); - break; - } - case M_DQT: - { - read_dqt_marker(); - break; - } - case M_DRI: - { - read_dri_marker(); - break; - } - //case M_APP0: /* no need to read the JFIF marker */ - - case M_JPG: - case M_RST0: /* no parameters */ - case M_RST1: - case M_RST2: - case M_RST3: - case M_RST4: - case M_RST5: - case M_RST6: - case M_RST7: - case M_TEM: - { - stop_decoding(JPGD_UNEXPECTED_MARKER); - break; - } - default: /* must be DNL, DHP, EXP, APPn, JPGn, COM, or RESn or APP0 */ - { - skip_variable_marker(); - break; - } - } - } -} - -// Finds the start of image (SOI) marker. -// This code is rather defensive: it only checks the first 512 bytes to avoid -// false positives. -void jpeg_decoder::locate_soi_marker() -{ - uint lastchar, thischar; - uint bytesleft; - - lastchar = get_bits(8); - - thischar = get_bits(8); - - /* ok if it's a normal JPEG file without a special header */ - - if ((lastchar == 0xFF) && (thischar == M_SOI)) - return; - - bytesleft = 4096; //512; - - for ( ; ; ) - { - if (--bytesleft == 0) - stop_decoding(JPGD_NOT_JPEG); - - lastchar = thischar; - - thischar = get_bits(8); - - if (lastchar == 0xFF) - { - if (thischar == M_SOI) - break; - else if (thischar == M_EOI) // get_bits will keep returning M_EOI if we read past the end - stop_decoding(JPGD_NOT_JPEG); - } - } - - // Check the next character after marker: if it's not 0xFF, it can't be the start of the next marker, so the file is bad. - thischar = (m_bit_buf >> 24) & 0xFF; - - if (thischar != 0xFF) - stop_decoding(JPGD_NOT_JPEG); -} - -// Find a start of frame (SOF) marker. -void jpeg_decoder::locate_sof_marker() -{ - locate_soi_marker(); - - int c = process_markers(); - - switch (c) - { - case M_SOF2: - m_progressive_flag = JPGD_TRUE; - case M_SOF0: /* baseline DCT */ - case M_SOF1: /* extended sequential DCT */ - { - read_sof_marker(); - break; - } - case M_SOF9: /* Arithmitic coding */ - { - stop_decoding(JPGD_NO_ARITHMITIC_SUPPORT); - break; - } - default: - { - stop_decoding(JPGD_UNSUPPORTED_MARKER); - break; - } - } -} - -// Find a start of scan (SOS) marker. -int jpeg_decoder::locate_sos_marker() -{ - int c; - - c = process_markers(); - - if (c == M_EOI) - return JPGD_FALSE; - else if (c != M_SOS) - stop_decoding(JPGD_UNEXPECTED_MARKER); - - read_sos_marker(); - - return JPGD_TRUE; -} - -// Reset everything to default/uninitialized state. -void jpeg_decoder::init(jpeg_decoder_stream *pStream) -{ - m_pMem_blocks = NULL; - m_error_code = JPGD_SUCCESS; - m_ready_flag = false; - m_image_x_size = m_image_y_size = 0; - m_pStream = pStream; - m_progressive_flag = JPGD_FALSE; - - memset(m_huff_ac, 0, sizeof(m_huff_ac)); - memset(m_huff_num, 0, sizeof(m_huff_num)); - memset(m_huff_val, 0, sizeof(m_huff_val)); - memset(m_quant, 0, sizeof(m_quant)); - - m_scan_type = 0; - m_comps_in_frame = 0; - - memset(m_comp_h_samp, 0, sizeof(m_comp_h_samp)); - memset(m_comp_v_samp, 0, sizeof(m_comp_v_samp)); - memset(m_comp_quant, 0, sizeof(m_comp_quant)); - memset(m_comp_ident, 0, sizeof(m_comp_ident)); - memset(m_comp_h_blocks, 0, sizeof(m_comp_h_blocks)); - memset(m_comp_v_blocks, 0, sizeof(m_comp_v_blocks)); - - m_comps_in_scan = 0; - memset(m_comp_list, 0, sizeof(m_comp_list)); - memset(m_comp_dc_tab, 0, sizeof(m_comp_dc_tab)); - memset(m_comp_ac_tab, 0, sizeof(m_comp_ac_tab)); - - m_spectral_start = 0; - m_spectral_end = 0; - m_successive_low = 0; - m_successive_high = 0; - m_max_mcu_x_size = 0; - m_max_mcu_y_size = 0; - m_blocks_per_mcu = 0; - m_max_blocks_per_row = 0; - m_mcus_per_row = 0; - m_mcus_per_col = 0; - m_expanded_blocks_per_component = 0; - m_expanded_blocks_per_mcu = 0; - m_expanded_blocks_per_row = 0; - m_freq_domain_chroma_upsample = false; - - memset(m_mcu_org, 0, sizeof(m_mcu_org)); - - m_total_lines_left = 0; - m_mcu_lines_left = 0; - m_real_dest_bytes_per_scan_line = 0; - m_dest_bytes_per_scan_line = 0; - m_dest_bytes_per_pixel = 0; - - memset(m_pHuff_tabs, 0, sizeof(m_pHuff_tabs)); - - memset(m_dc_coeffs, 0, sizeof(m_dc_coeffs)); - memset(m_ac_coeffs, 0, sizeof(m_ac_coeffs)); - memset(m_block_y_mcu, 0, sizeof(m_block_y_mcu)); - - m_eob_run = 0; - - memset(m_block_y_mcu, 0, sizeof(m_block_y_mcu)); - - m_pIn_buf_ofs = m_in_buf; - m_in_buf_left = 0; - m_eof_flag = false; - m_tem_flag = 0; - - memset(m_in_buf_pad_start, 0, sizeof(m_in_buf_pad_start)); - memset(m_in_buf, 0, sizeof(m_in_buf)); - memset(m_in_buf_pad_end, 0, sizeof(m_in_buf_pad_end)); - - m_restart_interval = 0; - m_restarts_left = 0; - m_next_restart_num = 0; - - m_max_mcus_per_row = 0; - m_max_blocks_per_mcu = 0; - m_max_mcus_per_col = 0; - - memset(m_last_dc_val, 0, sizeof(m_last_dc_val)); - m_pMCU_coefficients = NULL; - m_pSample_buf = NULL; - - m_total_bytes_read = 0; - - m_pScan_line_0 = NULL; - m_pScan_line_1 = NULL; - - // Ready the input buffer. - prep_in_buffer(); - - // Prime the bit buffer. - m_bits_left = 16; - m_bit_buf = 0; - - get_bits(16); - get_bits(16); - - for (int i = 0; i < JPGD_MAX_BLOCKS_PER_MCU; i++) - m_mcu_block_max_zag[i] = 64; -} - -#define SCALEBITS 16 -#define ONE_HALF ((int) 1 << (SCALEBITS-1)) -#define FIX(x) ((int) ((x) * (1L<> SCALEBITS; - m_cbb[i] = ( FIX(1.77200f) * k + ONE_HALF) >> SCALEBITS; - m_crg[i] = (-FIX(0.71414f)) * k; - m_cbg[i] = (-FIX(0.34414f)) * k + ONE_HALF; - } -} - -// This method throws back into the stream any bytes that where read -// into the bit buffer during initial marker scanning. -void jpeg_decoder::fix_in_buffer() -{ - // In case any 0xFF's where pulled into the buffer during marker scanning. - JPGD_ASSERT((m_bits_left & 7) == 0); - - if (m_bits_left == 16) - stuff_char( (uint8)(m_bit_buf & 0xFF)); - - if (m_bits_left >= 8) - stuff_char( (uint8)((m_bit_buf >> 8) & 0xFF)); - - stuff_char((uint8)((m_bit_buf >> 16) & 0xFF)); - stuff_char((uint8)((m_bit_buf >> 24) & 0xFF)); - - m_bits_left = 16; - get_bits_no_markers(16); - get_bits_no_markers(16); -} - -void jpeg_decoder::transform_mcu(int mcu_row) -{ - jpgd_block_t* pSrc_ptr = m_pMCU_coefficients; - uint8* pDst_ptr = m_pSample_buf + mcu_row * m_blocks_per_mcu * 64; - - for (int mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) - { - idct(pSrc_ptr, pDst_ptr, m_mcu_block_max_zag[mcu_block]); - pSrc_ptr += 64; - pDst_ptr += 64; - } -} - -static const uint8 s_max_rc[64] = -{ - 17, 18, 34, 50, 50, 51, 52, 52, 52, 68, 84, 84, 84, 84, 85, 86, 86, 86, 86, 86, - 102, 118, 118, 118, 118, 118, 118, 119, 120, 120, 120, 120, 120, 120, 120, 136, - 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, - 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136, 136 -}; - -void jpeg_decoder::transform_mcu_expand(int mcu_row) -{ - jpgd_block_t* pSrc_ptr = m_pMCU_coefficients; - uint8* pDst_ptr = m_pSample_buf + mcu_row * m_expanded_blocks_per_mcu * 64; - - // Y IDCT - int mcu_block; - for (mcu_block = 0; mcu_block < m_expanded_blocks_per_component; mcu_block++) - { - idct(pSrc_ptr, pDst_ptr, m_mcu_block_max_zag[mcu_block]); - pSrc_ptr += 64; - pDst_ptr += 64; - } - - // Chroma IDCT, with upsampling - jpgd_block_t temp_block[64]; - - for (int i = 0; i < 2; i++) - { - DCT_Upsample::Matrix44 P, Q, R, S; - - JPGD_ASSERT(m_mcu_block_max_zag[mcu_block] >= 1); - JPGD_ASSERT(m_mcu_block_max_zag[mcu_block] <= 64); - - int max_zag = m_mcu_block_max_zag[mcu_block++] - 1; - if (max_zag <= 0) max_zag = 0; // should never happen, only here to shut up static analysis - switch (s_max_rc[max_zag]) - { - case 1*16+1: - DCT_Upsample::P_Q<1, 1>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<1, 1>::calc(R, S, pSrc_ptr); - break; - case 1*16+2: - DCT_Upsample::P_Q<1, 2>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<1, 2>::calc(R, S, pSrc_ptr); - break; - case 2*16+2: - DCT_Upsample::P_Q<2, 2>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<2, 2>::calc(R, S, pSrc_ptr); - break; - case 3*16+2: - DCT_Upsample::P_Q<3, 2>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<3, 2>::calc(R, S, pSrc_ptr); - break; - case 3*16+3: - DCT_Upsample::P_Q<3, 3>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<3, 3>::calc(R, S, pSrc_ptr); - break; - case 3*16+4: - DCT_Upsample::P_Q<3, 4>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<3, 4>::calc(R, S, pSrc_ptr); - break; - case 4*16+4: - DCT_Upsample::P_Q<4, 4>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<4, 4>::calc(R, S, pSrc_ptr); - break; - case 5*16+4: - DCT_Upsample::P_Q<5, 4>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<5, 4>::calc(R, S, pSrc_ptr); - break; - case 5*16+5: - DCT_Upsample::P_Q<5, 5>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<5, 5>::calc(R, S, pSrc_ptr); - break; - case 5*16+6: - DCT_Upsample::P_Q<5, 6>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<5, 6>::calc(R, S, pSrc_ptr); - break; - case 6*16+6: - DCT_Upsample::P_Q<6, 6>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<6, 6>::calc(R, S, pSrc_ptr); - break; - case 7*16+6: - DCT_Upsample::P_Q<7, 6>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<7, 6>::calc(R, S, pSrc_ptr); - break; - case 7*16+7: - DCT_Upsample::P_Q<7, 7>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<7, 7>::calc(R, S, pSrc_ptr); - break; - case 7*16+8: - DCT_Upsample::P_Q<7, 8>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<7, 8>::calc(R, S, pSrc_ptr); - break; - case 8*16+8: - DCT_Upsample::P_Q<8, 8>::calc(P, Q, pSrc_ptr); - DCT_Upsample::R_S<8, 8>::calc(R, S, pSrc_ptr); - break; - default: - JPGD_ASSERT(false); - } - - DCT_Upsample::Matrix44 a(P + Q); P -= Q; - DCT_Upsample::Matrix44& b = P; - DCT_Upsample::Matrix44 c(R + S); R -= S; - DCT_Upsample::Matrix44& d = R; - - DCT_Upsample::Matrix44::add_and_store(temp_block, a, c); - idct_4x4(temp_block, pDst_ptr); - pDst_ptr += 64; - - DCT_Upsample::Matrix44::sub_and_store(temp_block, a, c); - idct_4x4(temp_block, pDst_ptr); - pDst_ptr += 64; - - DCT_Upsample::Matrix44::add_and_store(temp_block, b, d); - idct_4x4(temp_block, pDst_ptr); - pDst_ptr += 64; - - DCT_Upsample::Matrix44::sub_and_store(temp_block, b, d); - idct_4x4(temp_block, pDst_ptr); - pDst_ptr += 64; - - pSrc_ptr += 64; - } -} - -// Loads and dequantizes the next row of (already decoded) coefficients. -// Progressive images only. -void jpeg_decoder::load_next_row() -{ - int i; - jpgd_block_t *p; - jpgd_quant_t *q; - int mcu_row, mcu_block, row_block = 0; - int component_num, component_id; - int block_x_mcu[JPGD_MAX_COMPONENTS]; - - memset(block_x_mcu, 0, JPGD_MAX_COMPONENTS * sizeof(int)); - - for (mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) - { - int block_x_mcu_ofs = 0, block_y_mcu_ofs = 0; - - for (mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) - { - component_id = m_mcu_org[mcu_block]; - q = m_quant[m_comp_quant[component_id]]; - - p = m_pMCU_coefficients + 64 * mcu_block; - - jpgd_block_t* pAC = coeff_buf_getp(m_ac_coeffs[component_id], block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs); - jpgd_block_t* pDC = coeff_buf_getp(m_dc_coeffs[component_id], block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs); - p[0] = pDC[0]; - memcpy(&p[1], &pAC[1], 63 * sizeof(jpgd_block_t)); - - for (i = 63; i > 0; i--) - if (p[g_ZAG[i]]) - break; - - m_mcu_block_max_zag[mcu_block] = i + 1; - - for ( ; i >= 0; i--) - if (p[g_ZAG[i]]) - p[g_ZAG[i]] = static_cast(p[g_ZAG[i]] * q[i]); - - row_block++; - - if (m_comps_in_scan == 1) - block_x_mcu[component_id]++; - else - { - if (++block_x_mcu_ofs == m_comp_h_samp[component_id]) - { - block_x_mcu_ofs = 0; - - if (++block_y_mcu_ofs == m_comp_v_samp[component_id]) - { - block_y_mcu_ofs = 0; - - block_x_mcu[component_id] += m_comp_h_samp[component_id]; - } - } - } - } - - if (m_freq_domain_chroma_upsample) - transform_mcu_expand(mcu_row); - else - transform_mcu(mcu_row); - } - - if (m_comps_in_scan == 1) - m_block_y_mcu[m_comp_list[0]]++; - else - { - for (component_num = 0; component_num < m_comps_in_scan; component_num++) - { - component_id = m_comp_list[component_num]; - - m_block_y_mcu[component_id] += m_comp_v_samp[component_id]; - } - } -} - -// Restart interval processing. -void jpeg_decoder::process_restart() -{ - int i; - int c = 0; - - // Align to a byte boundry - // FIXME: Is this really necessary? get_bits_no_markers() never reads in markers! - //get_bits_no_markers(m_bits_left & 7); - - // Let's scan a little bit to find the marker, but not _too_ far. - // 1536 is a "fudge factor" that determines how much to scan. - for (i = 1536; i > 0; i--) - if (get_char() == 0xFF) - break; - - if (i == 0) - stop_decoding(JPGD_BAD_RESTART_MARKER); - - for ( ; i > 0; i--) - if ((c = get_char()) != 0xFF) - break; - - if (i == 0) - stop_decoding(JPGD_BAD_RESTART_MARKER); - - // Is it the expected marker? If not, something bad happened. - if (c != (m_next_restart_num + M_RST0)) - stop_decoding(JPGD_BAD_RESTART_MARKER); - - // Reset each component's DC prediction values. - memset(&m_last_dc_val, 0, m_comps_in_frame * sizeof(uint)); - - m_eob_run = 0; - - m_restarts_left = m_restart_interval; - - m_next_restart_num = (m_next_restart_num + 1) & 7; - - // Get the bit buffer going again... - - m_bits_left = 16; - get_bits_no_markers(16); - get_bits_no_markers(16); -} - -static inline int dequantize_ac(int c, int q) { c *= q; return c; } - -// Decodes and dequantizes the next row of coefficients. -void jpeg_decoder::decode_next_row() -{ - int row_block = 0; - - for (int mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) - { - if ((m_restart_interval) && (m_restarts_left == 0)) - process_restart(); - - jpgd_block_t* p = m_pMCU_coefficients; - for (int mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++, p += 64) - { - int component_id = m_mcu_org[mcu_block]; - jpgd_quant_t* q = m_quant[m_comp_quant[component_id]]; - - int r, s; - s = huff_decode(m_pHuff_tabs[m_comp_dc_tab[component_id]], r); - s = JPGD_HUFF_EXTEND(r, s); - - m_last_dc_val[component_id] = (s += m_last_dc_val[component_id]); - - p[0] = static_cast(s * q[0]); - - int prev_num_set = m_mcu_block_max_zag[mcu_block]; - - huff_tables *pH = m_pHuff_tabs[m_comp_ac_tab[component_id]]; - - int k; - for (k = 1; k < 64; k++) - { - int extra_bits; - s = huff_decode(pH, extra_bits); - - r = s >> 4; - s &= 15; - - if (s) - { - if (r) - { - if ((k + r) > 63) - stop_decoding(JPGD_DECODE_ERROR); - - if (k < prev_num_set) - { - int n = JPGD_MIN(r, prev_num_set - k); - int kt = k; - while (n--) - p[g_ZAG[kt++]] = 0; - } - - k += r; - } - - s = JPGD_HUFF_EXTEND(extra_bits, s); - - JPGD_ASSERT(k < 64); - - p[g_ZAG[k]] = static_cast(dequantize_ac(s, q[k])); //s * q[k]; - } - else - { - if (r == 15) - { - if ((k + 16) > 64) - stop_decoding(JPGD_DECODE_ERROR); - - if (k < prev_num_set) - { - int n = JPGD_MIN(16, prev_num_set - k); - int kt = k; - while (n--) - { - JPGD_ASSERT(kt <= 63); - p[g_ZAG[kt++]] = 0; - } - } - - k += 16 - 1; // - 1 because the loop counter is k - JPGD_ASSERT(p[g_ZAG[k]] == 0); - } - else - break; - } - } - - if (k < prev_num_set) - { - int kt = k; - while (kt < prev_num_set) - p[g_ZAG[kt++]] = 0; - } - - m_mcu_block_max_zag[mcu_block] = k; - - row_block++; - } - - if (m_freq_domain_chroma_upsample) - transform_mcu_expand(mcu_row); - else - transform_mcu(mcu_row); - - m_restarts_left--; - } -} - -// YCbCr H1V1 (1x1:1:1, 3 m_blocks per MCU) to RGB -void jpeg_decoder::H1V1Convert() -{ - int row = m_max_mcu_y_size - m_mcu_lines_left; - uint8 *d = m_pScan_line_0; - uint8 *s = m_pSample_buf + row * 8; - - for (int i = m_max_mcus_per_row; i > 0; i--) - { - for (int j = 0; j < 8; j++) - { - int y = s[j]; - int cb = s[64+j]; - int cr = s[128+j]; - - d[0] = clamp(y + m_crr[cr]); - d[1] = clamp(y + ((m_crg[cr] + m_cbg[cb]) >> 16)); - d[2] = clamp(y + m_cbb[cb]); - d[3] = 255; - - d += 4; - } - - s += 64*3; - } -} - -// YCbCr H2V1 (2x1:1:1, 4 m_blocks per MCU) to RGB -void jpeg_decoder::H2V1Convert() -{ - int row = m_max_mcu_y_size - m_mcu_lines_left; - uint8 *d0 = m_pScan_line_0; - uint8 *y = m_pSample_buf + row * 8; - uint8 *c = m_pSample_buf + 2*64 + row * 8; - - for (int i = m_max_mcus_per_row; i > 0; i--) - { - for (int l = 0; l < 2; l++) - { - for (int j = 0; j < 4; j++) - { - int cb = c[0]; - int cr = c[64]; - - int rc = m_crr[cr]; - int gc = ((m_crg[cr] + m_cbg[cb]) >> 16); - int bc = m_cbb[cb]; - - int yy = y[j<<1]; - d0[0] = clamp(yy+rc); - d0[1] = clamp(yy+gc); - d0[2] = clamp(yy+bc); - d0[3] = 255; - - yy = y[(j<<1)+1]; - d0[4] = clamp(yy+rc); - d0[5] = clamp(yy+gc); - d0[6] = clamp(yy+bc); - d0[7] = 255; - - d0 += 8; - - c++; - } - y += 64; - } - - y += 64*4 - 64*2; - c += 64*4 - 8; - } -} - -// YCbCr H2V1 (1x2:1:1, 4 m_blocks per MCU) to RGB -void jpeg_decoder::H1V2Convert() -{ - int row = m_max_mcu_y_size - m_mcu_lines_left; - uint8 *d0 = m_pScan_line_0; - uint8 *d1 = m_pScan_line_1; - uint8 *y; - uint8 *c; - - if (row < 8) - y = m_pSample_buf + row * 8; - else - y = m_pSample_buf + 64*1 + (row & 7) * 8; - - c = m_pSample_buf + 64*2 + (row >> 1) * 8; - - for (int i = m_max_mcus_per_row; i > 0; i--) - { - for (int j = 0; j < 8; j++) - { - int cb = c[0+j]; - int cr = c[64+j]; - - int rc = m_crr[cr]; - int gc = ((m_crg[cr] + m_cbg[cb]) >> 16); - int bc = m_cbb[cb]; - - int yy = y[j]; - d0[0] = clamp(yy+rc); - d0[1] = clamp(yy+gc); - d0[2] = clamp(yy+bc); - d0[3] = 255; - - yy = y[8+j]; - d1[0] = clamp(yy+rc); - d1[1] = clamp(yy+gc); - d1[2] = clamp(yy+bc); - d1[3] = 255; - - d0 += 4; - d1 += 4; - } - - y += 64*4; - c += 64*4; - } -} - -// YCbCr H2V2 (2x2:1:1, 6 m_blocks per MCU) to RGB -void jpeg_decoder::H2V2Convert() -{ - int row = m_max_mcu_y_size - m_mcu_lines_left; - uint8 *d0 = m_pScan_line_0; - uint8 *d1 = m_pScan_line_1; - uint8 *y; - uint8 *c; - - if (row < 8) - y = m_pSample_buf + row * 8; - else - y = m_pSample_buf + 64*2 + (row & 7) * 8; - - c = m_pSample_buf + 64*4 + (row >> 1) * 8; - - for (int i = m_max_mcus_per_row; i > 0; i--) - { - for (int l = 0; l < 2; l++) - { - for (int j = 0; j < 8; j += 2) - { - int cb = c[0]; - int cr = c[64]; - - int rc = m_crr[cr]; - int gc = ((m_crg[cr] + m_cbg[cb]) >> 16); - int bc = m_cbb[cb]; - - int yy = y[j]; - d0[0] = clamp(yy+rc); - d0[1] = clamp(yy+gc); - d0[2] = clamp(yy+bc); - d0[3] = 255; - - yy = y[j+1]; - d0[4] = clamp(yy+rc); - d0[5] = clamp(yy+gc); - d0[6] = clamp(yy+bc); - d0[7] = 255; - - yy = y[j+8]; - d1[0] = clamp(yy+rc); - d1[1] = clamp(yy+gc); - d1[2] = clamp(yy+bc); - d1[3] = 255; - - yy = y[j+8+1]; - d1[4] = clamp(yy+rc); - d1[5] = clamp(yy+gc); - d1[6] = clamp(yy+bc); - d1[7] = 255; - - d0 += 8; - d1 += 8; - - c++; - } - y += 64; - } - - y += 64*6 - 64*2; - c += 64*6 - 8; - } -} - -// Y (1 block per MCU) to 8-bit grayscale -void jpeg_decoder::gray_convert() -{ - int row = m_max_mcu_y_size - m_mcu_lines_left; - uint8 *d = m_pScan_line_0; - uint8 *s = m_pSample_buf + row * 8; - - for (int i = m_max_mcus_per_row; i > 0; i--) - { - *(uint *)d = *(uint *)s; - *(uint *)(&d[4]) = *(uint *)(&s[4]); - - s += 64; - d += 8; - } -} - -void jpeg_decoder::expanded_convert() -{ - int row = m_max_mcu_y_size - m_mcu_lines_left; - - uint8* Py = m_pSample_buf + (row / 8) * 64 * m_comp_h_samp[0] + (row & 7) * 8; - - uint8* d = m_pScan_line_0; - - for (int i = m_max_mcus_per_row; i > 0; i--) - { - for (int k = 0; k < m_max_mcu_x_size; k += 8) - { - const int Y_ofs = k * 8; - const int Cb_ofs = Y_ofs + 64 * m_expanded_blocks_per_component; - const int Cr_ofs = Y_ofs + 64 * m_expanded_blocks_per_component * 2; - for (int j = 0; j < 8; j++) - { - int y = Py[Y_ofs + j]; - int cb = Py[Cb_ofs + j]; - int cr = Py[Cr_ofs + j]; - - d[0] = clamp(y + m_crr[cr]); - d[1] = clamp(y + ((m_crg[cr] + m_cbg[cb]) >> 16)); - d[2] = clamp(y + m_cbb[cb]); - d[3] = 255; - - d += 4; - } - } - - Py += 64 * m_expanded_blocks_per_mcu; - } -} - -// Find end of image (EOI) marker, so we can return to the user the exact size of the input stream. -void jpeg_decoder::find_eoi() -{ - if (!m_progressive_flag) - { - // Attempt to read the EOI marker. - //get_bits_no_markers(m_bits_left & 7); - - // Prime the bit buffer - m_bits_left = 16; - get_bits(16); - get_bits(16); - - // The next marker _should_ be EOI - process_markers(); - } - - m_total_bytes_read -= m_in_buf_left; -} - -int jpeg_decoder::decode(const void** pScan_line, uint* pScan_line_len) -{ - if ((m_error_code) || (!m_ready_flag)) - return JPGD_FAILED; - - if (m_total_lines_left == 0) - return JPGD_DONE; - - if (m_mcu_lines_left == 0) - { - if (setjmp(m_jmp_state)) - return JPGD_FAILED; - - if (m_progressive_flag) - load_next_row(); - else - decode_next_row(); - - // Find the EOI marker if that was the last row. - if (m_total_lines_left <= m_max_mcu_y_size) - find_eoi(); - - m_mcu_lines_left = m_max_mcu_y_size; - } - - if (m_freq_domain_chroma_upsample) - { - expanded_convert(); - *pScan_line = m_pScan_line_0; - } - else - { - switch (m_scan_type) - { - case JPGD_YH2V2: - { - if ((m_mcu_lines_left & 1) == 0) - { - H2V2Convert(); - *pScan_line = m_pScan_line_0; - } - else - *pScan_line = m_pScan_line_1; - - break; - } - case JPGD_YH2V1: - { - H2V1Convert(); - *pScan_line = m_pScan_line_0; - break; - } - case JPGD_YH1V2: - { - if ((m_mcu_lines_left & 1) == 0) - { - H1V2Convert(); - *pScan_line = m_pScan_line_0; - } - else - *pScan_line = m_pScan_line_1; - - break; - } - case JPGD_YH1V1: - { - H1V1Convert(); - *pScan_line = m_pScan_line_0; - break; - } - case JPGD_GRAYSCALE: - { - gray_convert(); - *pScan_line = m_pScan_line_0; - - break; - } - } - } - - *pScan_line_len = m_real_dest_bytes_per_scan_line; - - m_mcu_lines_left--; - m_total_lines_left--; - - return JPGD_SUCCESS; -} - -// Creates the tables needed for efficient Huffman decoding. -void jpeg_decoder::make_huff_table(int index, huff_tables *pH) -{ - int p, i, l, si; - uint8 huffsize[257]; - uint huffcode[257]; - uint code; - uint subtree; - int code_size; - int lastp; - int nextfreeentry; - int currententry; - - pH->ac_table = m_huff_ac[index] != 0; - - p = 0; - - for (l = 1; l <= 16; l++) - { - for (i = 1; i <= m_huff_num[index][l]; i++) - huffsize[p++] = static_cast(l); - } - - huffsize[p] = 0; - - lastp = p; - - code = 0; - si = huffsize[0]; - p = 0; - - while (huffsize[p]) - { - while (huffsize[p] == si) - { - huffcode[p++] = code; - code++; - } - - code <<= 1; - si++; - } - - memset(pH->look_up, 0, sizeof(pH->look_up)); - memset(pH->look_up2, 0, sizeof(pH->look_up2)); - memset(pH->tree, 0, sizeof(pH->tree)); - memset(pH->code_size, 0, sizeof(pH->code_size)); - - nextfreeentry = -1; - - p = 0; - - while (p < lastp) - { - i = m_huff_val[index][p]; - code = huffcode[p]; - code_size = huffsize[p]; - - pH->code_size[i] = static_cast(code_size); - - if (code_size <= 8) - { - code <<= (8 - code_size); - - for (l = 1 << (8 - code_size); l > 0; l--) - { - JPGD_ASSERT(i < 256); - - pH->look_up[code] = i; - - bool has_extrabits = false; - int extra_bits = 0; - int num_extra_bits = i & 15; - - int bits_to_fetch = code_size; - if (num_extra_bits) - { - int total_codesize = code_size + num_extra_bits; - if (total_codesize <= 8) - { - has_extrabits = true; - extra_bits = ((1 << num_extra_bits) - 1) & (code >> (8 - total_codesize)); - JPGD_ASSERT(extra_bits <= 0x7FFF); - bits_to_fetch += num_extra_bits; - } - } - - if (!has_extrabits) - pH->look_up2[code] = i | (bits_to_fetch << 8); - else - pH->look_up2[code] = i | 0x8000 | (extra_bits << 16) | (bits_to_fetch << 8); - - code++; - } - } - else - { - subtree = (code >> (code_size - 8)) & 0xFF; - - currententry = pH->look_up[subtree]; - - if (currententry == 0) - { - pH->look_up[subtree] = currententry = nextfreeentry; - pH->look_up2[subtree] = currententry = nextfreeentry; - - nextfreeentry -= 2; - } - - code <<= (16 - (code_size - 8)); - - for (l = code_size; l > 9; l--) - { - if ((code & 0x8000) == 0) - currententry--; - - if (pH->tree[-currententry - 1] == 0) - { - pH->tree[-currententry - 1] = nextfreeentry; - - currententry = nextfreeentry; - - nextfreeentry -= 2; - } - else - currententry = pH->tree[-currententry - 1]; - - code <<= 1; - } - - if ((code & 0x8000) == 0) - currententry--; - - pH->tree[-currententry - 1] = i; - } - - p++; - } -} - -// Verifies the quantization tables needed for this scan are available. -void jpeg_decoder::check_quant_tables() -{ - for (int i = 0; i < m_comps_in_scan; i++) - if (m_quant[m_comp_quant[m_comp_list[i]]] == NULL) - stop_decoding(JPGD_UNDEFINED_QUANT_TABLE); -} - -// Verifies that all the Huffman tables needed for this scan are available. -void jpeg_decoder::check_huff_tables() -{ - for (int i = 0; i < m_comps_in_scan; i++) - { - if ((m_spectral_start == 0) && (m_huff_num[m_comp_dc_tab[m_comp_list[i]]] == NULL)) - stop_decoding(JPGD_UNDEFINED_HUFF_TABLE); - - if ((m_spectral_end > 0) && (m_huff_num[m_comp_ac_tab[m_comp_list[i]]] == NULL)) - stop_decoding(JPGD_UNDEFINED_HUFF_TABLE); - } - - for (int i = 0; i < JPGD_MAX_HUFF_TABLES; i++) - if (m_huff_num[i]) - { - if (!m_pHuff_tabs[i]) - m_pHuff_tabs[i] = (huff_tables *)alloc(sizeof(huff_tables)); - - make_huff_table(i, m_pHuff_tabs[i]); - } -} - -// Determines the component order inside each MCU. -// Also calcs how many MCU's are on each row, etc. -void jpeg_decoder::calc_mcu_block_order() -{ - int component_num, component_id; - int max_h_samp = 0, max_v_samp = 0; - - for (component_id = 0; component_id < m_comps_in_frame; component_id++) - { - if (m_comp_h_samp[component_id] > max_h_samp) - max_h_samp = m_comp_h_samp[component_id]; - - if (m_comp_v_samp[component_id] > max_v_samp) - max_v_samp = m_comp_v_samp[component_id]; - } - - for (component_id = 0; component_id < m_comps_in_frame; component_id++) - { - m_comp_h_blocks[component_id] = ((((m_image_x_size * m_comp_h_samp[component_id]) + (max_h_samp - 1)) / max_h_samp) + 7) / 8; - m_comp_v_blocks[component_id] = ((((m_image_y_size * m_comp_v_samp[component_id]) + (max_v_samp - 1)) / max_v_samp) + 7) / 8; - } - - if (m_comps_in_scan == 1) - { - m_mcus_per_row = m_comp_h_blocks[m_comp_list[0]]; - m_mcus_per_col = m_comp_v_blocks[m_comp_list[0]]; - } - else - { - m_mcus_per_row = (((m_image_x_size + 7) / 8) + (max_h_samp - 1)) / max_h_samp; - m_mcus_per_col = (((m_image_y_size + 7) / 8) + (max_v_samp - 1)) / max_v_samp; - } - - if (m_comps_in_scan == 1) - { - m_mcu_org[0] = m_comp_list[0]; - - m_blocks_per_mcu = 1; - } - else - { - m_blocks_per_mcu = 0; - - for (component_num = 0; component_num < m_comps_in_scan; component_num++) - { - int num_blocks; - - component_id = m_comp_list[component_num]; - - num_blocks = m_comp_h_samp[component_id] * m_comp_v_samp[component_id]; - - while (num_blocks--) - m_mcu_org[m_blocks_per_mcu++] = component_id; - } - } -} - -// Starts a new scan. -int jpeg_decoder::init_scan() -{ - if (!locate_sos_marker()) - return JPGD_FALSE; - - calc_mcu_block_order(); - - check_huff_tables(); - - check_quant_tables(); - - memset(m_last_dc_val, 0, m_comps_in_frame * sizeof(uint)); - - m_eob_run = 0; - - if (m_restart_interval) - { - m_restarts_left = m_restart_interval; - m_next_restart_num = 0; - } - - fix_in_buffer(); - - return JPGD_TRUE; -} - -// Starts a frame. Determines if the number of components or sampling factors -// are supported. -void jpeg_decoder::init_frame() -{ - int i; - - if (m_comps_in_frame == 1) - { - if ((m_comp_h_samp[0] != 1) || (m_comp_v_samp[0] != 1)) - stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS); - - m_scan_type = JPGD_GRAYSCALE; - m_max_blocks_per_mcu = 1; - m_max_mcu_x_size = 8; - m_max_mcu_y_size = 8; - } - else if (m_comps_in_frame == 3) - { - if ( ((m_comp_h_samp[1] != 1) || (m_comp_v_samp[1] != 1)) || - ((m_comp_h_samp[2] != 1) || (m_comp_v_samp[2] != 1)) ) - stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS); - - if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 1)) - { - m_scan_type = JPGD_YH1V1; - - m_max_blocks_per_mcu = 3; - m_max_mcu_x_size = 8; - m_max_mcu_y_size = 8; - } - else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 1)) - { - m_scan_type = JPGD_YH2V1; - m_max_blocks_per_mcu = 4; - m_max_mcu_x_size = 16; - m_max_mcu_y_size = 8; - } - else if ((m_comp_h_samp[0] == 1) && (m_comp_v_samp[0] == 2)) - { - m_scan_type = JPGD_YH1V2; - m_max_blocks_per_mcu = 4; - m_max_mcu_x_size = 8; - m_max_mcu_y_size = 16; - } - else if ((m_comp_h_samp[0] == 2) && (m_comp_v_samp[0] == 2)) - { - m_scan_type = JPGD_YH2V2; - m_max_blocks_per_mcu = 6; - m_max_mcu_x_size = 16; - m_max_mcu_y_size = 16; - } - else - stop_decoding(JPGD_UNSUPPORTED_SAMP_FACTORS); - } - else - stop_decoding(JPGD_UNSUPPORTED_COLORSPACE); - - m_max_mcus_per_row = (m_image_x_size + (m_max_mcu_x_size - 1)) / m_max_mcu_x_size; - m_max_mcus_per_col = (m_image_y_size + (m_max_mcu_y_size - 1)) / m_max_mcu_y_size; - - // These values are for the *destination* pixels: after conversion. - if (m_scan_type == JPGD_GRAYSCALE) - m_dest_bytes_per_pixel = 1; - else - m_dest_bytes_per_pixel = 4; - - m_dest_bytes_per_scan_line = ((m_image_x_size + 15) & 0xFFF0) * m_dest_bytes_per_pixel; - - m_real_dest_bytes_per_scan_line = (m_image_x_size * m_dest_bytes_per_pixel); - - // Initialize two scan line buffers. - m_pScan_line_0 = (uint8 *)alloc(m_dest_bytes_per_scan_line, true); - if ((m_scan_type == JPGD_YH1V2) || (m_scan_type == JPGD_YH2V2)) - m_pScan_line_1 = (uint8 *)alloc(m_dest_bytes_per_scan_line, true); - - m_max_blocks_per_row = m_max_mcus_per_row * m_max_blocks_per_mcu; - - // Should never happen - if (m_max_blocks_per_row > JPGD_MAX_BLOCKS_PER_ROW) - stop_decoding(JPGD_ASSERTION_ERROR); - - // Allocate the coefficient buffer, enough for one MCU - m_pMCU_coefficients = (jpgd_block_t*)alloc(m_max_blocks_per_mcu * 64 * sizeof(jpgd_block_t)); - - for (i = 0; i < m_max_blocks_per_mcu; i++) - m_mcu_block_max_zag[i] = 64; - - m_expanded_blocks_per_component = m_comp_h_samp[0] * m_comp_v_samp[0]; - m_expanded_blocks_per_mcu = m_expanded_blocks_per_component * m_comps_in_frame; - m_expanded_blocks_per_row = m_max_mcus_per_row * m_expanded_blocks_per_mcu; - // Freq. domain chroma upsampling is only supported for H2V2 subsampling factor (the most common one I've seen). - m_freq_domain_chroma_upsample = false; -#if JPGD_SUPPORT_FREQ_DOMAIN_UPSAMPLING - m_freq_domain_chroma_upsample = (m_expanded_blocks_per_mcu == 4*3); -#endif - - if (m_freq_domain_chroma_upsample) - m_pSample_buf = (uint8 *)alloc(m_expanded_blocks_per_row * 64); - else - m_pSample_buf = (uint8 *)alloc(m_max_blocks_per_row * 64); - - m_total_lines_left = m_image_y_size; - - m_mcu_lines_left = 0; - - create_look_ups(); -} - -// The coeff_buf series of methods originally stored the coefficients -// into a "virtual" file which was located in EMS, XMS, or a disk file. A cache -// was used to make this process more efficient. Now, we can store the entire -// thing in RAM. -jpeg_decoder::coeff_buf* jpeg_decoder::coeff_buf_open(int block_num_x, int block_num_y, int block_len_x, int block_len_y) -{ - coeff_buf* cb = (coeff_buf*)alloc(sizeof(coeff_buf)); - - cb->block_num_x = block_num_x; - cb->block_num_y = block_num_y; - cb->block_len_x = block_len_x; - cb->block_len_y = block_len_y; - cb->block_size = (block_len_x * block_len_y) * sizeof(jpgd_block_t); - cb->pData = (uint8 *)alloc(cb->block_size * block_num_x * block_num_y, true); - return cb; -} - -inline jpgd_block_t *jpeg_decoder::coeff_buf_getp(coeff_buf *cb, int block_x, int block_y) -{ - JPGD_ASSERT((block_x < cb->block_num_x) && (block_y < cb->block_num_y)); - return (jpgd_block_t *)(cb->pData + block_x * cb->block_size + block_y * (cb->block_size * cb->block_num_x)); -} - -// The following methods decode the various types of m_blocks encountered -// in progressively encoded images. -void jpeg_decoder::decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y) -{ - int s, r; - jpgd_block_t *p = pD->coeff_buf_getp(pD->m_dc_coeffs[component_id], block_x, block_y); - - if ((s = pD->huff_decode(pD->m_pHuff_tabs[pD->m_comp_dc_tab[component_id]])) != 0) - { - r = pD->get_bits_no_markers(s); - s = JPGD_HUFF_EXTEND(r, s); - } - - pD->m_last_dc_val[component_id] = (s += pD->m_last_dc_val[component_id]); - - p[0] = static_cast(s << pD->m_successive_low); -} - -void jpeg_decoder::decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y) -{ - if (pD->get_bits_no_markers(1)) - { - jpgd_block_t *p = pD->coeff_buf_getp(pD->m_dc_coeffs[component_id], block_x, block_y); - - p[0] |= (1 << pD->m_successive_low); - } -} - -void jpeg_decoder::decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y) -{ - int k, s, r; - - if (pD->m_eob_run) - { - pD->m_eob_run--; - return; - } - - jpgd_block_t *p = pD->coeff_buf_getp(pD->m_ac_coeffs[component_id], block_x, block_y); - - for (k = pD->m_spectral_start; k <= pD->m_spectral_end; k++) - { - s = pD->huff_decode(pD->m_pHuff_tabs[pD->m_comp_ac_tab[component_id]]); - - r = s >> 4; - s &= 15; - - if (s) - { - if ((k += r) > 63) - pD->stop_decoding(JPGD_DECODE_ERROR); - - r = pD->get_bits_no_markers(s); - s = JPGD_HUFF_EXTEND(r, s); - - p[g_ZAG[k]] = static_cast(s << pD->m_successive_low); - } - else - { - if (r == 15) - { - if ((k += 15) > 63) - pD->stop_decoding(JPGD_DECODE_ERROR); - } - else - { - pD->m_eob_run = 1 << r; - - if (r) - pD->m_eob_run += pD->get_bits_no_markers(r); - - pD->m_eob_run--; - - break; - } - } - } -} - -void jpeg_decoder::decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y) -{ - int s, k, r; - int p1 = 1 << pD->m_successive_low; - int m1 = (-1) << pD->m_successive_low; - jpgd_block_t *p = pD->coeff_buf_getp(pD->m_ac_coeffs[component_id], block_x, block_y); - - JPGD_ASSERT(pD->m_spectral_end <= 63); - - k = pD->m_spectral_start; - - if (pD->m_eob_run == 0) - { - for ( ; k <= pD->m_spectral_end; k++) - { - s = pD->huff_decode(pD->m_pHuff_tabs[pD->m_comp_ac_tab[component_id]]); - - r = s >> 4; - s &= 15; - - if (s) - { - if (s != 1) - pD->stop_decoding(JPGD_DECODE_ERROR); - - if (pD->get_bits_no_markers(1)) - s = p1; - else - s = m1; - } - else - { - if (r != 15) - { - pD->m_eob_run = 1 << r; - - if (r) - pD->m_eob_run += pD->get_bits_no_markers(r); - - break; - } - } - - do - { - jpgd_block_t *this_coef = p + g_ZAG[k & 63]; - - if (*this_coef != 0) - { - if (pD->get_bits_no_markers(1)) - { - if ((*this_coef & p1) == 0) - { - if (*this_coef >= 0) - *this_coef = static_cast(*this_coef + p1); - else - *this_coef = static_cast(*this_coef + m1); - } - } - } - else - { - if (--r < 0) - break; - } - - k++; - - } while (k <= pD->m_spectral_end); - - if ((s) && (k < 64)) - { - p[g_ZAG[k]] = static_cast(s); - } - } - } - - if (pD->m_eob_run > 0) - { - for ( ; k <= pD->m_spectral_end; k++) - { - jpgd_block_t *this_coef = p + g_ZAG[k & 63]; // logical AND to shut up static code analysis - - if (*this_coef != 0) - { - if (pD->get_bits_no_markers(1)) - { - if ((*this_coef & p1) == 0) - { - if (*this_coef >= 0) - *this_coef = static_cast(*this_coef + p1); - else - *this_coef = static_cast(*this_coef + m1); - } - } - } - } - - pD->m_eob_run--; - } -} - -// Decode a scan in a progressively encoded image. -void jpeg_decoder::decode_scan(pDecode_block_func decode_block_func) -{ - int mcu_row, mcu_col, mcu_block; - int block_x_mcu[JPGD_MAX_COMPONENTS], m_block_y_mcu[JPGD_MAX_COMPONENTS]; - - memset(m_block_y_mcu, 0, sizeof(m_block_y_mcu)); - - for (mcu_col = 0; mcu_col < m_mcus_per_col; mcu_col++) - { - int component_num, component_id; - - memset(block_x_mcu, 0, sizeof(block_x_mcu)); - - for (mcu_row = 0; mcu_row < m_mcus_per_row; mcu_row++) - { - int block_x_mcu_ofs = 0, block_y_mcu_ofs = 0; - - if ((m_restart_interval) && (m_restarts_left == 0)) - process_restart(); - - for (mcu_block = 0; mcu_block < m_blocks_per_mcu; mcu_block++) - { - component_id = m_mcu_org[mcu_block]; - - decode_block_func(this, component_id, block_x_mcu[component_id] + block_x_mcu_ofs, m_block_y_mcu[component_id] + block_y_mcu_ofs); - - if (m_comps_in_scan == 1) - block_x_mcu[component_id]++; - else - { - if (++block_x_mcu_ofs == m_comp_h_samp[component_id]) - { - block_x_mcu_ofs = 0; - - if (++block_y_mcu_ofs == m_comp_v_samp[component_id]) - { - block_y_mcu_ofs = 0; - block_x_mcu[component_id] += m_comp_h_samp[component_id]; - } - } - } - } - - m_restarts_left--; - } - - if (m_comps_in_scan == 1) - m_block_y_mcu[m_comp_list[0]]++; - else - { - for (component_num = 0; component_num < m_comps_in_scan; component_num++) - { - component_id = m_comp_list[component_num]; - m_block_y_mcu[component_id] += m_comp_v_samp[component_id]; - } - } - } -} - -// Decode a progressively encoded image. -void jpeg_decoder::init_progressive() -{ - int i; - - if (m_comps_in_frame == 4) - stop_decoding(JPGD_UNSUPPORTED_COLORSPACE); - - // Allocate the coefficient buffers. - for (i = 0; i < m_comps_in_frame; i++) - { - m_dc_coeffs[i] = coeff_buf_open(m_max_mcus_per_row * m_comp_h_samp[i], m_max_mcus_per_col * m_comp_v_samp[i], 1, 1); - m_ac_coeffs[i] = coeff_buf_open(m_max_mcus_per_row * m_comp_h_samp[i], m_max_mcus_per_col * m_comp_v_samp[i], 8, 8); - } - - for ( ; ; ) - { - int dc_only_scan, refinement_scan; - pDecode_block_func decode_block_func; - - if (!init_scan()) - break; - - dc_only_scan = (m_spectral_start == 0); - refinement_scan = (m_successive_high != 0); - - if ((m_spectral_start > m_spectral_end) || (m_spectral_end > 63)) - stop_decoding(JPGD_BAD_SOS_SPECTRAL); - - if (dc_only_scan) - { - if (m_spectral_end) - stop_decoding(JPGD_BAD_SOS_SPECTRAL); - } - else if (m_comps_in_scan != 1) /* AC scans can only contain one component */ - stop_decoding(JPGD_BAD_SOS_SPECTRAL); - - if ((refinement_scan) && (m_successive_low != m_successive_high - 1)) - stop_decoding(JPGD_BAD_SOS_SUCCESSIVE); - - if (dc_only_scan) - { - if (refinement_scan) - decode_block_func = decode_block_dc_refine; - else - decode_block_func = decode_block_dc_first; - } - else - { - if (refinement_scan) - decode_block_func = decode_block_ac_refine; - else - decode_block_func = decode_block_ac_first; - } - - decode_scan(decode_block_func); - - m_bits_left = 16; - get_bits(16); - get_bits(16); - } - - m_comps_in_scan = m_comps_in_frame; - - for (i = 0; i < m_comps_in_frame; i++) - m_comp_list[i] = i; - - calc_mcu_block_order(); -} - -void jpeg_decoder::init_sequential() -{ - if (!init_scan()) - stop_decoding(JPGD_UNEXPECTED_MARKER); -} - -void jpeg_decoder::decode_start() -{ - init_frame(); - - if (m_progressive_flag) - init_progressive(); - else - init_sequential(); -} - -void jpeg_decoder::decode_init(jpeg_decoder_stream *pStream) -{ - init(pStream); - locate_sof_marker(); -} - -jpeg_decoder::jpeg_decoder(jpeg_decoder_stream *pStream) -{ - if (setjmp(m_jmp_state)) - return; - decode_init(pStream); -} - -int jpeg_decoder::begin_decoding() -{ - if (m_ready_flag) - return JPGD_SUCCESS; - - if (m_error_code) - return JPGD_FAILED; - - if (setjmp(m_jmp_state)) - return JPGD_FAILED; - - decode_start(); - - m_ready_flag = true; - - return JPGD_SUCCESS; -} - -jpeg_decoder::~jpeg_decoder() -{ - free_all_blocks(); -} - -jpeg_decoder_file_stream::jpeg_decoder_file_stream() -{ - m_pFile = NULL; - m_eof_flag = false; - m_error_flag = false; -} - -void jpeg_decoder_file_stream::close() -{ - if (m_pFile) - { - fclose(m_pFile); - m_pFile = NULL; - } - - m_eof_flag = false; - m_error_flag = false; -} - -jpeg_decoder_file_stream::~jpeg_decoder_file_stream() -{ - close(); -} - -bool jpeg_decoder_file_stream::open(const char *Pfilename) -{ - close(); - - m_eof_flag = false; - m_error_flag = false; - -#if defined(_MSC_VER) - m_pFile = NULL; - fopen_s(&m_pFile, Pfilename, "rb"); -#else - m_pFile = fopen(Pfilename, "rb"); -#endif - return m_pFile != NULL; -} - -int jpeg_decoder_file_stream::read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) -{ - if (!m_pFile) - return -1; - - if (m_eof_flag) - { - *pEOF_flag = true; - return 0; - } - - if (m_error_flag) - return -1; - - int bytes_read = static_cast(fread(pBuf, 1, max_bytes_to_read, m_pFile)); - if (bytes_read < max_bytes_to_read) - { - if (ferror(m_pFile)) - { - m_error_flag = true; - return -1; - } - - m_eof_flag = true; - *pEOF_flag = true; - } - - return bytes_read; -} - -bool jpeg_decoder_mem_stream::open(const uint8 *pSrc_data, uint size) -{ - close(); - m_pSrc_data = pSrc_data; - m_ofs = 0; - m_size = size; - return true; -} - -int jpeg_decoder_mem_stream::read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) -{ - *pEOF_flag = false; - - if (!m_pSrc_data) - return -1; - - uint bytes_remaining = m_size - m_ofs; - if ((uint)max_bytes_to_read > bytes_remaining) - { - max_bytes_to_read = bytes_remaining; - *pEOF_flag = true; - } - - memcpy(pBuf, m_pSrc_data + m_ofs, max_bytes_to_read); - m_ofs += max_bytes_to_read; - - return max_bytes_to_read; -} - -unsigned char *decompress_jpeg_image_from_stream(jpeg_decoder_stream *pStream, int *width, int *height, int *actual_comps, int req_comps) -{ - if (!actual_comps) - return NULL; - *actual_comps = 0; - - if ((!pStream) || (!width) || (!height) || (!req_comps)) - return NULL; - - if ((req_comps != 1) && (req_comps != 3) && (req_comps != 4)) - return NULL; - - jpeg_decoder decoder(pStream); - if (decoder.get_error_code() != JPGD_SUCCESS) - return NULL; - - const int image_width = decoder.get_width(), image_height = decoder.get_height(); - *width = image_width; - *height = image_height; - *actual_comps = decoder.get_num_components(); - - if (decoder.begin_decoding() != JPGD_SUCCESS) - return NULL; - - const int dst_bpl = image_width * req_comps; - - uint8 *pImage_data = (uint8*)jpgd_malloc(dst_bpl * image_height); - if (!pImage_data) - return NULL; - - for (int y = 0; y < image_height; y++) - { - const uint8* pScan_line; - uint scan_line_len; - if (decoder.decode((const void**)&pScan_line, &scan_line_len) != JPGD_SUCCESS) - { - jpgd_free(pImage_data); - return NULL; - } - - uint8 *pDst = pImage_data + y * dst_bpl; - - if (((req_comps == 1) && (decoder.get_num_components() == 1)) || ((req_comps == 4) && (decoder.get_num_components() == 3))) - memcpy(pDst, pScan_line, dst_bpl); - else if (decoder.get_num_components() == 1) - { - if (req_comps == 3) - { - for (int x = 0; x < image_width; x++) - { - uint8 luma = pScan_line[x]; - pDst[0] = luma; - pDst[1] = luma; - pDst[2] = luma; - pDst += 3; - } - } - else - { - for (int x = 0; x < image_width; x++) - { - uint8 luma = pScan_line[x]; - pDst[0] = luma; - pDst[1] = luma; - pDst[2] = luma; - pDst[3] = 255; - pDst += 4; - } - } - } - else if (decoder.get_num_components() == 3) - { - if (req_comps == 1) - { - const int YR = 19595, YG = 38470, YB = 7471; - for (int x = 0; x < image_width; x++) - { - int r = pScan_line[x*4+0]; - int g = pScan_line[x*4+1]; - int b = pScan_line[x*4+2]; - *pDst++ = static_cast((r * YR + g * YG + b * YB + 32768) >> 16); - } - } - else - { - for (int x = 0; x < image_width; x++) - { - pDst[0] = pScan_line[x*4+0]; - pDst[1] = pScan_line[x*4+1]; - pDst[2] = pScan_line[x*4+2]; - pDst += 3; - } - } - } - } - - return pImage_data; -} - -unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, int *width, int *height, int *actual_comps, int req_comps) -{ - jpgd::jpeg_decoder_mem_stream mem_stream(pSrc_data, src_data_size); - return decompress_jpeg_image_from_stream(&mem_stream, width, height, actual_comps, req_comps); -} - -unsigned char *decompress_jpeg_image_from_file(const char *pSrc_filename, int *width, int *height, int *actual_comps, int req_comps) -{ - jpgd::jpeg_decoder_file_stream file_stream; - if (!file_stream.open(pSrc_filename)) - return NULL; - return decompress_jpeg_image_from_stream(&file_stream, width, height, actual_comps, req_comps); -} - -} // namespace jpgd \ No newline at end of file diff --git a/jpeg-compressor/jpgd.h b/jpeg-compressor/jpgd.h deleted file mode 100644 index 150b9a0b26..0000000000 --- a/jpeg-compressor/jpgd.h +++ /dev/null @@ -1,319 +0,0 @@ -// jpgd.h - C++ class for JPEG decompression. -// Public domain, Rich Geldreich -#ifndef JPEG_DECODER_H -#define JPEG_DECODER_H - -#include -#include -#include - -#ifdef _MSC_VER - #define JPGD_NORETURN __declspec(noreturn) -#elif defined(__GNUC__) - #define JPGD_NORETURN __attribute__ ((noreturn)) -#else - #define JPGD_NORETURN -#endif - -namespace jpgd -{ - typedef unsigned char uint8; - typedef signed short int16; - typedef unsigned short uint16; - typedef unsigned int uint; - typedef signed int int32; - - // Loads a JPEG image from a memory buffer or a file. - // req_comps can be 1 (grayscale), 3 (RGB), or 4 (RGBA). - // On return, width/height will be set to the image's dimensions, and actual_comps will be set to the either 1 (grayscale) or 3 (RGB). - // Notes: For more control over where and how the source data is read, see the decompress_jpeg_image_from_stream() function below, or call the jpeg_decoder class directly. - // Requesting a 8 or 32bpp image is currently a little faster than 24bpp because the jpeg_decoder class itself currently always unpacks to either 8 or 32bpp. - unsigned char *decompress_jpeg_image_from_memory(const unsigned char *pSrc_data, int src_data_size, int *width, int *height, int *actual_comps, int req_comps); - unsigned char *decompress_jpeg_image_from_file(const char *pSrc_filename, int *width, int *height, int *actual_comps, int req_comps); - - // Success/failure error codes. - enum jpgd_status - { - JPGD_SUCCESS = 0, JPGD_FAILED = -1, JPGD_DONE = 1, - JPGD_BAD_DHT_COUNTS = -256, JPGD_BAD_DHT_INDEX, JPGD_BAD_DHT_MARKER, JPGD_BAD_DQT_MARKER, JPGD_BAD_DQT_TABLE, - JPGD_BAD_PRECISION, JPGD_BAD_HEIGHT, JPGD_BAD_WIDTH, JPGD_TOO_MANY_COMPONENTS, - JPGD_BAD_SOF_LENGTH, JPGD_BAD_VARIABLE_MARKER, JPGD_BAD_DRI_LENGTH, JPGD_BAD_SOS_LENGTH, - JPGD_BAD_SOS_COMP_ID, JPGD_W_EXTRA_BYTES_BEFORE_MARKER, JPGD_NO_ARITHMITIC_SUPPORT, JPGD_UNEXPECTED_MARKER, - JPGD_NOT_JPEG, JPGD_UNSUPPORTED_MARKER, JPGD_BAD_DQT_LENGTH, JPGD_TOO_MANY_BLOCKS, - JPGD_UNDEFINED_QUANT_TABLE, JPGD_UNDEFINED_HUFF_TABLE, JPGD_NOT_SINGLE_SCAN, JPGD_UNSUPPORTED_COLORSPACE, - JPGD_UNSUPPORTED_SAMP_FACTORS, JPGD_DECODE_ERROR, JPGD_BAD_RESTART_MARKER, JPGD_ASSERTION_ERROR, - JPGD_BAD_SOS_SPECTRAL, JPGD_BAD_SOS_SUCCESSIVE, JPGD_STREAM_READ, JPGD_NOTENOUGHMEM - }; - - // Input stream interface. - // Derive from this class to read input data from sources other than files or memory. Set m_eof_flag to true when no more data is available. - // The decoder is rather greedy: it will keep on calling this method until its internal input buffer is full, or until the EOF flag is set. - // It the input stream contains data after the JPEG stream's EOI (end of image) marker it will probably be pulled into the internal buffer. - // Call the get_total_bytes_read() method to determine the actual size of the JPEG stream after successful decoding. - class jpeg_decoder_stream - { - public: - jpeg_decoder_stream() { } - virtual ~jpeg_decoder_stream() { } - - // The read() method is called when the internal input buffer is empty. - // Parameters: - // pBuf - input buffer - // max_bytes_to_read - maximum bytes that can be written to pBuf - // pEOF_flag - set this to true if at end of stream (no more bytes remaining) - // Returns -1 on error, otherwise return the number of bytes actually written to the buffer (which may be 0). - // Notes: This method will be called in a loop until you set *pEOF_flag to true or the internal buffer is full. - virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag) = 0; - }; - - // stdio FILE stream class. - class jpeg_decoder_file_stream : public jpeg_decoder_stream - { - jpeg_decoder_file_stream(const jpeg_decoder_file_stream &); - jpeg_decoder_file_stream &operator =(const jpeg_decoder_file_stream &); - - FILE *m_pFile; - bool m_eof_flag, m_error_flag; - - public: - jpeg_decoder_file_stream(); - virtual ~jpeg_decoder_file_stream(); - - bool open(const char *Pfilename); - void close(); - - virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag); - }; - - // Memory stream class. - class jpeg_decoder_mem_stream : public jpeg_decoder_stream - { - const uint8 *m_pSrc_data; - uint m_ofs, m_size; - - public: - jpeg_decoder_mem_stream() : m_pSrc_data(NULL), m_ofs(0), m_size(0) { } - jpeg_decoder_mem_stream(const uint8 *pSrc_data, uint size) : m_pSrc_data(pSrc_data), m_ofs(0), m_size(size) { } - - virtual ~jpeg_decoder_mem_stream() { } - - bool open(const uint8 *pSrc_data, uint size); - void close() { m_pSrc_data = NULL; m_ofs = 0; m_size = 0; } - - virtual int read(uint8 *pBuf, int max_bytes_to_read, bool *pEOF_flag); - }; - - // Loads JPEG file from a jpeg_decoder_stream. - unsigned char *decompress_jpeg_image_from_stream(jpeg_decoder_stream *pStream, int *width, int *height, int *actual_comps, int req_comps); - - enum - { - JPGD_IN_BUF_SIZE = 8192, JPGD_MAX_BLOCKS_PER_MCU = 10, JPGD_MAX_HUFF_TABLES = 8, JPGD_MAX_QUANT_TABLES = 4, - JPGD_MAX_COMPONENTS = 4, JPGD_MAX_COMPS_IN_SCAN = 4, JPGD_MAX_BLOCKS_PER_ROW = 8192, JPGD_MAX_HEIGHT = 16384, JPGD_MAX_WIDTH = 16384 - }; - - typedef int16 jpgd_quant_t; - typedef int16 jpgd_block_t; - - class jpeg_decoder - { - public: - // Call get_error_code() after constructing to determine if the stream is valid or not. You may call the get_width(), get_height(), etc. - // methods after the constructor is called. You may then either destruct the object, or begin decoding the image by calling begin_decoding(), then decode() on each scanline. - jpeg_decoder(jpeg_decoder_stream *pStream); - - ~jpeg_decoder(); - - // Call this method after constructing the object to begin decompression. - // If JPGD_SUCCESS is returned you may then call decode() on each scanline. - int begin_decoding(); - - // Returns the next scan line. - // For grayscale images, pScan_line will point to a buffer containing 8-bit pixels (get_bytes_per_pixel() will return 1). - // Otherwise, it will always point to a buffer containing 32-bit RGBA pixels (A will always be 255, and get_bytes_per_pixel() will return 4). - // Returns JPGD_SUCCESS if a scan line has been returned. - // Returns JPGD_DONE if all scan lines have been returned. - // Returns JPGD_FAILED if an error occurred. Call get_error_code() for a more info. - int decode(const void** pScan_line, uint* pScan_line_len); - - inline jpgd_status get_error_code() const { return m_error_code; } - - inline int get_width() const { return m_image_x_size; } - inline int get_height() const { return m_image_y_size; } - - inline int get_num_components() const { return m_comps_in_frame; } - - inline int get_bytes_per_pixel() const { return m_dest_bytes_per_pixel; } - inline int get_bytes_per_scan_line() const { return m_image_x_size * get_bytes_per_pixel(); } - - // Returns the total number of bytes actually consumed by the decoder (which should equal the actual size of the JPEG file). - inline int get_total_bytes_read() const { return m_total_bytes_read; } - - private: - jpeg_decoder(const jpeg_decoder &); - jpeg_decoder &operator =(const jpeg_decoder &); - - typedef void (*pDecode_block_func)(jpeg_decoder *, int, int, int); - - struct huff_tables - { - bool ac_table; - uint look_up[256]; - uint look_up2[256]; - uint8 code_size[256]; - uint tree[512]; - }; - - struct coeff_buf - { - uint8 *pData; - int block_num_x, block_num_y; - int block_len_x, block_len_y; - int block_size; - }; - - struct mem_block - { - mem_block *m_pNext; - size_t m_used_count; - size_t m_size; - char m_data[1]; - }; - - jmp_buf m_jmp_state; - mem_block *m_pMem_blocks; - int m_image_x_size; - int m_image_y_size; - jpeg_decoder_stream *m_pStream; - int m_progressive_flag; - uint8 m_huff_ac[JPGD_MAX_HUFF_TABLES]; - uint8* m_huff_num[JPGD_MAX_HUFF_TABLES]; // pointer to number of Huffman codes per bit size - uint8* m_huff_val[JPGD_MAX_HUFF_TABLES]; // pointer to Huffman codes per bit size - jpgd_quant_t* m_quant[JPGD_MAX_QUANT_TABLES]; // pointer to quantization tables - int m_scan_type; // Gray, Yh1v1, Yh1v2, Yh2v1, Yh2v2 (CMYK111, CMYK4114 no longer supported) - int m_comps_in_frame; // # of components in frame - int m_comp_h_samp[JPGD_MAX_COMPONENTS]; // component's horizontal sampling factor - int m_comp_v_samp[JPGD_MAX_COMPONENTS]; // component's vertical sampling factor - int m_comp_quant[JPGD_MAX_COMPONENTS]; // component's quantization table selector - int m_comp_ident[JPGD_MAX_COMPONENTS]; // component's ID - int m_comp_h_blocks[JPGD_MAX_COMPONENTS]; - int m_comp_v_blocks[JPGD_MAX_COMPONENTS]; - int m_comps_in_scan; // # of components in scan - int m_comp_list[JPGD_MAX_COMPS_IN_SCAN]; // components in this scan - int m_comp_dc_tab[JPGD_MAX_COMPONENTS]; // component's DC Huffman coding table selector - int m_comp_ac_tab[JPGD_MAX_COMPONENTS]; // component's AC Huffman coding table selector - int m_spectral_start; // spectral selection start - int m_spectral_end; // spectral selection end - int m_successive_low; // successive approximation low - int m_successive_high; // successive approximation high - int m_max_mcu_x_size; // MCU's max. X size in pixels - int m_max_mcu_y_size; // MCU's max. Y size in pixels - int m_blocks_per_mcu; - int m_max_blocks_per_row; - int m_mcus_per_row, m_mcus_per_col; - int m_mcu_org[JPGD_MAX_BLOCKS_PER_MCU]; - int m_total_lines_left; // total # lines left in image - int m_mcu_lines_left; // total # lines left in this MCU - int m_real_dest_bytes_per_scan_line; - int m_dest_bytes_per_scan_line; // rounded up - int m_dest_bytes_per_pixel; // 4 (RGB) or 1 (Y) - huff_tables* m_pHuff_tabs[JPGD_MAX_HUFF_TABLES]; - coeff_buf* m_dc_coeffs[JPGD_MAX_COMPONENTS]; - coeff_buf* m_ac_coeffs[JPGD_MAX_COMPONENTS]; - int m_eob_run; - int m_block_y_mcu[JPGD_MAX_COMPONENTS]; - uint8* m_pIn_buf_ofs; - int m_in_buf_left; - int m_tem_flag; - bool m_eof_flag; - uint8 m_in_buf_pad_start[128]; - uint8 m_in_buf[JPGD_IN_BUF_SIZE + 128]; - uint8 m_in_buf_pad_end[128]; - int m_bits_left; - uint m_bit_buf; - int m_restart_interval; - int m_restarts_left; - int m_next_restart_num; - int m_max_mcus_per_row; - int m_max_blocks_per_mcu; - int m_expanded_blocks_per_mcu; - int m_expanded_blocks_per_row; - int m_expanded_blocks_per_component; - bool m_freq_domain_chroma_upsample; - int m_max_mcus_per_col; - uint m_last_dc_val[JPGD_MAX_COMPONENTS]; - jpgd_block_t* m_pMCU_coefficients; - int m_mcu_block_max_zag[JPGD_MAX_BLOCKS_PER_MCU]; - uint8* m_pSample_buf; - int m_crr[256]; - int m_cbb[256]; - int m_crg[256]; - int m_cbg[256]; - uint8* m_pScan_line_0; - uint8* m_pScan_line_1; - jpgd_status m_error_code; - bool m_ready_flag; - int m_total_bytes_read; - - void free_all_blocks(); - JPGD_NORETURN void stop_decoding(jpgd_status status); - void *alloc(size_t n, bool zero = false); - void word_clear(void *p, uint16 c, uint n); - void prep_in_buffer(); - void read_dht_marker(); - void read_dqt_marker(); - void read_sof_marker(); - void skip_variable_marker(); - void read_dri_marker(); - void read_sos_marker(); - int next_marker(); - int process_markers(); - void locate_soi_marker(); - void locate_sof_marker(); - int locate_sos_marker(); - void init(jpeg_decoder_stream * pStream); - void create_look_ups(); - void fix_in_buffer(); - void transform_mcu(int mcu_row); - void transform_mcu_expand(int mcu_row); - coeff_buf* coeff_buf_open(int block_num_x, int block_num_y, int block_len_x, int block_len_y); - inline jpgd_block_t *coeff_buf_getp(coeff_buf *cb, int block_x, int block_y); - void load_next_row(); - void decode_next_row(); - void make_huff_table(int index, huff_tables *pH); - void check_quant_tables(); - void check_huff_tables(); - void calc_mcu_block_order(); - int init_scan(); - void init_frame(); - void process_restart(); - void decode_scan(pDecode_block_func decode_block_func); - void init_progressive(); - void init_sequential(); - void decode_start(); - void decode_init(jpeg_decoder_stream * pStream); - void H2V2Convert(); - void H2V1Convert(); - void H1V2Convert(); - void H1V1Convert(); - void gray_convert(); - void expanded_convert(); - void find_eoi(); - inline uint get_char(); - inline uint get_char(bool *pPadding_flag); - inline void stuff_char(uint8 q); - inline uint8 get_octet(); - inline uint get_bits(int num_bits); - inline uint get_bits_no_markers(int numbits); - inline int huff_decode(huff_tables *pH); - inline int huff_decode(huff_tables *pH, int& extrabits); - static inline uint8 clamp(int i); - static void decode_block_dc_first(jpeg_decoder *pD, int component_id, int block_x, int block_y); - static void decode_block_dc_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y); - static void decode_block_ac_first(jpeg_decoder *pD, int component_id, int block_x, int block_y); - static void decode_block_ac_refine(jpeg_decoder *pD, int component_id, int block_x, int block_y); - }; - -} // namespace jpgd - -#endif // JPEG_DECODER_H diff --git a/lodepng/lodepng.cpp b/lodepng/lodepng.cpp deleted file mode 100644 index 6fbd6acb4c..0000000000 --- a/lodepng/lodepng.cpp +++ /dev/null @@ -1,6267 +0,0 @@ -/* -LodePNG version 20130831 - -Copyright (c) 2005-2013 Lode Vandevenne - -This software is provided 'as-is', without any express or implied -warranty. In no event will the authors be held liable for any damages -arising from the use of this software. - -Permission is granted to anyone to use this software for any purpose, -including commercial applications, and to alter it and redistribute it -freely, subject to the following restrictions: - - 1. The origin of this software must not be misrepresented; you must not - claim that you wrote the original software. If you use this software - in a product, an acknowledgment in the product documentation would be - appreciated but is not required. - - 2. Altered source versions must be plainly marked as such, and must not be - misrepresented as being the original software. - - 3. This notice may not be removed or altered from any source - distribution. -*/ - -/* -The manual and changelog are in the header file "lodepng.h" -Rename this file to lodepng.cpp to use it for C++, or to lodepng.c to use it for C. -*/ - -#include "lodepng.h" - -#include -#include - -#ifdef LODEPNG_COMPILE_CPP -#include -#endif /*LODEPNG_COMPILE_CPP*/ - -#define VERSION_STRING "20130831" - -/* -This source file is built up in the following large parts. The code sections -with the "LODEPNG_COMPILE_" #defines divide this up further in an intermixed way. --Tools for C and common code for PNG and Zlib --C Code for Zlib (huffman, deflate, ...) --C Code for PNG (file format chunks, adam7, PNG filters, color conversions, ...) --The C++ wrapper around all of the above -*/ - -/*The malloc, realloc and free functions defined here with "lodepng_" in front -of the name, so that you can easily change them to others related to your -platform if needed. Everything else in the code calls these. Pass --DLODEPNG_NO_COMPILE_ALLOCATORS to the compiler, or comment out -#define LODEPNG_COMPILE_ALLOCATORS in the header, to disable the ones here and -define them in your own project's source files without needing to change -lodepng source code. Don't forget to remove "static" if you copypaste them -from here.*/ - -#ifdef LODEPNG_COMPILE_ALLOCATORS -static void* lodepng_malloc(size_t size) -{ - return malloc(size); -} - -static void* lodepng_realloc(void* ptr, size_t new_size) -{ - return realloc(ptr, new_size); -} - -static void lodepng_free(void* ptr) -{ - free(ptr); -} -#else /*LODEPNG_COMPILE_ALLOCATORS*/ -void* lodepng_malloc(size_t size); -void* lodepng_realloc(void* ptr, size_t new_size); -void lodepng_free(void* ptr); -#endif /*LODEPNG_COMPILE_ALLOCATORS*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // Tools for C, and common code for PNG and Zlib. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/* -Often in case of an error a value is assigned to a variable and then it breaks -out of a loop (to go to the cleanup phase of a function). This macro does that. -It makes the error handling code shorter and more readable. - -Example: if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83); -*/ -#define CERROR_BREAK(errorvar, code)\ -{\ - errorvar = code;\ - break;\ -} - -/*version of CERROR_BREAK that assumes the common case where the error variable is named "error"*/ -#define ERROR_BREAK(code) CERROR_BREAK(error, code) - -/*Set error var to the error code, and return it.*/ -#define CERROR_RETURN_ERROR(errorvar, code)\ -{\ - errorvar = code;\ - return code;\ -} - -/*Try the code, if it returns error, also return the error.*/ -#define CERROR_TRY_RETURN(call)\ -{\ - unsigned error = call;\ - if(error) return error;\ -} - -/* -About uivector, ucvector and string: --All of them wrap dynamic arrays or text strings in a similar way. --LodePNG was originally written in C++. The vectors replace the std::vectors that were used in the C++ version. --The string tools are made to avoid problems with compilers that declare things like strncat as deprecated. --They're not used in the interface, only internally in this file as static functions. --As with many other structs in this file, the init and cleanup functions serve as ctor and dtor. -*/ - -#ifdef LODEPNG_COMPILE_ZLIB -/*dynamic vector of unsigned ints*/ -typedef struct uivector -{ - unsigned* data; - size_t size; /*size in number of unsigned longs*/ - size_t allocsize; /*allocated size in bytes*/ -} uivector; - -static void uivector_cleanup(void* p) -{ - ((uivector*)p)->size = ((uivector*)p)->allocsize = 0; - lodepng_free(((uivector*)p)->data); - ((uivector*)p)->data = NULL; -} - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_resize(uivector* p, size_t size) -{ - if(size * sizeof(unsigned) > p->allocsize) - { - size_t newsize = size * sizeof(unsigned) * 2; - void* data = lodepng_realloc(p->data, newsize); - if(data) - { - p->allocsize = newsize; - p->data = (unsigned*)data; - p->size = size; - } - else return 0; - } - else p->size = size; - return 1; -} - -/*resize and give all new elements the value*/ -static unsigned uivector_resizev(uivector* p, size_t size, unsigned value) -{ - size_t oldsize = p->size, i; - if(!uivector_resize(p, size)) return 0; - for(i = oldsize; i < size; i++) p->data[i] = value; - return 1; -} - -static void uivector_init(uivector* p) -{ - p->data = NULL; - p->size = p->allocsize = 0; -} - -#ifdef LODEPNG_COMPILE_ENCODER -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_push_back(uivector* p, unsigned c) -{ - if(!uivector_resize(p, p->size + 1)) return 0; - p->data[p->size - 1] = c; - return 1; -} - -/*copy q to p, returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned uivector_copy(uivector* p, const uivector* q) -{ - size_t i; - if(!uivector_resize(p, q->size)) return 0; - for(i = 0; i < q->size; i++) p->data[i] = q->data[i]; - return 1; -} - -static void uivector_swap(uivector* p, uivector* q) -{ - size_t tmp; - unsigned* tmpp; - tmp = p->size; p->size = q->size; q->size = tmp; - tmp = p->allocsize; p->allocsize = q->allocsize; q->allocsize = tmp; - tmpp = p->data; p->data = q->data; q->data = tmpp; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_ZLIB*/ - -/* /////////////////////////////////////////////////////////////////////////// */ - -/*dynamic vector of unsigned chars*/ -typedef struct ucvector -{ - unsigned char* data; - size_t size; /*used size*/ - size_t allocsize; /*allocated size*/ -} ucvector; - -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_resize(ucvector* p, size_t size) -{ - if(size * sizeof(unsigned char) > p->allocsize) - { - size_t newsize = size * sizeof(unsigned char) * 2; - void* data = lodepng_realloc(p->data, newsize); - if(data) - { - p->allocsize = newsize; - p->data = (unsigned char*)data; - p->size = size; - } - else return 0; /*error: not enough memory*/ - } - else p->size = size; - return 1; -} - -#ifdef LODEPNG_COMPILE_PNG - -static void ucvector_cleanup(void* p) -{ - ((ucvector*)p)->size = ((ucvector*)p)->allocsize = 0; - lodepng_free(((ucvector*)p)->data); - ((ucvector*)p)->data = NULL; -} - -static void ucvector_init(ucvector* p) -{ - p->data = NULL; - p->size = p->allocsize = 0; -} - -#ifdef LODEPNG_COMPILE_DECODER -/*resize and give all new elements the value*/ -static unsigned ucvector_resizev(ucvector* p, size_t size, unsigned char value) -{ - size_t oldsize = p->size, i; - if(!ucvector_resize(p, size)) return 0; - for(i = oldsize; i < size; i++) p->data[i] = value; - return 1; -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ZLIB -/*you can both convert from vector to buffer&size and vica versa. If you use -init_buffer to take over a buffer and size, it is not needed to use cleanup*/ -static void ucvector_init_buffer(ucvector* p, unsigned char* buffer, size_t size) -{ - p->data = buffer; - p->allocsize = p->size = size; -} -#endif /*LODEPNG_COMPILE_ZLIB*/ - -#if (defined(LODEPNG_COMPILE_PNG) && defined(LODEPNG_COMPILE_ANCILLARY_CHUNKS)) || defined(LODEPNG_COMPILE_ENCODER) -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned ucvector_push_back(ucvector* p, unsigned char c) -{ - if(!ucvector_resize(p, p->size + 1)) return 0; - p->data[p->size - 1] = c; - return 1; -} -#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ - - -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_PNG -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*returns 1 if success, 0 if failure ==> nothing done*/ -static unsigned string_resize(char** out, size_t size) -{ - char* data = (char*)lodepng_realloc(*out, size + 1); - if(data) - { - data[size] = 0; /*null termination char*/ - *out = data; - } - return data != 0; -} - -/*init a {char*, size_t} pair for use as string*/ -static void string_init(char** out) -{ - *out = NULL; - string_resize(out, 0); -} - -/*free the above pair again*/ -static void string_cleanup(char** out) -{ - lodepng_free(*out); - *out = NULL; -} - -static void string_set(char** out, const char* in) -{ - size_t insize = strlen(in), i = 0; - if(string_resize(out, insize)) - { - for(i = 0; i < insize; i++) - { - (*out)[i] = in[i]; - } - } -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -/* ////////////////////////////////////////////////////////////////////////// */ - -unsigned lodepng_read32bitInt(const unsigned char* buffer) -{ - return (buffer[0] << 24) | (buffer[1] << 16) | (buffer[2] << 8) | buffer[3]; -} - -#if defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER) -/*buffer must have at least 4 allocated bytes available*/ -static void lodepng_set32bitInt(unsigned char* buffer, unsigned value) -{ - buffer[0] = (unsigned char)((value >> 24) & 0xff); - buffer[1] = (unsigned char)((value >> 16) & 0xff); - buffer[2] = (unsigned char)((value >> 8) & 0xff); - buffer[3] = (unsigned char)((value ) & 0xff); -} -#endif /*defined(LODEPNG_COMPILE_PNG) || defined(LODEPNG_COMPILE_ENCODER)*/ - -#ifdef LODEPNG_COMPILE_ENCODER -static void lodepng_add32bitInt(ucvector* buffer, unsigned value) -{ - ucvector_resize(buffer, buffer->size + 4); /*todo: give error if resize failed*/ - lodepng_set32bitInt(&buffer->data[buffer->size - 4], value); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / File IO / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_DISK - -unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename) -{ - FILE* file; - long size; - - /*provide some proper output values if error will happen*/ - *out = 0; - *outsize = 0; - - file = fopen(filename, "rb"); - if(!file) return 78; - - /*get filesize:*/ - fseek(file , 0 , SEEK_END); - size = ftell(file); - rewind(file); - - /*read contents of the file into the vector*/ - *outsize = 0; - *out = (unsigned char*)lodepng_malloc((size_t)size); - if(size && (*out)) (*outsize) = fread(*out, 1, (size_t)size, file); - - fclose(file); - if(!(*out) && size) return 83; /*the above malloc failed*/ - return 0; -} - -/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ -unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename) -{ - FILE* file; - file = fopen(filename, "wb" ); - if(!file) return 79; - fwrite((char*)buffer , 1 , buffersize, file); - fclose(file); - return 0; -} - -#endif /*LODEPNG_COMPILE_DISK*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // End of common code and tools. Begin of Zlib related code. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_ENCODER -/*TODO: this ignores potential out of memory errors*/ -static void addBitToStream(size_t* bitpointer, ucvector* bitstream, unsigned char bit) -{ - /*add a new byte at the end*/ - if((*bitpointer) % 8 == 0) ucvector_push_back(bitstream, (unsigned char)0); - /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/ - (bitstream->data[bitstream->size - 1]) |= (bit << ((*bitpointer) & 0x7)); - (*bitpointer)++; -} - -static void addBitsToStream(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) -{ - size_t i; - for(i = 0; i < nbits; i++) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> i) & 1)); -} - -static void addBitsToStreamReversed(size_t* bitpointer, ucvector* bitstream, unsigned value, size_t nbits) -{ - size_t i; - for(i = 0; i < nbits; i++) addBitToStream(bitpointer, bitstream, (unsigned char)((value >> (nbits - 1 - i)) & 1)); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -#define READBIT(bitpointer, bitstream) ((bitstream[bitpointer >> 3] >> (bitpointer & 0x7)) & (unsigned char)1) - -static unsigned char readBitFromStream(size_t* bitpointer, const unsigned char* bitstream) -{ - unsigned char result = (unsigned char)(READBIT(*bitpointer, bitstream)); - (*bitpointer)++; - return result; -} - -static unsigned readBitsFromStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) -{ - unsigned result = 0, i; - for(i = 0; i < nbits; i++) - { - result += ((unsigned)READBIT(*bitpointer, bitstream)) << i; - (*bitpointer)++; - } - return result; -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Deflate - Huffman / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#define FIRST_LENGTH_CODE_INDEX 257 -#define LAST_LENGTH_CODE_INDEX 285 -/*256 literals, the end code, some length codes, and 2 unused codes*/ -#define NUM_DEFLATE_CODE_SYMBOLS 288 -/*the distance codes have their own symbols, 30 used, 2 unused*/ -#define NUM_DISTANCE_SYMBOLS 32 -/*the code length codes. 0-15: code lengths, 16: copy previous 3-6 times, 17: 3-10 zeros, 18: 11-138 zeros*/ -#define NUM_CODE_LENGTH_CODES 19 - -/*the base lengths represented by codes 257-285*/ -static const unsigned LENGTHBASE[29] - = {3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51, 59, - 67, 83, 99, 115, 131, 163, 195, 227, 258}; - -/*the extra bits used by codes 257-285 (added to base length)*/ -static const unsigned LENGTHEXTRA[29] - = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, - 4, 4, 4, 4, 5, 5, 5, 5, 0}; - -/*the base backwards distances (the bits of distance codes appear after length codes and use their own huffman tree)*/ -static const unsigned DISTANCEBASE[30] - = {1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513, - 769, 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577}; - -/*the extra bits of backwards distances (added to base)*/ -static const unsigned DISTANCEEXTRA[30] - = {0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, - 8, 9, 9, 10, 10, 11, 11, 12, 12, 13, 13}; - -/*the order in which "code length alphabet code lengths" are stored, out of this -the huffman tree of the dynamic huffman tree lengths is generated*/ -static const unsigned CLCL_ORDER[NUM_CODE_LENGTH_CODES] - = {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; - -/* ////////////////////////////////////////////////////////////////////////// */ - -/* -Huffman tree struct, containing multiple representations of the tree -*/ -typedef struct HuffmanTree -{ - unsigned* tree2d; - unsigned* tree1d; - unsigned* lengths; /*the lengths of the codes of the 1d-tree*/ - unsigned maxbitlen; /*maximum number of bits a single code can get*/ - unsigned numcodes; /*number of symbols in the alphabet = number of codes*/ -} HuffmanTree; - -/*function used for debug purposes to draw the tree in ascii art with C++*/ -/* -static void HuffmanTree_draw(HuffmanTree* tree) -{ - std::cout << "tree. length: " << tree->numcodes << " maxbitlen: " << tree->maxbitlen << std::endl; - for(size_t i = 0; i < tree->tree1d.size; i++) - { - if(tree->lengths.data[i]) - std::cout << i << " " << tree->tree1d.data[i] << " " << tree->lengths.data[i] << std::endl; - } - std::cout << std::endl; -}*/ - -static void HuffmanTree_init(HuffmanTree* tree) -{ - tree->tree2d = 0; - tree->tree1d = 0; - tree->lengths = 0; -} - -static void HuffmanTree_cleanup(HuffmanTree* tree) -{ - lodepng_free(tree->tree2d); - lodepng_free(tree->tree1d); - lodepng_free(tree->lengths); -} - -/*the tree representation used by the decoder. return value is error*/ -static unsigned HuffmanTree_make2DTree(HuffmanTree* tree) -{ - unsigned nodefilled = 0; /*up to which node it is filled*/ - unsigned treepos = 0; /*position in the tree (1 of the numcodes columns)*/ - unsigned n, i; - - tree->tree2d = (unsigned*)lodepng_malloc(tree->numcodes * 2 * sizeof(unsigned)); - if(!tree->tree2d) return 83; /*alloc fail*/ - - /* - convert tree1d[] to tree2d[][]. In the 2D array, a value of 32767 means - uninited, a value >= numcodes is an address to another bit, a value < numcodes - is a code. The 2 rows are the 2 possible bit values (0 or 1), there are as - many columns as codes - 1. - A good huffmann tree has N * 2 - 1 nodes, of which N - 1 are internal nodes. - Here, the internal nodes are stored (what their 0 and 1 option point to). - There is only memory for such good tree currently, if there are more nodes - (due to too long length codes), error 55 will happen - */ - for(n = 0; n < tree->numcodes * 2; n++) - { - tree->tree2d[n] = 32767; /*32767 here means the tree2d isn't filled there yet*/ - } - - for(n = 0; n < tree->numcodes; n++) /*the codes*/ - { - for(i = 0; i < tree->lengths[n]; i++) /*the bits for this code*/ - { - unsigned char bit = (unsigned char)((tree->tree1d[n] >> (tree->lengths[n] - i - 1)) & 1); - if(treepos > tree->numcodes - 2) return 55; /*oversubscribed, see comment in lodepng_error_text*/ - if(tree->tree2d[2 * treepos + bit] == 32767) /*not yet filled in*/ - { - if(i + 1 == tree->lengths[n]) /*last bit*/ - { - tree->tree2d[2 * treepos + bit] = n; /*put the current code in it*/ - treepos = 0; - } - else - { - /*put address of the next step in here, first that address has to be found of course - (it's just nodefilled + 1)...*/ - nodefilled++; - /*addresses encoded with numcodes added to it*/ - tree->tree2d[2 * treepos + bit] = nodefilled + tree->numcodes; - treepos = nodefilled; - } - } - else treepos = tree->tree2d[2 * treepos + bit] - tree->numcodes; - } - } - - for(n = 0; n < tree->numcodes * 2; n++) - { - if(tree->tree2d[n] == 32767) tree->tree2d[n] = 0; /*remove possible remaining 32767's*/ - } - - return 0; -} - -/* -Second step for the ...makeFromLengths and ...makeFromFrequencies functions. -numcodes, lengths and maxbitlen must already be filled in correctly. return -value is error. -*/ -static unsigned HuffmanTree_makeFromLengths2(HuffmanTree* tree) -{ - uivector blcount; - uivector nextcode; - unsigned bits, n, error = 0; - - uivector_init(&blcount); - uivector_init(&nextcode); - - tree->tree1d = (unsigned*)lodepng_malloc(tree->numcodes * sizeof(unsigned)); - if(!tree->tree1d) error = 83; /*alloc fail*/ - - if(!uivector_resizev(&blcount, tree->maxbitlen + 1, 0) - || !uivector_resizev(&nextcode, tree->maxbitlen + 1, 0)) - error = 83; /*alloc fail*/ - - if(!error) - { - /*step 1: count number of instances of each code length*/ - for(bits = 0; bits < tree->numcodes; bits++) blcount.data[tree->lengths[bits]]++; - /*step 2: generate the nextcode values*/ - for(bits = 1; bits <= tree->maxbitlen; bits++) - { - nextcode.data[bits] = (nextcode.data[bits - 1] + blcount.data[bits - 1]) << 1; - } - /*step 3: generate all the codes*/ - for(n = 0; n < tree->numcodes; n++) - { - if(tree->lengths[n] != 0) tree->tree1d[n] = nextcode.data[tree->lengths[n]]++; - } - } - - uivector_cleanup(&blcount); - uivector_cleanup(&nextcode); - - if(!error) return HuffmanTree_make2DTree(tree); - else return error; -} - -/* -given the code lengths (as stored in the PNG file), generate the tree as defined -by Deflate. maxbitlen is the maximum bits that a code in the tree can have. -return value is error. -*/ -static unsigned HuffmanTree_makeFromLengths(HuffmanTree* tree, const unsigned* bitlen, - size_t numcodes, unsigned maxbitlen) -{ - unsigned i; - tree->lengths = (unsigned*)lodepng_malloc(numcodes * sizeof(unsigned)); - if(!tree->lengths) return 83; /*alloc fail*/ - for(i = 0; i < numcodes; i++) tree->lengths[i] = bitlen[i]; - tree->numcodes = (unsigned)numcodes; /*number of symbols*/ - tree->maxbitlen = maxbitlen; - return HuffmanTree_makeFromLengths2(tree); -} - -#ifdef LODEPNG_COMPILE_ENCODER - -/* -A coin, this is the terminology used for the package-merge algorithm and the -coin collector's problem. This is used to generate the huffman tree. -A coin can be multiple coins (when they're merged) -*/ -typedef struct Coin -{ - uivector symbols; - float weight; /*the sum of all weights in this coin*/ -} Coin; - -static void coin_init(Coin* c) -{ - uivector_init(&c->symbols); -} - -/*argument c is void* so that this dtor can be given as function pointer to the vector resize function*/ -static void coin_cleanup(void* c) -{ - uivector_cleanup(&((Coin*)c)->symbols); -} - -static void coin_copy(Coin* c1, const Coin* c2) -{ - c1->weight = c2->weight; - uivector_copy(&c1->symbols, &c2->symbols); -} - -static void add_coins(Coin* c1, const Coin* c2) -{ - size_t i; - for(i = 0; i < c2->symbols.size; i++) uivector_push_back(&c1->symbols, c2->symbols.data[i]); - c1->weight += c2->weight; -} - -static void init_coins(Coin* coins, size_t num) -{ - size_t i; - for(i = 0; i < num; i++) coin_init(&coins[i]); -} - -static void cleanup_coins(Coin* coins, size_t num) -{ - size_t i; - for(i = 0; i < num; i++) coin_cleanup(&coins[i]); -} - -/* -This uses a simple combsort to sort the data. This function is not critical for -overall encoding speed and the data amount isn't that large. -*/ -static void sort_coins(Coin* data, size_t amount) -{ - size_t gap = amount; - unsigned char swapped = 0; - while((gap > 1) || swapped) - { - size_t i; - gap = (gap * 10) / 13; /*shrink factor 1.3*/ - if(gap == 9 || gap == 10) gap = 11; /*combsort11*/ - if(gap < 1) gap = 1; - swapped = 0; - for(i = 0; i < amount - gap; i++) - { - size_t j = i + gap; - if(data[j].weight < data[i].weight) - { - float temp = data[j].weight; data[j].weight = data[i].weight; data[i].weight = temp; - uivector_swap(&data[i].symbols, &data[j].symbols); - swapped = 1; - } - } - } -} - -static unsigned append_symbol_coins(Coin* coins, const unsigned* frequencies, unsigned numcodes, size_t sum) -{ - unsigned i; - unsigned j = 0; /*index of present symbols*/ - for(i = 0; i < numcodes; i++) - { - if(frequencies[i] != 0) /*only include symbols that are present*/ - { - coins[j].weight = frequencies[i] / (float)sum; - uivector_push_back(&coins[j].symbols, i); - j++; - } - } - return 0; -} - -unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, - size_t numcodes, unsigned maxbitlen) -{ - unsigned i, j; - size_t sum = 0, numpresent = 0; - unsigned error = 0; - Coin* coins; /*the coins of the currently calculated row*/ - Coin* prev_row; /*the previous row of coins*/ - unsigned numcoins; - unsigned coinmem; - - if(numcodes == 0) return 80; /*error: a tree of 0 symbols is not supposed to be made*/ - - for(i = 0; i < numcodes; i++) - { - if(frequencies[i] > 0) - { - numpresent++; - sum += frequencies[i]; - } - } - - for(i = 0; i < numcodes; i++) lengths[i] = 0; - - /*ensure at least two present symbols. There should be at least one symbol - according to RFC 1951 section 3.2.7. To decoders incorrectly require two. To - make these work as well ensure there are at least two symbols. The - Package-Merge code below also doesn't work correctly if there's only one - symbol, it'd give it the theoritical 0 bits but in practice zlib wants 1 bit*/ - if(numpresent == 0) - { - lengths[0] = lengths[1] = 1; /*note that for RFC 1951 section 3.2.7, only lengths[0] = 1 is needed*/ - } - else if(numpresent == 1) - { - for(i = 0; i < numcodes; i++) - { - if(frequencies[i]) - { - lengths[i] = 1; - lengths[i == 0 ? 1 : 0] = 1; - break; - } - } - } - else - { - /*Package-Merge algorithm represented by coin collector's problem - For every symbol, maxbitlen coins will be created*/ - - coinmem = numpresent * 2; /*max amount of coins needed with the current algo*/ - coins = (Coin*)lodepng_malloc(sizeof(Coin) * coinmem); - prev_row = (Coin*)lodepng_malloc(sizeof(Coin) * coinmem); - if(!coins || !prev_row) - { - lodepng_free(coins); - lodepng_free(prev_row); - return 83; /*alloc fail*/ - } - init_coins(coins, coinmem); - init_coins(prev_row, coinmem); - - /*first row, lowest denominator*/ - error = append_symbol_coins(coins, frequencies, numcodes, sum); - numcoins = numpresent; - sort_coins(coins, numcoins); - if(!error) - { - unsigned numprev = 0; - for(j = 1; j <= maxbitlen && !error; j++) /*each of the remaining rows*/ - { - unsigned tempnum; - Coin* tempcoins; - /*swap prev_row and coins, and their amounts*/ - tempcoins = prev_row; prev_row = coins; coins = tempcoins; - tempnum = numprev; numprev = numcoins; numcoins = tempnum; - - cleanup_coins(coins, numcoins); - init_coins(coins, numcoins); - - numcoins = 0; - - /*fill in the merged coins of the previous row*/ - for(i = 0; i + 1 < numprev; i += 2) - { - /*merge prev_row[i] and prev_row[i + 1] into new coin*/ - Coin* coin = &coins[numcoins++]; - coin_copy(coin, &prev_row[i]); - add_coins(coin, &prev_row[i + 1]); - } - /*fill in all the original symbols again*/ - if(j < maxbitlen) - { - error = append_symbol_coins(coins + numcoins, frequencies, numcodes, sum); - numcoins += numpresent; - } - sort_coins(coins, numcoins); - } - } - - if(!error) - { - /*calculate the lenghts of each symbol, as the amount of times a coin of each symbol is used*/ - for(i = 0; i < numpresent - 1; i++) - { - Coin* coin = &coins[i]; - for(j = 0; j < coin->symbols.size; j++) lengths[coin->symbols.data[j]]++; - } - } - - cleanup_coins(coins, coinmem); - lodepng_free(coins); - cleanup_coins(prev_row, coinmem); - lodepng_free(prev_row); - } - - return error; -} - -/*Create the Huffman tree given the symbol frequencies*/ -static unsigned HuffmanTree_makeFromFrequencies(HuffmanTree* tree, const unsigned* frequencies, - size_t mincodes, size_t numcodes, unsigned maxbitlen) -{ - unsigned error = 0; - while(!frequencies[numcodes - 1] && numcodes > mincodes) numcodes--; /*trim zeroes*/ - tree->maxbitlen = maxbitlen; - tree->numcodes = (unsigned)numcodes; /*number of symbols*/ - tree->lengths = (unsigned*)lodepng_realloc(tree->lengths, numcodes * sizeof(unsigned)); - if(!tree->lengths) return 83; /*alloc fail*/ - /*initialize all lengths to 0*/ - memset(tree->lengths, 0, numcodes * sizeof(unsigned)); - - error = lodepng_huffman_code_lengths(tree->lengths, frequencies, numcodes, maxbitlen); - if(!error) error = HuffmanTree_makeFromLengths2(tree); - return error; -} - -static unsigned HuffmanTree_getCode(const HuffmanTree* tree, unsigned index) -{ - return tree->tree1d[index]; -} - -static unsigned HuffmanTree_getLength(const HuffmanTree* tree, unsigned index) -{ - return tree->lengths[index]; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/*get the literal and length code tree of a deflated block with fixed tree, as per the deflate specification*/ -static unsigned generateFixedLitLenTree(HuffmanTree* tree) -{ - unsigned i, error = 0; - unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); - if(!bitlen) return 83; /*alloc fail*/ - - /*288 possible codes: 0-255=literals, 256=endcode, 257-285=lengthcodes, 286-287=unused*/ - for(i = 0; i <= 143; i++) bitlen[i] = 8; - for(i = 144; i <= 255; i++) bitlen[i] = 9; - for(i = 256; i <= 279; i++) bitlen[i] = 7; - for(i = 280; i <= 287; i++) bitlen[i] = 8; - - error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DEFLATE_CODE_SYMBOLS, 15); - - lodepng_free(bitlen); - return error; -} - -/*get the distance code tree of a deflated block with fixed tree, as specified in the deflate specification*/ -static unsigned generateFixedDistanceTree(HuffmanTree* tree) -{ - unsigned i, error = 0; - unsigned* bitlen = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); - if(!bitlen) return 83; /*alloc fail*/ - - /*there are 32 distance codes, but 30-31 are unused*/ - for(i = 0; i < NUM_DISTANCE_SYMBOLS; i++) bitlen[i] = 5; - error = HuffmanTree_makeFromLengths(tree, bitlen, NUM_DISTANCE_SYMBOLS, 15); - - lodepng_free(bitlen); - return error; -} - -#ifdef LODEPNG_COMPILE_DECODER - -/* -returns the code, or (unsigned)(-1) if error happened -inbitlength is the length of the complete buffer, in bits (so its byte length times 8) -*/ -static unsigned huffmanDecodeSymbol(const unsigned char* in, size_t* bp, - const HuffmanTree* codetree, size_t inbitlength) -{ - unsigned treepos = 0, ct; - for(;;) - { - if(*bp >= inbitlength) return (unsigned)(-1); /*error: end of input memory reached without endcode*/ - /* - decode the symbol from the tree. The "readBitFromStream" code is inlined in - the expression below because this is the biggest bottleneck while decoding - */ - ct = codetree->tree2d[(treepos << 1) + READBIT(*bp, in)]; - (*bp)++; - if(ct < codetree->numcodes) return ct; /*the symbol is decoded, return it*/ - else treepos = ct - codetree->numcodes; /*symbol not yet decoded, instead move tree position*/ - - if(treepos >= codetree->numcodes) return (unsigned)(-1); /*error: it appeared outside the codetree*/ - } -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Inflator (Decompressor) / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*get the tree of a deflated block with fixed tree, as specified in the deflate specification*/ -static void getTreeInflateFixed(HuffmanTree* tree_ll, HuffmanTree* tree_d) -{ - /*TODO: check for out of memory errors*/ - generateFixedLitLenTree(tree_ll); - generateFixedDistanceTree(tree_d); -} - -/*get the tree of a deflated block with dynamic tree, the tree itself is also Huffman compressed with a known tree*/ -static unsigned getTreeInflateDynamic(HuffmanTree* tree_ll, HuffmanTree* tree_d, - const unsigned char* in, size_t* bp, size_t inlength) -{ - /*make sure that length values that aren't filled in will be 0, or a wrong tree will be generated*/ - unsigned error = 0; - unsigned n, HLIT, HDIST, HCLEN, i; - size_t inbitlength = inlength * 8; - - /*see comments in deflateDynamic for explanation of the context and these variables, it is analogous*/ - unsigned* bitlen_ll = 0; /*lit,len code lengths*/ - unsigned* bitlen_d = 0; /*dist code lengths*/ - /*code length code lengths ("clcl"), the bit lengths of the huffman tree used to compress bitlen_ll and bitlen_d*/ - unsigned* bitlen_cl = 0; - HuffmanTree tree_cl; /*the code tree for code length codes (the huffman tree for compressed huffman trees)*/ - - if((*bp) >> 3 >= inlength - 2) return 49; /*error: the bit pointer is or will go past the memory*/ - - /*number of literal/length codes + 257. Unlike the spec, the value 257 is added to it here already*/ - HLIT = readBitsFromStream(bp, in, 5) + 257; - /*number of distance codes. Unlike the spec, the value 1 is added to it here already*/ - HDIST = readBitsFromStream(bp, in, 5) + 1; - /*number of code length codes. Unlike the spec, the value 4 is added to it here already*/ - HCLEN = readBitsFromStream(bp, in, 4) + 4; - - HuffmanTree_init(&tree_cl); - - while(!error) - { - /*read the code length codes out of 3 * (amount of code length codes) bits*/ - - bitlen_cl = (unsigned*)lodepng_malloc(NUM_CODE_LENGTH_CODES * sizeof(unsigned)); - if(!bitlen_cl) ERROR_BREAK(83 /*alloc fail*/); - - for(i = 0; i < NUM_CODE_LENGTH_CODES; i++) - { - if(i < HCLEN) bitlen_cl[CLCL_ORDER[i]] = readBitsFromStream(bp, in, 3); - else bitlen_cl[CLCL_ORDER[i]] = 0; /*if not, it must stay 0*/ - } - - error = HuffmanTree_makeFromLengths(&tree_cl, bitlen_cl, NUM_CODE_LENGTH_CODES, 7); - if(error) break; - - /*now we can use this tree to read the lengths for the tree that this function will return*/ - bitlen_ll = (unsigned*)lodepng_malloc(NUM_DEFLATE_CODE_SYMBOLS * sizeof(unsigned)); - bitlen_d = (unsigned*)lodepng_malloc(NUM_DISTANCE_SYMBOLS * sizeof(unsigned)); - if(!bitlen_ll || !bitlen_d) ERROR_BREAK(83 /*alloc fail*/); - for(i = 0; i < NUM_DEFLATE_CODE_SYMBOLS; i++) bitlen_ll[i] = 0; - for(i = 0; i < NUM_DISTANCE_SYMBOLS; i++) bitlen_d[i] = 0; - - /*i is the current symbol we're reading in the part that contains the code lengths of lit/len and dist codes*/ - i = 0; - while(i < HLIT + HDIST) - { - unsigned code = huffmanDecodeSymbol(in, bp, &tree_cl, inbitlength); - if(code <= 15) /*a length code*/ - { - if(i < HLIT) bitlen_ll[i] = code; - else bitlen_d[i - HLIT] = code; - i++; - } - else if(code == 16) /*repeat previous*/ - { - unsigned replength = 3; /*read in the 2 bits that indicate repeat length (3-6)*/ - unsigned value; /*set value to the previous code*/ - - if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - if (i == 0) ERROR_BREAK(54); /*can't repeat previous if i is 0*/ - - replength += readBitsFromStream(bp, in, 2); - - if(i < HLIT + 1) value = bitlen_ll[i - 1]; - else value = bitlen_d[i - HLIT - 1]; - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; n++) - { - if(i >= HLIT + HDIST) ERROR_BREAK(13); /*error: i is larger than the amount of codes*/ - if(i < HLIT) bitlen_ll[i] = value; - else bitlen_d[i - HLIT] = value; - i++; - } - } - else if(code == 17) /*repeat "0" 3-10 times*/ - { - unsigned replength = 3; /*read in the bits that indicate repeat length*/ - if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - - replength += readBitsFromStream(bp, in, 3); - - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; n++) - { - if(i >= HLIT + HDIST) ERROR_BREAK(14); /*error: i is larger than the amount of codes*/ - - if(i < HLIT) bitlen_ll[i] = 0; - else bitlen_d[i - HLIT] = 0; - i++; - } - } - else if(code == 18) /*repeat "0" 11-138 times*/ - { - unsigned replength = 11; /*read in the bits that indicate repeat length*/ - if(*bp >= inbitlength) ERROR_BREAK(50); /*error, bit pointer jumps past memory*/ - - replength += readBitsFromStream(bp, in, 7); - - /*repeat this value in the next lengths*/ - for(n = 0; n < replength; n++) - { - if(i >= HLIT + HDIST) ERROR_BREAK(15); /*error: i is larger than the amount of codes*/ - - if(i < HLIT) bitlen_ll[i] = 0; - else bitlen_d[i - HLIT] = 0; - i++; - } - } - else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ - { - if(code == (unsigned)(-1)) - { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - error = (*bp) > inbitlength ? 10 : 11; - } - else error = 16; /*unexisting code, this can never happen*/ - break; - } - } - if(error) break; - - if(bitlen_ll[256] == 0) ERROR_BREAK(64); /*the length of the end code 256 must be larger than 0*/ - - /*now we've finally got HLIT and HDIST, so generate the code trees, and the function is done*/ - error = HuffmanTree_makeFromLengths(tree_ll, bitlen_ll, NUM_DEFLATE_CODE_SYMBOLS, 15); - if(error) break; - error = HuffmanTree_makeFromLengths(tree_d, bitlen_d, NUM_DISTANCE_SYMBOLS, 15); - - break; /*end of error-while*/ - } - - lodepng_free(bitlen_cl); - lodepng_free(bitlen_ll); - lodepng_free(bitlen_d); - HuffmanTree_cleanup(&tree_cl); - - return error; -} - -/*inflate a block with dynamic of fixed Huffman tree*/ -static unsigned inflateHuffmanBlock(ucvector* out, const unsigned char* in, size_t* bp, - size_t* pos, size_t inlength, unsigned btype) -{ - unsigned error = 0; - HuffmanTree tree_ll; /*the huffman tree for literal and length codes*/ - HuffmanTree tree_d; /*the huffman tree for distance codes*/ - size_t inbitlength = inlength * 8; - - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - - if(btype == 1) getTreeInflateFixed(&tree_ll, &tree_d); - else if(btype == 2) error = getTreeInflateDynamic(&tree_ll, &tree_d, in, bp, inlength); - - while(!error) /*decode all symbols until end reached, breaks at end code*/ - { - /*code_ll is literal, length or end code*/ - unsigned code_ll = huffmanDecodeSymbol(in, bp, &tree_ll, inbitlength); - if(code_ll <= 255) /*literal symbol*/ - { - if((*pos) >= out->size) - { - /*reserve more room at once*/ - if(!ucvector_resize(out, ((*pos) + 1) * 2)) ERROR_BREAK(83 /*alloc fail*/); - } - out->data[(*pos)] = (unsigned char)(code_ll); - (*pos)++; - } - else if(code_ll >= FIRST_LENGTH_CODE_INDEX && code_ll <= LAST_LENGTH_CODE_INDEX) /*length code*/ - { - unsigned code_d, distance; - unsigned numextrabits_l, numextrabits_d; /*extra bits for length and distance*/ - size_t start, forward, backward, length; - - /*part 1: get length base*/ - length = LENGTHBASE[code_ll - FIRST_LENGTH_CODE_INDEX]; - - /*part 2: get extra bits and add the value of that to length*/ - numextrabits_l = LENGTHEXTRA[code_ll - FIRST_LENGTH_CODE_INDEX]; - if(*bp >= inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ - length += readBitsFromStream(bp, in, numextrabits_l); - - /*part 3: get distance code*/ - code_d = huffmanDecodeSymbol(in, bp, &tree_d, inbitlength); - if(code_d > 29) - { - if(code_ll == (unsigned)(-1)) /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ - { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - error = (*bp) > inlength * 8 ? 10 : 11; - } - else error = 18; /*error: invalid distance code (30-31 are never used)*/ - break; - } - distance = DISTANCEBASE[code_d]; - - /*part 4: get extra bits from distance*/ - numextrabits_d = DISTANCEEXTRA[code_d]; - if(*bp >= inbitlength) ERROR_BREAK(51); /*error, bit pointer will jump past memory*/ - - distance += readBitsFromStream(bp, in, numextrabits_d); - - /*part 5: fill in all the out[n] values based on the length and dist*/ - start = (*pos); - if(distance > start) ERROR_BREAK(52); /*too long backward distance*/ - backward = start - distance; - if((*pos) + length >= out->size) - { - /*reserve more room at once*/ - if(!ucvector_resize(out, ((*pos) + length) * 2)) ERROR_BREAK(83 /*alloc fail*/); - } - - for(forward = 0; forward < length; forward++) - { - out->data[(*pos)] = out->data[backward]; - (*pos)++; - backward++; - if(backward >= start) backward = start - distance; - } - } - else if(code_ll == 256) - { - break; /*end code, break the loop*/ - } - else /*if(code == (unsigned)(-1))*/ /*huffmanDecodeSymbol returns (unsigned)(-1) in case of error*/ - { - /*return error code 10 or 11 depending on the situation that happened in huffmanDecodeSymbol - (10=no endcode, 11=wrong jump outside of tree)*/ - error = (*bp) > inlength * 8 ? 10 : 11; - break; - } - } - - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - - return error; -} - -static unsigned inflateNoCompression(ucvector* out, const unsigned char* in, size_t* bp, size_t* pos, size_t inlength) -{ - /*go to first boundary of byte*/ - size_t p; - unsigned LEN, NLEN, n, error = 0; - while(((*bp) & 0x7) != 0) (*bp)++; - p = (*bp) / 8; /*byte position*/ - - /*read LEN (2 bytes) and NLEN (2 bytes)*/ - if(p >= inlength - 4) return 52; /*error, bit pointer will jump past memory*/ - LEN = in[p] + 256 * in[p + 1]; p += 2; - NLEN = in[p] + 256 * in[p + 1]; p += 2; - - /*check if 16-bit NLEN is really the one's complement of LEN*/ - if(LEN + NLEN != 65535) return 21; /*error: NLEN is not one's complement of LEN*/ - - if((*pos) + LEN >= out->size) - { - if(!ucvector_resize(out, (*pos) + LEN)) return 83; /*alloc fail*/ - } - - /*read the literal data: LEN bytes are now stored in the out buffer*/ - if(p + LEN > inlength) return 23; /*error: reading outside of in buffer*/ - for(n = 0; n < LEN; n++) out->data[(*pos)++] = in[p++]; - - (*bp) = p * 8; - - return error; -} - -static unsigned lodepng_inflatev(ucvector* out, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) -{ - /*bit pointer in the "in" data, current byte is bp >> 3, current bit is bp & 0x7 (from lsb to msb of the byte)*/ - size_t bp = 0; - unsigned BFINAL = 0; - size_t pos = 0; /*byte position in the out buffer*/ - - unsigned error = 0; - - (void)settings; - - while(!BFINAL) - { - unsigned BTYPE; - if(bp + 2 >= insize * 8) return 52; /*error, bit pointer will jump past memory*/ - BFINAL = readBitFromStream(&bp, in); - BTYPE = 1 * readBitFromStream(&bp, in); - BTYPE += 2 * readBitFromStream(&bp, in); - - if(BTYPE == 3) return 20; /*error: invalid BTYPE*/ - else if(BTYPE == 0) error = inflateNoCompression(out, in, &bp, &pos, insize); /*no compression*/ - else error = inflateHuffmanBlock(out, in, &bp, &pos, insize, BTYPE); /*compression, BTYPE 01 or 10*/ - - if(error) return error; - } - - /*Only now we know the true size of out, resize it to that*/ - if(!ucvector_resize(out, pos)) error = 83; /*alloc fail*/ - - return error; -} - -unsigned lodepng_inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) -{ - unsigned error; - ucvector v; - ucvector_init_buffer(&v, *out, *outsize); - error = lodepng_inflatev(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -static unsigned inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings) -{ - if(settings->custom_inflate) - { - return settings->custom_inflate(out, outsize, in, insize, settings); - } - else - { - return lodepng_inflate(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Deflator (Compressor) / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static const size_t MAX_SUPPORTED_DEFLATE_LENGTH = 258; - -/*bitlen is the size in bits of the code*/ -static void addHuffmanSymbol(size_t* bp, ucvector* compressed, unsigned code, unsigned bitlen) -{ - addBitsToStreamReversed(bp, compressed, code, bitlen); -} - -/*search the index in the array, that has the largest value smaller than or equal to the given value, -given array must be sorted (if no value is smaller, it returns the size of the given array)*/ -static size_t searchCodeIndex(const unsigned* array, size_t array_size, size_t value) -{ - /*linear search implementation*/ - /*for(size_t i = 1; i < array_size; i++) if(array[i] > value) return i - 1; - return array_size - 1;*/ - - /*binary search implementation (not that much faster) (precondition: array_size > 0)*/ - size_t left = 1; - size_t right = array_size - 1; - while(left <= right) - { - size_t mid = (left + right) / 2; - if(array[mid] <= value) left = mid + 1; /*the value to find is more to the right*/ - else if(array[mid - 1] > value) right = mid - 1; /*the value to find is more to the left*/ - else return mid - 1; - } - return array_size - 1; -} - -static void addLengthDistance(uivector* values, size_t length, size_t distance) -{ - /*values in encoded vector are those used by deflate: - 0-255: literal bytes - 256: end - 257-285: length/distance pair (length code, followed by extra length bits, distance code, extra distance bits) - 286-287: invalid*/ - - unsigned length_code = (unsigned)searchCodeIndex(LENGTHBASE, 29, length); - unsigned extra_length = (unsigned)(length - LENGTHBASE[length_code]); - unsigned dist_code = (unsigned)searchCodeIndex(DISTANCEBASE, 30, distance); - unsigned extra_distance = (unsigned)(distance - DISTANCEBASE[dist_code]); - - uivector_push_back(values, length_code + FIRST_LENGTH_CODE_INDEX); - uivector_push_back(values, extra_length); - uivector_push_back(values, dist_code); - uivector_push_back(values, extra_distance); -} - -static const unsigned HASH_NUM_VALUES = 65536; -static const unsigned HASH_NUM_CHARACTERS = 3; -static const unsigned HASH_SHIFT = 2; -/* -The HASH_NUM_CHARACTERS value is used to make encoding faster by using longer -sequences to generate a hash value from the stream bytes. Setting it to 3 -gives exactly the same compression as the brute force method, since deflate's -run length encoding starts with lengths of 3. Setting it to higher values, -like 6, can make the encoding faster (not always though!), but will cause the -encoding to miss any length between 3 and this value, so that the compression -may be worse (but this can vary too depending on the image, sometimes it is -even a bit better instead). -The HASH_NUM_VALUES is the amount of unique possible hash values that -combinations of bytes can give, the higher it is the more memory is needed, but -if it's too low the advantage of hashing is gone. -*/ - -typedef struct Hash -{ - int* head; /*hash value to head circular pos*/ - int* val; /*circular pos to hash value*/ - /*circular pos to prev circular pos*/ - unsigned short* chain; - unsigned short* zeros; -} Hash; - -static unsigned hash_init(Hash* hash, unsigned windowsize) -{ - unsigned i; - hash->head = (int*)lodepng_malloc(sizeof(int) * HASH_NUM_VALUES); - hash->val = (int*)lodepng_malloc(sizeof(int) * windowsize); - hash->chain = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - hash->zeros = (unsigned short*)lodepng_malloc(sizeof(unsigned short) * windowsize); - - if(!hash->head || !hash->val || !hash->chain || !hash->zeros) return 83; /*alloc fail*/ - - /*initialize hash table*/ - for(i = 0; i < HASH_NUM_VALUES; i++) hash->head[i] = -1; - for(i = 0; i < windowsize; i++) hash->val[i] = -1; - for(i = 0; i < windowsize; i++) hash->chain[i] = i; /*same value as index indicates uninitialized*/ - - return 0; -} - -static void hash_cleanup(Hash* hash) -{ - lodepng_free(hash->head); - lodepng_free(hash->val); - lodepng_free(hash->chain); - lodepng_free(hash->zeros); -} - -static unsigned getHash(const unsigned char* data, size_t size, size_t pos) -{ - unsigned result = 0; - size_t amount, i; - if(pos >= size) return 0; - amount = HASH_NUM_CHARACTERS; - if(pos + amount >= size) amount = size - pos; - for(i = 0; i < amount; i++) result ^= (data[pos + i] << (i * HASH_SHIFT)); - return result % HASH_NUM_VALUES; -} - -static unsigned countZeros(const unsigned char* data, size_t size, size_t pos) -{ - const unsigned char* start = data + pos; - const unsigned char* end = start + MAX_SUPPORTED_DEFLATE_LENGTH; - if(end > data + size) end = data + size; - data = start; - while (data != end && *data == 0) data++; - /*subtracting two addresses returned as 32-bit number (max value is MAX_SUPPORTED_DEFLATE_LENGTH)*/ - return (unsigned)(data - start); -} - -static void updateHashChain(Hash* hash, size_t pos, int hashval, unsigned windowsize) -{ - unsigned wpos = pos % windowsize; - hash->val[wpos] = hashval; - if(hash->head[hashval] != -1) hash->chain[wpos] = hash->head[hashval]; - hash->head[hashval] = wpos; -} - -/* -LZ77-encode the data. Return value is error code. The input are raw bytes, the output -is in the form of unsigned integers with codes representing for example literal bytes, or -length/distance pairs. -It uses a hash table technique to let it encode faster. When doing LZ77 encoding, a -sliding window (of windowsize) is used, and all past bytes in that window can be used as -the "dictionary". A brute force search through all possible distances would be slow, and -this hash technique is one out of several ways to speed this up. -*/ -static unsigned encodeLZ77(uivector* out, Hash* hash, - const unsigned char* in, size_t inpos, size_t insize, unsigned windowsize, - unsigned minmatch, unsigned nicematch, unsigned lazymatching) -{ - unsigned short numzeros = 0; - int usezeros = windowsize >= 8192; /*for small window size, the 'max chain length' optimization does a better job*/ - unsigned pos, i, error = 0; - /*for large window lengths, assume the user wants no compression loss. Otherwise, max hash chain length speedup.*/ - unsigned maxchainlength = windowsize >= 8192 ? windowsize : windowsize / 8; - unsigned maxlazymatch = windowsize >= 8192 ? MAX_SUPPORTED_DEFLATE_LENGTH : 64; - - if(!error) - { - unsigned offset; /*the offset represents the distance in LZ77 terminology*/ - unsigned length; - unsigned lazy = 0; - unsigned lazylength = 0, lazyoffset = 0; - unsigned hashval; - unsigned current_offset, current_length; - const unsigned char *lastptr, *foreptr, *backptr; - unsigned short hashpos, prevpos; - - for(pos = inpos; pos < insize; pos++) - { - size_t wpos = pos % windowsize; /*position for in 'circular' hash buffers*/ - - hashval = getHash(in, insize, pos); - updateHashChain(hash, pos, hashval, windowsize); - - if(usezeros && hashval == 0) - { - numzeros = countZeros(in, insize, pos); - hash->zeros[wpos] = numzeros; - } - - /*the length and offset found for the current position*/ - length = 0; - offset = 0; - - prevpos = hash->head[hashval]; - hashpos = hash->chain[prevpos]; - - lastptr = &in[insize < pos + MAX_SUPPORTED_DEFLATE_LENGTH ? insize : pos + MAX_SUPPORTED_DEFLATE_LENGTH]; - - /*search for the longest string*/ - if(hash->val[wpos] == (int)hashval) - { - unsigned chainlength = 0; - for(;;) - { - /*stop when went completely around the circular buffer*/ - if(prevpos < wpos && hashpos > prevpos && hashpos <= wpos) break; - if(prevpos > wpos && (hashpos <= wpos || hashpos > prevpos)) break; - if(chainlength++ >= maxchainlength) break; - - current_offset = hashpos <= wpos ? wpos - hashpos : wpos - hashpos + windowsize; - if(current_offset > 0) - { - /*test the next characters*/ - foreptr = &in[pos]; - backptr = &in[pos - current_offset]; - - /*common case in PNGs is lots of zeros. Quickly skip over them as a speedup*/ - if(usezeros && hashval == 0 && hash->val[hashpos] == 0 /*hashval[hashpos] may be out of date*/) - { - unsigned short skip = hash->zeros[hashpos]; - if(skip > numzeros) skip = numzeros; - backptr += skip; - foreptr += skip; - } - - /* multiple checks at once per array bounds check */ - while(foreptr != lastptr && *backptr == *foreptr) /*maximum supported length by deflate is max length*/ - { - ++backptr; - ++foreptr; - } - current_length = (unsigned)(foreptr - &in[pos]); - - if(current_length > length) - { - length = current_length; /*the longest length*/ - offset = current_offset; /*the offset that is related to this longest length*/ - /*jump out once a length of max length is found (speed gain)*/ - if(current_length >= nicematch || current_length == MAX_SUPPORTED_DEFLATE_LENGTH) break; - } - } - - if(hashpos == hash->chain[hashpos]) break; - - prevpos = hashpos; - hashpos = hash->chain[hashpos]; - } - } - - if(lazymatching) - { - if(!lazy && length >= 3 && length <= maxlazymatch && length < MAX_SUPPORTED_DEFLATE_LENGTH) - { - lazy = 1; - lazylength = length; - lazyoffset = offset; - continue; /*try the next byte*/ - } - if(lazy) - { - lazy = 0; - if(pos == 0) ERROR_BREAK(81); - if(length > lazylength + 1) - { - /*push the previous character as literal*/ - if(!uivector_push_back(out, in[pos - 1])) ERROR_BREAK(83 /*alloc fail*/); - } - else - { - length = lazylength; - offset = lazyoffset; - hash->head[hashval] = -1; /*the same hashchain update will be done, this ensures no wrong alteration*/ - pos--; - } - } - } - if(length >= 3 && offset > windowsize) ERROR_BREAK(86 /*too big (or overflown negative) offset*/); - - /**encode it as length/distance pair or literal value**/ - if(length < 3) /*only lengths of 3 or higher are supported as length/distance pair*/ - { - if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); - } - else if(length < minmatch || (length == 3 && offset > 4096)) - { - /*compensate for the fact that longer offsets have more extra bits, a - length of only 3 may be not worth it then*/ - if(!uivector_push_back(out, in[pos])) ERROR_BREAK(83 /*alloc fail*/); - } - else - { - addLengthDistance(out, length, offset); - for(i = 1; i < length; i++) - { - pos++; - hashval = getHash(in, insize, pos); - updateHashChain(hash, pos, hashval, windowsize); - if(usezeros && hashval == 0) - { - hash->zeros[pos % windowsize] = countZeros(in, insize, pos); - } - } - } - - } /*end of the loop through each character of input*/ - } /*end of "if(!error)"*/ - - return error; -} - -/* /////////////////////////////////////////////////////////////////////////// */ - -static unsigned deflateNoCompression(ucvector* out, const unsigned char* data, size_t datasize) -{ - /*non compressed deflate block data: 1 bit BFINAL,2 bits BTYPE,(5 bits): it jumps to start of next byte, - 2 bytes LEN, 2 bytes NLEN, LEN bytes literal DATA*/ - - size_t i, j, numdeflateblocks = (datasize + 65534) / 65535; - unsigned datapos = 0; - for(i = 0; i < numdeflateblocks; i++) - { - unsigned BFINAL, BTYPE, LEN, NLEN; - unsigned char firstbyte; - - BFINAL = (i == numdeflateblocks - 1); - BTYPE = 0; - - firstbyte = (unsigned char)(BFINAL + ((BTYPE & 1) << 1) + ((BTYPE & 2) << 1)); - ucvector_push_back(out, firstbyte); - - LEN = 65535; - if(datasize - datapos < 65535) LEN = (unsigned)datasize - datapos; - NLEN = 65535 - LEN; - - ucvector_push_back(out, (unsigned char)(LEN % 256)); - ucvector_push_back(out, (unsigned char)(LEN / 256)); - ucvector_push_back(out, (unsigned char)(NLEN % 256)); - ucvector_push_back(out, (unsigned char)(NLEN / 256)); - - /*Decompressed data*/ - for(j = 0; j < 65535 && datapos < datasize; j++) - { - ucvector_push_back(out, data[datapos++]); - } - } - - return 0; -} - -/* -write the lz77-encoded data, which has lit, len and dist codes, to compressed stream using huffman trees. -tree_ll: the tree for lit and len codes. -tree_d: the tree for distance codes. -*/ -static void writeLZ77data(size_t* bp, ucvector* out, const uivector* lz77_encoded, - const HuffmanTree* tree_ll, const HuffmanTree* tree_d) -{ - size_t i = 0; - for(i = 0; i < lz77_encoded->size; i++) - { - unsigned val = lz77_encoded->data[i]; - addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_ll, val), HuffmanTree_getLength(tree_ll, val)); - if(val > 256) /*for a length code, 3 more things have to be added*/ - { - unsigned length_index = val - FIRST_LENGTH_CODE_INDEX; - unsigned n_length_extra_bits = LENGTHEXTRA[length_index]; - unsigned length_extra_bits = lz77_encoded->data[++i]; - - unsigned distance_code = lz77_encoded->data[++i]; - - unsigned distance_index = distance_code; - unsigned n_distance_extra_bits = DISTANCEEXTRA[distance_index]; - unsigned distance_extra_bits = lz77_encoded->data[++i]; - - addBitsToStream(bp, out, length_extra_bits, n_length_extra_bits); - addHuffmanSymbol(bp, out, HuffmanTree_getCode(tree_d, distance_code), - HuffmanTree_getLength(tree_d, distance_code)); - addBitsToStream(bp, out, distance_extra_bits, n_distance_extra_bits); - } - } -} - -/*Deflate for a block of type "dynamic", that is, with freely, optimally, created huffman trees*/ -static unsigned deflateDynamic(ucvector* out, size_t* bp, Hash* hash, - const unsigned char* data, size_t datapos, size_t dataend, - const LodePNGCompressSettings* settings, int final) -{ - unsigned error = 0; - - /* - A block is compressed as follows: The PNG data is lz77 encoded, resulting in - literal bytes and length/distance pairs. This is then huffman compressed with - two huffman trees. One huffman tree is used for the lit and len values ("ll"), - another huffman tree is used for the dist values ("d"). These two trees are - stored using their code lengths, and to compress even more these code lengths - are also run-length encoded and huffman compressed. This gives a huffman tree - of code lengths "cl". The code lenghts used to describe this third tree are - the code length code lengths ("clcl"). - */ - - /*The lz77 encoded data, represented with integers since there will also be length and distance codes in it*/ - uivector lz77_encoded; - HuffmanTree tree_ll; /*tree for lit,len values*/ - HuffmanTree tree_d; /*tree for distance codes*/ - HuffmanTree tree_cl; /*tree for encoding the code lengths representing tree_ll and tree_d*/ - uivector frequencies_ll; /*frequency of lit,len codes*/ - uivector frequencies_d; /*frequency of dist codes*/ - uivector frequencies_cl; /*frequency of code length codes*/ - uivector bitlen_lld; /*lit,len,dist code lenghts (int bits), literally (without repeat codes).*/ - uivector bitlen_lld_e; /*bitlen_lld encoded with repeat codes (this is a rudemtary run length compression)*/ - /*bitlen_cl is the code length code lengths ("clcl"). The bit lengths of codes to represent tree_cl - (these are written as is in the file, it would be crazy to compress these using yet another huffman - tree that needs to be represented by yet another set of code lengths)*/ - uivector bitlen_cl; - size_t datasize = dataend - datapos; - - /* - Due to the huffman compression of huffman tree representations ("two levels"), there are some anologies: - bitlen_lld is to tree_cl what data is to tree_ll and tree_d. - bitlen_lld_e is to bitlen_lld what lz77_encoded is to data. - bitlen_cl is to bitlen_lld_e what bitlen_lld is to lz77_encoded. - */ - - unsigned BFINAL = final; - size_t numcodes_ll, numcodes_d, i; - unsigned HLIT, HDIST, HCLEN; - - uivector_init(&lz77_encoded); - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - HuffmanTree_init(&tree_cl); - uivector_init(&frequencies_ll); - uivector_init(&frequencies_d); - uivector_init(&frequencies_cl); - uivector_init(&bitlen_lld); - uivector_init(&bitlen_lld_e); - uivector_init(&bitlen_cl); - - /*This while loop never loops due to a break at the end, it is here to - allow breaking out of it to the cleanup phase on error conditions.*/ - while(!error) - { - if(settings->use_lz77) - { - error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, - settings->minmatch, settings->nicematch, settings->lazymatching); - if(error) break; - } - else - { - if(!uivector_resize(&lz77_encoded, datasize)) ERROR_BREAK(83 /*alloc fail*/); - for(i = datapos; i < dataend; i++) lz77_encoded.data[i] = data[i]; /*no LZ77, but still will be Huffman compressed*/ - } - - if(!uivector_resizev(&frequencies_ll, 286, 0)) ERROR_BREAK(83 /*alloc fail*/); - if(!uivector_resizev(&frequencies_d, 30, 0)) ERROR_BREAK(83 /*alloc fail*/); - - /*Count the frequencies of lit, len and dist codes*/ - for(i = 0; i < lz77_encoded.size; i++) - { - unsigned symbol = lz77_encoded.data[i]; - frequencies_ll.data[symbol]++; - if(symbol > 256) - { - unsigned dist = lz77_encoded.data[i + 2]; - frequencies_d.data[dist]++; - i += 3; - } - } - frequencies_ll.data[256] = 1; /*there will be exactly 1 end code, at the end of the block*/ - - /*Make both huffman trees, one for the lit and len codes, one for the dist codes*/ - error = HuffmanTree_makeFromFrequencies(&tree_ll, frequencies_ll.data, 257, frequencies_ll.size, 15); - if(error) break; - /*2, not 1, is chosen for mincodes: some buggy PNG decoders require at least 2 symbols in the dist tree*/ - error = HuffmanTree_makeFromFrequencies(&tree_d, frequencies_d.data, 2, frequencies_d.size, 15); - if(error) break; - - numcodes_ll = tree_ll.numcodes; if(numcodes_ll > 286) numcodes_ll = 286; - numcodes_d = tree_d.numcodes; if(numcodes_d > 30) numcodes_d = 30; - /*store the code lengths of both generated trees in bitlen_lld*/ - for(i = 0; i < numcodes_ll; i++) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_ll, (unsigned)i)); - for(i = 0; i < numcodes_d; i++) uivector_push_back(&bitlen_lld, HuffmanTree_getLength(&tree_d, (unsigned)i)); - - /*run-length compress bitlen_ldd into bitlen_lld_e by using repeat codes 16 (copy length 3-6 times), - 17 (3-10 zeroes), 18 (11-138 zeroes)*/ - for(i = 0; i < (unsigned)bitlen_lld.size; i++) - { - unsigned j = 0; /*amount of repititions*/ - while(i + j + 1 < (unsigned)bitlen_lld.size && bitlen_lld.data[i + j + 1] == bitlen_lld.data[i]) j++; - - if(bitlen_lld.data[i] == 0 && j >= 2) /*repeat code for zeroes*/ - { - j++; /*include the first zero*/ - if(j <= 10) /*repeat code 17 supports max 10 zeroes*/ - { - uivector_push_back(&bitlen_lld_e, 17); - uivector_push_back(&bitlen_lld_e, j - 3); - } - else /*repeat code 18 supports max 138 zeroes*/ - { - if(j > 138) j = 138; - uivector_push_back(&bitlen_lld_e, 18); - uivector_push_back(&bitlen_lld_e, j - 11); - } - i += (j - 1); - } - else if(j >= 3) /*repeat code for value other than zero*/ - { - size_t k; - unsigned num = j / 6, rest = j % 6; - uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); - for(k = 0; k < num; k++) - { - uivector_push_back(&bitlen_lld_e, 16); - uivector_push_back(&bitlen_lld_e, 6 - 3); - } - if(rest >= 3) - { - uivector_push_back(&bitlen_lld_e, 16); - uivector_push_back(&bitlen_lld_e, rest - 3); - } - else j -= rest; - i += j; - } - else /*too short to benefit from repeat code*/ - { - uivector_push_back(&bitlen_lld_e, bitlen_lld.data[i]); - } - } - - /*generate tree_cl, the huffmantree of huffmantrees*/ - - if(!uivector_resizev(&frequencies_cl, NUM_CODE_LENGTH_CODES, 0)) ERROR_BREAK(83 /*alloc fail*/); - for(i = 0; i < bitlen_lld_e.size; i++) - { - frequencies_cl.data[bitlen_lld_e.data[i]]++; - /*after a repeat code come the bits that specify the number of repetitions, - those don't need to be in the frequencies_cl calculation*/ - if(bitlen_lld_e.data[i] >= 16) i++; - } - - error = HuffmanTree_makeFromFrequencies(&tree_cl, frequencies_cl.data, - frequencies_cl.size, frequencies_cl.size, 7); - if(error) break; - - if(!uivector_resize(&bitlen_cl, tree_cl.numcodes)) ERROR_BREAK(83 /*alloc fail*/); - for(i = 0; i < tree_cl.numcodes; i++) - { - /*lenghts of code length tree is in the order as specified by deflate*/ - bitlen_cl.data[i] = HuffmanTree_getLength(&tree_cl, CLCL_ORDER[i]); - } - while(bitlen_cl.data[bitlen_cl.size - 1] == 0 && bitlen_cl.size > 4) - { - /*remove zeros at the end, but minimum size must be 4*/ - if(!uivector_resize(&bitlen_cl, bitlen_cl.size - 1)) ERROR_BREAK(83 /*alloc fail*/); - } - if(error) break; - - /* - Write everything into the output - - After the BFINAL and BTYPE, the dynamic block consists out of the following: - - 5 bits HLIT, 5 bits HDIST, 4 bits HCLEN - - (HCLEN+4)*3 bits code lengths of code length alphabet - - HLIT + 257 code lenghts of lit/length alphabet (encoded using the code length - alphabet, + possible repetition codes 16, 17, 18) - - HDIST + 1 code lengths of distance alphabet (encoded using the code length - alphabet, + possible repetition codes 16, 17, 18) - - compressed data - - 256 (end code) - */ - - /*Write block type*/ - addBitToStream(bp, out, BFINAL); - addBitToStream(bp, out, 0); /*first bit of BTYPE "dynamic"*/ - addBitToStream(bp, out, 1); /*second bit of BTYPE "dynamic"*/ - - /*write the HLIT, HDIST and HCLEN values*/ - HLIT = (unsigned)(numcodes_ll - 257); - HDIST = (unsigned)(numcodes_d - 1); - HCLEN = (unsigned)bitlen_cl.size - 4; - /*trim zeroes for HCLEN. HLIT and HDIST were already trimmed at tree creation*/ - while(!bitlen_cl.data[HCLEN + 4 - 1] && HCLEN > 0) HCLEN--; - addBitsToStream(bp, out, HLIT, 5); - addBitsToStream(bp, out, HDIST, 5); - addBitsToStream(bp, out, HCLEN, 4); - - /*write the code lenghts of the code length alphabet*/ - for(i = 0; i < HCLEN + 4; i++) addBitsToStream(bp, out, bitlen_cl.data[i], 3); - - /*write the lenghts of the lit/len AND the dist alphabet*/ - for(i = 0; i < bitlen_lld_e.size; i++) - { - addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_cl, bitlen_lld_e.data[i]), - HuffmanTree_getLength(&tree_cl, bitlen_lld_e.data[i])); - /*extra bits of repeat codes*/ - if(bitlen_lld_e.data[i] == 16) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 2); - else if(bitlen_lld_e.data[i] == 17) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 3); - else if(bitlen_lld_e.data[i] == 18) addBitsToStream(bp, out, bitlen_lld_e.data[++i], 7); - } - - /*write the compressed data symbols*/ - writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); - /*error: the length of the end code 256 must be larger than 0*/ - if(HuffmanTree_getLength(&tree_ll, 256) == 0) ERROR_BREAK(64); - - /*write the end code*/ - addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); - - break; /*end of error-while*/ - } - - /*cleanup*/ - uivector_cleanup(&lz77_encoded); - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - HuffmanTree_cleanup(&tree_cl); - uivector_cleanup(&frequencies_ll); - uivector_cleanup(&frequencies_d); - uivector_cleanup(&frequencies_cl); - uivector_cleanup(&bitlen_lld_e); - uivector_cleanup(&bitlen_lld); - uivector_cleanup(&bitlen_cl); - - return error; -} - -static unsigned deflateFixed(ucvector* out, size_t* bp, Hash* hash, - const unsigned char* data, - size_t datapos, size_t dataend, - const LodePNGCompressSettings* settings, int final) -{ - HuffmanTree tree_ll; /*tree for literal values and length codes*/ - HuffmanTree tree_d; /*tree for distance codes*/ - - unsigned BFINAL = final; - unsigned error = 0; - size_t i; - - HuffmanTree_init(&tree_ll); - HuffmanTree_init(&tree_d); - - generateFixedLitLenTree(&tree_ll); - generateFixedDistanceTree(&tree_d); - - addBitToStream(bp, out, BFINAL); - addBitToStream(bp, out, 1); /*first bit of BTYPE*/ - addBitToStream(bp, out, 0); /*second bit of BTYPE*/ - - if(settings->use_lz77) /*LZ77 encoded*/ - { - uivector lz77_encoded; - uivector_init(&lz77_encoded); - error = encodeLZ77(&lz77_encoded, hash, data, datapos, dataend, settings->windowsize, - settings->minmatch, settings->nicematch, settings->lazymatching); - if(!error) writeLZ77data(bp, out, &lz77_encoded, &tree_ll, &tree_d); - uivector_cleanup(&lz77_encoded); - } - else /*no LZ77, but still will be Huffman compressed*/ - { - for(i = datapos; i < dataend; i++) - { - addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, data[i]), HuffmanTree_getLength(&tree_ll, data[i])); - } - } - /*add END code*/ - if(!error) addHuffmanSymbol(bp, out, HuffmanTree_getCode(&tree_ll, 256), HuffmanTree_getLength(&tree_ll, 256)); - - /*cleanup*/ - HuffmanTree_cleanup(&tree_ll); - HuffmanTree_cleanup(&tree_d); - - return error; -} - -static unsigned lodepng_deflatev(ucvector* out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) -{ - unsigned error = 0; - size_t i, blocksize, numdeflateblocks; - size_t bp = 0; /*the bit pointer*/ - Hash hash; - - if(settings->btype > 2) return 61; - else if(settings->btype == 0) return deflateNoCompression(out, in, insize); - else if(settings->btype == 1) blocksize = insize; - else /*if(settings->btype == 2)*/ - { - blocksize = insize / 8 + 8; - if(blocksize < 65535) blocksize = 65535; - } - - numdeflateblocks = (insize + blocksize - 1) / blocksize; - if(numdeflateblocks == 0) numdeflateblocks = 1; - - error = hash_init(&hash, settings->windowsize); - if(error) return error; - - for(i = 0; i < numdeflateblocks && !error; i++) - { - int final = i == numdeflateblocks - 1; - size_t start = i * blocksize; - size_t end = start + blocksize; - if(end > insize) end = insize; - - if(settings->btype == 1) error = deflateFixed(out, &bp, &hash, in, start, end, settings, final); - else if(settings->btype == 2) error = deflateDynamic(out, &bp, &hash, in, start, end, settings, final); - } - - hash_cleanup(&hash); - - return error; -} - -unsigned lodepng_deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) -{ - unsigned error; - ucvector v; - ucvector_init_buffer(&v, *out, *outsize); - error = lodepng_deflatev(&v, in, insize, settings); - *out = v.data; - *outsize = v.size; - return error; -} - -static unsigned deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings) -{ - if(settings->custom_deflate) - { - return settings->custom_deflate(out, outsize, in, insize, settings); - } - else - { - return lodepng_deflate(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Adler32 */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static unsigned update_adler32(unsigned adler, const unsigned char* data, unsigned len) -{ - unsigned s1 = adler & 0xffff; - unsigned s2 = (adler >> 16) & 0xffff; - - while(len > 0) - { - /*at least 5550 sums can be done before the sums overflow, saving a lot of module divisions*/ - unsigned amount = len > 5550 ? 5550 : len; - len -= amount; - while(amount > 0) - { - s1 += (*data++); - s2 += s1; - amount--; - } - s1 %= 65521; - s2 %= 65521; - } - - return (s2 << 16) | s1; -} - -/*Return the adler32 of the bytes data[0..len-1]*/ -static unsigned adler32(const unsigned char* data, unsigned len) -{ - return update_adler32(1L, data, len); -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Zlib / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_DECODER - -unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) -{ - unsigned error = 0; - unsigned CM, CINFO, FDICT; - - if(insize < 2) return 53; /*error, size of zlib data too small*/ - /*read information from zlib header*/ - if((in[0] * 256 + in[1]) % 31 != 0) - { - /*error: 256 * in[0] + in[1] must be a multiple of 31, the FCHECK value is supposed to be made that way*/ - return 24; - } - - CM = in[0] & 15; - CINFO = (in[0] >> 4) & 15; - /*FCHECK = in[1] & 31;*/ /*FCHECK is already tested above*/ - FDICT = (in[1] >> 5) & 1; - /*FLEVEL = (in[1] >> 6) & 3;*/ /*FLEVEL is not used here*/ - - if(CM != 8 || CINFO > 7) - { - /*error: only compression method 8: inflate with sliding window of 32k is supported by the PNG spec*/ - return 25; - } - if(FDICT != 0) - { - /*error: the specification of PNG says about the zlib stream: - "The additional flags shall not specify a preset dictionary."*/ - return 26; - } - - error = inflate(out, outsize, in + 2, insize - 2, settings); - if(error) return error; - - if(!settings->ignore_adler32) - { - unsigned ADLER32 = lodepng_read32bitInt(&in[insize - 4]); - unsigned checksum = adler32(*out, (unsigned)(*outsize)); - if(checksum != ADLER32) return 58; /*error, adler checksum not correct, data must be corrupted*/ - } - - return 0; /*no error*/ -} - -static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) -{ - if(settings->custom_zlib) - return settings->custom_zlib(out, outsize, in, insize, settings); - else - return lodepng_zlib_decompress(out, outsize, in, insize, settings); -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER - -unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) -{ - /*initially, *out must be NULL and outsize 0, if you just give some random *out - that's pointing to a non allocated buffer, this'll crash*/ - ucvector outv; - size_t i; - unsigned error; - unsigned char* deflatedata = 0; - size_t deflatesize = 0; - - unsigned ADLER32; - /*zlib data: 1 byte CMF (CM+CINFO), 1 byte FLG, deflate data, 4 byte ADLER32 checksum of the Decompressed data*/ - unsigned CMF = 120; /*0b01111000: CM 8, CINFO 7. With CINFO 7, any window size up to 32768 can be used.*/ - unsigned FLEVEL = 0; - unsigned FDICT = 0; - unsigned CMFFLG = 256 * CMF + FDICT * 32 + FLEVEL * 64; - unsigned FCHECK = 31 - CMFFLG % 31; - CMFFLG += FCHECK; - - /*ucvector-controlled version of the output buffer, for dynamic array*/ - ucvector_init_buffer(&outv, *out, *outsize); - - ucvector_push_back(&outv, (unsigned char)(CMFFLG / 256)); - ucvector_push_back(&outv, (unsigned char)(CMFFLG % 256)); - - error = deflate(&deflatedata, &deflatesize, in, insize, settings); - - if(!error) - { - ADLER32 = adler32(in, (unsigned)insize); - for(i = 0; i < deflatesize; i++) ucvector_push_back(&outv, deflatedata[i]); - lodepng_free(deflatedata); - lodepng_add32bitInt(&outv, ADLER32); - } - - *out = outv.data; - *outsize = outv.size; - - return error; -} - -/* compress using the default or custom zlib function */ -static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) -{ - if(settings->custom_zlib) - { - return settings->custom_zlib(out, outsize, in, insize, settings); - } - else - { - return lodepng_zlib_compress(out, outsize, in, insize, settings); - } -} - -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#else /*no LODEPNG_COMPILE_ZLIB*/ - -#ifdef LODEPNG_COMPILE_DECODER -static unsigned zlib_decompress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGDecompressSettings* settings) -{ - if (!settings->custom_zlib) return 87; /*no custom zlib function provided */ - return settings->custom_zlib(out, outsize, in, insize, settings); -} -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER -static unsigned zlib_compress(unsigned char** out, size_t* outsize, const unsigned char* in, - size_t insize, const LodePNGCompressSettings* settings) -{ - if (!settings->custom_zlib) return 87; /*no custom zlib function provided */ - return settings->custom_zlib(out, outsize, in, insize, settings); -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#endif /*LODEPNG_COMPILE_ZLIB*/ - -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_ENCODER - -/*this is a good tradeoff between speed and compression ratio*/ -#define DEFAULT_WINDOWSIZE 2048 - -void lodepng_compress_settings_init(LodePNGCompressSettings* settings) -{ - /*compress with dynamic huffman tree (not in the mathematical sense, just not the predefined one)*/ - settings->btype = 2; - settings->use_lz77 = 1; - settings->windowsize = DEFAULT_WINDOWSIZE; - settings->minmatch = 3; - settings->nicematch = 128; - settings->lazymatching = 1; - - settings->custom_zlib = 0; - settings->custom_deflate = 0; - settings->custom_context = 0; -} - -const LodePNGCompressSettings lodepng_default_compress_settings = {2, 1, DEFAULT_WINDOWSIZE, 3, 128, 1, 0, 0, 0}; - - -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DECODER - -void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings) -{ - settings->ignore_adler32 = 0; - - settings->custom_zlib = 0; - settings->custom_inflate = 0; - settings->custom_context = 0; -} - -const LodePNGDecompressSettings lodepng_default_decompress_settings = {0, 0, 0, 0}; - -#endif /*LODEPNG_COMPILE_DECODER*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // End of Zlib related code. Begin of PNG related code. // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - -#ifdef LODEPNG_COMPILE_PNG - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / CRC32 / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/* CRC polynomial: 0xedb88320 */ -static unsigned lodepng_crc32_table[256] = { - 0u, 1996959894u, 3993919788u, 2567524794u, 124634137u, 1886057615u, 3915621685u, 2657392035u, - 249268274u, 2044508324u, 3772115230u, 2547177864u, 162941995u, 2125561021u, 3887607047u, 2428444049u, - 498536548u, 1789927666u, 4089016648u, 2227061214u, 450548861u, 1843258603u, 4107580753u, 2211677639u, - 325883990u, 1684777152u, 4251122042u, 2321926636u, 335633487u, 1661365465u, 4195302755u, 2366115317u, - 997073096u, 1281953886u, 3579855332u, 2724688242u, 1006888145u, 1258607687u, 3524101629u, 2768942443u, - 901097722u, 1119000684u, 3686517206u, 2898065728u, 853044451u, 1172266101u, 3705015759u, 2882616665u, - 651767980u, 1373503546u, 3369554304u, 3218104598u, 565507253u, 1454621731u, 3485111705u, 3099436303u, - 671266974u, 1594198024u, 3322730930u, 2970347812u, 795835527u, 1483230225u, 3244367275u, 3060149565u, - 1994146192u, 31158534u, 2563907772u, 4023717930u, 1907459465u, 112637215u, 2680153253u, 3904427059u, - 2013776290u, 251722036u, 2517215374u, 3775830040u, 2137656763u, 141376813u, 2439277719u, 3865271297u, - 1802195444u, 476864866u, 2238001368u, 4066508878u, 1812370925u, 453092731u, 2181625025u, 4111451223u, - 1706088902u, 314042704u, 2344532202u, 4240017532u, 1658658271u, 366619977u, 2362670323u, 4224994405u, - 1303535960u, 984961486u, 2747007092u, 3569037538u, 1256170817u, 1037604311u, 2765210733u, 3554079995u, - 1131014506u, 879679996u, 2909243462u, 3663771856u, 1141124467u, 855842277u, 2852801631u, 3708648649u, - 1342533948u, 654459306u, 3188396048u, 3373015174u, 1466479909u, 544179635u, 3110523913u, 3462522015u, - 1591671054u, 702138776u, 2966460450u, 3352799412u, 1504918807u, 783551873u, 3082640443u, 3233442989u, - 3988292384u, 2596254646u, 62317068u, 1957810842u, 3939845945u, 2647816111u, 81470997u, 1943803523u, - 3814918930u, 2489596804u, 225274430u, 2053790376u, 3826175755u, 2466906013u, 167816743u, 2097651377u, - 4027552580u, 2265490386u, 503444072u, 1762050814u, 4150417245u, 2154129355u, 426522225u, 1852507879u, - 4275313526u, 2312317920u, 282753626u, 1742555852u, 4189708143u, 2394877945u, 397917763u, 1622183637u, - 3604390888u, 2714866558u, 953729732u, 1340076626u, 3518719985u, 2797360999u, 1068828381u, 1219638859u, - 3624741850u, 2936675148u, 906185462u, 1090812512u, 3747672003u, 2825379669u, 829329135u, 1181335161u, - 3412177804u, 3160834842u, 628085408u, 1382605366u, 3423369109u, 3138078467u, 570562233u, 1426400815u, - 3317316542u, 2998733608u, 733239954u, 1555261956u, 3268935591u, 3050360625u, 752459403u, 1541320221u, - 2607071920u, 3965973030u, 1969922972u, 40735498u, 2617837225u, 3943577151u, 1913087877u, 83908371u, - 2512341634u, 3803740692u, 2075208622u, 213261112u, 2463272603u, 3855990285u, 2094854071u, 198958881u, - 2262029012u, 4057260610u, 1759359992u, 534414190u, 2176718541u, 4139329115u, 1873836001u, 414664567u, - 2282248934u, 4279200368u, 1711684554u, 285281116u, 2405801727u, 4167216745u, 1634467795u, 376229701u, - 2685067896u, 3608007406u, 1308918612u, 956543938u, 2808555105u, 3495958263u, 1231636301u, 1047427035u, - 2932959818u, 3654703836u, 1088359270u, 936918000u, 2847714899u, 3736837829u, 1202900863u, 817233897u, - 3183342108u, 3401237130u, 1404277552u, 615818150u, 3134207493u, 3453421203u, 1423857449u, 601450431u, - 3009837614u, 3294710456u, 1567103746u, 711928724u, 3020668471u, 3272380065u, 1510334235u, 755167117u -}; - -/*Return the CRC of the bytes buf[0..len-1].*/ -unsigned lodepng_crc32(const unsigned char* buf, size_t len) -{ - unsigned c = 0xffffffffL; - size_t n; - - for(n = 0; n < len; n++) - { - c = lodepng_crc32_table[(c ^ buf[n]) & 0xff] ^ (c >> 8); - } - return c ^ 0xffffffffL; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Reading and writing single bits and bytes from/to stream for LodePNG / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -static unsigned char readBitFromReversedStream(size_t* bitpointer, const unsigned char* bitstream) -{ - unsigned char result = (unsigned char)((bitstream[(*bitpointer) >> 3] >> (7 - ((*bitpointer) & 0x7))) & 1); - (*bitpointer)++; - return result; -} - -static unsigned readBitsFromReversedStream(size_t* bitpointer, const unsigned char* bitstream, size_t nbits) -{ - unsigned result = 0; - size_t i; - for(i = nbits - 1; i < nbits; i--) - { - result += (unsigned)readBitFromReversedStream(bitpointer, bitstream) << i; - } - return result; -} - -#ifdef LODEPNG_COMPILE_DECODER -static void setBitOfReversedStream0(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) -{ - /*the current bit in bitstream must be 0 for this to work*/ - if(bit) - { - /*earlier bit of huffman code is in a lesser significant bit of an earlier byte*/ - bitstream[(*bitpointer) >> 3] |= (bit << (7 - ((*bitpointer) & 0x7))); - } - (*bitpointer)++; -} -#endif /*LODEPNG_COMPILE_DECODER*/ - -static void setBitOfReversedStream(size_t* bitpointer, unsigned char* bitstream, unsigned char bit) -{ - /*the current bit in bitstream may be 0 or 1 for this to work*/ - if(bit == 0) bitstream[(*bitpointer) >> 3] &= (unsigned char)(~(1 << (7 - ((*bitpointer) & 0x7)))); - else bitstream[(*bitpointer) >> 3] |= (1 << (7 - ((*bitpointer) & 0x7))); - (*bitpointer)++; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG chunks / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -unsigned lodepng_chunk_length(const unsigned char* chunk) -{ - return lodepng_read32bitInt(&chunk[0]); -} - -void lodepng_chunk_type(char type[5], const unsigned char* chunk) -{ - unsigned i; - for(i = 0; i < 4; i++) type[i] = chunk[4 + i]; - type[4] = 0; /*null termination char*/ -} - -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type) -{ - if(strlen(type) != 4) return 0; - return (chunk[4] == type[0] && chunk[5] == type[1] && chunk[6] == type[2] && chunk[7] == type[3]); -} - -unsigned char lodepng_chunk_ancillary(const unsigned char* chunk) -{ - return((chunk[4] & 32) != 0); -} - -unsigned char lodepng_chunk_private(const unsigned char* chunk) -{ - return((chunk[6] & 32) != 0); -} - -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk) -{ - return((chunk[7] & 32) != 0); -} - -unsigned char* lodepng_chunk_data(unsigned char* chunk) -{ - return &chunk[8]; -} - -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk) -{ - return &chunk[8]; -} - -unsigned lodepng_chunk_check_crc(const unsigned char* chunk) -{ - unsigned length = lodepng_chunk_length(chunk); - unsigned CRC = lodepng_read32bitInt(&chunk[length + 8]); - /*the CRC is taken of the data and the 4 chunk type letters, not the length*/ - unsigned checksum = lodepng_crc32(&chunk[4], length + 4); - if(CRC != checksum) return 1; - else return 0; -} - -void lodepng_chunk_generate_crc(unsigned char* chunk) -{ - unsigned length = lodepng_chunk_length(chunk); - unsigned CRC = lodepng_crc32(&chunk[4], length + 4); - lodepng_set32bitInt(chunk + 8 + length, CRC); -} - -unsigned char* lodepng_chunk_next(unsigned char* chunk) -{ - unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; - return &chunk[total_chunk_length]; -} - -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk) -{ - unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; - return &chunk[total_chunk_length]; -} - -unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk) -{ - unsigned i; - unsigned total_chunk_length = lodepng_chunk_length(chunk) + 12; - unsigned char *chunk_start, *new_buffer; - size_t new_length = (*outlength) + total_chunk_length; - if(new_length < total_chunk_length || new_length < (*outlength)) return 77; /*integer overflow happened*/ - - new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); - if(!new_buffer) return 83; /*alloc fail*/ - (*out) = new_buffer; - (*outlength) = new_length; - chunk_start = &(*out)[new_length - total_chunk_length]; - - for(i = 0; i < total_chunk_length; i++) chunk_start[i] = chunk[i]; - - return 0; -} - -unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, - const char* type, const unsigned char* data) -{ - unsigned i; - unsigned char *chunk, *new_buffer; - size_t new_length = (*outlength) + length + 12; - if(new_length < length + 12 || new_length < (*outlength)) return 77; /*integer overflow happened*/ - new_buffer = (unsigned char*)lodepng_realloc(*out, new_length); - if(!new_buffer) return 83; /*alloc fail*/ - (*out) = new_buffer; - (*outlength) = new_length; - chunk = &(*out)[(*outlength) - length - 12]; - - /*1: length*/ - lodepng_set32bitInt(chunk, (unsigned)length); - - /*2: chunk name (4 letters)*/ - chunk[4] = type[0]; - chunk[5] = type[1]; - chunk[6] = type[2]; - chunk[7] = type[3]; - - /*3: the data*/ - for(i = 0; i < length; i++) chunk[8 + i] = data[i]; - - /*4: CRC (of the chunkname characters and the data)*/ - lodepng_chunk_generate_crc(chunk); - - return 0; -} - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / Color types and such / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*return type is a LodePNG error code*/ -static unsigned checkColorValidity(LodePNGColorType colortype, unsigned bd) /*bd = bitdepth*/ -{ - switch(colortype) - { - case 0: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 || bd == 16)) return 37; break; /*grey*/ - case 2: if(!( bd == 8 || bd == 16)) return 37; break; /*RGB*/ - case 3: if(!(bd == 1 || bd == 2 || bd == 4 || bd == 8 )) return 37; break; /*palette*/ - case 4: if(!( bd == 8 || bd == 16)) return 37; break; /*grey + alpha*/ - case 6: if(!( bd == 8 || bd == 16)) return 37; break; /*RGBA*/ - default: return 31; - } - return 0; /*allowed color type / bits combination*/ -} - -static unsigned getNumColorChannels(LodePNGColorType colortype) -{ - switch(colortype) - { - case 0: return 1; /*grey*/ - case 2: return 3; /*RGB*/ - case 3: return 1; /*palette*/ - case 4: return 2; /*grey + alpha*/ - case 6: return 4; /*RGBA*/ - } - return 0; /*unexisting color type*/ -} - -static unsigned lodepng_get_bpp_lct(LodePNGColorType colortype, unsigned bitdepth) -{ - /*bits per pixel is amount of channels * bits per channel*/ - return getNumColorChannels(colortype) * bitdepth; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -void lodepng_color_mode_init(LodePNGColorMode* info) -{ - info->key_defined = 0; - info->key_r = info->key_g = info->key_b = 0; - info->colortype = LCT_RGBA; - info->bitdepth = 8; - info->palette = 0; - info->palettesize = 0; -} - -void lodepng_color_mode_cleanup(LodePNGColorMode* info) -{ - lodepng_palette_clear(info); -} - -unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source) -{ - size_t i; - lodepng_color_mode_cleanup(dest); - *dest = *source; - if(source->palette) - { - dest->palette = (unsigned char*)lodepng_malloc(1024); - if(!dest->palette && source->palettesize) return 83; /*alloc fail*/ - for(i = 0; i < source->palettesize * 4; i++) dest->palette[i] = source->palette[i]; - } - return 0; -} - -static int lodepng_color_mode_equal(const LodePNGColorMode* a, const LodePNGColorMode* b) -{ - size_t i; - if(a->colortype != b->colortype) return 0; - if(a->bitdepth != b->bitdepth) return 0; - if(a->key_defined != b->key_defined) return 0; - if(a->key_defined) - { - if(a->key_r != b->key_r) return 0; - if(a->key_g != b->key_g) return 0; - if(a->key_b != b->key_b) return 0; - } - if(a->palettesize != b->palettesize) return 0; - for(i = 0; i < a->palettesize * 4; i++) - { - if(a->palette[i] != b->palette[i]) return 0; - } - return 1; -} - -void lodepng_palette_clear(LodePNGColorMode* info) -{ - if(info->palette) lodepng_free(info->palette); - info->palette = 0; - info->palettesize = 0; -} - -unsigned lodepng_palette_add(LodePNGColorMode* info, - unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - unsigned char* data; - /*the same resize technique as C++ std::vectors is used, and here it's made so that for a palette with - the max of 256 colors, it'll have the exact alloc size*/ - if(!info->palette) /*allocate palette if empty*/ - { - /*room for 256 colors with 4 bytes each*/ - data = (unsigned char*)lodepng_realloc(info->palette, 1024); - if(!data) return 83; /*alloc fail*/ - else info->palette = data; - } - info->palette[4 * info->palettesize + 0] = r; - info->palette[4 * info->palettesize + 1] = g; - info->palette[4 * info->palettesize + 2] = b; - info->palette[4 * info->palettesize + 3] = a; - info->palettesize++; - return 0; -} - -unsigned lodepng_get_bpp(const LodePNGColorMode* info) -{ - /*calculate bits per pixel out of colortype and bitdepth*/ - return lodepng_get_bpp_lct(info->colortype, info->bitdepth); -} - -unsigned lodepng_get_channels(const LodePNGColorMode* info) -{ - return getNumColorChannels(info->colortype); -} - -unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info) -{ - return info->colortype == LCT_GREY || info->colortype == LCT_GREY_ALPHA; -} - -unsigned lodepng_is_alpha_type(const LodePNGColorMode* info) -{ - return (info->colortype & 4) != 0; /*4 or 6*/ -} - -unsigned lodepng_is_palette_type(const LodePNGColorMode* info) -{ - return info->colortype == LCT_PALETTE; -} - -unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info) -{ - size_t i; - for(i = 0; i < info->palettesize; i++) - { - if(info->palette[i * 4 + 3] < 255) return 1; - } - return 0; -} - -unsigned lodepng_can_have_alpha(const LodePNGColorMode* info) -{ - return info->key_defined - || lodepng_is_alpha_type(info) - || lodepng_has_palette_alpha(info); -} - -size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color) -{ - return (w * h * lodepng_get_bpp(color) + 7) / 8; -} - -size_t lodepng_get_raw_size_lct(unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) -{ - return (w * h * lodepng_get_bpp_lct(colortype, bitdepth) + 7) / 8; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -static void LodePNGUnknownChunks_init(LodePNGInfo* info) -{ - unsigned i; - for(i = 0; i < 3; i++) info->unknown_chunks_data[i] = 0; - for(i = 0; i < 3; i++) info->unknown_chunks_size[i] = 0; -} - -static void LodePNGUnknownChunks_cleanup(LodePNGInfo* info) -{ - unsigned i; - for(i = 0; i < 3; i++) lodepng_free(info->unknown_chunks_data[i]); -} - -static unsigned LodePNGUnknownChunks_copy(LodePNGInfo* dest, const LodePNGInfo* src) -{ - unsigned i; - - LodePNGUnknownChunks_cleanup(dest); - - for(i = 0; i < 3; i++) - { - size_t j; - dest->unknown_chunks_size[i] = src->unknown_chunks_size[i]; - dest->unknown_chunks_data[i] = (unsigned char*)lodepng_malloc(src->unknown_chunks_size[i]); - if(!dest->unknown_chunks_data[i] && dest->unknown_chunks_size[i]) return 83; /*alloc fail*/ - for(j = 0; j < src->unknown_chunks_size[i]; j++) - { - dest->unknown_chunks_data[i][j] = src->unknown_chunks_data[i][j]; - } - } - - return 0; -} - -/******************************************************************************/ - -static void LodePNGText_init(LodePNGInfo* info) -{ - info->text_num = 0; - info->text_keys = NULL; - info->text_strings = NULL; -} - -static void LodePNGText_cleanup(LodePNGInfo* info) -{ - size_t i; - for(i = 0; i < info->text_num; i++) - { - string_cleanup(&info->text_keys[i]); - string_cleanup(&info->text_strings[i]); - } - lodepng_free(info->text_keys); - lodepng_free(info->text_strings); -} - -static unsigned LodePNGText_copy(LodePNGInfo* dest, const LodePNGInfo* source) -{ - size_t i = 0; - dest->text_keys = 0; - dest->text_strings = 0; - dest->text_num = 0; - for(i = 0; i < source->text_num; i++) - { - CERROR_TRY_RETURN(lodepng_add_text(dest, source->text_keys[i], source->text_strings[i])); - } - return 0; -} - -void lodepng_clear_text(LodePNGInfo* info) -{ - LodePNGText_cleanup(info); -} - -unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str) -{ - char** new_keys = (char**)(lodepng_realloc(info->text_keys, sizeof(char*) * (info->text_num + 1))); - char** new_strings = (char**)(lodepng_realloc(info->text_strings, sizeof(char*) * (info->text_num + 1))); - if(!new_keys || !new_strings) - { - lodepng_free(new_keys); - lodepng_free(new_strings); - return 83; /*alloc fail*/ - } - - info->text_num++; - info->text_keys = new_keys; - info->text_strings = new_strings; - - string_init(&info->text_keys[info->text_num - 1]); - string_set(&info->text_keys[info->text_num - 1], key); - - string_init(&info->text_strings[info->text_num - 1]); - string_set(&info->text_strings[info->text_num - 1], str); - - return 0; -} - -/******************************************************************************/ - -static void LodePNGIText_init(LodePNGInfo* info) -{ - info->itext_num = 0; - info->itext_keys = NULL; - info->itext_langtags = NULL; - info->itext_transkeys = NULL; - info->itext_strings = NULL; -} - -static void LodePNGIText_cleanup(LodePNGInfo* info) -{ - size_t i; - for(i = 0; i < info->itext_num; i++) - { - string_cleanup(&info->itext_keys[i]); - string_cleanup(&info->itext_langtags[i]); - string_cleanup(&info->itext_transkeys[i]); - string_cleanup(&info->itext_strings[i]); - } - lodepng_free(info->itext_keys); - lodepng_free(info->itext_langtags); - lodepng_free(info->itext_transkeys); - lodepng_free(info->itext_strings); -} - -static unsigned LodePNGIText_copy(LodePNGInfo* dest, const LodePNGInfo* source) -{ - size_t i = 0; - dest->itext_keys = 0; - dest->itext_langtags = 0; - dest->itext_transkeys = 0; - dest->itext_strings = 0; - dest->itext_num = 0; - for(i = 0; i < source->itext_num; i++) - { - CERROR_TRY_RETURN(lodepng_add_itext(dest, source->itext_keys[i], source->itext_langtags[i], - source->itext_transkeys[i], source->itext_strings[i])); - } - return 0; -} - -void lodepng_clear_itext(LodePNGInfo* info) -{ - LodePNGIText_cleanup(info); -} - -unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str) -{ - char** new_keys = (char**)(lodepng_realloc(info->itext_keys, sizeof(char*) * (info->itext_num + 1))); - char** new_langtags = (char**)(lodepng_realloc(info->itext_langtags, sizeof(char*) * (info->itext_num + 1))); - char** new_transkeys = (char**)(lodepng_realloc(info->itext_transkeys, sizeof(char*) * (info->itext_num + 1))); - char** new_strings = (char**)(lodepng_realloc(info->itext_strings, sizeof(char*) * (info->itext_num + 1))); - if(!new_keys || !new_langtags || !new_transkeys || !new_strings) - { - lodepng_free(new_keys); - lodepng_free(new_langtags); - lodepng_free(new_transkeys); - lodepng_free(new_strings); - return 83; /*alloc fail*/ - } - - info->itext_num++; - info->itext_keys = new_keys; - info->itext_langtags = new_langtags; - info->itext_transkeys = new_transkeys; - info->itext_strings = new_strings; - - string_init(&info->itext_keys[info->itext_num - 1]); - string_set(&info->itext_keys[info->itext_num - 1], key); - - string_init(&info->itext_langtags[info->itext_num - 1]); - string_set(&info->itext_langtags[info->itext_num - 1], langtag); - - string_init(&info->itext_transkeys[info->itext_num - 1]); - string_set(&info->itext_transkeys[info->itext_num - 1], transkey); - - string_init(&info->itext_strings[info->itext_num - 1]); - string_set(&info->itext_strings[info->itext_num - 1], str); - - return 0; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -void lodepng_info_init(LodePNGInfo* info) -{ - lodepng_color_mode_init(&info->color); - info->interlace_method = 0; - info->compression_method = 0; - info->filter_method = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - info->background_defined = 0; - info->background_r = info->background_g = info->background_b = 0; - - LodePNGText_init(info); - LodePNGIText_init(info); - - info->time_defined = 0; - info->phys_defined = 0; - - LodePNGUnknownChunks_init(info); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -void lodepng_info_cleanup(LodePNGInfo* info) -{ - lodepng_color_mode_cleanup(&info->color); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - LodePNGText_cleanup(info); - LodePNGIText_cleanup(info); - - LodePNGUnknownChunks_cleanup(info); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source) -{ - lodepng_info_cleanup(dest); - *dest = *source; - lodepng_color_mode_init(&dest->color); - CERROR_TRY_RETURN(lodepng_color_mode_copy(&dest->color, &source->color)); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - CERROR_TRY_RETURN(LodePNGText_copy(dest, source)); - CERROR_TRY_RETURN(LodePNGIText_copy(dest, source)); - - LodePNGUnknownChunks_init(dest); - CERROR_TRY_RETURN(LodePNGUnknownChunks_copy(dest, source)); -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - return 0; -} - -void lodepng_info_swap(LodePNGInfo* a, LodePNGInfo* b) -{ - LodePNGInfo temp = *a; - *a = *b; - *b = temp; -} - -/* ////////////////////////////////////////////////////////////////////////// */ - -/*index: bitgroup index, bits: bitgroup size(1, 2 or 4, in: bitgroup value, out: octet array to add bits to*/ -static void addColorBits(unsigned char* out, size_t index, unsigned bits, unsigned in) -{ - /*p = the partial index in the byte, e.g. with 4 palettebits it is 0 for first half or 1 for second half*/ - unsigned p = index % (8 / bits); - in &= (1 << bits) - 1; /*filter out any other bits of the input value*/ - in = in << (bits * (8 / bits - p - 1)); - if(p == 0) out[index * bits / 8] = in; - else out[index * bits / 8] |= in; -} - -typedef struct ColorTree ColorTree; - -/* -One node of a color tree -This is the data structure used to count the number of unique colors and to get a palette -index for a color. It's like an octree, but because the alpha channel is used too, each -node has 16 instead of 8 children. -*/ -struct ColorTree -{ - ColorTree* children[16]; /*up to 16 pointers to ColorTree of next level*/ - int index; /*the payload. Only has a meaningful value if this is in the last level*/ -}; - -static void color_tree_init(ColorTree* tree) -{ - int i; - for(i = 0; i < 16; i++) tree->children[i] = 0; - tree->index = -1; -} - -static void color_tree_cleanup(ColorTree* tree) -{ - int i; - for(i = 0; i < 16; i++) - { - if(tree->children[i]) - { - color_tree_cleanup(tree->children[i]); - lodepng_free(tree->children[i]); - } - } -} - -/*returns -1 if color not present, its index otherwise*/ -static int color_tree_get(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - int bit = 0; - for(bit = 0; bit < 8; bit++) - { - int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); - if(!tree->children[i]) return -1; - else tree = tree->children[i]; - } - return tree ? tree->index : -1; -} - -#ifdef LODEPNG_COMPILE_ENCODER -static int color_tree_has(ColorTree* tree, unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - return color_tree_get(tree, r, g, b, a) >= 0; -} -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/*color is not allowed to already exist. -Index should be >= 0 (it's signed to be compatible with using -1 for "doesn't exist")*/ -static void color_tree_add(ColorTree* tree, - unsigned char r, unsigned char g, unsigned char b, unsigned char a, int index) -{ - int bit; - for(bit = 0; bit < 8; bit++) - { - int i = 8 * ((r >> bit) & 1) + 4 * ((g >> bit) & 1) + 2 * ((b >> bit) & 1) + 1 * ((a >> bit) & 1); - if(!tree->children[i]) - { - tree->children[i] = (ColorTree*)lodepng_malloc(sizeof(ColorTree)); - color_tree_init(tree->children[i]); - } - tree = tree->children[i]; - } - tree->index = index; -} - -/*put a pixel, given its RGBA color, into image of any color type*/ -static unsigned rgba8ToPixel(unsigned char* out, size_t i, - const LodePNGColorMode* mode, ColorTree* tree /*for palette*/, - unsigned char r, unsigned char g, unsigned char b, unsigned char a) -{ - if(mode->colortype == LCT_GREY) - { - unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; - if(mode->bitdepth == 8) out[i] = grey; - else if(mode->bitdepth == 16) out[i * 2 + 0] = out[i * 2 + 1] = grey; - else - { - /*take the most significant bits of grey*/ - grey = (grey >> (8 - mode->bitdepth)) & ((1 << mode->bitdepth) - 1); - addColorBits(out, i, mode->bitdepth, grey); - } - } - else if(mode->colortype == LCT_RGB) - { - if(mode->bitdepth == 8) - { - out[i * 3 + 0] = r; - out[i * 3 + 1] = g; - out[i * 3 + 2] = b; - } - else - { - out[i * 6 + 0] = out[i * 6 + 1] = r; - out[i * 6 + 2] = out[i * 6 + 3] = g; - out[i * 6 + 4] = out[i * 6 + 5] = b; - } - } - else if(mode->colortype == LCT_PALETTE) - { - int index = color_tree_get(tree, r, g, b, a); - if(index < 0) return 82; /*color not in palette*/ - if(mode->bitdepth == 8) out[i] = index; - else addColorBits(out, i, mode->bitdepth, index); - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - unsigned char grey = r; /*((unsigned short)r + g + b) / 3*/; - if(mode->bitdepth == 8) - { - out[i * 2 + 0] = grey; - out[i * 2 + 1] = a; - } - else if(mode->bitdepth == 16) - { - out[i * 4 + 0] = out[i * 4 + 1] = grey; - out[i * 4 + 2] = out[i * 4 + 3] = a; - } - } - else if(mode->colortype == LCT_RGBA) - { - if(mode->bitdepth == 8) - { - out[i * 4 + 0] = r; - out[i * 4 + 1] = g; - out[i * 4 + 2] = b; - out[i * 4 + 3] = a; - } - else - { - out[i * 8 + 0] = out[i * 8 + 1] = r; - out[i * 8 + 2] = out[i * 8 + 3] = g; - out[i * 8 + 4] = out[i * 8 + 5] = b; - out[i * 8 + 6] = out[i * 8 + 7] = a; - } - } - - return 0; /*no error*/ -} - -/*put a pixel, given its RGBA16 color, into image of any color 16-bitdepth type*/ -static unsigned rgba16ToPixel(unsigned char* out, size_t i, - const LodePNGColorMode* mode, - unsigned short r, unsigned short g, unsigned short b, unsigned short a) -{ - if(mode->bitdepth != 16) return 85; /*must be 16 for this function*/ - if(mode->colortype == LCT_GREY) - { - unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; - out[i * 2 + 0] = (grey >> 8) & 255; - out[i * 2 + 1] = grey & 255; - } - else if(mode->colortype == LCT_RGB) - { - out[i * 6 + 0] = (r >> 8) & 255; - out[i * 6 + 1] = r & 255; - out[i * 6 + 2] = (g >> 8) & 255; - out[i * 6 + 3] = g & 255; - out[i * 6 + 4] = (b >> 8) & 255; - out[i * 6 + 5] = b & 255; - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - unsigned short grey = r; /*((unsigned)r + g + b) / 3*/; - out[i * 4 + 0] = (grey >> 8) & 255; - out[i * 4 + 1] = grey & 255; - out[i * 4 + 2] = (a >> 8) & 255; - out[i * 4 + 3] = a & 255; - } - else if(mode->colortype == LCT_RGBA) - { - out[i * 8 + 0] = (r >> 8) & 255; - out[i * 8 + 1] = r & 255; - out[i * 8 + 2] = (g >> 8) & 255; - out[i * 8 + 3] = g & 255; - out[i * 8 + 4] = (b >> 8) & 255; - out[i * 8 + 5] = b & 255; - out[i * 8 + 6] = (a >> 8) & 255; - out[i * 8 + 7] = a & 255; - } - - return 0; /*no error*/ -} - -/*Get RGBA8 color of pixel with index i (y * width + x) from the raw image with given color type.*/ -static unsigned getPixelColorRGBA8(unsigned char* r, unsigned char* g, - unsigned char* b, unsigned char* a, - const unsigned char* in, size_t i, - const LodePNGColorMode* mode, - unsigned fix_png) -{ - if(mode->colortype == LCT_GREY) - { - if(mode->bitdepth == 8) - { - *r = *g = *b = in[i]; - if(mode->key_defined && *r == mode->key_r) *a = 0; - else *a = 255; - } - else if(mode->bitdepth == 16) - { - *r = *g = *b = in[i * 2 + 0]; - if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; - else *a = 255; - } - else - { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = i * mode->bitdepth; - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - *r = *g = *b = (value * 255) / highest; - if(mode->key_defined && value == mode->key_r) *a = 0; - else *a = 255; - } - } - else if(mode->colortype == LCT_RGB) - { - if(mode->bitdepth == 8) - { - *r = in[i * 3 + 0]; *g = in[i * 3 + 1]; *b = in[i * 3 + 2]; - if(mode->key_defined && *r == mode->key_r && *g == mode->key_g && *b == mode->key_b) *a = 0; - else *a = 255; - } - else - { - *r = in[i * 6 + 0]; - *g = in[i * 6 + 2]; - *b = in[i * 6 + 4]; - if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; - else *a = 255; - } - } - else if(mode->colortype == LCT_PALETTE) - { - unsigned index; - if(mode->bitdepth == 8) index = in[i]; - else - { - size_t j = i * mode->bitdepth; - index = readBitsFromReversedStream(&j, in, mode->bitdepth); - } - - if(index >= mode->palettesize) - { - /*This is an error according to the PNG spec, but fix_png can ignore it*/ - if(!fix_png) return (mode->bitdepth == 8 ? 46 : 47); /*index out of palette*/ - *r = *g = *b = 0; - *a = 255; - } - else - { - *r = mode->palette[index * 4 + 0]; - *g = mode->palette[index * 4 + 1]; - *b = mode->palette[index * 4 + 2]; - *a = mode->palette[index * 4 + 3]; - } - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - if(mode->bitdepth == 8) - { - *r = *g = *b = in[i * 2 + 0]; - *a = in[i * 2 + 1]; - } - else - { - *r = *g = *b = in[i * 4 + 0]; - *a = in[i * 4 + 2]; - } - } - else if(mode->colortype == LCT_RGBA) - { - if(mode->bitdepth == 8) - { - *r = in[i * 4 + 0]; - *g = in[i * 4 + 1]; - *b = in[i * 4 + 2]; - *a = in[i * 4 + 3]; - } - else - { - *r = in[i * 8 + 0]; - *g = in[i * 8 + 2]; - *b = in[i * 8 + 4]; - *a = in[i * 8 + 6]; - } - } - - return 0; /*no error*/ -} - -/*Similar to getPixelColorRGBA8, but with all the for loops inside of the color -mode test cases, optimized to convert the colors much faster, when converting -to RGBA or RGB with 8 bit per cannel. buffer must be RGBA or RGB output with -enough memory, if has_alpha is true the output is RGBA. mode has the color mode -of the input buffer.*/ -static unsigned getPixelColorsRGBA8(unsigned char* buffer, size_t numpixels, - unsigned has_alpha, const unsigned char* in, - const LodePNGColorMode* mode, - unsigned fix_png) -{ - unsigned num_channels = has_alpha ? 4 : 3; - size_t i; - if(mode->colortype == LCT_GREY) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i]; - if(has_alpha) buffer[3] = mode->key_defined && in[i] == mode->key_r ? 0 : 255; - } - } - else if(mode->bitdepth == 16) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i * 2]; - if(has_alpha) buffer[3] = mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r ? 0 : 255; - } - } - else - { - unsigned highest = ((1U << mode->bitdepth) - 1U); /*highest possible value for this bit depth*/ - size_t j = 0; - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - unsigned value = readBitsFromReversedStream(&j, in, mode->bitdepth); - buffer[0] = buffer[1] = buffer[2] = (value * 255) / highest; - if(has_alpha) buffer[3] = mode->key_defined && value == mode->key_r ? 0 : 255; - } - } - } - else if(mode->colortype == LCT_RGB) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 3 + 0]; - buffer[1] = in[i * 3 + 1]; - buffer[2] = in[i * 3 + 2]; - if(has_alpha) buffer[3] = mode->key_defined && buffer[0] == mode->key_r - && buffer[1]== mode->key_g && buffer[2] == mode->key_b ? 0 : 255; - } - } - else - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 6 + 0]; - buffer[1] = in[i * 6 + 2]; - buffer[2] = in[i * 6 + 4]; - if(has_alpha) buffer[3] = mode->key_defined - && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b ? 0 : 255; - } - } - } - else if(mode->colortype == LCT_PALETTE) - { - unsigned index; - size_t j = 0; - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - if(mode->bitdepth == 8) index = in[i]; - else index = readBitsFromReversedStream(&j, in, mode->bitdepth); - - if(index >= mode->palettesize) - { - /*This is an error according to the PNG spec, but fix_png can ignore it*/ - if(!fix_png) return (mode->bitdepth == 8 ? 46 : 47); /*index out of palette*/ - buffer[0] = buffer[1] = buffer[2] = 0; - if(has_alpha) buffer[3] = 255; - } - else - { - buffer[0] = mode->palette[index * 4 + 0]; - buffer[1] = mode->palette[index * 4 + 1]; - buffer[2] = mode->palette[index * 4 + 2]; - if(has_alpha) buffer[3] = mode->palette[index * 4 + 3]; - } - } - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i * 2 + 0]; - if(has_alpha) buffer[3] = in[i * 2 + 1]; - } - } - else - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = buffer[1] = buffer[2] = in[i * 4 + 0]; - if(has_alpha) buffer[3] = in[i * 4 + 2]; - } - } - } - else if(mode->colortype == LCT_RGBA) - { - if(mode->bitdepth == 8) - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 4 + 0]; - buffer[1] = in[i * 4 + 1]; - buffer[2] = in[i * 4 + 2]; - if(has_alpha) buffer[3] = in[i * 4 + 3]; - } - } - else - { - for(i = 0; i < numpixels; i++, buffer += num_channels) - { - buffer[0] = in[i * 8 + 0]; - buffer[1] = in[i * 8 + 2]; - buffer[2] = in[i * 8 + 4]; - if(has_alpha) buffer[3] = in[i * 8 + 6]; - } - } - } - - return 0; /*no error*/ -} - -/*Get RGBA16 color of pixel with index i (y * width + x) from the raw image with -given color type, but the given color type must be 16-bit itself.*/ -static unsigned getPixelColorRGBA16(unsigned short* r, unsigned short* g, unsigned short* b, unsigned short* a, - const unsigned char* in, size_t i, const LodePNGColorMode* mode) -{ - if(mode->bitdepth != 16) return 85; /*error: this function only supports 16-bit input*/ - - if(mode->colortype == LCT_GREY) - { - *r = *g = *b = 256 * in[i * 2 + 0] + in[i * 2 + 1]; - if(mode->key_defined && 256U * in[i * 2 + 0] + in[i * 2 + 1] == mode->key_r) *a = 0; - else *a = 65535; - } - else if(mode->colortype == LCT_RGB) - { - *r = 256 * in[i * 6 + 0] + in[i * 6 + 1]; - *g = 256 * in[i * 6 + 2] + in[i * 6 + 3]; - *b = 256 * in[i * 6 + 4] + in[i * 6 + 5]; - if(mode->key_defined && 256U * in[i * 6 + 0] + in[i * 6 + 1] == mode->key_r - && 256U * in[i * 6 + 2] + in[i * 6 + 3] == mode->key_g - && 256U * in[i * 6 + 4] + in[i * 6 + 5] == mode->key_b) *a = 0; - else *a = 65535; - } - else if(mode->colortype == LCT_GREY_ALPHA) - { - *r = *g = *b = 256 * in[i * 4 + 0] + in[i * 4 + 1]; - *a = 256 * in[i * 4 + 2] + in[i * 4 + 3]; - } - else if(mode->colortype == LCT_RGBA) - { - *r = 256 * in[i * 8 + 0] + in[i * 8 + 1]; - *g = 256 * in[i * 8 + 2] + in[i * 8 + 3]; - *b = 256 * in[i * 8 + 4] + in[i * 8 + 5]; - *a = 256 * in[i * 8 + 6] + in[i * 8 + 7]; - } - else return 85; /*error: this function only supports 16-bit input, not palettes*/ - - return 0; /*no error*/ -} - -/* -converts from any color type to 24-bit or 32-bit (later maybe more supported). return value = LodePNG error code -the out buffer must have (w * h * bpp + 7) / 8 bytes, where bpp is the bits per pixel of the output color type -(lodepng_get_bpp) for < 8 bpp images, there may _not_ be padding bits at the end of scanlines. -*/ -unsigned lodepng_convert(unsigned char* out, const unsigned char* in, - LodePNGColorMode* mode_out, LodePNGColorMode* mode_in, - unsigned w, unsigned h, unsigned fix_png) -{ - unsigned error = 0; - size_t i; - ColorTree tree; - size_t numpixels = w * h; - - if(lodepng_color_mode_equal(mode_out, mode_in)) - { - size_t numbytes = lodepng_get_raw_size(w, h, mode_in); - for(i = 0; i < numbytes; i++) out[i] = in[i]; - return error; - } - - if(mode_out->colortype == LCT_PALETTE) - { - size_t palsize = 1 << mode_out->bitdepth; - if(mode_out->palettesize < palsize) palsize = mode_out->palettesize; - color_tree_init(&tree); - for(i = 0; i < palsize; i++) - { - unsigned char* p = &mode_out->palette[i * 4]; - color_tree_add(&tree, p[0], p[1], p[2], p[3], i); - } - } - - if(mode_in->bitdepth == 16 && mode_out->bitdepth == 16) - { - for(i = 0; i < numpixels; i++) - { - unsigned short r = 0, g = 0, b = 0, a = 0; - error = getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode_in); - if(error) break; - error = rgba16ToPixel(out, i, mode_out, r, g, b, a); - if(error) break; - } - } - else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGBA) - { - error = getPixelColorsRGBA8(out, numpixels, 1, in, mode_in, fix_png); - } - else if(mode_out->bitdepth == 8 && mode_out->colortype == LCT_RGB) - { - error = getPixelColorsRGBA8(out, numpixels, 0, in, mode_in, fix_png); - } - else - { - unsigned char r = 0, g = 0, b = 0, a = 0; - for(i = 0; i < numpixels; i++) - { - error = getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode_in, fix_png); - if(error) break; - error = rgba8ToPixel(out, i, mode_out, &tree, r, g, b, a); - if(error) break; - } - } - - if(mode_out->colortype == LCT_PALETTE) - { - color_tree_cleanup(&tree); - } - - return error; -} - -#ifdef LODEPNG_COMPILE_ENCODER - -typedef struct ColorProfile -{ - unsigned char sixteenbit; /*needs more than 8 bits per channel*/ - unsigned char sixteenbit_done; - - - unsigned char colored; /*not greyscale*/ - unsigned char colored_done; - - unsigned char key; /*a color key is required, or more*/ - unsigned short key_r; /*these values are always in 16-bit bitdepth in the profile*/ - unsigned short key_g; - unsigned short key_b; - unsigned char alpha; /*alpha channel, or alpha palette, required*/ - unsigned char alpha_done; - - unsigned numcolors; - ColorTree tree; /*for listing the counted colors, up to 256*/ - unsigned char* palette; /*size 1024. Remember up to the first 256 RGBA colors*/ - unsigned maxnumcolors; /*if more than that amount counted*/ - unsigned char numcolors_done; - - unsigned greybits; /*amount of bits required for greyscale (1, 2, 4, 8). Does not take 16 bit into account.*/ - unsigned char greybits_done; - -} ColorProfile; - -static void color_profile_init(ColorProfile* profile, LodePNGColorMode* mode) -{ - profile->sixteenbit = 0; - profile->sixteenbit_done = mode->bitdepth == 16 ? 0 : 1; - - profile->colored = 0; - profile->colored_done = lodepng_is_greyscale_type(mode) ? 1 : 0; - - profile->key = 0; - profile->alpha = 0; - profile->alpha_done = lodepng_can_have_alpha(mode) ? 0 : 1; - - profile->numcolors = 0; - color_tree_init(&profile->tree); - profile->palette = (unsigned char*)lodepng_malloc(1024); - profile->maxnumcolors = 257; - if(lodepng_get_bpp(mode) <= 8) - { - int bpp = lodepng_get_bpp(mode); - profile->maxnumcolors = bpp == 1 ? 2 : (bpp == 2 ? 4 : (bpp == 4 ? 16 : 256)); - } - profile->numcolors_done = 0; - - profile->greybits = 1; - profile->greybits_done = lodepng_get_bpp(mode) == 1 ? 1 : 0; -} - -static void color_profile_cleanup(ColorProfile* profile) -{ - color_tree_cleanup(&profile->tree); - lodepng_free(profile->palette); -} - -/*function used for debug purposes with C++*/ -/*void printColorProfile(ColorProfile* p) -{ - std::cout << "sixteenbit: " << (int)p->sixteenbit << std::endl; - std::cout << "sixteenbit_done: " << (int)p->sixteenbit_done << std::endl; - std::cout << "colored: " << (int)p->colored << std::endl; - std::cout << "colored_done: " << (int)p->colored_done << std::endl; - std::cout << "key: " << (int)p->key << std::endl; - std::cout << "key_r: " << (int)p->key_r << std::endl; - std::cout << "key_g: " << (int)p->key_g << std::endl; - std::cout << "key_b: " << (int)p->key_b << std::endl; - std::cout << "alpha: " << (int)p->alpha << std::endl; - std::cout << "alpha_done: " << (int)p->alpha_done << std::endl; - std::cout << "numcolors: " << (int)p->numcolors << std::endl; - std::cout << "maxnumcolors: " << (int)p->maxnumcolors << std::endl; - std::cout << "numcolors_done: " << (int)p->numcolors_done << std::endl; - std::cout << "greybits: " << (int)p->greybits << std::endl; - std::cout << "greybits_done: " << (int)p->greybits_done << std::endl; -}*/ - -/*Returns how many bits needed to represent given value (max 8 bit)*/ -unsigned getValueRequiredBits(unsigned short value) -{ - if(value == 0 || value == 255) return 1; - /*The scaling of 2-bit and 4-bit values uses multiples of 85 and 17*/ - if(value % 17 == 0) return value % 85 == 0 ? 2 : 4; - return 8; -} - -/*profile must already have been inited with mode. -It's ok to set some parameters of profile to done already.*/ -static unsigned get_color_profile(ColorProfile* profile, - const unsigned char* in, - size_t numpixels /*must be full image size, for certain filesize based choices*/, - LodePNGColorMode* mode, - unsigned fix_png) -{ - unsigned error = 0; - size_t i; - - if(mode->bitdepth == 16) - { - for(i = 0; i < numpixels; i++) - { - unsigned short r, g, b, a; - error = getPixelColorRGBA16(&r, &g, &b, &a, in, i, mode); - if(error) break; - - /*a color is considered good for 8-bit if the first byte and the second byte are equal, - (so if it's divisible through 257), NOT necessarily if the second byte is 0*/ - if(!profile->sixteenbit_done - && (((r & 255) != ((r >> 8) & 255)) - || ((g & 255) != ((g >> 8) & 255)) - || ((b & 255) != ((b >> 8) & 255)))) - { - profile->sixteenbit = 1; - profile->sixteenbit_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore at 16-bit*/ - profile->numcolors_done = 1; /*counting colors no longer useful, palette doesn't support 16-bit*/ - } - - if(!profile->colored_done && (r != g || r != b)) - { - profile->colored = 1; - profile->colored_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore*/ - } - - if(!profile->alpha_done && a != 65535) - { - /*only use color key if numpixels large enough to justify tRNS chunk size*/ - if(a == 0 && numpixels > 16 && !(profile->key && (r != profile->key_r || g != profile->key_g || b != profile->key_b))) - { - if(!profile->alpha && !profile->key) - { - profile->key = 1; - profile->key_r = r; - profile->key_g = g; - profile->key_b = b; - } - } - else - { - profile->alpha = 1; - profile->alpha_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore*/ - } - } - - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - if(!profile->alpha_done && a == 65535 && profile->key - && r == profile->key_r && g == profile->key_g && b == profile->key_b) - { - profile->alpha = 1; - profile->alpha_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore*/ - } - - if(!profile->greybits_done) - { - /*assuming 8-bit r, this test does not care about 16-bit*/ - unsigned bits = getValueRequiredBits(r); - if(bits > profile->greybits) profile->greybits = bits; - if(profile->greybits >= 8) profile->greybits_done = 1; - } - - if(!profile->numcolors_done) - { - /*assuming 8-bit rgba, this test does not care about 16-bit*/ - if(!color_tree_has(&profile->tree, (unsigned char)r, (unsigned char)g, (unsigned char)b, (unsigned char)a)) - { - color_tree_add(&profile->tree, (unsigned char)r, (unsigned char)g, (unsigned char)b, (unsigned char)a, - profile->numcolors); - if(profile->numcolors < 256) - { - unsigned char* p = profile->palette; - unsigned i = profile->numcolors; - p[i * 4 + 0] = (unsigned char)r; - p[i * 4 + 1] = (unsigned char)g; - p[i * 4 + 2] = (unsigned char)b; - p[i * 4 + 3] = (unsigned char)a; - } - profile->numcolors++; - if(profile->numcolors >= profile->maxnumcolors) profile->numcolors_done = 1; - } - } - - if(profile->alpha_done && profile->numcolors_done - && profile->colored_done && profile->sixteenbit_done && profile->greybits_done) - { - break; - } - }; - } - else /* < 16-bit */ - { - for(i = 0; i < numpixels; i++) - { - unsigned char r = 0, g = 0, b = 0, a = 0; - error = getPixelColorRGBA8(&r, &g, &b, &a, in, i, mode, fix_png); - if(error) break; - - if(!profile->colored_done && (r != g || r != b)) - { - profile->colored = 1; - profile->colored_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore*/ - } - - if(!profile->alpha_done && a != 255) - { - if(a == 0 && !(profile->key && (r != profile->key_r || g != profile->key_g || b != profile->key_b))) - { - if(!profile->key) - { - profile->key = 1; - profile->key_r = r; - profile->key_g = g; - profile->key_b = b; - } - } - else - { - profile->alpha = 1; - profile->alpha_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore*/ - } - } - - /* Color key cannot be used if an opaque pixel also has that RGB color. */ - if(!profile->alpha_done && a == 255 && profile->key - && r == profile->key_r && g == profile->key_g && b == profile->key_b) - { - profile->alpha = 1; - profile->alpha_done = 1; - profile->greybits_done = 1; /*greybits is not applicable anymore*/ - } - - if(!profile->greybits_done) - { - unsigned bits = getValueRequiredBits(r); - if(bits > profile->greybits) profile->greybits = bits; - if(profile->greybits >= 8) profile->greybits_done = 1; - } - - if(!profile->numcolors_done) - { - if(!color_tree_has(&profile->tree, r, g, b, a)) - { - - color_tree_add(&profile->tree, r, g, b, a, profile->numcolors); - if(profile->numcolors < 256) - { - unsigned char* p = profile->palette; - unsigned i = profile->numcolors; - p[i * 4 + 0] = r; - p[i * 4 + 1] = g; - p[i * 4 + 2] = b; - p[i * 4 + 3] = a; - } - profile->numcolors++; - if(profile->numcolors >= profile->maxnumcolors) profile->numcolors_done = 1; - } - } - - if(profile->alpha_done && profile->numcolors_done && profile->colored_done && profile->greybits_done) - { - break; - } - }; - } - - /*make the profile's key always 16-bit for consistency*/ - if(mode->bitdepth < 16) - { - /*repeat each byte twice*/ - profile->key_r *= 257; - profile->key_g *= 257; - profile->key_b *= 257; - } - - return error; -} - -static void setColorKeyFrom16bit(LodePNGColorMode* mode_out, unsigned r, unsigned g, unsigned b, unsigned bitdepth) -{ - unsigned mask = (1 << bitdepth) - 1; - mode_out->key_defined = 1; - mode_out->key_r = r & mask; - mode_out->key_g = g & mask; - mode_out->key_b = b & mask; -} - -/*updates values of mode with a potentially smaller color model. mode_out should -contain the user chosen color model, but will be overwritten with the new chosen one.*/ -static unsigned doAutoChooseColor(LodePNGColorMode* mode_out, - const unsigned char* image, unsigned w, unsigned h, LodePNGColorMode* mode_in, - LodePNGAutoConvert auto_convert) -{ - ColorProfile profile; - unsigned error = 0; - int no_nibbles = auto_convert == LAC_AUTO_NO_NIBBLES || auto_convert == LAC_AUTO_NO_NIBBLES_NO_PALETTE; - int no_palette = auto_convert == LAC_AUTO_NO_PALETTE || auto_convert == LAC_AUTO_NO_NIBBLES_NO_PALETTE; - - if(auto_convert == LAC_ALPHA) - { - if(mode_out->colortype != LCT_RGBA && mode_out->colortype != LCT_GREY_ALPHA) return 0; - } - - color_profile_init(&profile, mode_in); - if(auto_convert == LAC_ALPHA) - { - profile.colored_done = 1; - profile.greybits_done = 1; - profile.numcolors_done = 1; - profile.sixteenbit_done = 1; - } - error = get_color_profile(&profile, image, w * h, mode_in, 0 /*fix_png*/); - if(!error && auto_convert == LAC_ALPHA) - { - if(!profile.alpha) - { - mode_out->colortype = (mode_out->colortype == LCT_RGBA ? LCT_RGB : LCT_GREY); - if(profile.key) setColorKeyFrom16bit(mode_out, profile.key_r, profile.key_g, profile.key_b, mode_out->bitdepth); - } - } - else if(!error && auto_convert != LAC_ALPHA) - { - mode_out->key_defined = 0; - - if(profile.sixteenbit) - { - mode_out->bitdepth = 16; - if(profile.alpha) - { - mode_out->colortype = profile.colored ? LCT_RGBA : LCT_GREY_ALPHA; - } - else - { - mode_out->colortype = profile.colored ? LCT_RGB : LCT_GREY; - if(profile.key) setColorKeyFrom16bit(mode_out, profile.key_r, profile.key_g, profile.key_b, mode_out->bitdepth); - } - } - else /*less than 16 bits per channel*/ - { - /*don't add palette overhead if image hasn't got a lot of pixels*/ - unsigned n = profile.numcolors; - int palette_ok = !no_palette && n <= 256 && (n * 2 < w * h); - unsigned palettebits = n <= 2 ? 1 : (n <= 4 ? 2 : (n <= 16 ? 4 : 8)); - int grey_ok = !profile.colored && !profile.alpha; /*grey without alpha, with potentially low bits*/ - if(palette_ok || grey_ok) - { - if(!palette_ok || (grey_ok && profile.greybits <= palettebits)) - { - unsigned grey = profile.key_r; - mode_out->colortype = LCT_GREY; - mode_out->bitdepth = profile.greybits; - if(profile.key) setColorKeyFrom16bit(mode_out, grey, grey, grey, mode_out->bitdepth); - } - else - { - /*fill in the palette*/ - unsigned i; - unsigned char* p = profile.palette; - /*remove potential earlier palette*/ - lodepng_palette_clear(mode_out); - for(i = 0; i < profile.numcolors; i++) - { - error = lodepng_palette_add(mode_out, p[i * 4 + 0], p[i * 4 + 1], p[i * 4 + 2], p[i * 4 + 3]); - if(error) break; - } - - mode_out->colortype = LCT_PALETTE; - mode_out->bitdepth = palettebits; - } - } - else /*8-bit per channel*/ - { - mode_out->bitdepth = 8; - if(profile.alpha) - { - mode_out->colortype = profile.colored ? LCT_RGBA : LCT_GREY_ALPHA; - } - else - { - mode_out->colortype = profile.colored ? LCT_RGB : LCT_GREY /*LCT_GREY normally won't occur, already done earlier*/; - if(profile.key) setColorKeyFrom16bit(mode_out, profile.key_r, profile.key_g, profile.key_b, mode_out->bitdepth); - } - } - } - } - - color_profile_cleanup(&profile); - - if(mode_out->colortype == LCT_PALETTE && mode_in->palettesize == mode_out->palettesize) - { - /*In this case keep the palette order of the input, so that the user can choose an optimal one*/ - size_t i; - for(i = 0; i < mode_in->palettesize * 4; i++) - { - mode_out->palette[i] = mode_in->palette[i]; - } - } - - if(no_nibbles && mode_out->bitdepth < 8) - { - /*palette can keep its small amount of colors, as long as no indices use it*/ - mode_out->bitdepth = 8; - } - - return error; -} - -#endif /* #ifdef LODEPNG_COMPILE_ENCODER */ - -/* -Paeth predicter, used by PNG filter type 4 -The parameters are of type short, but should come from unsigned chars, the shorts -are only needed to make the paeth calculation correct. -*/ -static unsigned char paethPredictor(short a, short b, short c) -{ - short pa = abs(b - c); - short pb = abs(a - c); - short pc = abs(a + b - c - c); - - if(pc < pa && pc < pb) return (unsigned char)c; - else if(pb < pa) return (unsigned char)b; - else return (unsigned char)a; -} - -/*shared values used by multiple Adam7 related functions*/ - -static const unsigned ADAM7_IX[7] = { 0, 4, 0, 2, 0, 1, 0 }; /*x start values*/ -static const unsigned ADAM7_IY[7] = { 0, 0, 4, 0, 2, 0, 1 }; /*y start values*/ -static const unsigned ADAM7_DX[7] = { 8, 8, 4, 4, 2, 2, 1 }; /*x delta values*/ -static const unsigned ADAM7_DY[7] = { 8, 8, 8, 4, 4, 2, 2 }; /*y delta values*/ - -/* -Outputs various dimensions and positions in the image related to the Adam7 reduced images. -passw: output containing the width of the 7 passes -passh: output containing the height of the 7 passes -filter_passstart: output containing the index of the start and end of each - reduced image with filter bytes -padded_passstart output containing the index of the start and end of each - reduced image when without filter bytes but with padded scanlines -passstart: output containing the index of the start and end of each reduced - image without padding between scanlines, but still padding between the images -w, h: width and height of non-interlaced image -bpp: bits per pixel -"padded" is only relevant if bpp is less than 8 and a scanline or image does not - end at a full byte -*/ -static void Adam7_getpassvalues(unsigned passw[7], unsigned passh[7], size_t filter_passstart[8], - size_t padded_passstart[8], size_t passstart[8], unsigned w, unsigned h, unsigned bpp) -{ - /*the passstart values have 8 values: the 8th one indicates the byte after the end of the 7th (= last) pass*/ - unsigned i; - - /*calculate width and height in pixels of each pass*/ - for(i = 0; i < 7; i++) - { - passw[i] = (w + ADAM7_DX[i] - ADAM7_IX[i] - 1) / ADAM7_DX[i]; - passh[i] = (h + ADAM7_DY[i] - ADAM7_IY[i] - 1) / ADAM7_DY[i]; - if(passw[i] == 0) passh[i] = 0; - if(passh[i] == 0) passw[i] = 0; - } - - filter_passstart[0] = padded_passstart[0] = passstart[0] = 0; - for(i = 0; i < 7; i++) - { - /*if passw[i] is 0, it's 0 bytes, not 1 (no filtertype-byte)*/ - filter_passstart[i + 1] = filter_passstart[i] - + ((passw[i] && passh[i]) ? passh[i] * (1 + (passw[i] * bpp + 7) / 8) : 0); - /*bits padded if needed to fill full byte at end of each scanline*/ - padded_passstart[i + 1] = padded_passstart[i] + passh[i] * ((passw[i] * bpp + 7) / 8); - /*only padded at end of reduced image*/ - passstart[i + 1] = passstart[i] + (passh[i] * passw[i] * bpp + 7) / 8; - } -} - -#ifdef LODEPNG_COMPILE_DECODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG Decoder / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*read the information from the header and store it in the LodePNGInfo. return value is error*/ -unsigned lodepng_inspect(unsigned* w, unsigned* h, LodePNGState* state, - const unsigned char* in, size_t insize) -{ - LodePNGInfo* info = &state->info_png; - if(insize == 0 || in == 0) - { - CERROR_RETURN_ERROR(state->error, 48); /*error: the given data is empty*/ - } - if(insize < 29) - { - CERROR_RETURN_ERROR(state->error, 27); /*error: the data length is smaller than the length of a PNG header*/ - } - - /*when decoding a new PNG image, make sure all parameters created after previous decoding are reset*/ - lodepng_info_cleanup(info); - lodepng_info_init(info); - - if(in[0] != 137 || in[1] != 80 || in[2] != 78 || in[3] != 71 - || in[4] != 13 || in[5] != 10 || in[6] != 26 || in[7] != 10) - { - CERROR_RETURN_ERROR(state->error, 28); /*error: the first 8 bytes are not the correct PNG signature*/ - } - if(in[12] != 'I' || in[13] != 'H' || in[14] != 'D' || in[15] != 'R') - { - CERROR_RETURN_ERROR(state->error, 29); /*error: it doesn't start with a IHDR chunk!*/ - } - - /*read the values given in the header*/ - *w = lodepng_read32bitInt(&in[16]); - *h = lodepng_read32bitInt(&in[20]); - info->color.bitdepth = in[24]; - info->color.colortype = (LodePNGColorType)in[25]; - info->compression_method = in[26]; - info->filter_method = in[27]; - info->interlace_method = in[28]; - - if(!state->decoder.ignore_crc) - { - unsigned CRC = lodepng_read32bitInt(&in[29]); - unsigned checksum = lodepng_crc32(&in[12], 17); - if(CRC != checksum) - { - CERROR_RETURN_ERROR(state->error, 57); /*invalid CRC*/ - } - } - - /*error: only compression method 0 is allowed in the specification*/ - if(info->compression_method != 0) CERROR_RETURN_ERROR(state->error, 32); - /*error: only filter method 0 is allowed in the specification*/ - if(info->filter_method != 0) CERROR_RETURN_ERROR(state->error, 33); - /*error: only interlace methods 0 and 1 exist in the specification*/ - if(info->interlace_method > 1) CERROR_RETURN_ERROR(state->error, 34); - - state->error = checkColorValidity(info->color.colortype, info->color.bitdepth); - return state->error; -} - -static unsigned unfilterScanline(unsigned char* recon, const unsigned char* scanline, const unsigned char* precon, - size_t bytewidth, unsigned char filterType, size_t length) -{ - /* - For PNG filter method 0 - unfilter a PNG image scanline by scanline. when the pixels are smaller than 1 byte, - the filter works byte per byte (bytewidth = 1) - precon is the previous unfiltered scanline, recon the result, scanline the current one - the incoming scanlines do NOT include the filtertype byte, that one is given in the parameter filterType instead - recon and scanline MAY be the same memory address! precon must be disjoint. - */ - - size_t i; - switch(filterType) - { - case 0: - for(i = 0; i < length; i++) recon[i] = scanline[i]; - break; - case 1: - for(i = 0; i < bytewidth; i++) recon[i] = scanline[i]; - for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + recon[i - bytewidth]; - break; - case 2: - if(precon) - { - for(i = 0; i < length; i++) recon[i] = scanline[i] + precon[i]; - } - else - { - for(i = 0; i < length; i++) recon[i] = scanline[i]; - } - break; - case 3: - if(precon) - { - for(i = 0; i < bytewidth; i++) recon[i] = scanline[i] + precon[i] / 2; - for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + ((recon[i - bytewidth] + precon[i]) / 2); - } - else - { - for(i = 0; i < bytewidth; i++) recon[i] = scanline[i]; - for(i = bytewidth; i < length; i++) recon[i] = scanline[i] + recon[i - bytewidth] / 2; - } - break; - case 4: - if(precon) - { - for(i = 0; i < bytewidth; i++) - { - recon[i] = (scanline[i] + precon[i]); /*paethPredictor(0, precon[i], 0) is always precon[i]*/ - } - for(i = bytewidth; i < length; i++) - { - recon[i] = (scanline[i] + paethPredictor(recon[i - bytewidth], precon[i], precon[i - bytewidth])); - } - } - else - { - for(i = 0; i < bytewidth; i++) - { - recon[i] = scanline[i]; - } - for(i = bytewidth; i < length; i++) - { - /*paethPredictor(recon[i - bytewidth], 0, 0) is always recon[i - bytewidth]*/ - recon[i] = (scanline[i] + recon[i - bytewidth]); - } - } - break; - default: return 36; /*error: unexisting filter type given*/ - } - return 0; -} - -static unsigned unfilter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) -{ - /* - For PNG filter method 0 - this function unfilters a single image (e.g. without interlacing this is called once, with Adam7 seven times) - out must have enough bytes allocated already, in must have the scanlines + 1 filtertype byte per scanline - w and h are image dimensions or dimensions of reduced image, bpp is bits per pixel - in and out are allowed to be the same memory address (but aren't the same size since in has the extra filter bytes) - */ - - unsigned y; - unsigned char* prevline = 0; - - /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ - size_t bytewidth = (bpp + 7) / 8; - size_t linebytes = (w * bpp + 7) / 8; - - for(y = 0; y < h; y++) - { - size_t outindex = linebytes * y; - size_t inindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - unsigned char filterType = in[inindex]; - - CERROR_TRY_RETURN(unfilterScanline(&out[outindex], &in[inindex + 1], prevline, bytewidth, filterType, linebytes)); - - prevline = &out[outindex]; - } - - return 0; -} - -/* -in: Adam7 interlaced image, with no padding bits between scanlines, but between - reduced images so that each reduced image starts at a byte. -out: the same pixels, but re-ordered so that they're now a non-interlaced image with size w*h -bpp: bits per pixel -out has the following size in bits: w * h * bpp. -in is possibly bigger due to padding bits between reduced images. -out must be big enough AND must be 0 everywhere if bpp < 8 in the current implementation -(because that's likely a little bit faster) -NOTE: comments about padding bits are only relevant if bpp < 8 -*/ -static void Adam7_deinterlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) -{ - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - if(bpp >= 8) - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - size_t bytewidth = bpp / 8; - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - size_t pixelinstart = passstart[i] + (y * passw[i] + x) * bytewidth; - size_t pixeloutstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; - for(b = 0; b < bytewidth; b++) - { - out[pixeloutstart + b] = in[pixelinstart + b]; - } - } - } - } - else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - unsigned ilinebits = bpp * passw[i]; - unsigned olinebits = bpp * w; - size_t obp, ibp; /*bit pointers (for out and in buffer)*/ - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - ibp = (8 * passstart[i]) + (y * ilinebits + x * bpp); - obp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; - for(b = 0; b < bpp; b++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - /*note that this function assumes the out buffer is completely 0, use setBitOfReversedStream otherwise*/ - setBitOfReversedStream0(&obp, out, bit); - } - } - } - } -} - -static void removePaddingBits(unsigned char* out, const unsigned char* in, - size_t olinebits, size_t ilinebits, unsigned h) -{ - /* - After filtering there are still padding bits if scanlines have non multiple of 8 bit amounts. They need - to be removed (except at last scanline of (Adam7-reduced) image) before working with pure image buffers - for the Adam7 code, the color convert code and the output to the user. - in and out are allowed to be the same buffer, in may also be higher but still overlapping; in must - have >= ilinebits*h bits, out must have >= olinebits*h bits, olinebits must be <= ilinebits - also used to move bits after earlier such operations happened, e.g. in a sequence of reduced images from Adam7 - only useful if (ilinebits - olinebits) is a value in the range 1..7 - */ - unsigned y; - size_t diff = ilinebits - olinebits; - size_t ibp = 0, obp = 0; /*input and output bit pointers*/ - for(y = 0; y < h; y++) - { - size_t x; - for(x = 0; x < olinebits; x++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - ibp += diff; - } -} - -/*out must be buffer big enough to contain full image, and in must contain the full decompressed data from -the IDAT chunks (with filter index bytes and possible padding bits) -return value is error*/ -static unsigned postProcessScanlines(unsigned char* out, unsigned char* in, - unsigned w, unsigned h, const LodePNGInfo* info_png) -{ - /* - This function converts the filtered-padded-interlaced data into pure 2D image buffer with the PNG's colortype. - Steps: - *) if no Adam7: 1) unfilter 2) remove padding bits (= posible extra bits per scanline if bpp < 8) - *) if adam7: 1) 7x unfilter 2) 7x remove padding bits 3) Adam7_deinterlace - NOTE: the in buffer will be overwritten with intermediate data! - */ - unsigned bpp = lodepng_get_bpp(&info_png->color); - if(bpp == 0) return 31; /*error: invalid colortype*/ - - if(info_png->interlace_method == 0) - { - if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) - { - CERROR_TRY_RETURN(unfilter(in, in, w, h, bpp)); - removePaddingBits(out, in, w * bpp, ((w * bpp + 7) / 8) * 8, h); - } - /*we can immediatly filter into the out buffer, no other steps needed*/ - else CERROR_TRY_RETURN(unfilter(out, in, w, h, bpp)); - } - else /*interlace_method is 1 (Adam7)*/ - { - unsigned passw[7], passh[7]; size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - for(i = 0; i < 7; i++) - { - CERROR_TRY_RETURN(unfilter(&in[padded_passstart[i]], &in[filter_passstart[i]], passw[i], passh[i], bpp)); - /*TODO: possible efficiency improvement: if in this reduced image the bits fit nicely in 1 scanline, - move bytes instead of bits or move not at all*/ - if(bpp < 8) - { - /*remove padding bits in scanlines; after this there still may be padding - bits between the different reduced images: each reduced image still starts nicely at a byte*/ - removePaddingBits(&in[passstart[i]], &in[padded_passstart[i]], passw[i] * bpp, - ((passw[i] * bpp + 7) / 8) * 8, passh[i]); - } - } - - Adam7_deinterlace(out, in, w, h, bpp); - } - - return 0; -} - -static unsigned readChunk_PLTE(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) -{ - unsigned pos = 0, i; - if(color->palette) lodepng_free(color->palette); - color->palettesize = chunkLength / 3; - color->palette = (unsigned char*)lodepng_malloc(4 * color->palettesize); - if(!color->palette && color->palettesize) - { - color->palettesize = 0; - return 83; /*alloc fail*/ - } - if(color->palettesize > 256) return 38; /*error: palette too big*/ - - for(i = 0; i < color->palettesize; i++) - { - color->palette[4 * i + 0] = data[pos++]; /*R*/ - color->palette[4 * i + 1] = data[pos++]; /*G*/ - color->palette[4 * i + 2] = data[pos++]; /*B*/ - color->palette[4 * i + 3] = 255; /*alpha*/ - } - - return 0; /* OK */ -} - -static unsigned readChunk_tRNS(LodePNGColorMode* color, const unsigned char* data, size_t chunkLength) -{ - unsigned i; - if(color->colortype == LCT_PALETTE) - { - /*error: more alpha values given than there are palette entries*/ - if(chunkLength > color->palettesize) return 38; - - for(i = 0; i < chunkLength; i++) color->palette[4 * i + 3] = data[i]; - } - else if(color->colortype == LCT_GREY) - { - /*error: this chunk must be 2 bytes for greyscale image*/ - if(chunkLength != 2) return 30; - - color->key_defined = 1; - color->key_r = color->key_g = color->key_b = 256 * data[0] + data[1]; - } - else if(color->colortype == LCT_RGB) - { - /*error: this chunk must be 6 bytes for RGB image*/ - if(chunkLength != 6) return 41; - - color->key_defined = 1; - color->key_r = 256 * data[0] + data[1]; - color->key_g = 256 * data[2] + data[3]; - color->key_b = 256 * data[4] + data[5]; - } - else return 42; /*error: tRNS chunk not allowed for other color models*/ - - return 0; /* OK */ -} - - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*background color chunk (bKGD)*/ -static unsigned readChunk_bKGD(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - if(info->color.colortype == LCT_PALETTE) - { - /*error: this chunk must be 1 byte for indexed color image*/ - if(chunkLength != 1) return 43; - - info->background_defined = 1; - info->background_r = info->background_g = info->background_b = data[0]; - } - else if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) - { - /*error: this chunk must be 2 bytes for greyscale image*/ - if(chunkLength != 2) return 44; - - info->background_defined = 1; - info->background_r = info->background_g = info->background_b - = 256 * data[0] + data[1]; - } - else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) - { - /*error: this chunk must be 6 bytes for greyscale image*/ - if(chunkLength != 6) return 45; - - info->background_defined = 1; - info->background_r = 256 * data[0] + data[1]; - info->background_g = 256 * data[2] + data[3]; - info->background_b = 256 * data[4] + data[5]; - } - - return 0; /* OK */ -} - -/*text chunk (tEXt)*/ -static unsigned readChunk_tEXt(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - unsigned error = 0; - char *key = 0, *str = 0; - unsigned i; - - while(!error) /*not really a while loop, only used to break on error*/ - { - unsigned length, string2_begin; - - length = 0; - while(length < chunkLength && data[length] != 0) length++; - /*even though it's not allowed by the standard, no error is thrown if - there's no null termination char, if the text is empty*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - key[length] = 0; - for(i = 0; i < length; i++) key[i] = data[i]; - - string2_begin = length + 1; /*skip keyword null terminator*/ - - length = chunkLength < string2_begin ? 0 : chunkLength - string2_begin; - str = (char*)lodepng_malloc(length + 1); - if(!str) CERROR_BREAK(error, 83); /*alloc fail*/ - - str[length] = 0; - for(i = 0; i < length; i++) str[i] = data[string2_begin + i]; - - error = lodepng_add_text(info, key, str); - - break; - } - - lodepng_free(key); - lodepng_free(str); - - return error; -} - -/*compressed text chunk (zTXt)*/ -static unsigned readChunk_zTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, - const unsigned char* data, size_t chunkLength) -{ - unsigned error = 0; - unsigned i; - - unsigned length, string2_begin; - char *key = 0; - ucvector decoded; - - ucvector_init(&decoded); - - while(!error) /*not really a while loop, only used to break on error*/ - { - for(length = 0; length < chunkLength && data[length] != 0; length++) ; - if(length + 2 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - key[length] = 0; - for(i = 0; i < length; i++) key[i] = data[i]; - - if(data[length + 1] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ - - string2_begin = length + 2; - if(string2_begin > chunkLength) CERROR_BREAK(error, 75); /*no null termination, corrupt?*/ - - length = chunkLength - string2_begin; - /*will fail if zlib error, e.g. if length is too small*/ - error = zlib_decompress(&decoded.data, &decoded.size, - (unsigned char*)(&data[string2_begin]), - length, zlibsettings); - if(error) break; - ucvector_push_back(&decoded, 0); - - error = lodepng_add_text(info, key, (char*)decoded.data); - - break; - } - - lodepng_free(key); - ucvector_cleanup(&decoded); - - return error; -} - -/*international text chunk (iTXt)*/ -static unsigned readChunk_iTXt(LodePNGInfo* info, const LodePNGDecompressSettings* zlibsettings, - const unsigned char* data, size_t chunkLength) -{ - unsigned error = 0; - unsigned i; - - unsigned length, begin, compressed; - char *key = 0, *langtag = 0, *transkey = 0; - ucvector decoded; - ucvector_init(&decoded); - - while(!error) /*not really a while loop, only used to break on error*/ - { - /*Quick check if the chunk length isn't too small. Even without check - it'd still fail with other error checks below if it's too short. This just gives a different error code.*/ - if(chunkLength < 5) CERROR_BREAK(error, 30); /*iTXt chunk too short*/ - - /*read the key*/ - for(length = 0; length < chunkLength && data[length] != 0; length++) ; - if(length + 3 >= chunkLength) CERROR_BREAK(error, 75); /*no null termination char, corrupt?*/ - if(length < 1 || length > 79) CERROR_BREAK(error, 89); /*keyword too short or long*/ - - key = (char*)lodepng_malloc(length + 1); - if(!key) CERROR_BREAK(error, 83); /*alloc fail*/ - - key[length] = 0; - for(i = 0; i < length; i++) key[i] = data[i]; - - /*read the compression method*/ - compressed = data[length + 1]; - if(data[length + 2] != 0) CERROR_BREAK(error, 72); /*the 0 byte indicating compression must be 0*/ - - /*even though it's not allowed by the standard, no error is thrown if - there's no null termination char, if the text is empty for the next 3 texts*/ - - /*read the langtag*/ - begin = length + 3; - length = 0; - for(i = begin; i < chunkLength && data[i] != 0; i++) length++; - - langtag = (char*)lodepng_malloc(length + 1); - if(!langtag) CERROR_BREAK(error, 83); /*alloc fail*/ - - langtag[length] = 0; - for(i = 0; i < length; i++) langtag[i] = data[begin + i]; - - /*read the transkey*/ - begin += length + 1; - length = 0; - for(i = begin; i < chunkLength && data[i] != 0; i++) length++; - - transkey = (char*)lodepng_malloc(length + 1); - if(!transkey) CERROR_BREAK(error, 83); /*alloc fail*/ - - transkey[length] = 0; - for(i = 0; i < length; i++) transkey[i] = data[begin + i]; - - /*read the actual text*/ - begin += length + 1; - - length = chunkLength < begin ? 0 : chunkLength - begin; - - if(compressed) - { - /*will fail if zlib error, e.g. if length is too small*/ - error = zlib_decompress(&decoded.data, &decoded.size, - (unsigned char*)(&data[begin]), - length, zlibsettings); - if(error) break; - if(decoded.allocsize < decoded.size) decoded.allocsize = decoded.size; - ucvector_push_back(&decoded, 0); - } - else - { - if(!ucvector_resize(&decoded, length + 1)) CERROR_BREAK(error, 83 /*alloc fail*/); - - decoded.data[length] = 0; - for(i = 0; i < length; i++) decoded.data[i] = data[begin + i]; - } - - error = lodepng_add_itext(info, key, langtag, transkey, (char*)decoded.data); - - break; - } - - lodepng_free(key); - lodepng_free(langtag); - lodepng_free(transkey); - ucvector_cleanup(&decoded); - - return error; -} - -static unsigned readChunk_tIME(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - if(chunkLength != 7) return 73; /*invalid tIME chunk size*/ - - info->time_defined = 1; - info->time.year = 256 * data[0] + data[+ 1]; - info->time.month = data[2]; - info->time.day = data[3]; - info->time.hour = data[4]; - info->time.minute = data[5]; - info->time.second = data[6]; - - return 0; /* OK */ -} - -static unsigned readChunk_pHYs(LodePNGInfo* info, const unsigned char* data, size_t chunkLength) -{ - if(chunkLength != 9) return 74; /*invalid pHYs chunk size*/ - - info->phys_defined = 1; - info->phys_x = 16777216 * data[0] + 65536 * data[1] + 256 * data[2] + data[3]; - info->phys_y = 16777216 * data[4] + 65536 * data[5] + 256 * data[6] + data[7]; - info->phys_unit = data[8]; - - return 0; /* OK */ -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/*read a PNG, the result will be in the same color type as the PNG (hence "generic")*/ -static void decodeGeneric(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize) -{ - unsigned char IEND = 0; - const unsigned char* chunk; - size_t i; - ucvector idat; /*the data from idat chunks*/ - - /*for unknown chunk order*/ - unsigned unknown = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - unsigned critical_pos = 1; /*1 = after IHDR, 2 = after PLTE, 3 = after IDAT*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - - /*provide some proper output values if error will happen*/ - *out = 0; - - state->error = lodepng_inspect(w, h, state, in, insize); /*reads header and resets other parameters in state->info_png*/ - if(state->error) return; - - ucvector_init(&idat); - chunk = &in[33]; /*first byte of the first chunk after the header*/ - - /*loop through the chunks, ignoring unknown chunks and stopping at IEND chunk. - IDAT data is put at the start of the in buffer*/ - while(!IEND && !state->error) - { - unsigned chunkLength; - const unsigned char* data; /*the data in the chunk*/ - - /*error: size of the in buffer too small to contain next chunk*/ - if((size_t)((chunk - in) + 12) > insize || chunk < in) CERROR_BREAK(state->error, 30); - - /*length of the data of the chunk, excluding the length bytes, chunk type and CRC bytes*/ - chunkLength = lodepng_chunk_length(chunk); - /*error: chunk length larger than the max PNG chunk size*/ - if(chunkLength > 2147483647) CERROR_BREAK(state->error, 63); - - if((size_t)((chunk - in) + chunkLength + 12) > insize || (chunk + chunkLength + 12) < in) - { - CERROR_BREAK(state->error, 64); /*error: size of the in buffer too small to contain next chunk*/ - } - - data = lodepng_chunk_data_const(chunk); - - /*IDAT chunk, containing compressed image data*/ - if(lodepng_chunk_type_equals(chunk, "IDAT")) - { - size_t oldsize = idat.size; - if(!ucvector_resize(&idat, oldsize + chunkLength)) CERROR_BREAK(state->error, 83 /*alloc fail*/); - for(i = 0; i < chunkLength; i++) idat.data[oldsize + i] = data[i]; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - critical_pos = 3; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - /*IEND chunk*/ - else if(lodepng_chunk_type_equals(chunk, "IEND")) - { - IEND = 1; - } - /*palette chunk (PLTE)*/ - else if(lodepng_chunk_type_equals(chunk, "PLTE")) - { - state->error = readChunk_PLTE(&state->info_png.color, data, chunkLength); - if(state->error) break; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - critical_pos = 2; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - /*palette transparency chunk (tRNS)*/ - else if(lodepng_chunk_type_equals(chunk, "tRNS")) - { - state->error = readChunk_tRNS(&state->info_png.color, data, chunkLength); - if(state->error) break; - } -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*background color chunk (bKGD)*/ - else if(lodepng_chunk_type_equals(chunk, "bKGD")) - { - state->error = readChunk_bKGD(&state->info_png, data, chunkLength); - if(state->error) break; - } - /*text chunk (tEXt)*/ - else if(lodepng_chunk_type_equals(chunk, "tEXt")) - { - if(state->decoder.read_text_chunks) - { - state->error = readChunk_tEXt(&state->info_png, data, chunkLength); - if(state->error) break; - } - } - /*compressed text chunk (zTXt)*/ - else if(lodepng_chunk_type_equals(chunk, "zTXt")) - { - if(state->decoder.read_text_chunks) - { - state->error = readChunk_zTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); - if(state->error) break; - } - } - /*international text chunk (iTXt)*/ - else if(lodepng_chunk_type_equals(chunk, "iTXt")) - { - if(state->decoder.read_text_chunks) - { - state->error = readChunk_iTXt(&state->info_png, &state->decoder.zlibsettings, data, chunkLength); - if(state->error) break; - } - } - else if(lodepng_chunk_type_equals(chunk, "tIME")) - { - state->error = readChunk_tIME(&state->info_png, data, chunkLength); - if(state->error) break; - } - else if(lodepng_chunk_type_equals(chunk, "pHYs")) - { - state->error = readChunk_pHYs(&state->info_png, data, chunkLength); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - else /*it's not an implemented chunk type, so ignore it: skip over the data*/ - { - /*error: unknown critical chunk (5th bit of first byte of chunk type is 0)*/ - if(!lodepng_chunk_ancillary(chunk)) CERROR_BREAK(state->error, 69); - - unknown = 1; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - if(state->decoder.remember_unknown_chunks) - { - state->error = lodepng_chunk_append(&state->info_png.unknown_chunks_data[critical_pos - 1], - &state->info_png.unknown_chunks_size[critical_pos - 1], chunk); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - } - - if(!state->decoder.ignore_crc && !unknown) /*check CRC if wanted, only on known chunk types*/ - { - if(lodepng_chunk_check_crc(chunk)) CERROR_BREAK(state->error, 57); /*invalid CRC*/ - } - - if(!IEND) chunk = lodepng_chunk_next_const(chunk); - } - - if(!state->error) - { - ucvector scanlines; - ucvector_init(&scanlines); - - /*maximum final image length is already reserved in the vector's length - this is not really necessary*/ - if(!ucvector_resize(&scanlines, lodepng_get_raw_size(*w, *h, &state->info_png.color) + *h)) - { - state->error = 83; /*alloc fail*/ - } - if(!state->error) - { - /*decompress with the Zlib decompressor*/ - state->error = zlib_decompress(&scanlines.data, &scanlines.size, idat.data, - idat.size, &state->decoder.zlibsettings); - } - - if(!state->error) - { - ucvector outv; - ucvector_init(&outv); - if(!ucvector_resizev(&outv, - lodepng_get_raw_size(*w, *h, &state->info_png.color), 0)) state->error = 83; /*alloc fail*/ - if(!state->error) state->error = postProcessScanlines(outv.data, scanlines.data, *w, *h, &state->info_png); - *out = outv.data; - } - ucvector_cleanup(&scanlines); - } - - ucvector_cleanup(&idat); -} - -unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize) -{ - *out = 0; - decodeGeneric(out, w, h, state, in, insize); - if(state->error) return state->error; - if(!state->decoder.color_convert || lodepng_color_mode_equal(&state->info_raw, &state->info_png.color)) - { - /*same color type, no copying or converting of data needed*/ - /*store the info_png color settings on the info_raw so that the info_raw still reflects what colortype - the raw image has to the end user*/ - if(!state->decoder.color_convert) - { - state->error = lodepng_color_mode_copy(&state->info_raw, &state->info_png.color); - if(state->error) return state->error; - } - } - else - { - /*color conversion needed; sort of copy of the data*/ - unsigned char* data = *out; - size_t outsize; - - /*TODO: check if this works according to the statement in the documentation: "The converter can convert - from greyscale input color type, to 8-bit greyscale or greyscale with alpha"*/ - if(!(state->info_raw.colortype == LCT_RGB || state->info_raw.colortype == LCT_RGBA) - && !(state->info_raw.bitdepth == 8)) - { - return 56; /*unsupported color mode conversion*/ - } - - outsize = lodepng_get_raw_size(*w, *h, &state->info_raw); - *out = (unsigned char*)lodepng_malloc(outsize); - if(!(*out)) - { - state->error = 83; /*alloc fail*/ - } - else state->error = lodepng_convert(*out, data, &state->info_raw, &state->info_png.color, *w, *h, state->decoder.fix_png); - lodepng_free(data); - } - return state->error; -} - -unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, - size_t insize, LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned error; - LodePNGState state; - lodepng_state_init(&state); - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - error = lodepng_decode(out, w, h, &state, in, insize); - lodepng_state_cleanup(&state); - return error; -} - -unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) -{ - return lodepng_decode_memory(out, w, h, in, insize, LCT_RGBA, 8); -} - -unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, const unsigned char* in, size_t insize) -{ - return lodepng_decode_memory(out, w, h, in, insize, LCT_RGB, 8); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename, - LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned char* buffer; - size_t buffersize; - unsigned error; - error = lodepng_load_file(&buffer, &buffersize, filename); - if(!error) error = lodepng_decode_memory(out, w, h, buffer, buffersize, colortype, bitdepth); - lodepng_free(buffer); - return error; -} - -unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) -{ - return lodepng_decode_file(out, w, h, filename, LCT_RGBA, 8); -} - -unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, const char* filename) -{ - return lodepng_decode_file(out, w, h, filename, LCT_RGB, 8); -} -#endif /*LODEPNG_COMPILE_DISK*/ - -void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings) -{ - settings->color_convert = 1; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - settings->read_text_chunks = 1; - settings->remember_unknown_chunks = 0; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - settings->ignore_crc = 0; - settings->fix_png = 0; - lodepng_decompress_settings_init(&settings->zlibsettings); -} - -#endif /*LODEPNG_COMPILE_DECODER*/ - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) - -void lodepng_state_init(LodePNGState* state) -{ -#ifdef LODEPNG_COMPILE_DECODER - lodepng_decoder_settings_init(&state->decoder); -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER - lodepng_encoder_settings_init(&state->encoder); -#endif /*LODEPNG_COMPILE_ENCODER*/ - lodepng_color_mode_init(&state->info_raw); - lodepng_info_init(&state->info_png); - state->error = 1; -} - -void lodepng_state_cleanup(LodePNGState* state) -{ - lodepng_color_mode_cleanup(&state->info_raw); - lodepng_info_cleanup(&state->info_png); -} - -void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source) -{ - lodepng_state_cleanup(dest); - *dest = *source; - lodepng_color_mode_init(&dest->info_raw); - lodepng_info_init(&dest->info_png); - dest->error = lodepng_color_mode_copy(&dest->info_raw, &source->info_raw); if(dest->error) return; - dest->error = lodepng_info_copy(&dest->info_png, &source->info_png); if(dest->error) return; -} - -#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ - -#ifdef LODEPNG_COMPILE_ENCODER - -/* ////////////////////////////////////////////////////////////////////////// */ -/* / PNG Encoder / */ -/* ////////////////////////////////////////////////////////////////////////// */ - -/*chunkName must be string of 4 characters*/ -static unsigned addChunk(ucvector* out, const char* chunkName, const unsigned char* data, size_t length) -{ - CERROR_TRY_RETURN(lodepng_chunk_create(&out->data, &out->size, (unsigned)length, chunkName, data)); - out->allocsize = out->size; /*fix the allocsize again*/ - return 0; -} - -static void writeSignature(ucvector* out) -{ - /*8 bytes PNG signature, aka the magic bytes*/ - ucvector_push_back(out, 137); - ucvector_push_back(out, 80); - ucvector_push_back(out, 78); - ucvector_push_back(out, 71); - ucvector_push_back(out, 13); - ucvector_push_back(out, 10); - ucvector_push_back(out, 26); - ucvector_push_back(out, 10); -} - -static unsigned addChunk_IHDR(ucvector* out, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth, unsigned interlace_method) -{ - unsigned error = 0; - ucvector header; - ucvector_init(&header); - - lodepng_add32bitInt(&header, w); /*width*/ - lodepng_add32bitInt(&header, h); /*height*/ - ucvector_push_back(&header, (unsigned char)bitdepth); /*bit depth*/ - ucvector_push_back(&header, (unsigned char)colortype); /*color type*/ - ucvector_push_back(&header, 0); /*compression method*/ - ucvector_push_back(&header, 0); /*filter method*/ - ucvector_push_back(&header, interlace_method); /*interlace method*/ - - error = addChunk(out, "IHDR", header.data, header.size); - ucvector_cleanup(&header); - - return error; -} - -static unsigned addChunk_PLTE(ucvector* out, const LodePNGColorMode* info) -{ - unsigned error = 0; - size_t i; - ucvector PLTE; - ucvector_init(&PLTE); - for(i = 0; i < info->palettesize * 4; i++) - { - /*add all channels except alpha channel*/ - if(i % 4 != 3) ucvector_push_back(&PLTE, info->palette[i]); - } - error = addChunk(out, "PLTE", PLTE.data, PLTE.size); - ucvector_cleanup(&PLTE); - - return error; -} - -static unsigned addChunk_tRNS(ucvector* out, const LodePNGColorMode* info) -{ - unsigned error = 0; - size_t i; - ucvector tRNS; - ucvector_init(&tRNS); - if(info->colortype == LCT_PALETTE) - { - size_t amount = info->palettesize; - /*the tail of palette values that all have 255 as alpha, does not have to be encoded*/ - for(i = info->palettesize; i > 0; i--) - { - if(info->palette[4 * (i - 1) + 3] == 255) amount--; - else break; - } - /*add only alpha channel*/ - for(i = 0; i < amount; i++) ucvector_push_back(&tRNS, info->palette[4 * i + 3]); - } - else if(info->colortype == LCT_GREY) - { - if(info->key_defined) - { - ucvector_push_back(&tRNS, (unsigned char)(info->key_r / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_r % 256)); - } - } - else if(info->colortype == LCT_RGB) - { - if(info->key_defined) - { - ucvector_push_back(&tRNS, (unsigned char)(info->key_r / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_r % 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_g / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_g % 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_b / 256)); - ucvector_push_back(&tRNS, (unsigned char)(info->key_b % 256)); - } - } - - error = addChunk(out, "tRNS", tRNS.data, tRNS.size); - ucvector_cleanup(&tRNS); - - return error; -} - -static unsigned addChunk_IDAT(ucvector* out, const unsigned char* data, size_t datasize, - LodePNGCompressSettings* zlibsettings) -{ - ucvector zlibdata; - unsigned error = 0; - - /*compress with the Zlib compressor*/ - ucvector_init(&zlibdata); - error = zlib_compress(&zlibdata.data, &zlibdata.size, data, datasize, zlibsettings); - if(!error) error = addChunk(out, "IDAT", zlibdata.data, zlibdata.size); - ucvector_cleanup(&zlibdata); - - return error; -} - -static unsigned addChunk_IEND(ucvector* out) -{ - unsigned error = 0; - error = addChunk(out, "IEND", 0, 0); - return error; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - -static unsigned addChunk_tEXt(ucvector* out, const char* keyword, const char* textstring) -{ - unsigned error = 0; - size_t i; - ucvector text; - ucvector_init(&text); - for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&text, (unsigned char)keyword[i]); - if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ - ucvector_push_back(&text, 0); /*0 termination char*/ - for(i = 0; textstring[i] != 0; i++) ucvector_push_back(&text, (unsigned char)textstring[i]); - error = addChunk(out, "tEXt", text.data, text.size); - ucvector_cleanup(&text); - - return error; -} - -static unsigned addChunk_zTXt(ucvector* out, const char* keyword, const char* textstring, - LodePNGCompressSettings* zlibsettings) -{ - unsigned error = 0; - ucvector data, compressed; - size_t i, textsize = strlen(textstring); - - ucvector_init(&data); - ucvector_init(&compressed); - for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&data, (unsigned char)keyword[i]); - if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ - ucvector_push_back(&data, 0); /*0 termination char*/ - ucvector_push_back(&data, 0); /*compression method: 0*/ - - error = zlib_compress(&compressed.data, &compressed.size, - (unsigned char*)textstring, textsize, zlibsettings); - if(!error) - { - for(i = 0; i < compressed.size; i++) ucvector_push_back(&data, compressed.data[i]); - error = addChunk(out, "zTXt", data.data, data.size); - } - - ucvector_cleanup(&compressed); - ucvector_cleanup(&data); - return error; -} - -static unsigned addChunk_iTXt(ucvector* out, unsigned compressed, const char* keyword, const char* langtag, - const char* transkey, const char* textstring, LodePNGCompressSettings* zlibsettings) -{ - unsigned error = 0; - ucvector data; - size_t i, textsize = strlen(textstring); - - ucvector_init(&data); - - for(i = 0; keyword[i] != 0; i++) ucvector_push_back(&data, (unsigned char)keyword[i]); - if(i < 1 || i > 79) return 89; /*error: invalid keyword size*/ - ucvector_push_back(&data, 0); /*null termination char*/ - ucvector_push_back(&data, compressed ? 1 : 0); /*compression flag*/ - ucvector_push_back(&data, 0); /*compression method*/ - for(i = 0; langtag[i] != 0; i++) ucvector_push_back(&data, (unsigned char)langtag[i]); - ucvector_push_back(&data, 0); /*null termination char*/ - for(i = 0; transkey[i] != 0; i++) ucvector_push_back(&data, (unsigned char)transkey[i]); - ucvector_push_back(&data, 0); /*null termination char*/ - - if(compressed) - { - ucvector compressed_data; - ucvector_init(&compressed_data); - error = zlib_compress(&compressed_data.data, &compressed_data.size, - (unsigned char*)textstring, textsize, zlibsettings); - if(!error) - { - for(i = 0; i < compressed_data.size; i++) ucvector_push_back(&data, compressed_data.data[i]); - } - ucvector_cleanup(&compressed_data); - } - else /*not compressed*/ - { - for(i = 0; textstring[i] != 0; i++) ucvector_push_back(&data, (unsigned char)textstring[i]); - } - - if(!error) error = addChunk(out, "iTXt", data.data, data.size); - ucvector_cleanup(&data); - return error; -} - -static unsigned addChunk_bKGD(ucvector* out, const LodePNGInfo* info) -{ - unsigned error = 0; - ucvector bKGD; - ucvector_init(&bKGD); - if(info->color.colortype == LCT_GREY || info->color.colortype == LCT_GREY_ALPHA) - { - ucvector_push_back(&bKGD, (unsigned char)(info->background_r / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); - } - else if(info->color.colortype == LCT_RGB || info->color.colortype == LCT_RGBA) - { - ucvector_push_back(&bKGD, (unsigned char)(info->background_r / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_g / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_g % 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_b / 256)); - ucvector_push_back(&bKGD, (unsigned char)(info->background_b % 256)); - } - else if(info->color.colortype == LCT_PALETTE) - { - ucvector_push_back(&bKGD, (unsigned char)(info->background_r % 256)); /*palette index*/ - } - - error = addChunk(out, "bKGD", bKGD.data, bKGD.size); - ucvector_cleanup(&bKGD); - - return error; -} - -static unsigned addChunk_tIME(ucvector* out, const LodePNGTime* time) -{ - unsigned error = 0; - unsigned char* data = (unsigned char*)lodepng_malloc(7); - if(!data) return 83; /*alloc fail*/ - data[0] = (unsigned char)(time->year / 256); - data[1] = (unsigned char)(time->year % 256); - data[2] = time->month; - data[3] = time->day; - data[4] = time->hour; - data[5] = time->minute; - data[6] = time->second; - error = addChunk(out, "tIME", data, 7); - lodepng_free(data); - return error; -} - -static unsigned addChunk_pHYs(ucvector* out, const LodePNGInfo* info) -{ - unsigned error = 0; - ucvector data; - ucvector_init(&data); - - lodepng_add32bitInt(&data, info->phys_x); - lodepng_add32bitInt(&data, info->phys_y); - ucvector_push_back(&data, info->phys_unit); - - error = addChunk(out, "pHYs", data.data, data.size); - ucvector_cleanup(&data); - - return error; -} - -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -static void filterScanline(unsigned char* out, const unsigned char* scanline, const unsigned char* prevline, - size_t length, size_t bytewidth, unsigned char filterType) -{ - size_t i; - switch(filterType) - { - case 0: /*None*/ - for(i = 0; i < length; i++) out[i] = scanline[i]; - break; - case 1: /*Sub*/ - if(prevline) - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth]; - } - else - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth]; - } - break; - case 2: /*Up*/ - if(prevline) - { - for(i = 0; i < length; i++) out[i] = scanline[i] - prevline[i]; - } - else - { - for(i = 0; i < length; i++) out[i] = scanline[i]; - } - break; - case 3: /*Average*/ - if(prevline) - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i] - prevline[i] / 2; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - ((scanline[i - bytewidth] + prevline[i]) / 2); - } - else - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - for(i = bytewidth; i < length; i++) out[i] = scanline[i] - scanline[i - bytewidth] / 2; - } - break; - case 4: /*Paeth*/ - if(prevline) - { - /*paethPredictor(0, prevline[i], 0) is always prevline[i]*/ - for(i = 0; i < bytewidth; i++) out[i] = (scanline[i] - prevline[i]); - for(i = bytewidth; i < length; i++) - { - out[i] = (scanline[i] - paethPredictor(scanline[i - bytewidth], prevline[i], prevline[i - bytewidth])); - } - } - else - { - for(i = 0; i < bytewidth; i++) out[i] = scanline[i]; - /*paethPredictor(scanline[i - bytewidth], 0, 0) is always scanline[i - bytewidth]*/ - for(i = bytewidth; i < length; i++) out[i] = (scanline[i] - scanline[i - bytewidth]); - } - break; - default: return; /*unexisting filter type given*/ - } -} - -/* log2 approximation. A slight bit faster than std::log. */ -static float flog2(float f) -{ - float result = 0; - while(f > 32) { result += 4; f /= 16; } - while(f > 2) { result++; f /= 2; } - return result + 1.442695f * (f * f * f / 3 - 3 * f * f / 2 + 3 * f - 1.83333f); -} - -static unsigned filter(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, - const LodePNGColorMode* info, const LodePNGEncoderSettings* settings) -{ - /* - For PNG filter method 0 - out must be a buffer with as size: h + (w * h * bpp + 7) / 8, because there are - the scanlines with 1 extra byte per scanline - */ - - unsigned bpp = lodepng_get_bpp(info); - /*the width of a scanline in bytes, not including the filter type*/ - size_t linebytes = (w * bpp + 7) / 8; - /*bytewidth is used for filtering, is 1 when bpp < 8, number of bytes per pixel otherwise*/ - size_t bytewidth = (bpp + 7) / 8; - const unsigned char* prevline = 0; - unsigned x, y; - unsigned error = 0; - LodePNGFilterStrategy strategy = settings->filter_strategy; - - /* - There is a heuristic called the minimum sum of absolute differences heuristic, suggested by the PNG standard: - * If the image type is Palette, or the bit depth is smaller than 8, then do not filter the image (i.e. - use fixed filtering, with the filter None). - * (The other case) If the image type is Grayscale or RGB (with or without Alpha), and the bit depth is - not smaller than 8, then use adaptive filtering heuristic as follows: independently for each row, apply - all five filters and select the filter that produces the smallest sum of absolute values per row. - This heuristic is used if filter strategy is LFS_MINSUM and filter_palette_zero is true. - - If filter_palette_zero is true and filter_strategy is not LFS_MINSUM, the above heuristic is followed, - but for "the other case", whatever strategy filter_strategy is set to instead of the minimum sum - heuristic is used. - */ - if(settings->filter_palette_zero && - (info->colortype == LCT_PALETTE || info->bitdepth < 8)) strategy = LFS_ZERO; - - if(bpp == 0) return 31; /*error: invalid color type*/ - - if(strategy == LFS_ZERO) - { - for(y = 0; y < h; y++) - { - size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - size_t inindex = linebytes * y; - out[outindex] = 0; /*filter type byte*/ - filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, 0); - prevline = &in[inindex]; - } - } - else if(strategy == LFS_MINSUM) - { - /*adaptive filtering*/ - size_t sum[5]; - ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t smallest = 0; - unsigned type, bestType = 0; - - for(type = 0; type < 5; type++) - { - ucvector_init(&attempt[type]); - if(!ucvector_resize(&attempt[type], linebytes)) return 83; /*alloc fail*/ - } - - if(!error) - { - for(y = 0; y < h; y++) - { - /*try the 5 filter types*/ - for(type = 0; type < 5; type++) - { - filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); - - /*calculate the sum of the result*/ - sum[type] = 0; - if(type == 0) - { - for(x = 0; x < linebytes; x++) sum[type] += (unsigned char)(attempt[type].data[x]); - } - else - { - for(x = 0; x < linebytes; x++) - { - /*For differences, each byte should be treated as signed, values above 127 are negative - (converted to signed char). Filtertype 0 isn't a difference though, so use unsigned there. - This means filtertype 0 is almost never chosen, but that is justified.*/ - signed char s = (signed char)(attempt[type].data[x]); - sum[type] += s < 0 ? -s : s; - } - } - - /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || sum[type] < smallest) - { - bestType = type; - smallest = sum[type]; - } - } - - prevline = &in[y * linebytes]; - - /*now fill the out values*/ - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; - } - } - - for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); - } - else if(strategy == LFS_ENTROPY) - { - float sum[5]; - ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ - float smallest = 0; - unsigned type, bestType = 0; - unsigned count[256]; - - for(type = 0; type < 5; type++) - { - ucvector_init(&attempt[type]); - if(!ucvector_resize(&attempt[type], linebytes)) return 83; /*alloc fail*/ - } - - for(y = 0; y < h; y++) - { - /*try the 5 filter types*/ - for(type = 0; type < 5; type++) - { - filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); - for(x = 0; x < 256; x++) count[x] = 0; - for(x = 0; x < linebytes; x++) count[attempt[type].data[x]]++; - count[type]++; /*the filter type itself is part of the scanline*/ - sum[type] = 0; - for(x = 0; x < 256; x++) - { - float p = count[x] / (float)(linebytes + 1); - sum[type] += count[x] == 0 ? 0 : flog2(1 / p) * p; - } - /*check if this is smallest sum (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || sum[type] < smallest) - { - bestType = type; - smallest = sum[type]; - } - } - - prevline = &in[y * linebytes]; - - /*now fill the out values*/ - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; - } - - for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); - } - else if(strategy == LFS_PREDEFINED) - { - for(y = 0; y < h; y++) - { - size_t outindex = (1 + linebytes) * y; /*the extra filterbyte added to each row*/ - size_t inindex = linebytes * y; - unsigned type = settings->predefined_filters[y]; - out[outindex] = type; /*filter type byte*/ - filterScanline(&out[outindex + 1], &in[inindex], prevline, linebytes, bytewidth, type); - prevline = &in[inindex]; - } - } - else if(strategy == LFS_BRUTE_FORCE) - { - /*brute force filter chooser. - deflate the scanline after every filter attempt to see which one deflates best. - This is very slow and gives only slightly smaller, sometimes even larger, result*/ - size_t size[5]; - ucvector attempt[5]; /*five filtering attempts, one for each filter type*/ - size_t smallest = 0; - unsigned type = 0, bestType = 0; - unsigned char* dummy; - LodePNGCompressSettings zlibsettings = settings->zlibsettings; - /*use fixed tree on the attempts so that the tree is not adapted to the filtertype on purpose, - to simulate the true case where the tree is the same for the whole image. Sometimes it gives - better result with dynamic tree anyway. Using the fixed tree sometimes gives worse, but in rare - cases better compression. It does make this a bit less slow, so it's worth doing this.*/ - zlibsettings.btype = 1; - /*a custom encoder likely doesn't read the btype setting and is optimized for complete PNG - images only, so disable it*/ - zlibsettings.custom_zlib = 0; - zlibsettings.custom_deflate = 0; - for(type = 0; type < 5; type++) - { - ucvector_init(&attempt[type]); - ucvector_resize(&attempt[type], linebytes); /*todo: give error if resize failed*/ - } - for(y = 0; y < h; y++) /*try the 5 filter types*/ - { - for(type = 0; type < 5; type++) - { - unsigned testsize = attempt[type].size; - /*if(testsize > 8) testsize /= 8;*/ /*it already works good enough by testing a part of the row*/ - - filterScanline(attempt[type].data, &in[y * linebytes], prevline, linebytes, bytewidth, type); - size[type] = 0; - dummy = 0; - zlib_compress(&dummy, &size[type], attempt[type].data, testsize, &zlibsettings); - lodepng_free(dummy); - /*check if this is smallest size (or if type == 0 it's the first case so always store the values)*/ - if(type == 0 || size[type] < smallest) - { - bestType = type; - smallest = size[type]; - } - } - prevline = &in[y * linebytes]; - out[y * (linebytes + 1)] = bestType; /*the first byte of a scanline will be the filter type*/ - for(x = 0; x < linebytes; x++) out[y * (linebytes + 1) + 1 + x] = attempt[bestType].data[x]; - } - for(type = 0; type < 5; type++) ucvector_cleanup(&attempt[type]); - } - else return 88; /* unknown filter strategy */ - - return error; -} - -static void addPaddingBits(unsigned char* out, const unsigned char* in, - size_t olinebits, size_t ilinebits, unsigned h) -{ - /*The opposite of the removePaddingBits function - olinebits must be >= ilinebits*/ - unsigned y; - size_t diff = olinebits - ilinebits; - size_t obp = 0, ibp = 0; /*bit pointers*/ - for(y = 0; y < h; y++) - { - size_t x; - for(x = 0; x < ilinebits; x++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - /*obp += diff; --> no, fill in some value in the padding bits too, to avoid - "Use of uninitialised value of size ###" warning from valgrind*/ - for(x = 0; x < diff; x++) setBitOfReversedStream(&obp, out, 0); - } -} - -/* -in: non-interlaced image with size w*h -out: the same pixels, but re-ordered according to PNG's Adam7 interlacing, with - no padding bits between scanlines, but between reduced images so that each - reduced image starts at a byte. -bpp: bits per pixel -there are no padding bits, not between scanlines, not between reduced images -in has the following size in bits: w * h * bpp. -out is possibly bigger due to padding bits between reduced images -NOTE: comments about padding bits are only relevant if bpp < 8 -*/ -static void Adam7_interlace(unsigned char* out, const unsigned char* in, unsigned w, unsigned h, unsigned bpp) -{ - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned i; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - if(bpp >= 8) - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - size_t bytewidth = bpp / 8; - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - size_t pixelinstart = ((ADAM7_IY[i] + y * ADAM7_DY[i]) * w + ADAM7_IX[i] + x * ADAM7_DX[i]) * bytewidth; - size_t pixeloutstart = passstart[i] + (y * passw[i] + x) * bytewidth; - for(b = 0; b < bytewidth; b++) - { - out[pixeloutstart + b] = in[pixelinstart + b]; - } - } - } - } - else /*bpp < 8: Adam7 with pixels < 8 bit is a bit trickier: with bit pointers*/ - { - for(i = 0; i < 7; i++) - { - unsigned x, y, b; - unsigned ilinebits = bpp * passw[i]; - unsigned olinebits = bpp * w; - size_t obp, ibp; /*bit pointers (for out and in buffer)*/ - for(y = 0; y < passh[i]; y++) - for(x = 0; x < passw[i]; x++) - { - ibp = (ADAM7_IY[i] + y * ADAM7_DY[i]) * olinebits + (ADAM7_IX[i] + x * ADAM7_DX[i]) * bpp; - obp = (8 * passstart[i]) + (y * ilinebits + x * bpp); - for(b = 0; b < bpp; b++) - { - unsigned char bit = readBitFromReversedStream(&ibp, in); - setBitOfReversedStream(&obp, out, bit); - } - } - } - } -} - -/*out must be buffer big enough to contain uncompressed IDAT chunk data, and in must contain the full image. -return value is error**/ -static unsigned preProcessScanlines(unsigned char** out, size_t* outsize, const unsigned char* in, - unsigned w, unsigned h, - const LodePNGInfo* info_png, const LodePNGEncoderSettings* settings) -{ - /* - This function converts the pure 2D image with the PNG's colortype, into filtered-padded-interlaced data. Steps: - *) if no Adam7: 1) add padding bits (= posible extra bits per scanline if bpp < 8) 2) filter - *) if adam7: 1) Adam7_interlace 2) 7x add padding bits 3) 7x filter - */ - unsigned bpp = lodepng_get_bpp(&info_png->color); - unsigned error = 0; - - if(info_png->interlace_method == 0) - { - *outsize = h + (h * ((w * bpp + 7) / 8)); /*image size plus an extra byte per scanline + possible padding bits*/ - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!(*out) && (*outsize)) error = 83; /*alloc fail*/ - - if(!error) - { - /*non multiple of 8 bits per scanline, padding bits needed per scanline*/ - if(bpp < 8 && w * bpp != ((w * bpp + 7) / 8) * 8) - { - unsigned char* padded = (unsigned char*)lodepng_malloc(h * ((w * bpp + 7) / 8)); - if(!padded) error = 83; /*alloc fail*/ - if(!error) - { - addPaddingBits(padded, in, ((w * bpp + 7) / 8) * 8, w * bpp, h); - error = filter(*out, padded, w, h, &info_png->color, settings); - } - lodepng_free(padded); - } - else - { - /*we can immediatly filter into the out buffer, no other steps needed*/ - error = filter(*out, in, w, h, &info_png->color, settings); - } - } - } - else /*interlace_method is 1 (Adam7)*/ - { - unsigned passw[7], passh[7]; - size_t filter_passstart[8], padded_passstart[8], passstart[8]; - unsigned char* adam7; - - Adam7_getpassvalues(passw, passh, filter_passstart, padded_passstart, passstart, w, h, bpp); - - *outsize = filter_passstart[7]; /*image size plus an extra byte per scanline + possible padding bits*/ - *out = (unsigned char*)lodepng_malloc(*outsize); - if(!(*out)) error = 83; /*alloc fail*/ - - adam7 = (unsigned char*)lodepng_malloc(passstart[7]); - if(!adam7 && passstart[7]) error = 83; /*alloc fail*/ - - if(!error) - { - unsigned i; - - Adam7_interlace(adam7, in, w, h, bpp); - for(i = 0; i < 7; i++) - { - if(bpp < 8) - { - unsigned char* padded = (unsigned char*)lodepng_malloc(padded_passstart[i + 1] - padded_passstart[i]); - if(!padded) ERROR_BREAK(83); /*alloc fail*/ - addPaddingBits(padded, &adam7[passstart[i]], - ((passw[i] * bpp + 7) / 8) * 8, passw[i] * bpp, passh[i]); - error = filter(&(*out)[filter_passstart[i]], padded, - passw[i], passh[i], &info_png->color, settings); - lodepng_free(padded); - } - else - { - error = filter(&(*out)[filter_passstart[i]], &adam7[padded_passstart[i]], - passw[i], passh[i], &info_png->color, settings); - } - - if(error) break; - } - } - - lodepng_free(adam7); - } - - return error; -} - -/* -palette must have 4 * palettesize bytes allocated, and given in format RGBARGBARGBARGBA... -returns 0 if the palette is opaque, -returns 1 if the palette has a single color with alpha 0 ==> color key -returns 2 if the palette is semi-translucent. -*/ -static unsigned getPaletteTranslucency(const unsigned char* palette, size_t palettesize) -{ - size_t i, key = 0; - unsigned r = 0, g = 0, b = 0; /*the value of the color with alpha 0, so long as color keying is possible*/ - for(i = 0; i < palettesize; i++) - { - if(!key && palette[4 * i + 3] == 0) - { - r = palette[4 * i + 0]; g = palette[4 * i + 1]; b = palette[4 * i + 2]; - key = 1; - i = (size_t)(-1); /*restart from beginning, to detect earlier opaque colors with key's value*/ - } - else if(palette[4 * i + 3] != 255) return 2; - /*when key, no opaque RGB may have key's RGB*/ - else if(key && r == palette[i * 4 + 0] && g == palette[i * 4 + 1] && b == palette[i * 4 + 2]) return 2; - } - return key; -} - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -static unsigned addUnknownChunks(ucvector* out, unsigned char* data, size_t datasize) -{ - unsigned char* inchunk = data; - while((size_t)(inchunk - data) < datasize) - { - CERROR_TRY_RETURN(lodepng_chunk_append(&out->data, &out->size, inchunk)); - out->allocsize = out->size; /*fix the allocsize again*/ - inchunk = lodepng_chunk_next(inchunk); - } - return 0; -} -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -unsigned lodepng_encode(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGState* state) -{ - LodePNGInfo info; - ucvector outv; - unsigned char* data = 0; /*uncompressed version of the IDAT chunk data*/ - size_t datasize = 0; - - /*provide some proper output values if error will happen*/ - *out = 0; - *outsize = 0; - state->error = 0; - - lodepng_info_init(&info); - lodepng_info_copy(&info, &state->info_png); - - if((info.color.colortype == LCT_PALETTE || state->encoder.force_palette) - && (info.color.palettesize == 0 || info.color.palettesize > 256)) - { - state->error = 68; /*invalid palette size, it is only allowed to be 1-256*/ - return state->error; - } - - if(state->encoder.auto_convert != LAC_NO) - { - state->error = doAutoChooseColor(&info.color, image, w, h, &state->info_raw, - state->encoder.auto_convert); - } - if(state->error) return state->error; - - if(state->encoder.zlibsettings.windowsize > 32768) - { - CERROR_RETURN_ERROR(state->error, 60); /*error: windowsize larger than allowed*/ - } - if(state->encoder.zlibsettings.btype > 2) - { - CERROR_RETURN_ERROR(state->error, 61); /*error: unexisting btype*/ - } - if(state->info_png.interlace_method > 1) - { - CERROR_RETURN_ERROR(state->error, 71); /*error: unexisting interlace mode*/ - } - - state->error = checkColorValidity(info.color.colortype, info.color.bitdepth); - if(state->error) return state->error; /*error: unexisting color type given*/ - state->error = checkColorValidity(state->info_raw.colortype, state->info_raw.bitdepth); - if(state->error) return state->error; /*error: unexisting color type given*/ - - if(!lodepng_color_mode_equal(&state->info_raw, &info.color)) - { - unsigned char* converted; - size_t size = (w * h * lodepng_get_bpp(&info.color) + 7) / 8; - - converted = (unsigned char*)lodepng_malloc(size); - if(!converted && size) state->error = 83; /*alloc fail*/ - if(!state->error) - { - state->error = lodepng_convert(converted, image, &info.color, &state->info_raw, w, h, 0 /*fix_png*/); - } - if(!state->error) preProcessScanlines(&data, &datasize, converted, w, h, &info, &state->encoder); - lodepng_free(converted); - } - else preProcessScanlines(&data, &datasize, image, w, h, &info, &state->encoder); - - ucvector_init(&outv); - while(!state->error) /*while only executed once, to break on error*/ - { -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - size_t i; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*write signature and chunks*/ - writeSignature(&outv); - /*IHDR*/ - addChunk_IHDR(&outv, w, h, info.color.colortype, info.color.bitdepth, info.interlace_method); -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*unknown chunks between IHDR and PLTE*/ - if(info.unknown_chunks_data[0]) - { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[0], info.unknown_chunks_size[0]); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*PLTE*/ - if(info.color.colortype == LCT_PALETTE) - { - addChunk_PLTE(&outv, &info.color); - } - if(state->encoder.force_palette && (info.color.colortype == LCT_RGB || info.color.colortype == LCT_RGBA)) - { - addChunk_PLTE(&outv, &info.color); - } - /*tRNS*/ - if(info.color.colortype == LCT_PALETTE && getPaletteTranslucency(info.color.palette, info.color.palettesize) != 0) - { - addChunk_tRNS(&outv, &info.color); - } - if((info.color.colortype == LCT_GREY || info.color.colortype == LCT_RGB) && info.color.key_defined) - { - addChunk_tRNS(&outv, &info.color); - } -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*bKGD (must come between PLTE and the IDAt chunks*/ - if(info.background_defined) addChunk_bKGD(&outv, &info); - /*pHYs (must come before the IDAT chunks)*/ - if(info.phys_defined) addChunk_pHYs(&outv, &info); - - /*unknown chunks between PLTE and IDAT*/ - if(info.unknown_chunks_data[1]) - { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[1], info.unknown_chunks_size[1]); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*IDAT (multiple IDAT chunks must be consecutive)*/ - state->error = addChunk_IDAT(&outv, data, datasize, &state->encoder.zlibsettings); - if(state->error) break; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*tIME*/ - if(info.time_defined) addChunk_tIME(&outv, &info.time); - /*tEXt and/or zTXt*/ - for(i = 0; i < info.text_num; i++) - { - if(strlen(info.text_keys[i]) > 79) - { - state->error = 66; /*text chunk too large*/ - break; - } - if(strlen(info.text_keys[i]) < 1) - { - state->error = 67; /*text chunk too small*/ - break; - } - if(state->encoder.text_compression) - addChunk_zTXt(&outv, info.text_keys[i], info.text_strings[i], &state->encoder.zlibsettings); - else - addChunk_tEXt(&outv, info.text_keys[i], info.text_strings[i]); - } - /*LodePNG version id in text chunk*/ - if(state->encoder.add_id) - { - unsigned alread_added_id_text = 0; - for(i = 0; i < info.text_num; i++) - { - if(!strcmp(info.text_keys[i], "LodePNG")) - { - alread_added_id_text = 1; - break; - } - } - if(alread_added_id_text == 0) - addChunk_tEXt(&outv, "LodePNG", VERSION_STRING); /*it's shorter as tEXt than as zTXt chunk*/ - } - /*iTXt*/ - for(i = 0; i < info.itext_num; i++) - { - if(strlen(info.itext_keys[i]) > 79) - { - state->error = 66; /*text chunk too large*/ - break; - } - if(strlen(info.itext_keys[i]) < 1) - { - state->error = 67; /*text chunk too small*/ - break; - } - addChunk_iTXt(&outv, state->encoder.text_compression, - info.itext_keys[i], info.itext_langtags[i], info.itext_transkeys[i], info.itext_strings[i], - &state->encoder.zlibsettings); - } - - /*unknown chunks between IDAT and IEND*/ - if(info.unknown_chunks_data[2]) - { - state->error = addUnknownChunks(&outv, info.unknown_chunks_data[2], info.unknown_chunks_size[2]); - if(state->error) break; - } -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - /*IEND*/ - addChunk_IEND(&outv); - - break; /*this isn't really a while loop; no error happened so break out now!*/ - } - - lodepng_info_cleanup(&info); - lodepng_free(data); - /*instead of cleaning the vector up, give it to the output*/ - *out = outv.data; - *outsize = outv.size; - - return state->error; -} - -unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, const unsigned char* image, - unsigned w, unsigned h, LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned error; - LodePNGState state; - lodepng_state_init(&state); - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - state.info_png.color.colortype = colortype; - state.info_png.color.bitdepth = bitdepth; - lodepng_encode(out, outsize, image, w, h, &state); - error = state.error; - lodepng_state_cleanup(&state); - return error; -} - -unsigned lodepng_encode32(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGBA, 8); -} - -unsigned lodepng_encode24(unsigned char** out, size_t* outsize, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_memory(out, outsize, image, w, h, LCT_RGB, 8); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned lodepng_encode_file(const char* filename, const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode_memory(&buffer, &buffersize, image, w, h, colortype, bitdepth); - if(!error) error = lodepng_save_file(buffer, buffersize, filename); - lodepng_free(buffer); - return error; -} - -unsigned lodepng_encode32_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_file(filename, image, w, h, LCT_RGBA, 8); -} - -unsigned lodepng_encode24_file(const char* filename, const unsigned char* image, unsigned w, unsigned h) -{ - return lodepng_encode_file(filename, image, w, h, LCT_RGB, 8); -} -#endif /*LODEPNG_COMPILE_DISK*/ - -void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings) -{ - lodepng_compress_settings_init(&settings->zlibsettings); - settings->filter_palette_zero = 1; - settings->filter_strategy = LFS_MINSUM; - settings->auto_convert = LAC_AUTO; - settings->force_palette = 0; - settings->predefined_filters = 0; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - settings->add_id = 0; - settings->text_compression = 1; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} - -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ERROR_TEXT -/* -This returns the description of a numerical error code in English. This is also -the documentation of all the error codes. -*/ -const char* lodepng_error_text(unsigned code) -{ - switch(code) - { - case 0: return "no error, everything went ok"; - case 1: return "nothing done yet"; /*the Encoder/Decoder has done nothing yet, error checking makes no sense yet*/ - case 10: return "end of input memory reached without huffman end code"; /*while huffman decoding*/ - case 11: return "error in code tree made it jump outside of huffman tree"; /*while huffman decoding*/ - case 13: return "problem while processing dynamic deflate block"; - case 14: return "problem while processing dynamic deflate block"; - case 15: return "problem while processing dynamic deflate block"; - case 16: return "unexisting code while processing dynamic deflate block"; - case 17: return "end of out buffer memory reached while inflating"; - case 18: return "invalid distance code while inflating"; - case 19: return "end of out buffer memory reached while inflating"; - case 20: return "invalid deflate block BTYPE encountered while decoding"; - case 21: return "NLEN is not ones complement of LEN in a deflate block"; - /*end of out buffer memory reached while inflating: - This can happen if the inflated deflate data is longer than the amount of bytes required to fill up - all the pixels of the image, given the color depth and image dimensions. Something that doesn't - happen in a normal, well encoded, PNG image.*/ - case 22: return "end of out buffer memory reached while inflating"; - case 23: return "end of in buffer memory reached while inflating"; - case 24: return "invalid FCHECK in zlib header"; - case 25: return "invalid compression method in zlib header"; - case 26: return "FDICT encountered in zlib header while it's not used for PNG"; - case 27: return "PNG file is smaller than a PNG header"; - /*Checks the magic file header, the first 8 bytes of the PNG file*/ - case 28: return "incorrect PNG signature, it's no PNG or corrupted"; - case 29: return "first chunk is not the header chunk"; - case 30: return "chunk length too large, chunk broken off at end of file"; - case 31: return "illegal PNG color type or bpp"; - case 32: return "illegal PNG compression method"; - case 33: return "illegal PNG filter method"; - case 34: return "illegal PNG interlace method"; - case 35: return "chunk length of a chunk is too large or the chunk too small"; - case 36: return "illegal PNG filter type encountered"; - case 37: return "illegal bit depth for this color type given"; - case 38: return "the palette is too big"; /*more than 256 colors*/ - case 39: return "more palette alpha values given in tRNS chunk than there are colors in the palette"; - case 40: return "tRNS chunk has wrong size for greyscale image"; - case 41: return "tRNS chunk has wrong size for RGB image"; - case 42: return "tRNS chunk appeared while it was not allowed for this color type"; - case 43: return "bKGD chunk has wrong size for palette image"; - case 44: return "bKGD chunk has wrong size for greyscale image"; - case 45: return "bKGD chunk has wrong size for RGB image"; - /*Is the palette too small?*/ - case 46: return "a value in indexed image is larger than the palette size (bitdepth = 8)"; - /*Is the palette too small?*/ - case 47: return "a value in indexed image is larger than the palette size (bitdepth < 8)"; - /*the input data is empty, maybe a PNG file doesn't exist or is in the wrong path*/ - case 48: return "empty input or file doesn't exist"; - case 49: return "jumped past memory while generating dynamic huffman tree"; - case 50: return "jumped past memory while generating dynamic huffman tree"; - case 51: return "jumped past memory while inflating huffman block"; - case 52: return "jumped past memory while inflating"; - case 53: return "size of zlib data too small"; - case 54: return "repeat symbol in tree while there was no value symbol yet"; - /*jumped past tree while generating huffman tree, this could be when the - tree will have more leaves than symbols after generating it out of the - given lenghts. They call this an oversubscribed dynamic bit lengths tree in zlib.*/ - case 55: return "jumped past tree while generating huffman tree"; - case 56: return "given output image colortype or bitdepth not supported for color conversion"; - case 57: return "invalid CRC encountered (checking CRC can be disabled)"; - case 58: return "invalid ADLER32 encountered (checking ADLER32 can be disabled)"; - case 59: return "requested color conversion not supported"; - case 60: return "invalid window size given in the settings of the encoder (must be 0-32768)"; - case 61: return "invalid BTYPE given in the settings of the encoder (only 0, 1 and 2 are allowed)"; - /*LodePNG leaves the choice of RGB to greyscale conversion formula to the user.*/ - case 62: return "conversion from color to greyscale not supported"; - case 63: return "length of a chunk too long, max allowed for PNG is 2147483647 bytes per chunk"; /*(2^31-1)*/ - /*this would result in the inability of a deflated block to ever contain an end code. It must be at least 1.*/ - case 64: return "the length of the END symbol 256 in the Huffman tree is 0"; - case 66: return "the length of a text chunk keyword given to the encoder is longer than the maximum of 79 bytes"; - case 67: return "the length of a text chunk keyword given to the encoder is smaller than the minimum of 1 byte"; - case 68: return "tried to encode a PLTE chunk with a palette that has less than 1 or more than 256 colors"; - case 69: return "unknown chunk type with 'critical' flag encountered by the decoder"; - case 71: return "unexisting interlace mode given to encoder (must be 0 or 1)"; - case 72: return "while decoding, unexisting compression method encountering in zTXt or iTXt chunk (it must be 0)"; - case 73: return "invalid tIME chunk size"; - case 74: return "invalid pHYs chunk size"; - /*length could be wrong, or data chopped off*/ - case 75: return "no null termination char found while decoding text chunk"; - case 76: return "iTXt chunk too short to contain required bytes"; - case 77: return "integer overflow in buffer size"; - case 78: return "failed to open file for reading"; /*file doesn't exist or couldn't be opened for reading*/ - case 79: return "failed to open file for writing"; - case 80: return "tried creating a tree of 0 symbols"; - case 81: return "lazy matching at pos 0 is impossible"; - case 82: return "color conversion to palette requested while a color isn't in palette"; - case 83: return "memory allocation failed"; - case 84: return "given image too small to contain all pixels to be encoded"; - case 85: return "internal color conversion bug"; - case 86: return "impossible offset in lz77 encoding (internal bug)"; - case 87: return "must provide custom zlib function pointer if LODEPNG_COMPILE_ZLIB is not defined"; - case 88: return "invalid filter strategy given for LodePNGEncoderSettings.filter_strategy"; - case 89: return "text chunk keyword too short or long: must have size 1-79"; - } - return "unknown error code"; -} -#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ - -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* // C++ Wrapper // */ -/* ////////////////////////////////////////////////////////////////////////// */ -/* ////////////////////////////////////////////////////////////////////////// */ - - -#ifdef LODEPNG_COMPILE_CPP -namespace lodepng -{ - -#ifdef LODEPNG_COMPILE_DISK -void load_file(std::vector& buffer, const std::string& filename) -{ - std::ifstream file(filename.c_str(), std::ios::in|std::ios::binary|std::ios::ate); - - /*get filesize*/ - std::streamsize size = 0; - if(file.seekg(0, std::ios::end).good()) size = file.tellg(); - if(file.seekg(0, std::ios::beg).good()) size -= file.tellg(); - - /*read contents of the file into the vector*/ - buffer.resize(size_t(size)); - if(size > 0) file.read((char*)(&buffer[0]), size); -} - -/*write given buffer to the file, overwriting the file, it doesn't append to it.*/ -void save_file(const std::vector& buffer, const std::string& filename) -{ - std::ofstream file(filename.c_str(), std::ios::out|std::ios::binary); - file.write(buffer.empty() ? 0 : (char*)&buffer[0], std::streamsize(buffer.size())); -} -#endif //LODEPNG_COMPILE_DISK - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_DECODER -unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings& settings) -{ - unsigned char* buffer = 0; - size_t buffersize = 0; - unsigned error = zlib_decompress(&buffer, &buffersize, in, insize, &settings); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned decompress(std::vector& out, const std::vector& in, - const LodePNGDecompressSettings& settings) -{ - return decompress(out, in.empty() ? 0 : &in[0], in.size(), settings); -} -#endif //LODEPNG_COMPILE_DECODER - -#ifdef LODEPNG_COMPILE_ENCODER -unsigned compress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings& settings) -{ - unsigned char* buffer = 0; - size_t buffersize = 0; - unsigned error = zlib_compress(&buffer, &buffersize, in, insize, &settings); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned compress(std::vector& out, const std::vector& in, - const LodePNGCompressSettings& settings) -{ - return compress(out, in.empty() ? 0 : &in[0], in.size(), settings); -} -#endif //LODEPNG_COMPILE_ENCODER -#endif //LODEPNG_COMPILE_ZLIB - - -#ifdef LODEPNG_COMPILE_PNG - -State::State() -{ - lodepng_state_init(this); -} - -State::State(const State& other) -{ - lodepng_state_init(this); - lodepng_state_copy(this, &other); -} - -State::~State() -{ - lodepng_state_cleanup(this); -} - -State& State::operator=(const State& other) -{ - lodepng_state_copy(this, &other); - return *this; -} - -#ifdef LODEPNG_COMPILE_DECODER - -unsigned decode(std::vector& out, unsigned& w, unsigned& h, const unsigned char* in, - size_t insize, LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned char* buffer; - unsigned error = lodepng_decode_memory(&buffer, &w, &h, in, insize, colortype, bitdepth); - if(buffer && !error) - { - State state; - state.info_raw.colortype = colortype; - state.info_raw.bitdepth = bitdepth; - size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const std::vector& in, LodePNGColorType colortype, unsigned bitdepth) -{ - return decode(out, w, h, in.empty() ? 0 : &in[0], (unsigned)in.size(), colortype, bitdepth); -} - -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const unsigned char* in, size_t insize) -{ - unsigned char* buffer; - unsigned error = lodepng_decode(&buffer, &w, &h, &state, in, insize); - if(buffer && !error) - { - size_t buffersize = lodepng_get_raw_size(w, h, &state.info_raw); - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const std::vector& in) -{ - return decode(out, w, h, state, in.empty() ? 0 : &in[0], in.size()); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned decode(std::vector& out, unsigned& w, unsigned& h, const std::string& filename, - LodePNGColorType colortype, unsigned bitdepth) -{ - std::vector buffer; - load_file(buffer, filename); - return decode(out, w, h, buffer, colortype, bitdepth); -} -#endif //LODEPNG_COMPILE_DECODER -#endif //LODEPNG_COMPILE_DISK - -#ifdef LODEPNG_COMPILE_ENCODER -unsigned encode(std::vector& out, const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) -{ - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode_memory(&buffer, &buffersize, in, w, h, colortype, bitdepth); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) -{ - if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; - return encode(out, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); -} - -unsigned encode(std::vector& out, - const unsigned char* in, unsigned w, unsigned h, - State& state) -{ - unsigned char* buffer; - size_t buffersize; - unsigned error = lodepng_encode(&buffer, &buffersize, in, w, h, &state); - if(buffer) - { - out.insert(out.end(), &buffer[0], &buffer[buffersize]); - lodepng_free(buffer); - } - return error; -} - -unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - State& state) -{ - if(lodepng_get_raw_size(w, h, &state.info_raw) > in.size()) return 84; - return encode(out, in.empty() ? 0 : &in[0], w, h, state); -} - -#ifdef LODEPNG_COMPILE_DISK -unsigned encode(const std::string& filename, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) -{ - std::vector buffer; - unsigned error = encode(buffer, in, w, h, colortype, bitdepth); - if(!error) save_file(buffer, filename); - return error; -} - -unsigned encode(const std::string& filename, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth) -{ - if(lodepng_get_raw_size_lct(w, h, colortype, bitdepth) > in.size()) return 84; - return encode(filename, in.empty() ? 0 : &in[0], w, h, colortype, bitdepth); -} -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_ENCODER -#endif //LODEPNG_COMPILE_PNG -} //namespace lodepng -#endif /*LODEPNG_COMPILE_CPP*/ diff --git a/lodepng/lodepng.h b/lodepng/lodepng.h deleted file mode 100644 index 3c1c392820..0000000000 --- a/lodepng/lodepng.h +++ /dev/null @@ -1,1705 +0,0 @@ -/* -LodePNG version 20130831 - -Copyright (c) 2005-2013 Lode Vandevenne - -This software is provided 'as-is', without any express or implied -warranty. In no event will the authors be held liable for any damages -arising from the use of this software. - -Permission is granted to anyone to use this software for any purpose, -including commercial applications, and to alter it and redistribute it -freely, subject to the following restrictions: - - 1. The origin of this software must not be misrepresented; you must not - claim that you wrote the original software. If you use this software - in a product, an acknowledgment in the product documentation would be - appreciated but is not required. - - 2. Altered source versions must be plainly marked as such, and must not be - misrepresented as being the original software. - - 3. This notice may not be removed or altered from any source - distribution. -*/ - -#ifndef LODEPNG_H -#define LODEPNG_H - -#include /*for size_t*/ - -#ifdef __cplusplus -#include -#include -#endif /*__cplusplus*/ - -/* -The following #defines are used to create code sections. They can be disabled -to disable code sections, which can give faster compile time and smaller binary. -The "NO_COMPILE" defines are designed to be used to pass as defines to the -compiler command to disable them without modifying this header, e.g. --DLODEPNG_NO_COMPILE_ZLIB for gcc. -*/ -/*deflate & zlib. If disabled, you must specify alternative zlib functions in -the custom_zlib field of the compress and decompress settings*/ -#ifndef LODEPNG_NO_COMPILE_ZLIB -#define LODEPNG_COMPILE_ZLIB -#endif -/*png encoder and png decoder*/ -#ifndef LODEPNG_NO_COMPILE_PNG -#define LODEPNG_COMPILE_PNG -#endif -/*deflate&zlib decoder and png decoder*/ -#ifndef LODEPNG_NO_COMPILE_DECODER -#define LODEPNG_COMPILE_DECODER -#endif -/*deflate&zlib encoder and png encoder*/ -#ifndef LODEPNG_NO_COMPILE_ENCODER -#define LODEPNG_COMPILE_ENCODER -#endif -/*the optional built in harddisk file loading and saving functions*/ -#ifndef LODEPNG_NO_COMPILE_DISK -#define LODEPNG_COMPILE_DISK -#endif -/*support for chunks other than IHDR, IDAT, PLTE, tRNS, IEND: ancillary and unknown chunks*/ -#ifndef LODEPNG_NO_COMPILE_ANCILLARY_CHUNKS -#define LODEPNG_COMPILE_ANCILLARY_CHUNKS -#endif -/*ability to convert error numerical codes to English text string*/ -#ifndef LODEPNG_NO_COMPILE_ERROR_TEXT -#define LODEPNG_COMPILE_ERROR_TEXT -#endif -/*Compile the default allocators (C's free, malloc and realloc). If you disable this, -you can define the functions lodepng_free, lodepng_malloc and lodepng_realloc in your -source files with custom allocators.*/ -#ifndef LODEPNG_NO_COMPILE_ALLOCATORS -#define LODEPNG_COMPILE_ALLOCATORS -#endif -/*compile the C++ version (you can disable the C++ wrapper here even when compiling for C++)*/ -#ifdef __cplusplus -#ifndef LODEPNG_NO_COMPILE_CPP -#define LODEPNG_COMPILE_CPP -#endif -#endif - -#ifdef LODEPNG_COMPILE_PNG -/*The PNG color types (also used for raw).*/ -typedef enum LodePNGColorType -{ - LCT_GREY = 0, /*greyscale: 1,2,4,8,16 bit*/ - LCT_RGB = 2, /*RGB: 8,16 bit*/ - LCT_PALETTE = 3, /*palette: 1,2,4,8 bit*/ - LCT_GREY_ALPHA = 4, /*greyscale with alpha: 8,16 bit*/ - LCT_RGBA = 6 /*RGB with alpha: 8,16 bit*/ -} LodePNGColorType; - -#ifdef LODEPNG_COMPILE_DECODER -/* -Converts PNG data in memory to raw pixel data. -out: Output parameter. Pointer to buffer that will contain the raw pixel data. - After decoding, its size is w * h * (bytes per pixel) bytes larger than - initially. Bytes per pixel depends on colortype and bitdepth. - Must be freed after usage with free(*out). - Note: for 16-bit per channel colors, uses big endian format like PNG does. -w: Output parameter. Pointer to width of pixel data. -h: Output parameter. Pointer to height of pixel data. -in: Memory buffer with the PNG file. -insize: size of the in buffer. -colortype: the desired color type for the raw output image. See explanation on PNG color types. -bitdepth: the desired bit depth for the raw output image. See explanation on PNG color types. -Return value: LodePNG error code (0 means no error). -*/ -unsigned lodepng_decode_memory(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_decode_memory, but always decodes to 32-bit RGBA raw image*/ -unsigned lodepng_decode32(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize); - -/*Same as lodepng_decode_memory, but always decodes to 24-bit RGB raw image*/ -unsigned lodepng_decode24(unsigned char** out, unsigned* w, unsigned* h, - const unsigned char* in, size_t insize); - -#ifdef LODEPNG_COMPILE_DISK -/* -Load PNG from disk, from file with given name. -Same as the other decode functions, but instead takes a filename as input. -*/ -unsigned lodepng_decode_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_decode_file, but always decodes to 32-bit RGBA raw image.*/ -unsigned lodepng_decode32_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename); - -/*Same as lodepng_decode_file, but always decodes to 24-bit RGB raw image.*/ -unsigned lodepng_decode24_file(unsigned char** out, unsigned* w, unsigned* h, - const char* filename); -#endif /*LODEPNG_COMPILE_DISK*/ -#endif /*LODEPNG_COMPILE_DECODER*/ - - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Converts raw pixel data into a PNG image in memory. The colortype and bitdepth - of the output PNG image cannot be chosen, they are automatically determined - by the colortype, bitdepth and content of the input pixel data. - Note: for 16-bit per channel colors, needs big endian format like PNG does. -out: Output parameter. Pointer to buffer that will contain the PNG image data. - Must be freed after usage with free(*out). -outsize: Output parameter. Pointer to the size in bytes of the out buffer. -image: The raw pixel data to encode. The size of this buffer should be - w * h * (bytes per pixel), bytes per pixel depends on colortype and bitdepth. -w: width of the raw pixel data in pixels. -h: height of the raw pixel data in pixels. -colortype: the color type of the raw input image. See explanation on PNG color types. -bitdepth: the bit depth of the raw input image. See explanation on PNG color types. -Return value: LodePNG error code (0 means no error). -*/ -unsigned lodepng_encode_memory(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_encode_memory, but always encodes from 32-bit RGBA raw image.*/ -unsigned lodepng_encode32(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h); - -/*Same as lodepng_encode_memory, but always encodes from 24-bit RGB raw image.*/ -unsigned lodepng_encode24(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h); - -#ifdef LODEPNG_COMPILE_DISK -/* -Converts raw pixel data into a PNG file on disk. -Same as the other encode functions, but instead takes a filename as output. -NOTE: This overwrites existing files without warning! -*/ -unsigned lodepng_encode_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h, - LodePNGColorType colortype, unsigned bitdepth); - -/*Same as lodepng_encode_file, but always encodes from 32-bit RGBA raw image.*/ -unsigned lodepng_encode32_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h); - -/*Same as lodepng_encode_file, but always encodes from 24-bit RGB raw image.*/ -unsigned lodepng_encode24_file(const char* filename, - const unsigned char* image, unsigned w, unsigned h); -#endif /*LODEPNG_COMPILE_DISK*/ -#endif /*LODEPNG_COMPILE_ENCODER*/ - - -#ifdef LODEPNG_COMPILE_CPP -namespace lodepng -{ -#ifdef LODEPNG_COMPILE_DECODER -/*Same as lodepng_decode_memory, but decodes to an std::vector.*/ -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const unsigned char* in, size_t insize, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const std::vector& in, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#ifdef LODEPNG_COMPILE_DISK -/* -Converts PNG file from disk to raw pixel data in memory. -Same as the other decode functions, but instead takes a filename as input. -*/ -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - const std::string& filename, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_DECODER - -#ifdef LODEPNG_COMPILE_ENCODER -/*Same as lodepng_encode_memory, but encodes to an std::vector.*/ -unsigned encode(std::vector& out, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#ifdef LODEPNG_COMPILE_DISK -/* -Converts 32-bit RGBA raw pixel data into a PNG file on disk. -Same as the other encode functions, but instead takes a filename as output. -NOTE: This overwrites existing files without warning! -*/ -unsigned encode(const std::string& filename, - const unsigned char* in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -unsigned encode(const std::string& filename, - const std::vector& in, unsigned w, unsigned h, - LodePNGColorType colortype = LCT_RGBA, unsigned bitdepth = 8); -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_ENCODER -} //namespace lodepng -#endif /*LODEPNG_COMPILE_CPP*/ -#endif /*LODEPNG_COMPILE_PNG*/ - -#ifdef LODEPNG_COMPILE_ERROR_TEXT -/*Returns an English description of the numerical error code.*/ -const char* lodepng_error_text(unsigned code); -#endif /*LODEPNG_COMPILE_ERROR_TEXT*/ - -#ifdef LODEPNG_COMPILE_DECODER -/*Settings for zlib decompression*/ -typedef struct LodePNGDecompressSettings LodePNGDecompressSettings; -struct LodePNGDecompressSettings -{ - unsigned ignore_adler32; /*if 1, continue and don't give an error message if the Adler32 checksum is corrupted*/ - - /*use custom zlib decoder instead of built in one (default: null)*/ - unsigned (*custom_zlib)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGDecompressSettings*); - /*use custom deflate decoder instead of built in one (default: null) - if custom_zlib is used, custom_deflate is ignored since only the built in - zlib function will call custom_deflate*/ - unsigned (*custom_inflate)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGDecompressSettings*); - - void* custom_context; /*optional custom settings for custom functions*/ -}; - -extern const LodePNGDecompressSettings lodepng_default_decompress_settings; -void lodepng_decompress_settings_init(LodePNGDecompressSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Settings for zlib compression. Tweaking these settings tweaks the balance -between speed and compression ratio. -*/ -typedef struct LodePNGCompressSettings LodePNGCompressSettings; -struct LodePNGCompressSettings /*deflate = compress*/ -{ - /*LZ77 related settings*/ - unsigned btype; /*the block type for LZ (0, 1, 2 or 3, see zlib standard). Should be 2 for proper compression.*/ - unsigned use_lz77; /*whether or not to use LZ77. Should be 1 for proper compression.*/ - unsigned windowsize; /*the maximum is 32768, higher gives more compression but is slower. Typical value: 2048.*/ - unsigned minmatch; /*mininum lz77 length. 3 is normally best, 6 can be better for some PNGs. Default: 0*/ - unsigned nicematch; /*stop searching if >= this length found. Set to 258 for best compression. Default: 128*/ - unsigned lazymatching; /*use lazy matching: better compression but a bit slower. Default: true*/ - - /*use custom zlib encoder instead of built in one (default: null)*/ - unsigned (*custom_zlib)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGCompressSettings*); - /*use custom deflate encoder instead of built in one (default: null) - if custom_zlib is used, custom_deflate is ignored since only the built in - zlib function will call custom_deflate*/ - unsigned (*custom_deflate)(unsigned char**, size_t*, - const unsigned char*, size_t, - const LodePNGCompressSettings*); - - void* custom_context; /*optional custom settings for custom functions*/ -}; - -extern const LodePNGCompressSettings lodepng_default_compress_settings; -void lodepng_compress_settings_init(LodePNGCompressSettings* settings); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_PNG -/* -Color mode of an image. Contains all information required to decode the pixel -bits to RGBA colors. This information is the same as used in the PNG file -format, and is used both for PNG and raw image data in LodePNG. -*/ -typedef struct LodePNGColorMode -{ - /*header (IHDR)*/ - LodePNGColorType colortype; /*color type, see PNG standard or documentation further in this header file*/ - unsigned bitdepth; /*bits per sample, see PNG standard or documentation further in this header file*/ - - /* - palette (PLTE and tRNS) - - Dynamically allocated with the colors of the palette, including alpha. - When encoding a PNG, to store your colors in the palette of the LodePNGColorMode, first use - lodepng_palette_clear, then for each color use lodepng_palette_add. - If you encode an image without alpha with palette, don't forget to put value 255 in each A byte of the palette. - - When decoding, by default you can ignore this palette, since LodePNG already - fills the palette colors in the pixels of the raw RGBA output. - - The palette is only supported for color type 3. - */ - unsigned char* palette; /*palette in RGBARGBA... order. When allocated, must be either 0, or have size 1024*/ - size_t palettesize; /*palette size in number of colors (amount of bytes is 4 * palettesize)*/ - - /* - transparent color key (tRNS) - - This color uses the same bit depth as the bitdepth value in this struct, which can be 1-bit to 16-bit. - For greyscale PNGs, r, g and b will all 3 be set to the same. - - When decoding, by default you can ignore this information, since LodePNG sets - pixels with this key to transparent already in the raw RGBA output. - - The color key is only supported for color types 0 and 2. - */ - unsigned key_defined; /*is a transparent color key given? 0 = false, 1 = true*/ - unsigned key_r; /*red/greyscale component of color key*/ - unsigned key_g; /*green component of color key*/ - unsigned key_b; /*blue component of color key*/ -} LodePNGColorMode; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_color_mode_init(LodePNGColorMode* info); -void lodepng_color_mode_cleanup(LodePNGColorMode* info); -/*return value is error code (0 means no error)*/ -unsigned lodepng_color_mode_copy(LodePNGColorMode* dest, const LodePNGColorMode* source); - -void lodepng_palette_clear(LodePNGColorMode* info); -/*add 1 color to the palette*/ -unsigned lodepng_palette_add(LodePNGColorMode* info, - unsigned char r, unsigned char g, unsigned char b, unsigned char a); - -/*get the total amount of bits per pixel, based on colortype and bitdepth in the struct*/ -unsigned lodepng_get_bpp(const LodePNGColorMode* info); -/*get the amount of color channels used, based on colortype in the struct. -If a palette is used, it counts as 1 channel.*/ -unsigned lodepng_get_channels(const LodePNGColorMode* info); -/*is it a greyscale type? (only colortype 0 or 4)*/ -unsigned lodepng_is_greyscale_type(const LodePNGColorMode* info); -/*has it got an alpha channel? (only colortype 2 or 6)*/ -unsigned lodepng_is_alpha_type(const LodePNGColorMode* info); -/*has it got a palette? (only colortype 3)*/ -unsigned lodepng_is_palette_type(const LodePNGColorMode* info); -/*only returns true if there is a palette and there is a value in the palette with alpha < 255. -Loops through the palette to check this.*/ -unsigned lodepng_has_palette_alpha(const LodePNGColorMode* info); -/* -Check if the given color info indicates the possibility of having non-opaque pixels in the PNG image. -Returns true if the image can have translucent or invisible pixels (it still be opaque if it doesn't use such pixels). -Returns false if the image can only have opaque pixels. -In detail, it returns true only if it's a color type with alpha, or has a palette with non-opaque values, -or if "key_defined" is true. -*/ -unsigned lodepng_can_have_alpha(const LodePNGColorMode* info); -/*Returns the byte size of a raw image buffer with given width, height and color mode*/ -size_t lodepng_get_raw_size(unsigned w, unsigned h, const LodePNGColorMode* color); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -/*The information of a Time chunk in PNG.*/ -typedef struct LodePNGTime -{ - unsigned year; /*2 bytes used (0-65535)*/ - unsigned month; /*1-12*/ - unsigned day; /*1-31*/ - unsigned hour; /*0-23*/ - unsigned minute; /*0-59*/ - unsigned second; /*0-60 (to allow for leap seconds)*/ -} LodePNGTime; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/*Information about the PNG image, except pixels, width and height.*/ -typedef struct LodePNGInfo -{ - /*header (IHDR), palette (PLTE) and transparency (tRNS) chunks*/ - unsigned compression_method;/*compression method of the original file. Always 0.*/ - unsigned filter_method; /*filter method of the original file*/ - unsigned interlace_method; /*interlace method of the original file*/ - LodePNGColorMode color; /*color type and bits, palette and transparency of the PNG file*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /* - suggested background color chunk (bKGD) - This color uses the same color mode as the PNG (except alpha channel), which can be 1-bit to 16-bit. - - For greyscale PNGs, r, g and b will all 3 be set to the same. When encoding - the encoder writes the red one. For palette PNGs: When decoding, the RGB value - will be stored, not a palette index. But when encoding, specify the index of - the palette in background_r, the other two are then ignored. - - The decoder does not use this background color to edit the color of pixels. - */ - unsigned background_defined; /*is a suggested background color given?*/ - unsigned background_r; /*red component of suggested background color*/ - unsigned background_g; /*green component of suggested background color*/ - unsigned background_b; /*blue component of suggested background color*/ - - /* - non-international text chunks (tEXt and zTXt) - - The char** arrays each contain num strings. The actual messages are in - text_strings, while text_keys are keywords that give a short description what - the actual text represents, e.g. Title, Author, Description, or anything else. - - A keyword is minimum 1 character and maximum 79 characters long. It's - discouraged to use a single line length longer than 79 characters for texts. - - Don't allocate these text buffers yourself. Use the init/cleanup functions - correctly and use lodepng_add_text and lodepng_clear_text. - */ - size_t text_num; /*the amount of texts in these char** buffers (there may be more texts in itext)*/ - char** text_keys; /*the keyword of a text chunk (e.g. "Comment")*/ - char** text_strings; /*the actual text*/ - - /* - international text chunks (iTXt) - Similar to the non-international text chunks, but with additional strings - "langtags" and "transkeys". - */ - size_t itext_num; /*the amount of international texts in this PNG*/ - char** itext_keys; /*the English keyword of the text chunk (e.g. "Comment")*/ - char** itext_langtags; /*language tag for this text's language, ISO/IEC 646 string, e.g. ISO 639 language tag*/ - char** itext_transkeys; /*keyword translated to the international language - UTF-8 string*/ - char** itext_strings; /*the actual international text - UTF-8 string*/ - - /*time chunk (tIME)*/ - unsigned time_defined; /*set to 1 to make the encoder generate a tIME chunk*/ - LodePNGTime time; - - /*phys chunk (pHYs)*/ - unsigned phys_defined; /*if 0, there is no pHYs chunk and the values below are undefined, if 1 else there is one*/ - unsigned phys_x; /*pixels per unit in x direction*/ - unsigned phys_y; /*pixels per unit in y direction*/ - unsigned phys_unit; /*may be 0 (unknown unit) or 1 (metre)*/ - - /* - unknown chunks - There are 3 buffers, one for each position in the PNG where unknown chunks can appear - each buffer contains all unknown chunks for that position consecutively - The 3 buffers are the unknown chunks between certain critical chunks: - 0: IHDR-PLTE, 1: PLTE-IDAT, 2: IDAT-IEND - Do not allocate or traverse this data yourself. Use the chunk traversing functions declared - later, such as lodepng_chunk_next and lodepng_chunk_append, to read/write this struct. - */ - unsigned char* unknown_chunks_data[3]; - size_t unknown_chunks_size[3]; /*size in bytes of the unknown chunks, given for protection*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGInfo; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_info_init(LodePNGInfo* info); -void lodepng_info_cleanup(LodePNGInfo* info); -/*return value is error code (0 means no error)*/ -unsigned lodepng_info_copy(LodePNGInfo* dest, const LodePNGInfo* source); - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS -void lodepng_clear_text(LodePNGInfo* info); /*use this to clear the texts again after you filled them in*/ -unsigned lodepng_add_text(LodePNGInfo* info, const char* key, const char* str); /*push back both texts at once*/ - -void lodepng_clear_itext(LodePNGInfo* info); /*use this to clear the itexts again after you filled them in*/ -unsigned lodepng_add_itext(LodePNGInfo* info, const char* key, const char* langtag, - const char* transkey, const char* str); /*push back the 4 texts of 1 chunk at once*/ -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ - -/* -Converts raw buffer from one color type to another color type, based on -LodePNGColorMode structs to describe the input and output color type. -See the reference manual at the end of this header file to see which color conversions are supported. -return value = LodePNG error code (0 if all went ok, an error if the conversion isn't supported) -The out buffer must have size (w * h * bpp + 7) / 8, where bpp is the bits per pixel -of the output color type (lodepng_get_bpp) -The fix_png value works as described in struct LodePNGDecoderSettings. -Note: for 16-bit per channel colors, uses big endian format like PNG does. -*/ -unsigned lodepng_convert(unsigned char* out, const unsigned char* in, - LodePNGColorMode* mode_out, LodePNGColorMode* mode_in, - unsigned w, unsigned h, unsigned fix_png); - - -#ifdef LODEPNG_COMPILE_DECODER -/* -Settings for the decoder. This contains settings for the PNG and the Zlib -decoder, but not the Info settings from the Info structs. -*/ -typedef struct LodePNGDecoderSettings -{ - LodePNGDecompressSettings zlibsettings; /*in here is the setting to ignore Adler32 checksums*/ - - unsigned ignore_crc; /*ignore CRC checksums*/ - /* - The fix_png setting, if 1, makes the decoder tolerant towards some PNG images - that do not correctly follow the PNG specification. This only supports errors - that are fixable, were found in images that are actually used on the web, and - are silently tolerated by other decoders as well. Currently only one such fix - is implemented: if a palette index is out of bounds given the palette size, - interpret it as opaque black. - By default this value is 0, which makes it stop with an error on such images. - */ - unsigned fix_png; - unsigned color_convert; /*whether to convert the PNG to the color type you want. Default: yes*/ - -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - unsigned read_text_chunks; /*if false but remember_unknown_chunks is true, they're stored in the unknown chunks*/ - /*store all bytes from unknown chunks in the LodePNGInfo (off by default, useful for a png editor)*/ - unsigned remember_unknown_chunks; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGDecoderSettings; - -void lodepng_decoder_settings_init(LodePNGDecoderSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/*automatically use color type with less bits per pixel if losslessly possible. Default: AUTO*/ -typedef enum LodePNGFilterStrategy -{ - /*every filter at zero*/ - LFS_ZERO, - /*Use filter that gives minumum sum, as described in the official PNG filter heuristic.*/ - LFS_MINSUM, - /*Use the filter type that gives smallest Shannon entropy for this scanline. Depending - on the image, this is better or worse than minsum.*/ - LFS_ENTROPY, - /* - Brute-force-search PNG filters by compressing each filter for each scanline. - Experimental, very slow, and only rarely gives better compression than MINSUM. - */ - LFS_BRUTE_FORCE, - /*use predefined_filters buffer: you specify the filter type for each scanline*/ - LFS_PREDEFINED -} LodePNGFilterStrategy; - -/*automatically use color type with less bits per pixel if losslessly possible. Default: LAC_AUTO*/ -typedef enum LodePNGAutoConvert -{ - LAC_NO, /*use color type user requested*/ - LAC_ALPHA, /*use color type user requested, but if only opaque pixels and RGBA or grey+alpha, use RGB or grey*/ - LAC_AUTO, /*use PNG color type that can losslessly represent the uncompressed image the smallest possible*/ - /* - like AUTO, but do not choose 1, 2 or 4 bit per pixel types. - sometimes a PNG image compresses worse if less than 8 bits per pixels. - */ - LAC_AUTO_NO_NIBBLES, - /* - like AUTO, but never choose palette color type. For small images, encoding - the palette may take more bytes than what is gained. Note that AUTO also - already prevents encoding the palette for extremely small images, but that may - not be sufficient because due to the compression it cannot predict when to - switch. - */ - LAC_AUTO_NO_PALETTE, - LAC_AUTO_NO_NIBBLES_NO_PALETTE -} LodePNGAutoConvert; - - -/*Settings for the encoder.*/ -typedef struct LodePNGEncoderSettings -{ - LodePNGCompressSettings zlibsettings; /*settings for the zlib encoder, such as window size, ...*/ - - LodePNGAutoConvert auto_convert; /*how to automatically choose output PNG color type, if at all*/ - - /*If true, follows the official PNG heuristic: if the PNG uses a palette or lower than - 8 bit depth, set all filters to zero. Otherwise use the filter_strategy. Note that to - completely follow the official PNG heuristic, filter_palette_zero must be true and - filter_strategy must be LFS_MINSUM*/ - unsigned filter_palette_zero; - /*Which filter strategy to use when not using zeroes due to filter_palette_zero. - Set filter_palette_zero to 0 to ensure always using your chosen strategy. Default: LFS_MINSUM*/ - LodePNGFilterStrategy filter_strategy; - /*used if filter_strategy is LFS_PREDEFINED. In that case, this must point to a buffer with - the same length as the amount of scanlines in the image, and each value must <= 5. You - have to cleanup this buffer, LodePNG will never free it. Don't forget that filter_palette_zero - must be set to 0 to ensure this is also used on palette or low bitdepth images.*/ - unsigned char* predefined_filters; - - /*force creating a PLTE chunk if colortype is 2 or 6 (= a suggested palette). - If colortype is 3, PLTE is _always_ created.*/ - unsigned force_palette; -#ifdef LODEPNG_COMPILE_ANCILLARY_CHUNKS - /*add LodePNG identifier and version as a text chunk, for debugging*/ - unsigned add_id; - /*encode text chunks as zTXt chunks instead of tEXt chunks, and use compression in iTXt chunks*/ - unsigned text_compression; -#endif /*LODEPNG_COMPILE_ANCILLARY_CHUNKS*/ -} LodePNGEncoderSettings; - -void lodepng_encoder_settings_init(LodePNGEncoderSettings* settings); -#endif /*LODEPNG_COMPILE_ENCODER*/ - - -#if defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) -/*The settings, state and information for extended encoding and decoding.*/ -typedef struct LodePNGState -{ -#ifdef LODEPNG_COMPILE_DECODER - LodePNGDecoderSettings decoder; /*the decoding settings*/ -#endif /*LODEPNG_COMPILE_DECODER*/ -#ifdef LODEPNG_COMPILE_ENCODER - LodePNGEncoderSettings encoder; /*the encoding settings*/ -#endif /*LODEPNG_COMPILE_ENCODER*/ - LodePNGColorMode info_raw; /*specifies the format in which you would like to get the raw pixel buffer*/ - LodePNGInfo info_png; /*info of the PNG image obtained after decoding*/ - unsigned error; -#ifdef LODEPNG_COMPILE_CPP - //For the lodepng::State subclass. - virtual ~LodePNGState(){} -#endif -} LodePNGState; - -/*init, cleanup and copy functions to use with this struct*/ -void lodepng_state_init(LodePNGState* state); -void lodepng_state_cleanup(LodePNGState* state); -void lodepng_state_copy(LodePNGState* dest, const LodePNGState* source); -#endif /* defined(LODEPNG_COMPILE_DECODER) || defined(LODEPNG_COMPILE_ENCODER) */ - -#ifdef LODEPNG_COMPILE_DECODER -/* -Same as lodepng_decode_memory, but uses a LodePNGState to allow custom settings and -getting much more information about the PNG image and color mode. -*/ -unsigned lodepng_decode(unsigned char** out, unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize); - -/* -Read the PNG header, but not the actual data. This returns only the information -that is in the header chunk of the PNG, such as width, height and color type. The -information is placed in the info_png field of the LodePNGState. -*/ -unsigned lodepng_inspect(unsigned* w, unsigned* h, - LodePNGState* state, - const unsigned char* in, size_t insize); -#endif /*LODEPNG_COMPILE_DECODER*/ - - -#ifdef LODEPNG_COMPILE_ENCODER -/*This function allocates the out buffer with standard malloc and stores the size in *outsize.*/ -unsigned lodepng_encode(unsigned char** out, size_t* outsize, - const unsigned char* image, unsigned w, unsigned h, - LodePNGState* state); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -/* -The lodepng_chunk functions are normally not needed, except to traverse the -unknown chunks stored in the LodePNGInfo struct, or add new ones to it. -It also allows traversing the chunks of an encoded PNG file yourself. - -PNG standard chunk naming conventions: -First byte: uppercase = critical, lowercase = ancillary -Second byte: uppercase = public, lowercase = private -Third byte: must be uppercase -Fourth byte: uppercase = unsafe to copy, lowercase = safe to copy -*/ - -/*get the length of the data of the chunk. Total chunk length has 12 bytes more.*/ -unsigned lodepng_chunk_length(const unsigned char* chunk); - -/*puts the 4-byte type in null terminated string*/ -void lodepng_chunk_type(char type[5], const unsigned char* chunk); - -/*check if the type is the given type*/ -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type); - -/*0: it's one of the critical chunk types, 1: it's an ancillary chunk (see PNG standard)*/ -unsigned char lodepng_chunk_ancillary(const unsigned char* chunk); - -/*0: public, 1: private (see PNG standard)*/ -unsigned char lodepng_chunk_private(const unsigned char* chunk); - -/*0: the chunk is unsafe to copy, 1: the chunk is safe to copy (see PNG standard)*/ -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk); - -/*get pointer to the data of the chunk, where the input points to the header of the chunk*/ -unsigned char* lodepng_chunk_data(unsigned char* chunk); -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk); - -/*returns 0 if the crc is correct, 1 if it's incorrect (0 for OK as usual!)*/ -unsigned lodepng_chunk_check_crc(const unsigned char* chunk); - -/*generates the correct CRC from the data and puts it in the last 4 bytes of the chunk*/ -void lodepng_chunk_generate_crc(unsigned char* chunk); - -/*iterate to next chunks. don't use on IEND chunk, as there is no next chunk then*/ -unsigned char* lodepng_chunk_next(unsigned char* chunk); -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk); - -/* -Appends chunk to the data in out. The given chunk should already have its chunk header. -The out variable and outlength are updated to reflect the new reallocated buffer. -Returns error code (0 if it went ok) -*/ -unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk); - -/* -Appends new chunk to out. The chunk to append is given by giving its length, type -and data separately. The type is a 4-letter string. -The out variable and outlength are updated to reflect the new reallocated buffer. -Returne error code (0 if it went ok) -*/ -unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, - const char* type, const unsigned char* data); - - -/*Calculate CRC32 of buffer*/ -unsigned lodepng_crc32(const unsigned char* buf, size_t len); -#endif /*LODEPNG_COMPILE_PNG*/ - - -#ifdef LODEPNG_COMPILE_ZLIB -/* -This zlib part can be used independently to zlib compress and decompress a -buffer. It cannot be used to create gzip files however, and it only supports the -part of zlib that is required for PNG, it does not support dictionaries. -*/ - -#ifdef LODEPNG_COMPILE_DECODER -/*Inflate a buffer. Inflate is the decompression step of deflate. Out buffer must be freed after use.*/ -unsigned lodepng_inflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings); - -/* -Decompresses Zlib data. Reallocates the out buffer and appends the data. The -data must be according to the zlib specification. -Either, *out must be NULL and *outsize must be 0, or, *out must be a valid -buffer and *outsize its size in bytes. out must be freed by user after usage. -*/ -unsigned lodepng_zlib_decompress(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGDecompressSettings* settings); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -/* -Compresses data with Zlib. Reallocates the out buffer and appends the data. -Zlib adds a small header and trailer around the deflate data. -The data is output in the format of the zlib specification. -Either, *out must be NULL and *outsize must be 0, or, *out must be a valid -buffer and *outsize its size in bytes. out must be freed by user after usage. -*/ -unsigned lodepng_zlib_compress(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings); - -/* -Find length-limited Huffman code for given frequencies. This function is in the -public interface only for tests, it's used internally by lodepng_deflate. -*/ -unsigned lodepng_huffman_code_lengths(unsigned* lengths, const unsigned* frequencies, - size_t numcodes, unsigned maxbitlen); - -/*Compress a buffer with deflate. See RFC 1951. Out buffer must be freed after use.*/ -unsigned lodepng_deflate(unsigned char** out, size_t* outsize, - const unsigned char* in, size_t insize, - const LodePNGCompressSettings* settings); - -#endif /*LODEPNG_COMPILE_ENCODER*/ -#endif /*LODEPNG_COMPILE_ZLIB*/ - -#ifdef LODEPNG_COMPILE_DISK -/* -Load a file from disk into buffer. The function allocates the out buffer, and -after usage you should free it. -out: output parameter, contains pointer to loaded buffer. -outsize: output parameter, size of the allocated out buffer -filename: the path to the file to load -return value: error code (0 means ok) -*/ -unsigned lodepng_load_file(unsigned char** out, size_t* outsize, const char* filename); - -/* -Save a file from buffer to disk. Warning, if it exists, this function overwrites -the file without warning! -buffer: the buffer to write -buffersize: size of the buffer to write -filename: the path to the file to save to -return value: error code (0 means ok) -*/ -unsigned lodepng_save_file(const unsigned char* buffer, size_t buffersize, const char* filename); -#endif /*LODEPNG_COMPILE_DISK*/ - -#ifdef LODEPNG_COMPILE_CPP -//The LodePNG C++ wrapper uses std::vectors instead of manually allocated memory buffers. -namespace lodepng -{ -#ifdef LODEPNG_COMPILE_PNG -class State : public LodePNGState -{ - public: - State(); - State(const State& other); - virtual ~State(); - State& operator=(const State& other); -}; - -#ifdef LODEPNG_COMPILE_DECODER -//Same as other lodepng::decode, but using a State for more settings and information. -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const unsigned char* in, size_t insize); -unsigned decode(std::vector& out, unsigned& w, unsigned& h, - State& state, - const std::vector& in); -#endif /*LODEPNG_COMPILE_DECODER*/ - -#ifdef LODEPNG_COMPILE_ENCODER -//Same as other lodepng::encode, but using a State for more settings and information. -unsigned encode(std::vector& out, - const unsigned char* in, unsigned w, unsigned h, - State& state); -unsigned encode(std::vector& out, - const std::vector& in, unsigned w, unsigned h, - State& state); -#endif /*LODEPNG_COMPILE_ENCODER*/ - -#ifdef LODEPNG_COMPILE_DISK -/* -Load a file from disk into an std::vector. If the vector is empty, then either -the file doesn't exist or is an empty file. -*/ -void load_file(std::vector& buffer, const std::string& filename); - -/* -Save the binary data in an std::vector to a file on disk. The file is overwritten -without warning. -*/ -void save_file(const std::vector& buffer, const std::string& filename); -#endif //LODEPNG_COMPILE_DISK -#endif //LODEPNG_COMPILE_PNG - -#ifdef LODEPNG_COMPILE_ZLIB -#ifdef LODEPNG_COMPILE_DECODER -//Zlib-decompress an unsigned char buffer -unsigned decompress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); - -//Zlib-decompress an std::vector -unsigned decompress(std::vector& out, const std::vector& in, - const LodePNGDecompressSettings& settings = lodepng_default_decompress_settings); -#endif //LODEPNG_COMPILE_DECODER - -#ifdef LODEPNG_COMPILE_ENCODER -//Zlib-compress an unsigned char buffer -unsigned compress(std::vector& out, const unsigned char* in, size_t insize, - const LodePNGCompressSettings& settings = lodepng_default_compress_settings); - -//Zlib-compress an std::vector -unsigned compress(std::vector& out, const std::vector& in, - const LodePNGCompressSettings& settings = lodepng_default_compress_settings); -#endif //LODEPNG_COMPILE_ENCODER -#endif //LODEPNG_COMPILE_ZLIB -} //namespace lodepng -#endif /*LODEPNG_COMPILE_CPP*/ - -/* -TODO: -[.] test if there are no memory leaks or security exploits - done a lot but needs to be checked often -[.] check compatibility with vareous compilers - done but needs to be redone for every newer version -[X] converting color to 16-bit per channel types -[ ] read all public PNG chunk types (but never let the color profile and gamma ones touch RGB values) -[ ] make sure encoder generates no chunks with size > (2^31)-1 -[ ] partial decoding (stream processing) -[X] let the "isFullyOpaque" function check color keys and transparent palettes too -[X] better name for the variables "codes", "codesD", "codelengthcodes", "clcl" and "lldl" -[ ] don't stop decoding on errors like 69, 57, 58 (make warnings) -[ ] make option to choose if the raw image with non multiple of 8 bits per scanline should have padding bits or not -[ ] let the C++ wrapper catch exceptions coming from the standard library and return LodePNG error codes -*/ - -#endif /*LODEPNG_H inclusion guard*/ - -/* -LodePNG Documentation ---------------------- - -0. table of contents --------------------- - - 1. about - 1.1. supported features - 1.2. features not supported - 2. C and C++ version - 3. security - 4. decoding - 5. encoding - 6. color conversions - 6.1. PNG color types - 6.2. color conversions - 6.3. padding bits - 6.4. A note about 16-bits per channel and endianness - 7. error values - 8. chunks and PNG editing - 9. compiler support - 10. examples - 10.1. decoder C++ example - 10.2. decoder C example - 11. changes - 12. contact information - - -1. about --------- - -PNG is a file format to store raster images losslessly with good compression, -supporting different color types and alpha channel. - -LodePNG is a PNG codec according to the Portable Network Graphics (PNG) -Specification (Second Edition) - W3C Recommendation 10 November 2003. - -The specifications used are: - -*) Portable Network Graphics (PNG) Specification (Second Edition): - http://www.w3.org/TR/2003/REC-PNG-20031110 -*) RFC 1950 ZLIB Compressed Data Format version 3.3: - http://www.gzip.org/zlib/rfc-zlib.html -*) RFC 1951 DEFLATE Compressed Data Format Specification ver 1.3: - http://www.gzip.org/zlib/rfc-deflate.html - -The most recent version of LodePNG can currently be found at -http://lodev.org/lodepng/ - -LodePNG works both in C (ISO C90) and C++, with a C++ wrapper that adds -extra functionality. - -LodePNG exists out of two files: --lodepng.h: the header file for both C and C++ --lodepng.c(pp): give it the name lodepng.c or lodepng.cpp (or .cc) depending on your usage - -If you want to start using LodePNG right away without reading this doc, get the -examples from the LodePNG website to see how to use it in code, or check the -smaller examples in chapter 13 here. - -LodePNG is simple but only supports the basic requirements. To achieve -simplicity, the following design choices were made: There are no dependencies -on any external library. There are functions to decode and encode a PNG with -a single function call, and extended versions of these functions taking a -LodePNGState struct allowing to specify or get more information. By default -the colors of the raw image are always RGB or RGBA, no matter what color type -the PNG file uses. To read and write files, there are simple functions to -convert the files to/from buffers in memory. - -This all makes LodePNG suitable for loading textures in games, demos and small -programs, ... It's less suitable for full fledged image editors, loading PNGs -over network (it requires all the image data to be available before decoding can -begin), life-critical systems, ... - -1.1. supported features ------------------------ - -The following features are supported by the decoder: - -*) decoding of PNGs with any color type, bit depth and interlace mode, to a 24- or 32-bit color raw image, - or the same color type as the PNG -*) encoding of PNGs, from any raw image to 24- or 32-bit color, or the same color type as the raw image -*) Adam7 interlace and deinterlace for any color type -*) loading the image from harddisk or decoding it from a buffer from other sources than harddisk -*) support for alpha channels, including RGBA color model, translucent palettes and color keying -*) zlib decompression (inflate) -*) zlib compression (deflate) -*) CRC32 and ADLER32 checksums -*) handling of unknown chunks, allowing making a PNG editor that stores custom and unknown chunks. -*) the following chunks are supported (generated/interpreted) by both encoder and decoder: - IHDR: header information - PLTE: color palette - IDAT: pixel data - IEND: the final chunk - tRNS: transparency for palettized images - tEXt: textual information - zTXt: compressed textual information - iTXt: international textual information - bKGD: suggested background color - pHYs: physical dimensions - tIME: modification time - -1.2. features not supported ---------------------------- - -The following features are _not_ supported: - -*) some features needed to make a conformant PNG-Editor might be still missing. -*) partial loading/stream processing. All data must be available and is processed in one call. -*) The following public chunks are not supported but treated as unknown chunks by LodePNG - cHRM, gAMA, iCCP, sRGB, sBIT, hIST, sPLT - Some of these are not supported on purpose: LodePNG wants to provide the RGB values - stored in the pixels, not values modified by system dependent gamma or color models. - - -2. C and C++ version --------------------- - -The C version uses buffers allocated with alloc that you need to free() -yourself. You need to use init and cleanup functions for each struct whenever -using a struct from the C version to avoid exploits and memory leaks. - -The C++ version has extra functions with std::vectors in the interface and the -lodepng::State class which is a LodePNGState with constructor and destructor. - -These files work without modification for both C and C++ compilers because all -the additional C++ code is in "#ifdef __cplusplus" blocks that make C-compilers -ignore it, and the C code is made to compile both with strict ISO C90 and C++. - -To use the C++ version, you need to rename the source file to lodepng.cpp -(instead of lodepng.c), and compile it with a C++ compiler. - -To use the C version, you need to rename the source file to lodepng.c (instead -of lodepng.cpp), and compile it with a C compiler. - - -3. Security ------------ - -Even if carefully designed, it's always possible that LodePNG contains possible -exploits. If you discover one, please let me know, and it will be fixed. - -When using LodePNG, care has to be taken with the C version of LodePNG, as well -as the C-style structs when working with C++. The following conventions are used -for all C-style structs: - --if a struct has a corresponding init function, always call the init function when making a new one --if a struct has a corresponding cleanup function, call it before the struct disappears to avoid memory leaks --if a struct has a corresponding copy function, use the copy function instead of "=". - The destination must also be inited already. - - -4. Decoding ------------ - -Decoding converts a PNG compressed image to a raw pixel buffer. - -Most documentation on using the decoder is at its declarations in the header -above. For C, simple decoding can be done with functions such as -lodepng_decode32, and more advanced decoding can be done with the struct -LodePNGState and lodepng_decode. For C++, all decoding can be done with the -various lodepng::decode functions, and lodepng::State can be used for advanced -features. - -When using the LodePNGState, it uses the following fields for decoding: -*) LodePNGInfo info_png: it stores extra information about the PNG (the input) in here -*) LodePNGColorMode info_raw: here you can say what color mode of the raw image (the output) you want to get -*) LodePNGDecoderSettings decoder: you can specify a few extra settings for the decoder to use - -LodePNGInfo info_png --------------------- - -After decoding, this contains extra information of the PNG image, except the actual -pixels, width and height because these are already gotten directly from the decoder -functions. - -It contains for example the original color type of the PNG image, text comments, -suggested background color, etc... More details about the LodePNGInfo struct are -at its declaration documentation. - -LodePNGColorMode info_raw -------------------------- - -When decoding, here you can specify which color type you want -the resulting raw image to be. If this is different from the colortype of the -PNG, then the decoder will automatically convert the result. This conversion -always works, except if you want it to convert a color PNG to greyscale or to -a palette with missing colors. - -By default, 32-bit color is used for the result. - -LodePNGDecoderSettings decoder ------------------------------- - -The settings can be used to ignore the errors created by invalid CRC and Adler32 -chunks, and to disable the decoding of tEXt chunks. - -There's also a setting color_convert, true by default. If false, no conversion -is done, the resulting data will be as it was in the PNG (after decompression) -and you'll have to puzzle the colors of the pixels together yourself using the -color type information in the LodePNGInfo. - - -5. Encoding ------------ - -Encoding converts a raw pixel buffer to a PNG compressed image. - -Most documentation on using the encoder is at its declarations in the header -above. For C, simple encoding can be done with functions such as -lodepng_encode32, and more advanced decoding can be done with the struct -LodePNGState and lodepng_encode. For C++, all encoding can be done with the -various lodepng::encode functions, and lodepng::State can be used for advanced -features. - -Like the decoder, the encoder can also give errors. However it gives less errors -since the encoder input is trusted, the decoder input (a PNG image that could -be forged by anyone) is not trusted. - -When using the LodePNGState, it uses the following fields for encoding: -*) LodePNGInfo info_png: here you specify how you want the PNG (the output) to be. -*) LodePNGColorMode info_raw: here you say what color type of the raw image (the input) has -*) LodePNGEncoderSettings encoder: you can specify a few settings for the encoder to use - -LodePNGInfo info_png --------------------- - -When encoding, you use this the opposite way as when decoding: for encoding, -you fill in the values you want the PNG to have before encoding. By default it's -not needed to specify a color type for the PNG since it's automatically chosen, -but it's possible to choose it yourself given the right settings. - -The encoder will not always exactly match the LodePNGInfo struct you give, -it tries as close as possible. Some things are ignored by the encoder. The -encoder uses, for example, the following settings from it when applicable: -colortype and bitdepth, text chunks, time chunk, the color key, the palette, the -background color, the interlace method, unknown chunks, ... - -When encoding to a PNG with colortype 3, the encoder will generate a PLTE chunk. -If the palette contains any colors for which the alpha channel is not 255 (so -there are translucent colors in the palette), it'll add a tRNS chunk. - -LodePNGColorMode info_raw -------------------------- - -You specify the color type of the raw image that you give to the input here, -including a possible transparent color key and palette you happen to be using in -your raw image data. - -By default, 32-bit color is assumed, meaning your input has to be in RGBA -format with 4 bytes (unsigned chars) per pixel. - -LodePNGEncoderSettings encoder ------------------------------- - -The following settings are supported (some are in sub-structs): -*) auto_convert: when this option is enabled, the encoder will -automatically choose the smallest possible color mode (including color key) that -can encode the colors of all pixels without information loss. -*) btype: the block type for LZ77. 0 = uncompressed, 1 = fixed huffman tree, - 2 = dynamic huffman tree (best compression). Should be 2 for proper - compression. -*) use_lz77: whether or not to use LZ77 for compressed block types. Should be - true for proper compression. -*) windowsize: the window size used by the LZ77 encoder (1 - 32768). Has value - 2048 by default, but can be set to 32768 for better, but slow, compression. -*) force_palette: if colortype is 2 or 6, you can make the encoder write a PLTE - chunk if force_palette is true. This can used as suggested palette to convert - to by viewers that don't support more than 256 colors (if those still exist) -*) add_id: add text chunk "Encoder: LodePNG " to the image. -*) text_compression: default 1. If 1, it'll store texts as zTXt instead of tEXt chunks. - zTXt chunks use zlib compression on the text. This gives a smaller result on - large texts but a larger result on small texts (such as a single program name). - It's all tEXt or all zTXt though, there's no separate setting per text yet. - - -6. color conversions --------------------- - -An important thing to note about LodePNG, is that the color type of the PNG, and -the color type of the raw image, are completely independent. By default, when -you decode a PNG, you get the result as a raw image in the color type you want, -no matter whether the PNG was encoded with a palette, greyscale or RGBA color. -And if you encode an image, by default LodePNG will automatically choose the PNG -color type that gives good compression based on the values of colors and amount -of colors in the image. It can be configured to let you control it instead as -well, though. - -To be able to do this, LodePNG does conversions from one color mode to another. -It can convert from almost any color type to any other color type, except the -following conversions: RGB to greyscale is not supported, and converting to a -palette when the palette doesn't have a required color is not supported. This is -not supported on purpose: this is information loss which requires a color -reduction algorithm that is beyong the scope of a PNG encoder (yes, RGB to grey -is easy, but there are multiple ways if you want to give some channels more -weight). - -By default, when decoding, you get the raw image in 32-bit RGBA or 24-bit RGB -color, no matter what color type the PNG has. And by default when encoding, -LodePNG automatically picks the best color model for the output PNG, and expects -the input image to be 32-bit RGBA or 24-bit RGB. So, unless you want to control -the color format of the images yourself, you can skip this chapter. - -6.1. PNG color types --------------------- - -A PNG image can have many color types, ranging from 1-bit color to 64-bit color, -as well as palettized color modes. After the zlib decompression and unfiltering -in the PNG image is done, the raw pixel data will have that color type and thus -a certain amount of bits per pixel. If you want the output raw image after -decoding to have another color type, a conversion is done by LodePNG. - -The PNG specification gives the following color types: - -0: greyscale, bit depths 1, 2, 4, 8, 16 -2: RGB, bit depths 8 and 16 -3: palette, bit depths 1, 2, 4 and 8 -4: greyscale with alpha, bit depths 8 and 16 -6: RGBA, bit depths 8 and 16 - -Bit depth is the amount of bits per pixel per color channel. So the total amount -of bits per pixel is: amount of channels * bitdepth. - -6.2. color conversions ----------------------- - -As explained in the sections about the encoder and decoder, you can specify -color types and bit depths in info_png and info_raw to change the default -behaviour. - -If, when decoding, you want the raw image to be something else than the default, -you need to set the color type and bit depth you want in the LodePNGColorMode, -or the parameters of the simple function of LodePNG you're using. - -If, when encoding, you use another color type than the default in the input -image, you need to specify its color type and bit depth in the LodePNGColorMode -of the raw image, or use the parameters of the simplefunction of LodePNG you're -using. - -If, when encoding, you don't want LodePNG to choose the output PNG color type -but control it yourself, you need to set auto_convert in the encoder settings -to LAC_NONE, and specify the color type you want in the LodePNGInfo of the -encoder. - -If you do any of the above, LodePNG may need to do a color conversion, which -follows the rules below, and may sometimes not be allowed. - -To avoid some confusion: --the decoder converts from PNG to raw image --the encoder converts from raw image to PNG --the colortype and bitdepth in LodePNGColorMode info_raw, are those of the raw image --the colortype and bitdepth in the color field of LodePNGInfo info_png, are those of the PNG --when encoding, the color type in LodePNGInfo is ignored if auto_convert - is enabled, it is automatically generated instead --when decoding, the color type in LodePNGInfo is set by the decoder to that of the original - PNG image, but it can be ignored since the raw image has the color type you requested instead --if the color type of the LodePNGColorMode and PNG image aren't the same, a conversion - between the color types is done if the color types are supported. If it is not - supported, an error is returned. If the types are the same, no conversion is done. --even though some conversions aren't supported, LodePNG supports loading PNGs from any - colortype and saving PNGs to any colortype, sometimes it just requires preparing - the raw image correctly before encoding. --both encoder and decoder use the same color converter. - -Non supported color conversions: --color to greyscale: no error is thrown, but the result will look ugly because -only the red channel is taken --anything, to palette when that palette does not have that color in it: in this -case an error is thrown - -Supported color conversions: --anything to 8-bit RGB, 8-bit RGBA, 16-bit RGB, 16-bit RGBA --any grey or grey+alpha, to grey or grey+alpha --anything to a palette, as long as the palette has the requested colors in it --removing alpha channel --higher to smaller bitdepth, and vice versa - -If you want no color conversion to be done: --In the encoder, you can make it save a PNG with any color type by giving the -raw color mode and LodePNGInfo the same color mode, and setting auto_convert to -LAC_NO. --In the decoder, you can make it store the pixel data in the same color type -as the PNG has, by setting the color_convert setting to false. Settings in -info_raw are then ignored. - -The function lodepng_convert does the color conversion. It is available in the -interface but normally isn't needed since the encoder and decoder already call -it. - -6.3. padding bits ------------------ - -In the PNG file format, if a less than 8-bit per pixel color type is used and the scanlines -have a bit amount that isn't a multiple of 8, then padding bits are used so that each -scanline starts at a fresh byte. But that is NOT true for the LodePNG raw input and output. -The raw input image you give to the encoder, and the raw output image you get from the decoder -will NOT have these padding bits, e.g. in the case of a 1-bit image with a width -of 7 pixels, the first pixel of the second scanline will the the 8th bit of the first byte, -not the first bit of a new byte. - -6.4. A note about 16-bits per channel and endianness ----------------------------------------------------- - -LodePNG uses unsigned char arrays for 16-bit per channel colors too, just like -for any other color format. The 16-bit values are stored in big endian (most -significant byte first) in these arrays. This is the opposite order of the -little endian used by x86 CPU's. - -LodePNG always uses big endian because the PNG file format does so internally. -Conversions to other formats than PNG uses internally are not supported by -LodePNG on purpose, there are myriads of formats, including endianness of 16-bit -colors, the order in which you store R, G, B and A, and so on. Supporting and -converting to/from all that is outside the scope of LodePNG. - -This may mean that, depending on your use case, you may want to convert the big -endian output of LodePNG to little endian with a for loop. This is certainly not -always needed, many applications and libraries support big endian 16-bit colors -anyway, but it means you cannot simply cast the unsigned char* buffer to an -unsigned short* buffer on x86 CPUs. - - -7. error values ---------------- - -All functions in LodePNG that return an error code, return 0 if everything went -OK, or a non-zero code if there was an error. - -The meaning of the LodePNG error values can be retrieved with the function -lodepng_error_text: given the numerical error code, it returns a description -of the error in English as a string. - -Check the implementation of lodepng_error_text to see the meaning of each code. - - -8. chunks and PNG editing -------------------------- - -If you want to add extra chunks to a PNG you encode, or use LodePNG for a PNG -editor that should follow the rules about handling of unknown chunks, or if your -program is able to read other types of chunks than the ones handled by LodePNG, -then that's possible with the chunk functions of LodePNG. - -A PNG chunk has the following layout: - -4 bytes length -4 bytes type name -length bytes data -4 bytes CRC - -8.1. iterating through chunks ------------------------------ - -If you have a buffer containing the PNG image data, then the first chunk (the -IHDR chunk) starts at byte number 8 of that buffer. The first 8 bytes are the -signature of the PNG and are not part of a chunk. But if you start at byte 8 -then you have a chunk, and can check the following things of it. - -NOTE: none of these functions check for memory buffer boundaries. To avoid -exploits, always make sure the buffer contains all the data of the chunks. -When using lodepng_chunk_next, make sure the returned value is within the -allocated memory. - -unsigned lodepng_chunk_length(const unsigned char* chunk): - -Get the length of the chunk's data. The total chunk length is this length + 12. - -void lodepng_chunk_type(char type[5], const unsigned char* chunk): -unsigned char lodepng_chunk_type_equals(const unsigned char* chunk, const char* type): - -Get the type of the chunk or compare if it's a certain type - -unsigned char lodepng_chunk_critical(const unsigned char* chunk): -unsigned char lodepng_chunk_private(const unsigned char* chunk): -unsigned char lodepng_chunk_safetocopy(const unsigned char* chunk): - -Check if the chunk is critical in the PNG standard (only IHDR, PLTE, IDAT and IEND are). -Check if the chunk is private (public chunks are part of the standard, private ones not). -Check if the chunk is safe to copy. If it's not, then, when modifying data in a critical -chunk, unsafe to copy chunks of the old image may NOT be saved in the new one if your -program doesn't handle that type of unknown chunk. - -unsigned char* lodepng_chunk_data(unsigned char* chunk): -const unsigned char* lodepng_chunk_data_const(const unsigned char* chunk): - -Get a pointer to the start of the data of the chunk. - -unsigned lodepng_chunk_check_crc(const unsigned char* chunk): -void lodepng_chunk_generate_crc(unsigned char* chunk): - -Check if the crc is correct or generate a correct one. - -unsigned char* lodepng_chunk_next(unsigned char* chunk): -const unsigned char* lodepng_chunk_next_const(const unsigned char* chunk): - -Iterate to the next chunk. This works if you have a buffer with consecutive chunks. Note that these -functions do no boundary checking of the allocated data whatsoever, so make sure there is enough -data available in the buffer to be able to go to the next chunk. - -unsigned lodepng_chunk_append(unsigned char** out, size_t* outlength, const unsigned char* chunk): -unsigned lodepng_chunk_create(unsigned char** out, size_t* outlength, unsigned length, - const char* type, const unsigned char* data): - -These functions are used to create new chunks that are appended to the data in *out that has -length *outlength. The append function appends an existing chunk to the new data. The create -function creates a new chunk with the given parameters and appends it. Type is the 4-letter -name of the chunk. - -8.2. chunks in info_png ------------------------ - -The LodePNGInfo struct contains fields with the unknown chunk in it. It has 3 -buffers (each with size) to contain 3 types of unknown chunks: -the ones that come before the PLTE chunk, the ones that come between the PLTE -and the IDAT chunks, and the ones that come after the IDAT chunks. -It's necessary to make the distionction between these 3 cases because the PNG -standard forces to keep the ordering of unknown chunks compared to the critical -chunks, but does not force any other ordering rules. - -info_png.unknown_chunks_data[0] is the chunks before PLTE -info_png.unknown_chunks_data[1] is the chunks after PLTE, before IDAT -info_png.unknown_chunks_data[2] is the chunks after IDAT - -The chunks in these 3 buffers can be iterated through and read by using the same -way described in the previous subchapter. - -When using the decoder to decode a PNG, you can make it store all unknown chunks -if you set the option settings.remember_unknown_chunks to 1. By default, this -option is off (0). - -The encoder will always encode unknown chunks that are stored in the info_png. -If you need it to add a particular chunk that isn't known by LodePNG, you can -use lodepng_chunk_append or lodepng_chunk_create to the chunk data in -info_png.unknown_chunks_data[x]. - -Chunks that are known by LodePNG should not be added in that way. E.g. to make -LodePNG add a bKGD chunk, set background_defined to true and add the correct -parameters there instead. - - -9. compiler support -------------------- - -No libraries other than the current standard C library are needed to compile -LodePNG. For the C++ version, only the standard C++ library is needed on top. -Add the files lodepng.c(pp) and lodepng.h to your project, include -lodepng.h where needed, and your program can read/write PNG files. - -If performance is important, use optimization when compiling! For both the -encoder and decoder, this makes a large difference. - -Make sure that LodePNG is compiled with the same compiler of the same version -and with the same settings as the rest of the program, or the interfaces with -std::vectors and std::strings in C++ can be incompatible. - -CHAR_BITS must be 8 or higher, because LodePNG uses unsigned chars for octets. - -*) gcc and g++ - -LodePNG is developed in gcc so this compiler is natively supported. It gives no -warnings with compiler options "-Wall -Wextra -pedantic -ansi", with gcc and g++ -version 4.7.1 on Linux, 32-bit and 64-bit. - -*) Mingw - -The Mingw compiler (a port of gcc) for Windows is fully supported by LodePNG. - -*) Visual Studio 2005 and up, Visual C++ Express Edition 2005 and up - -Visual Studio may give warnings about 'fopen' being deprecated. A multiplatform library -can't support the proposed Visual Studio alternative however, so LodePNG keeps using -fopen. If you don't want to see the deprecated warnings, put this on top of lodepng.h -before the inclusions: -#define _CRT_SECURE_NO_DEPRECATE - -Other than the above warnings, LodePNG should be warning-free with warning -level 3 (W3). Warning level 4 (W4) will give warnings about integer conversions. -I'm not planning to resolve these warnings. To get rid of them, let Visual -Studio use warning level W3 for lodepng.cpp only: right click lodepng.cpp, -Properties, C/C++, General, Warning Level: Level 3 (/W3). - -Visual Studio may want "stdafx.h" files to be included in each source file and -give an error "unexpected end of file while looking for precompiled header". -That is not standard C++ and will not be added to the stock LodePNG. You can -disable it for lodepng.cpp only by right clicking it, Properties, C/C++, -Precompiled Headers, and set it to Not Using Precompiled Headers there. - -*) Visual Studio 6.0 - -LodePNG support for Visual Studio 6.0 is not guaranteed because VS6 doesn't -follow the C++ standard correctly. - -*) Comeau C/C++ - -Vesion 20070107 compiles without problems on the Comeau C/C++ Online Test Drive -at http://www.comeaucomputing.com/tryitout in both C90 and C++ mode. - -*) Compilers on Macintosh - -LodePNG has been reported to work both with the gcc and LLVM for Macintosh, both -for C and C++. - -*) Other Compilers - -If you encounter problems on other compilers, feel free to let me know and I may -try to fix it if the compiler is modern standards complient. - - -10. examples ------------- - -This decoder example shows the most basic usage of LodePNG. More complex -examples can be found on the LodePNG website. - -10.1. decoder C++ example -------------------------- - -#include "lodepng.h" -#include - -int main(int argc, char *argv[]) -{ - const char* filename = argc > 1 ? argv[1] : "test.png"; - - //load and decode - std::vector image; - unsigned width, height; - unsigned error = lodepng::decode(image, width, height, filename); - - //if there's an error, display it - if(error) std::cout << "decoder error " << error << ": " << lodepng_error_text(error) << std::endl; - - //the pixels are now in the vector "image", 4 bytes per pixel, ordered RGBARGBA..., use it as texture, draw it, ... -} - -10.2. decoder C example ------------------------ - -#include "lodepng.h" - -int main(int argc, char *argv[]) -{ - unsigned error; - unsigned char* image; - size_t width, height; - const char* filename = argc > 1 ? argv[1] : "test.png"; - - error = lodepng_decode32_file(&image, &width, &height, filename); - - if(error) printf("decoder error %u: %s\n", error, lodepng_error_text(error)); - - / * use image here * / - - free(image); - return 0; -} - - -11. changes ------------ - -The version number of LodePNG is the date of the change given in the format -yyyymmdd. - -Some changes aren't backwards compatible. Those are indicated with a (!) -symbol. - -*) 15 apr 2013: Fixed bug with LAC_ALPHA and color key. -*) 25 mar 2013: Added an optional feature to ignore some PNG errors (fix_png). -*) 11 mar 2013 (!): Bugfix with custom free. Changed from "my" to "lodepng_" - prefix for the custom allocators and made it possible with a new #define to - use custom ones in your project without needing to change lodepng's code. -*) 28 jan 2013: Bugfix with color key. -*) 27 okt 2012: Tweaks in text chunk keyword length error handling. -*) 8 okt 2012 (!): Added new filter strategy (entropy) and new auto color mode. - (no palette). Better deflate tree encoding. New compression tweak settings. - Faster color conversions while decoding. Some internal cleanups. -*) 23 sep 2012: Reduced warnings in Visual Studio a little bit. -*) 1 sep 2012 (!): Removed #define's for giving custom (de)compression functions - and made it work with function pointers instead. -*) 23 jun 2012: Added more filter strategies. Made it easier to use custom alloc - and free functions and toggle #defines from compiler flags. Small fixes. -*) 6 may 2012 (!): Made plugging in custom zlib/deflate functions more flexible. -*) 22 apr 2012 (!): Made interface more consistent, renaming a lot. Removed - redundant C++ codec classes. Reduced amount of structs. Everything changed, - but it is cleaner now imho and functionality remains the same. Also fixed - several bugs and shrinked the implementation code. Made new samples. -*) 6 nov 2011 (!): By default, the encoder now automatically chooses the best - PNG color model and bit depth, based on the amount and type of colors of the - raw image. For this, autoLeaveOutAlphaChannel replaced by auto_choose_color. -*) 9 okt 2011: simpler hash chain implementation for the encoder. -*) 8 sep 2011: lz77 encoder lazy matching instead of greedy matching. -*) 23 aug 2011: tweaked the zlib compression parameters after benchmarking. - A bug with the PNG filtertype heuristic was fixed, so that it chooses much - better ones (it's quite significant). A setting to do an experimental, slow, - brute force search for PNG filter types is added. -*) 17 aug 2011 (!): changed some C zlib related function names. -*) 16 aug 2011: made the code less wide (max 120 characters per line). -*) 17 apr 2011: code cleanup. Bugfixes. Convert low to 16-bit per sample colors. -*) 21 feb 2011: fixed compiling for C90. Fixed compiling with sections disabled. -*) 11 dec 2010: encoding is made faster, based on suggestion by Peter Eastman - to optimize long sequences of zeros. -*) 13 nov 2010: added LodePNG_InfoColor_hasPaletteAlpha and - LodePNG_InfoColor_canHaveAlpha functions for convenience. -*) 7 nov 2010: added LodePNG_error_text function to get error code description. -*) 30 okt 2010: made decoding slightly faster -*) 26 okt 2010: (!) changed some C function and struct names (more consistent). - Reorganized the documentation and the declaration order in the header. -*) 08 aug 2010: only changed some comments and external samples. -*) 05 jul 2010: fixed bug thanks to warnings in the new gcc version. -*) 14 mar 2010: fixed bug where too much memory was allocated for char buffers. -*) 02 sep 2008: fixed bug where it could create empty tree that linux apps could - read by ignoring the problem but windows apps couldn't. -*) 06 jun 2008: added more error checks for out of memory cases. -*) 26 apr 2008: added a few more checks here and there to ensure more safety. -*) 06 mar 2008: crash with encoding of strings fixed -*) 02 feb 2008: support for international text chunks added (iTXt) -*) 23 jan 2008: small cleanups, and #defines to divide code in sections -*) 20 jan 2008: support for unknown chunks allowing using LodePNG for an editor. -*) 18 jan 2008: support for tIME and pHYs chunks added to encoder and decoder. -*) 17 jan 2008: ability to encode and decode compressed zTXt chunks added - Also vareous fixes, such as in the deflate and the padding bits code. -*) 13 jan 2008: Added ability to encode Adam7-interlaced images. Improved - filtering code of encoder. -*) 07 jan 2008: (!) changed LodePNG to use ISO C90 instead of C++. A - C++ wrapper around this provides an interface almost identical to before. - Having LodePNG be pure ISO C90 makes it more portable. The C and C++ code - are together in these files but it works both for C and C++ compilers. -*) 29 dec 2007: (!) changed most integer types to unsigned int + other tweaks -*) 30 aug 2007: bug fixed which makes this Borland C++ compatible -*) 09 aug 2007: some VS2005 warnings removed again -*) 21 jul 2007: deflate code placed in new namespace separate from zlib code -*) 08 jun 2007: fixed bug with 2- and 4-bit color, and small interlaced images -*) 04 jun 2007: improved support for Visual Studio 2005: crash with accessing - invalid std::vector element [0] fixed, and level 3 and 4 warnings removed -*) 02 jun 2007: made the encoder add a tag with version by default -*) 27 may 2007: zlib and png code separated (but still in the same file), - simple encoder/decoder functions added for more simple usage cases -*) 19 may 2007: minor fixes, some code cleaning, new error added (error 69), - moved some examples from here to lodepng_examples.cpp -*) 12 may 2007: palette decoding bug fixed -*) 24 apr 2007: changed the license from BSD to the zlib license -*) 11 mar 2007: very simple addition: ability to encode bKGD chunks. -*) 04 mar 2007: (!) tEXt chunk related fixes, and support for encoding - palettized PNG images. Plus little interface change with palette and texts. -*) 03 mar 2007: Made it encode dynamic Huffman shorter with repeat codes. - Fixed a bug where the end code of a block had length 0 in the Huffman tree. -*) 26 feb 2007: Huffman compression with dynamic trees (BTYPE 2) now implemented - and supported by the encoder, resulting in smaller PNGs at the output. -*) 27 jan 2007: Made the Adler-32 test faster so that a timewaste is gone. -*) 24 jan 2007: gave encoder an error interface. Added color conversion from any - greyscale type to 8-bit greyscale with or without alpha. -*) 21 jan 2007: (!) Totally changed the interface. It allows more color types - to convert to and is more uniform. See the manual for how it works now. -*) 07 jan 2007: Some cleanup & fixes, and a few changes over the last days: - encode/decode custom tEXt chunks, separate classes for zlib & deflate, and - at last made the decoder give errors for incorrect Adler32 or Crc. -*) 01 jan 2007: Fixed bug with encoding PNGs with less than 8 bits per channel. -*) 29 dec 2006: Added support for encoding images without alpha channel, and - cleaned out code as well as making certain parts faster. -*) 28 dec 2006: Added "Settings" to the encoder. -*) 26 dec 2006: The encoder now does LZ77 encoding and produces much smaller files now. - Removed some code duplication in the decoder. Fixed little bug in an example. -*) 09 dec 2006: (!) Placed output parameters of public functions as first parameter. - Fixed a bug of the decoder with 16-bit per color. -*) 15 okt 2006: Changed documentation structure -*) 09 okt 2006: Encoder class added. It encodes a valid PNG image from the - given image buffer, however for now it's not compressed. -*) 08 sep 2006: (!) Changed to interface with a Decoder class -*) 30 jul 2006: (!) LodePNG_InfoPng , width and height are now retrieved in different - way. Renamed decodePNG to decodePNGGeneric. -*) 29 jul 2006: (!) Changed the interface: image info is now returned as a - struct of type LodePNG::LodePNG_Info, instead of a vector, which was a bit clumsy. -*) 28 jul 2006: Cleaned the code and added new error checks. - Corrected terminology "deflate" into "inflate". -*) 23 jun 2006: Added SDL example in the documentation in the header, this - example allows easy debugging by displaying the PNG and its transparency. -*) 22 jun 2006: (!) Changed way to obtain error value. Added - loadFile function for convenience. Made decodePNG32 faster. -*) 21 jun 2006: (!) Changed type of info vector to unsigned. - Changed position of palette in info vector. Fixed an important bug that - happened on PNGs with an uncompressed block. -*) 16 jun 2006: Internally changed unsigned into unsigned where - needed, and performed some optimizations. -*) 07 jun 2006: (!) Renamed functions to decodePNG and placed them - in LodePNG namespace. Changed the order of the parameters. Rewrote the - documentation in the header. Renamed files to lodepng.cpp and lodepng.h -*) 22 apr 2006: Optimized and improved some code -*) 07 sep 2005: (!) Changed to std::vector interface -*) 12 aug 2005: Initial release (C++, decoder only) - - -12. contact information ------------------------ - -Feel free to contact me with suggestions, problems, comments, ... concerning -LodePNG. If you encounter a PNG image that doesn't work properly with this -decoder, feel free to send it and I'll use it to find and fix the problem. - -My email address is (puzzle the account and domain together with an @ symbol): -Domain: gmail dot com. -Account: lode dot vandevenne. - - -Copyright (c) 2005-2013 Lode Vandevenne -*/ diff --git a/rpcs3.sln b/rpcs3.sln index 2d6dce268f..a67317b751 100644 --- a/rpcs3.sln +++ b/rpcs3.sln @@ -57,16 +57,10 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "xml", "wxWidgets\build\msw\ EndProject Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "xrc", "wxWidgets\build\msw\wx_xrc.vcxproj", "{8BECCA95-C7D7-CFF8-FDB1-4950E9F8E8E6}" EndProject -Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "lodepng", "lodepng", "{48675C2D-D949-400D-8758-E22FB6DF3D58}" +Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "stblib", "stblib", "{9D839DFB-76E6-4F10-8EED-BA2AC7CC3FB6}" ProjectSection(SolutionItems) = preProject - lodepng\lodepng.cpp = lodepng\lodepng.cpp - lodepng\lodepng.h = lodepng\lodepng.h - EndProjectSection -EndProject -Project("{2150E333-8FDC-42A3-9474-1A3956D46DE8}") = "jpeg-compressor", "jpeg-compressor", "{9D9C8EDB-4469-42B3-A3C7-712CDABC43B0}" - ProjectSection(SolutionItems) = preProject - jpeg-compressor\jpgd.cpp = jpeg-compressor\jpgd.cpp - jpeg-compressor\jpgd.h = jpeg-compressor\jpgd.h + stblib\stb_image.c = stblib\stb_image.c + stblib\stb_image.h = stblib\stb_image.h EndProjectSection EndProject Global diff --git a/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp new file mode 100644 index 0000000000..ee46e323da --- /dev/null +++ b/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp @@ -0,0 +1,206 @@ +#include "stdafx.h" +#include "Emu/SysCalls/SysCalls.h" +#include "Emu/SysCalls/SC_FUNC.h" + +#include "stblib/stb_image.h" +#include "stblib/stb_image.c" // (TODO: Should we put this elsewhere?) + +void cellGifDec_init(); +Module cellGifDec(0xf010, cellGifDec_init); + +//Return Codes +enum +{ + CELL_GIFDEC_ERROR_OPEN_FILE = 0x80611300, + CELL_GIFDEC_ERROR_STREAM_FORMAT = 0x80611301, + CELL_GIFDEC_ERROR_SEQ = 0x80611302, + CELL_GIFDEC_ERROR_ARG = 0x80611303, + CELL_GIFDEC_ERROR_FATAL = 0x80611304, + CELL_GIFDEC_ERROR_SPU_UNSUPPORT = 0x80611305, + CELL_GIFDEC_ERROR_SPU_ERROR = 0x80611306, + CELL_GIFDEC_ERROR_CB_PARAM = 0x80611307, +}; + +struct CellGifDecInfo +{ + u32 SWidth; + u32 SHeight; + u32 SGlobalColorTableFlag; + u32 SColorResolution; + u32 SSortFlag; + u32 SSizeOfGlobalColorTable; + u32 SBackGroundColor; + u32 SPixelAspectRatio; +}; + +struct CellGifDecSrc +{ + u32 srcSelect; // CellGifDecStreamSrcSel + u32 fileName; // const char* + u64 fileOffset; // int64_t + u32 fileSize; + u32 streamPtr; + u32 streamSize; + u32 spuThreadEnable; // CellGifDecSpuThreadEna +}; + +CellGifDecInfo current_info; +CellGifDecSrc current_src; + + +int cellGifDecCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam) +{ + UNIMPLEMENTED_FUNC(cellGifDec); + return CELL_OK; +} + +int cellGifDecExtCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam, u32 extThreadInParam, u32 extThreadOutParam) +{ + UNIMPLEMENTED_FUNC(cellGifDec); + return CELL_OK; +} + +int cellGifDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInfo) +{ + //current_src.srcSelect = Memory.Read32(src_addr); + current_src.fileName = Memory.Read32(src_addr+4); + //current_src.fileOffset = Memory.Read32(src_addr+8); + //current_src.fileSize = Memory.Read32(src_addr+12); + //current_src.streamPtr = Memory.Read32(src_addr+16); + //current_src.streamSize = Memory.Read32(src_addr+20); + //current_src.spuThreadEnable = Memory.Read32(src_addr+24); + + u32& fd_addr = subHandle_addr; // Set file descriptor as sub handler of the decoder + int ret = cellFsOpen(current_src.fileName, 0, fd_addr, NULL, 0); + if(ret != 0) return CELL_GIFDEC_ERROR_OPEN_FILE; + + return CELL_OK; +} + +int cellGifDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) +{ + u32& fd = subHandle; + + //Check size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(fd, sb_addr); + u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); + + //Write the header to buffer + u32 buffer = Memory.Alloc(13,1); // Alloc buffer for GIF header + u32 pos_addr = Memory.Alloc(8,1); + cellFsLseek(fd, 0, 0, pos_addr); + cellFsRead(fd, buffer, 13, NULL); + Memory.Free(pos_addr); + + if (Memory.Read32(buffer) != 0x47494638 || + (Memory.Read16(buffer+4) != 0x3961 && + Memory.Read16(buffer+4) != 0x3761)) // Error: The first 6 bytes are not a valid GIF signature + { + Memory.Free(buffer); + return CELL_GIFDEC_ERROR_STREAM_FORMAT; // Surprisingly there is no error code related with headerss + } + + u8 packedField = Memory.Read8(buffer+10); + current_info.SWidth = Memory.Read8(buffer+6) + Memory.Read8(buffer+7) * 256; + current_info.SHeight = Memory.Read8(buffer+8) + Memory.Read8(buffer+9) * 256; + current_info.SGlobalColorTableFlag = packedField >> 7; + current_info.SColorResolution = (packedField >> 4) & 7; + current_info.SSortFlag = (packedField >> 3) & 1; + current_info.SSizeOfGlobalColorTable = packedField & 7; + current_info.SBackGroundColor = Memory.Read8(buffer+11); + current_info.SPixelAspectRatio = Memory.Read8(buffer+12); + + mem_class_t info(info_addr); + info += current_info.SWidth; + info += current_info.SHeight; + info += current_info.SGlobalColorTableFlag; + info += current_info.SColorResolution; + info += current_info.SSortFlag; + info += current_info.SSizeOfGlobalColorTable; + info += current_info.SBackGroundColor; + info += current_info.SPixelAspectRatio; + Memory.Free(buffer); + + return CELL_OK; +} + +int cellGifDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outParam) +{ + UNIMPLEMENTED_FUNC(cellGifDec); + return CELL_OK; +} + +int cellGifDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataCtrlParam_addr, u32 dataOutInfo_addr) +{ + u32& fd = subHandle; + + //Get size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(fd, sb_addr); + u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); + + //Copy the GIF file to a buffer + u32 buffer = Memory.Alloc(fileSize,1); + u32 pos_addr = Memory.Alloc(8,1); + cellFsLseek(fd, 0, 0, pos_addr); + cellFsRead(fd, buffer, fileSize, NULL); + Memory.Free(pos_addr); + + //Decode GIF file. (TODO: Is there any faster alternative? Can we do it without external libraries?) + int width, height, actual_components; + unsigned char *gif = new unsigned char [fileSize]; + for(u32 i = 0; i < fileSize; i++){ + gif[i] = Memory.Read8(buffer+i); + } + + unsigned char *image = stbi_load_from_memory((const unsigned char*)gif, fileSize, &width, &height, &actual_components, 4); + if (!image) + { + Memory.Free(buffer); + return CELL_GIFDEC_ERROR_STREAM_FORMAT; + } + u32 image_size = width * height * 4; + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+3]); + Memory.Write8(data_addr+i+1, image[i+0]); + Memory.Write8(data_addr+i+2, image[i+1]); + Memory.Write8(data_addr+i+3, image[i+2]); + } + Memory.Free(buffer); + + return CELL_OK; +} + +int cellGifDecClose(u32 mainHandle, u32 subHandle) +{ + u32& fd = subHandle; + cellFsClose(fd); + + return CELL_OK; +} + +int cellGifDecDestroy(u32 mainHandle) +{ + UNIMPLEMENTED_FUNC(cellGifDec); + return CELL_OK; +} + +void cellGifDec_init() +{ + cellGifDec.AddFunc(0xb60d42a5, cellGifDecCreate); + cellGifDec.AddFunc(0x4711cb7f, cellGifDecExtCreate); + cellGifDec.AddFunc(0x75745079, cellGifDecOpen); + cellGifDec.AddFunc(0xf0da95de, cellGifDecReadHeader); + cellGifDec.AddFunc(0x41a90dc4, cellGifDecSetParameter); + cellGifDec.AddFunc(0x44b1bc61, cellGifDecDecodeData); + cellGifDec.AddFunc(0x116a7da9, cellGifDecClose); + cellGifDec.AddFunc(0xe74b2cb1, cellGifDecDestroy); + + /*cellGifDec.AddFunc(0x17fb83c1, cellGifDecExtOpen); + cellGifDec.AddFunc(0xe53f91f2, cellGifDecExtReadHeader); + cellGifDec.AddFunc(0x95cae771, cellGifDecExtSetParameter); + cellGifDec.AddFunc(0x02e7e03e, cellGifDecExtDecodeData);*/ +} \ No newline at end of file diff --git a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp index 5aefb12eba..d8bac54859 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp @@ -2,7 +2,7 @@ #include "Emu/SysCalls/SysCalls.h" #include "Emu/SysCalls/SC_FUNC.h" -#include "jpeg-compressor/jpgd.cpp" +#include "stblib/stb_image.h" void cellJpgDec_init(); Module cellJpgDec(0x000f, cellJpgDec_init); @@ -50,7 +50,7 @@ int cellJpgDecCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam) return CELL_OK; } -int cellJpgDecExtCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam) +int cellJpgDecExtCreate(u32 mainHandle, u32 threadInParam, u32 threadOutParam, u32 extThreadInParam, u32 extThreadOutParam) { UNIMPLEMENTED_FUNC(cellJpgDec); return CELL_OK; @@ -95,18 +95,18 @@ int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct cellFsFstat(fd, sb_addr); u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); //Copy the JPG file to a buffer u32 buffer = Memory.Alloc(fileSize,1); u32 pos_addr = Memory.Alloc(8,1); cellFsLseek(fd, 0, 0, pos_addr); cellFsRead(fd, buffer, fileSize, NULL); + Memory.Free(pos_addr); if (Memory.Read32(buffer) != 0xFFD8FFE0 || // Error: Not a valid SOI header Memory.Read32(buffer+6) != 0x4A464946) // Error: Not a valid JFIF string { - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_JPGDEC_ERROR_HEADER; } @@ -116,9 +116,17 @@ int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) while(i < fileSize) { i += block_length; // Increase the file index to get to the next block - if (i >= fileSize) return CELL_JPGDEC_ERROR_HEADER; // Check to protect against segmentation faults - if(Memory.Read8(buffer+i) != 0xFF) return CELL_JPGDEC_ERROR_HEADER; // Check that we are truly at the start of another block - if(Memory.Read8(buffer+i+1) == 0xC0) break; // 0xFFC0 is the "Start of frame" marker which contains the file size + if (i >= fileSize){ + Memory.Free(buffer); + return CELL_JPGDEC_ERROR_HEADER; // Check to protect against segmentation faults + } + if(Memory.Read8(buffer+i) != 0xFF){ + Memory.Free(buffer); + return CELL_JPGDEC_ERROR_HEADER; // Check that we are truly at the start of another block + } + if(Memory.Read8(buffer+i+1) == 0xC0){ + break; // 0xFFC0 is the "Start of frame" marker which contains the file size + } i += 2; // Skip the block marker block_length = Memory.Read8(buffer+i)*0xFF + Memory.Read8(buffer+i+1); // Go to the next block } @@ -133,9 +141,6 @@ int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) info += current_info.imageHeight; info += current_info.numComponents; info += current_info.colorSpace; - - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_OK; @@ -149,20 +154,28 @@ int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct cellFsFstat(fd, sb_addr); u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); //Copy the JPG file to a buffer u32 buffer = Memory.Alloc(fileSize,1); u32 pos_addr = Memory.Alloc(8,1); cellFsLseek(fd, 0, 0, pos_addr); cellFsRead(fd, buffer, fileSize, NULL); + Memory.Free(pos_addr); + //Decode JPG file. (TODO: Is there any faster alternative? Can we do it without external libraries?) int width, height, actual_components; unsigned char *jpg = new unsigned char [fileSize]; for(u32 i = 0; i < fileSize; i++){ jpg[i] = Memory.Read8(buffer+i); } - unsigned char *image = jpgd::decompress_jpeg_image_from_memory((const unsigned char*)jpg, fileSize, &width, &height, &actual_components, 4); + unsigned char *image = stbi_load_from_memory((const unsigned char*)jpg, fileSize, &width, &height, &actual_components, 4); + if (!image) + { + Memory.Free(buffer); + return CELL_JPGDEC_ERROR_STREAM_FORMAT; + } u32 image_size = width * height * 4; for(u32 i = 0; i < image_size; i+=4){ Memory.Write8(data_addr+i+0, image[i+3]); @@ -170,9 +183,6 @@ int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC Memory.Write8(data_addr+i+2, image[i+1]); Memory.Write8(data_addr+i+3, image[i+2]); } - - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_OK; diff --git a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp index c570e5230b..70659c9fed 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp @@ -2,7 +2,7 @@ #include "Emu/SysCalls/SysCalls.h" #include "Emu/SysCalls/SC_FUNC.h" -#include "lodepng/lodepng.cpp" +#include "stblib/stb_image.h" void cellPngDec_init(); Module cellPngDec(0x0018, cellPngDec_init); @@ -93,6 +93,7 @@ int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct cellFsFstat(fd, sb_addr); u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); if(fileSize < 29) return CELL_PNGDEC_ERROR_HEADER; // Error: The file is smaller than the length of a PNG header //Write the header to buffer @@ -100,13 +101,12 @@ int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) u32 pos_addr = Memory.Alloc(8,1); cellFsLseek(fd, 0, 0, pos_addr); cellFsRead(fd, buffer, 34, NULL); + Memory.Free(pos_addr); if (Memory.Read32(buffer) != 0x89504E47 || Memory.Read32(buffer+4) != 0x0D0A1A0A || // Error: The first 8 bytes are not a valid PNG signature Memory.Read32(buffer+12) != 0x49484452) // Error: The PNG file does not start with an IHDR chunk { - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_PNGDEC_ERROR_HEADER; } @@ -127,9 +127,6 @@ int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) info += current_info.bitDepth; info += current_info.interlaceMethod; info += current_info.chunkInformation; - - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_OK; @@ -143,44 +140,35 @@ int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct cellFsFstat(fd, sb_addr); u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); //Copy the PNG file to a buffer - u32 buffer = Memory.Alloc(fileSize,1); // Alloc buffer for PNG header + u32 buffer = Memory.Alloc(fileSize,1); u32 pos_addr = Memory.Alloc(8,1); cellFsLseek(fd, 0, 0, pos_addr); cellFsRead(fd, buffer, fileSize, NULL); + Memory.Free(pos_addr); //Decode PNG file. (TODO: Is there any faster alternative? Can we do it without external libraries?) - std::vector png; // PNG buffer - std::vector image; // Raw buffer - - //Load contents in png buffer - png.resize(size_t(fileSize)); + int width, height, actual_components; + unsigned char *png = new unsigned char [fileSize]; for(u32 i = 0; i < fileSize; i++){ png[i] = Memory.Read8(buffer+i); } - - //Decode - unsigned width, height; - unsigned error = lodepng::decode(image, width, height, png); - if (error) + + unsigned char *image = stbi_load_from_memory((const unsigned char*)png, fileSize, &width, &height, &actual_components, 4); + if (!image) { - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_PNGDEC_ERROR_STREAM_FORMAT; } - - u32 image_size = image.size(); + u32 image_size = width * height * 4; for(u32 i = 0; i < image_size; i+=4){ Memory.Write8(data_addr+i+0, image[i+3]); Memory.Write8(data_addr+i+1, image[i+0]); Memory.Write8(data_addr+i+2, image[i+1]); Memory.Write8(data_addr+i+3, image[i+2]); } - - Memory.Free(sb_addr); - Memory.Free(pos_addr); Memory.Free(buffer); return CELL_OK; diff --git a/rpcs3/rpcs3.vcxproj b/rpcs3/rpcs3.vcxproj index e55b47a184..fe1af9b6ff 100644 --- a/rpcs3/rpcs3.vcxproj +++ b/rpcs3/rpcs3.vcxproj @@ -245,6 +245,7 @@ + diff --git a/rpcs3/rpcs3.vcxproj.filters b/rpcs3/rpcs3.vcxproj.filters index 53aacbdd8e..09c648cf32 100644 --- a/rpcs3/rpcs3.vcxproj.filters +++ b/rpcs3/rpcs3.vcxproj.filters @@ -292,6 +292,9 @@ Emu\SysCalls\Modules + + Emu\SysCalls\Modules + diff --git a/stblib/stb_image.c b/stblib/stb_image.c new file mode 100644 index 0000000000..c1d494f8dd --- /dev/null +++ b/stblib/stb_image.c @@ -0,0 +1,4401 @@ +/* stbi-1.33 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c + when you control the images you're loading + no warranty implied; use at your own risk + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline (no JPEG progressive) + PNG 8-bit only + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - overridable dequantizing-IDCT, YCbCr-to-RGB conversion (define STBI_SIMD) + + Latest revisions: + 1.33 (2011-07-14) minor fixes suggested by Dave Moore + 1.32 (2011-07-13) info support for all filetypes (SpartanJ) + 1.31 (2011-06-19) a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) added ability to load files via io callbacks (Ben Wenger) + 1.29 (2010-08-16) various warning fixes from Aurelien Pocheville + + See end of file for full revision history. + + TODO: + stbi_info support for BMP,PSD,HDR,PIC + + + ============================ Contributors ========================= + + Image formats Optimizations & bugfixes + Sean Barrett (jpeg, png, bmp) Fabian "ryg" Giesen + Nicolas Schulz (hdr, psd) + Jonathan Dummer (tga) Bug fixes & warning fixes + Jean-Marc Lienher (gif) Marc LeBlanc + Tom Seddon (pic) Christpher Lloyd + Thatcher Ulrich (psd) Dave Moore + Won Chun + the Horde3D community + Extensions, features Janez Zemva + Jetro Lauha (stbi_info) Jonathan Blow + James "moose2000" Brown (iPhone PNG) Laurent Gomila + Ben "Disch" Wenger (io callbacks) Aruelien Pocheville + Martin "SpartanJ" Golini Ryamond Barbiero + David Woo + + + If your name should be here but isn't, let Sean know. + +*/ + +#ifndef STBI_HEADER_FILE_ONLY + +#ifndef STBI_NO_HDR +#include // ldexp +#include // strcmp, strtok +#endif + +#ifndef STBI_NO_STDIO +#include +#endif +#include +#include +#include +#include + +#ifndef _MSC_VER + #ifdef __cplusplus + #define stbi_inline inline + #else + #define stbi_inline + #endif +#else + #define stbi_inline __forceinline +#endif + + +// implementation: +typedef unsigned char uint8; +typedef unsigned short uint16; +typedef signed short int16; +typedef unsigned int uint32; +typedef signed int int32; +typedef unsigned int uint; + +// should produce compiler error if size is wrong +typedef unsigned char validate_uint32[sizeof(uint32)==4 ? 1 : -1]; + +#if defined(STBI_NO_STDIO) && !defined(STBI_NO_WRITE) +#define STBI_NO_WRITE +#endif + +#define STBI_NOTUSED(v) (void)sizeof(v) + +#ifdef _MSC_VER +#define STBI_HAS_LROTL +#endif + +#ifdef STBI_HAS_LROTL + #define stbi_lrot(x,y) _lrotl(x,y) +#else + #define stbi_lrot(x,y) (((x) << (y)) | ((x) >> (32 - (y)))) +#endif + +/////////////////////////////////////////////// +// +// stbi struct and start_xxx functions + +// stbi structure is our basic context used by all images, so it +// contains all the IO context, plus some basic image information +typedef struct +{ + uint32 img_x, img_y; + int img_n, img_out_n; + + stbi_io_callbacks io; + void *io_user_data; + + int read_from_callbacks; + int buflen; + uint8 buffer_start[128]; + + uint8 *img_buffer, *img_buffer_end; + uint8 *img_buffer_original; +} stbi; + + +static void refill_buffer(stbi *s); + +// initialize a memory-decode context +static void start_mem(stbi *s, uint8 const *buffer, int len) +{ + s->io.read = NULL; + s->read_from_callbacks = 0; + s->img_buffer = s->img_buffer_original = (uint8 *) buffer; + s->img_buffer_end = (uint8 *) buffer+len; +} + +// initialize a callback-based context +static void start_callbacks(stbi *s, stbi_io_callbacks *c, void *user) +{ + s->io = *c; + s->io_user_data = user; + s->buflen = sizeof(s->buffer_start); + s->read_from_callbacks = 1; + s->img_buffer_original = s->buffer_start; + refill_buffer(s); +} + +#ifndef STBI_NO_STDIO + +static int stdio_read(void *user, char *data, int size) +{ + return (int) fread(data,1,size,(FILE*) user); +} + +static void stdio_skip(void *user, unsigned n) +{ + fseek((FILE*) user, n, SEEK_CUR); +} + +static int stdio_eof(void *user) +{ + return feof((FILE*) user); +} + +static stbi_io_callbacks stbi_stdio_callbacks = +{ + stdio_read, + stdio_skip, + stdio_eof, +}; + +static void start_file(stbi *s, FILE *f) +{ + start_callbacks(s, &stbi_stdio_callbacks, (void *) f); +} + +//static void stop_file(stbi *s) { } + +#endif // !STBI_NO_STDIO + +static void stbi_rewind(stbi *s) +{ + // conceptually rewind SHOULD rewind to the beginning of the stream, + // but we just rewind to the beginning of the initial buffer, because + // we only use it after doing 'test', which only ever looks at at most 92 bytes + s->img_buffer = s->img_buffer_original; +} + +static int stbi_jpeg_test(stbi *s); +static stbi_uc *stbi_jpeg_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_jpeg_info(stbi *s, int *x, int *y, int *comp); +static int stbi_png_test(stbi *s); +static stbi_uc *stbi_png_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_png_info(stbi *s, int *x, int *y, int *comp); +static int stbi_bmp_test(stbi *s); +static stbi_uc *stbi_bmp_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_tga_test(stbi *s); +static stbi_uc *stbi_tga_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_tga_info(stbi *s, int *x, int *y, int *comp); +static int stbi_psd_test(stbi *s); +static stbi_uc *stbi_psd_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_hdr_test(stbi *s); +static float *stbi_hdr_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_pic_test(stbi *s); +static stbi_uc *stbi_pic_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_gif_test(stbi *s); +static stbi_uc *stbi_gif_load(stbi *s, int *x, int *y, int *comp, int req_comp); +static int stbi_gif_info(stbi *s, int *x, int *y, int *comp); + + +// this is not threadsafe +static const char *failure_reason; + +const char *stbi_failure_reason(void) +{ + return failure_reason; +} + +static int e(const char *str) +{ + failure_reason = str; + return 0; +} + +// e - error +// epf - error returning pointer to float +// epuc - error returning pointer to unsigned char + +#ifdef STBI_NO_FAILURE_STRINGS + #define e(x,y) 0 +#elif defined(STBI_FAILURE_USERMSG) + #define e(x,y) e(y) +#else + #define e(x,y) e(x) +#endif + +#define epf(x,y) ((float *) (e(x,y)?NULL:NULL)) +#define epuc(x,y) ((unsigned char *) (e(x,y)?NULL:NULL)) + +void stbi_image_free(void *retval_from_stbi_load) +{ + free(retval_from_stbi_load); +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp); +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp); +#endif + +static unsigned char *stbi_load_main(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + if (stbi_jpeg_test(s)) return stbi_jpeg_load(s,x,y,comp,req_comp); + if (stbi_png_test(s)) return stbi_png_load(s,x,y,comp,req_comp); + if (stbi_bmp_test(s)) return stbi_bmp_load(s,x,y,comp,req_comp); + if (stbi_gif_test(s)) return stbi_gif_load(s,x,y,comp,req_comp); + if (stbi_psd_test(s)) return stbi_psd_load(s,x,y,comp,req_comp); + if (stbi_pic_test(s)) return stbi_pic_load(s,x,y,comp,req_comp); + + #ifndef STBI_NO_HDR + if (stbi_hdr_test(s)) { + float *hdr = stbi_hdr_load(s, x,y,comp,req_comp); + return hdr_to_ldr(hdr, *x, *y, req_comp ? req_comp : *comp); + } + #endif + + // test tga last because it's a crappy test! + if (stbi_tga_test(s)) + return stbi_tga_load(s,x,y,comp,req_comp); + return epuc("unknown image type", "Image not of any known type, or corrupt"); +} + +#ifndef STBI_NO_STDIO +unsigned char *stbi_load(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + unsigned char *result; + if (!f) return epuc("can't fopen", "Unable to open file"); + result = stbi_load_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +unsigned char *stbi_load_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s,f); + return stbi_load_main(&s,x,y,comp,req_comp); +} +#endif //!STBI_NO_STDIO + +unsigned char *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s,buffer,len); + return stbi_load_main(&s,x,y,comp,req_comp); +} + +unsigned char *stbi_load_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi_load_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_HDR + +float *stbi_loadf_main(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + unsigned char *data; + #ifndef STBI_NO_HDR + if (stbi_hdr_test(s)) + return stbi_hdr_load(s,x,y,comp,req_comp); + #endif + data = stbi_load_main(s, x, y, comp, req_comp); + if (data) + return ldr_to_hdr(data, *x, *y, req_comp ? req_comp : *comp); + return epf("unknown image type", "Image not of any known type, or corrupt"); +} + +float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_mem(&s,buffer,len); + return stbi_loadf_main(&s,x,y,comp,req_comp); +} + +float *stbi_loadf_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi_loadf_main(&s,x,y,comp,req_comp); +} + +#ifndef STBI_NO_STDIO +float *stbi_loadf(char const *filename, int *x, int *y, int *comp, int req_comp) +{ + FILE *f = fopen(filename, "rb"); + float *result; + if (!f) return epf("can't fopen", "Unable to open file"); + result = stbi_loadf_from_file(f,x,y,comp,req_comp); + fclose(f); + return result; +} + +float *stbi_loadf_from_file(FILE *f, int *x, int *y, int *comp, int req_comp) +{ + stbi s; + start_file(&s,f); + return stbi_loadf_main(&s,x,y,comp,req_comp); +} +#endif // !STBI_NO_STDIO + +#endif // !STBI_NO_HDR + +// these is-hdr-or-not is defined independent of whether STBI_NO_HDR is +// defined, for API simplicity; if STBI_NO_HDR is defined, it always +// reports false! + +int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len) +{ + #ifndef STBI_NO_HDR + stbi s; + start_mem(&s,buffer,len); + return stbi_hdr_test(&s); + #else + STBI_NOTUSED(buffer); + STBI_NOTUSED(len); + return 0; + #endif +} + +#ifndef STBI_NO_STDIO +extern int stbi_is_hdr (char const *filename) +{ + FILE *f = fopen(filename, "rb"); + int result=0; + if (f) { + result = stbi_is_hdr_from_file(f); + fclose(f); + } + return result; +} + +extern int stbi_is_hdr_from_file(FILE *f) +{ + #ifndef STBI_NO_HDR + stbi s; + start_file(&s,f); + return stbi_hdr_test(&s); + #else + return 0; + #endif +} +#endif // !STBI_NO_STDIO + +extern int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user) +{ + #ifndef STBI_NO_HDR + stbi s; + start_callbacks(&s, (stbi_io_callbacks *) clbk, user); + return stbi_hdr_test(&s); + #else + return 0; + #endif +} + +#ifndef STBI_NO_HDR +static float h2l_gamma_i=1.0f/2.2f, h2l_scale_i=1.0f; +static float l2h_gamma=2.2f, l2h_scale=1.0f; + +void stbi_hdr_to_ldr_gamma(float gamma) { h2l_gamma_i = 1/gamma; } +void stbi_hdr_to_ldr_scale(float scale) { h2l_scale_i = 1/scale; } + +void stbi_ldr_to_hdr_gamma(float gamma) { l2h_gamma = gamma; } +void stbi_ldr_to_hdr_scale(float scale) { l2h_scale = scale; } +#endif + + +////////////////////////////////////////////////////////////////////////////// +// +// Common code used by all image loaders +// + +enum +{ + SCAN_load=0, + SCAN_type, + SCAN_header +}; + +static void refill_buffer(stbi *s) +{ + int n = (s->io.read)(s->io_user_data,(char*)s->buffer_start,s->buflen); + if (n == 0) { + // at end of file, treat same as if from memory + s->read_from_callbacks = 0; + s->img_buffer = s->img_buffer_end-1; + *s->img_buffer = 0; + } else { + s->img_buffer = s->buffer_start; + s->img_buffer_end = s->buffer_start + n; + } +} + +stbi_inline static int get8(stbi *s) +{ + if (s->img_buffer < s->img_buffer_end) + return *s->img_buffer++; + if (s->read_from_callbacks) { + refill_buffer(s); + return *s->img_buffer++; + } + return 0; +} + +stbi_inline static int at_eof(stbi *s) +{ + if (s->io.read) { + if (!(s->io.eof)(s->io_user_data)) return 0; + // if feof() is true, check if buffer = end + // special case: we've only got the special 0 character at the end + if (s->read_from_callbacks == 0) return 1; + } + + return s->img_buffer >= s->img_buffer_end; +} + +stbi_inline static uint8 get8u(stbi *s) +{ + return (uint8) get8(s); +} + +static void skip(stbi *s, int n) +{ + if (s->io.read) { + int blen = s->img_buffer_end - s->img_buffer; + if (blen < n) { + s->img_buffer = s->img_buffer_end; + (s->io.skip)(s->io_user_data, n - blen); + return; + } + } + s->img_buffer += n; +} + +static int getn(stbi *s, stbi_uc *buffer, int n) +{ + if (s->io.read) { + int blen = s->img_buffer_end - s->img_buffer; + if (blen < n) { + int res, count; + + memcpy(buffer, s->img_buffer, blen); + + count = (s->io.read)(s->io_user_data, (char*) buffer + blen, n - blen); + res = (count == (n-blen)); + s->img_buffer = s->img_buffer_end; + return res; + } + } + + if (s->img_buffer+n <= s->img_buffer_end) { + memcpy(buffer, s->img_buffer, n); + s->img_buffer += n; + return 1; + } else + return 0; +} + +static int get16(stbi *s) +{ + int z = get8(s); + return (z << 8) + get8(s); +} + +static uint32 get32(stbi *s) +{ + uint32 z = get16(s); + return (z << 16) + get16(s); +} + +static int get16le(stbi *s) +{ + int z = get8(s); + return z + (get8(s) << 8); +} + +static uint32 get32le(stbi *s) +{ + uint32 z = get16le(s); + return z + (get16le(s) << 16); +} + +////////////////////////////////////////////////////////////////////////////// +// +// generic converter from built-in img_n to req_comp +// individual types do this automatically as much as possible (e.g. jpeg +// does all cases internally since it needs to colorspace convert anyway, +// and it never has alpha, so very few cases ). png can automatically +// interleave an alpha=255 channel, but falls back to this for other cases +// +// assume data buffer is malloced, so malloc a new one and free that one +// only failure mode is malloc failing + +static uint8 compute_y(int r, int g, int b) +{ + return (uint8) (((r*77) + (g*150) + (29*b)) >> 8); +} + +static unsigned char *convert_format(unsigned char *data, int img_n, int req_comp, uint x, uint y) +{ + int i,j; + unsigned char *good; + + if (req_comp == img_n) return data; + assert(req_comp >= 1 && req_comp <= 4); + + good = (unsigned char *) malloc(req_comp * x * y); + if (good == NULL) { + free(data); + return epuc("outofmem", "Out of memory"); + } + + for (j=0; j < (int) y; ++j) { + unsigned char *src = data + j * x * img_n ; + unsigned char *dest = good + j * x * req_comp; + + #define COMBO(a,b) ((a)*8+(b)) + #define CASE(a,b) case COMBO(a,b): for(i=x-1; i >= 0; --i, src += a, dest += b) + // convert source image with img_n components to one with req_comp components; + // avoid switch per pixel, so use switch per scanline and massive macros + switch (COMBO(img_n, req_comp)) { + CASE(1,2) dest[0]=src[0], dest[1]=255; break; + CASE(1,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(1,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=255; break; + CASE(2,1) dest[0]=src[0]; break; + CASE(2,3) dest[0]=dest[1]=dest[2]=src[0]; break; + CASE(2,4) dest[0]=dest[1]=dest[2]=src[0], dest[3]=src[1]; break; + CASE(3,4) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2],dest[3]=255; break; + CASE(3,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(3,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = 255; break; + CASE(4,1) dest[0]=compute_y(src[0],src[1],src[2]); break; + CASE(4,2) dest[0]=compute_y(src[0],src[1],src[2]), dest[1] = src[3]; break; + CASE(4,3) dest[0]=src[0],dest[1]=src[1],dest[2]=src[2]; break; + default: assert(0); + } + #undef CASE + } + + free(data); + return good; +} + +#ifndef STBI_NO_HDR +static float *ldr_to_hdr(stbi_uc *data, int x, int y, int comp) +{ + int i,k,n; + float *output = (float *) malloc(x * y * comp * sizeof(float)); + if (output == NULL) { free(data); return epf("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + output[i*comp + k] = (float) pow(data[i*comp+k]/255.0f, l2h_gamma) * l2h_scale; + } + if (k < comp) output[i*comp + k] = data[i*comp+k]/255.0f; + } + free(data); + return output; +} + +#define float2int(x) ((int) (x)) +static stbi_uc *hdr_to_ldr(float *data, int x, int y, int comp) +{ + int i,k,n; + stbi_uc *output = (stbi_uc *) malloc(x * y * comp); + if (output == NULL) { free(data); return epuc("outofmem", "Out of memory"); } + // compute number of non-alpha components + if (comp & 1) n = comp; else n = comp-1; + for (i=0; i < x*y; ++i) { + for (k=0; k < n; ++k) { + float z = (float) pow(data[i*comp+k]*h2l_scale_i, h2l_gamma_i) * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (uint8) float2int(z); + } + if (k < comp) { + float z = data[i*comp+k] * 255 + 0.5f; + if (z < 0) z = 0; + if (z > 255) z = 255; + output[i*comp + k] = (uint8) float2int(z); + } + } + free(data); + return output; +} +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// "baseline" JPEG/JFIF decoder (not actually fully baseline implementation) +// +// simple implementation +// - channel subsampling of at most 2 in each dimension +// - doesn't support delayed output of y-dimension +// - simple interface (only one output format: 8-bit interleaved RGB) +// - doesn't try to recover corrupt jpegs +// - doesn't allow partial loading, loading multiple at once +// - still fast on x86 (copying globals into locals doesn't help x86) +// - allocates lots of intermediate memory (full size of all components) +// - non-interleaved case requires this anyway +// - allows good upsampling (see next) +// high-quality +// - upsampled channels are bilinearly interpolated, even across blocks +// - quality integer IDCT derived from IJG's 'slow' +// performance +// - fast huffman; reasonable integer IDCT +// - uses a lot of intermediate memory, could cache poorly +// - load http://nothings.org/remote/anemones.jpg 3 times on 2.8Ghz P4 +// stb_jpeg: 1.34 seconds (MSVC6, default release build) +// stb_jpeg: 1.06 seconds (MSVC6, processor = Pentium Pro) +// IJL11.dll: 1.08 seconds (compiled by intel) +// IJG 1998: 0.98 seconds (MSVC6, makefile provided by IJG) +// IJG 1998: 0.95 seconds (MSVC6, makefile + proc=PPro) + +// huffman decoding acceleration +#define FAST_BITS 9 // larger handles more cases; smaller stomps less cache + +typedef struct +{ + uint8 fast[1 << FAST_BITS]; + // weirdly, repacking this into AoS is a 10% speed loss, instead of a win + uint16 code[256]; + uint8 values[256]; + uint8 size[257]; + unsigned int maxcode[18]; + int delta[17]; // old 'firstsymbol' - old 'firstcode' +} huffman; + +typedef struct +{ + #ifdef STBI_SIMD + unsigned short dequant2[4][64]; + #endif + stbi *s; + huffman huff_dc[4]; + huffman huff_ac[4]; + uint8 dequant[4][64]; + +// sizes for components, interleaved MCUs + int img_h_max, img_v_max; + int img_mcu_x, img_mcu_y; + int img_mcu_w, img_mcu_h; + +// definition of jpeg image component + struct + { + int id; + int h,v; + int tq; + int hd,ha; + int dc_pred; + + int x,y,w2,h2; + uint8 *data; + void *raw_data; + uint8 *linebuf; + } img_comp[4]; + + uint32 code_buffer; // jpeg entropy-coded buffer + int code_bits; // number of valid bits + unsigned char marker; // marker seen while filling entropy buffer + int nomore; // flag if we saw a marker so must stop + + int scan_n, order[4]; + int restart_interval, todo; +} jpeg; + +static int build_huffman(huffman *h, int *count) +{ + int i,j,k=0,code; + // build size list for each symbol (from JPEG spec) + for (i=0; i < 16; ++i) + for (j=0; j < count[i]; ++j) + h->size[k++] = (uint8) (i+1); + h->size[k] = 0; + + // compute actual symbols (from jpeg spec) + code = 0; + k = 0; + for(j=1; j <= 16; ++j) { + // compute delta to add to code to compute symbol id + h->delta[j] = k - code; + if (h->size[k] == j) { + while (h->size[k] == j) + h->code[k++] = (uint16) (code++); + if (code-1 >= (1 << j)) return e("bad code lengths","Corrupt JPEG"); + } + // compute largest code + 1 for this size, preshifted as needed later + h->maxcode[j] = code << (16-j); + code <<= 1; + } + h->maxcode[j] = 0xffffffff; + + // build non-spec acceleration table; 255 is flag for not-accelerated + memset(h->fast, 255, 1 << FAST_BITS); + for (i=0; i < k; ++i) { + int s = h->size[i]; + if (s <= FAST_BITS) { + int c = h->code[i] << (FAST_BITS-s); + int m = 1 << (FAST_BITS-s); + for (j=0; j < m; ++j) { + h->fast[c+j] = (uint8) i; + } + } + } + return 1; +} + +static void grow_buffer_unsafe(jpeg *j) +{ + do { + int b = j->nomore ? 0 : get8(j->s); + if (b == 0xff) { + int c = get8(j->s); + if (c != 0) { + j->marker = (unsigned char) c; + j->nomore = 1; + return; + } + } + j->code_buffer |= b << (24 - j->code_bits); + j->code_bits += 8; + } while (j->code_bits <= 24); +} + +// (1 << n) - 1 +static uint32 bmask[17]={0,1,3,7,15,31,63,127,255,511,1023,2047,4095,8191,16383,32767,65535}; + +// decode a jpeg huffman value from the bitstream +stbi_inline static int decode(jpeg *j, huffman *h) +{ + unsigned int temp; + int c,k; + + if (j->code_bits < 16) grow_buffer_unsafe(j); + + // look at the top FAST_BITS and determine what symbol ID it is, + // if the code is <= FAST_BITS + c = (j->code_buffer >> (32 - FAST_BITS)) & ((1 << FAST_BITS)-1); + k = h->fast[c]; + if (k < 255) { + int s = h->size[k]; + if (s > j->code_bits) + return -1; + j->code_buffer <<= s; + j->code_bits -= s; + return h->values[k]; + } + + // naive test is to shift the code_buffer down so k bits are + // valid, then test against maxcode. To speed this up, we've + // preshifted maxcode left so that it has (16-k) 0s at the + // end; in other words, regardless of the number of bits, it + // wants to be compared against something shifted to have 16; + // that way we don't need to shift inside the loop. + temp = j->code_buffer >> 16; + for (k=FAST_BITS+1 ; ; ++k) + if (temp < h->maxcode[k]) + break; + if (k == 17) { + // error! code not found + j->code_bits -= 16; + return -1; + } + + if (k > j->code_bits) + return -1; + + // convert the huffman code to the symbol id + c = ((j->code_buffer >> (32 - k)) & bmask[k]) + h->delta[k]; + assert((((j->code_buffer) >> (32 - h->size[c])) & bmask[h->size[c]]) == h->code[c]); + + // convert the id to a symbol + j->code_bits -= k; + j->code_buffer <<= k; + return h->values[c]; +} + +// combined JPEG 'receive' and JPEG 'extend', since baseline +// always extends everything it receives. +stbi_inline static int extend_receive(jpeg *j, int n) +{ + unsigned int m = 1 << (n-1); + unsigned int k; + if (j->code_bits < n) grow_buffer_unsafe(j); + + #if 1 + k = stbi_lrot(j->code_buffer, n); + j->code_buffer = k & ~bmask[n]; + k &= bmask[n]; + j->code_bits -= n; + #else + k = (j->code_buffer >> (32 - n)) & bmask[n]; + j->code_bits -= n; + j->code_buffer <<= n; + #endif + // the following test is probably a random branch that won't + // predict well. I tried to table accelerate it but failed. + // maybe it's compiling as a conditional move? + if (k < m) + return (-1 << n) + k + 1; + else + return k; +} + +// given a value that's at position X in the zigzag stream, +// where does it appear in the 8x8 matrix coded as row-major? +static uint8 dezigzag[64+15] = +{ + 0, 1, 8, 16, 9, 2, 3, 10, + 17, 24, 32, 25, 18, 11, 4, 5, + 12, 19, 26, 33, 40, 48, 41, 34, + 27, 20, 13, 6, 7, 14, 21, 28, + 35, 42, 49, 56, 57, 50, 43, 36, + 29, 22, 15, 23, 30, 37, 44, 51, + 58, 59, 52, 45, 38, 31, 39, 46, + 53, 60, 61, 54, 47, 55, 62, 63, + // let corrupt input sample past end + 63, 63, 63, 63, 63, 63, 63, 63, + 63, 63, 63, 63, 63, 63, 63 +}; + +// decode one 64-entry block-- +static int decode_block(jpeg *j, short data[64], huffman *hdc, huffman *hac, int b) +{ + int diff,dc,k; + int t = decode(j, hdc); + if (t < 0) return e("bad huffman code","Corrupt JPEG"); + + // 0 all the ac values now so we can do it 32-bits at a time + memset(data,0,64*sizeof(data[0])); + + diff = t ? extend_receive(j, t) : 0; + dc = j->img_comp[b].dc_pred + diff; + j->img_comp[b].dc_pred = dc; + data[0] = (short) dc; + + // decode AC components, see JPEG spec + k = 1; + do { + int r,s; + int rs = decode(j, hac); + if (rs < 0) return e("bad huffman code","Corrupt JPEG"); + s = rs & 15; + r = rs >> 4; + if (s == 0) { + if (rs != 0xf0) break; // end block + k += 16; + } else { + k += r; + // decode into unzigzag'd location + data[dezigzag[k++]] = (short) extend_receive(j,s); + } + } while (k < 64); + return 1; +} + +// take a -128..127 value and clamp it and convert to 0..255 +stbi_inline static uint8 clamp(int x) +{ + // trick to use a single test to catch both cases + if ((unsigned int) x > 255) { + if (x < 0) return 0; + if (x > 255) return 255; + } + return (uint8) x; +} + +#define f2f(x) (int) (((x) * 4096 + 0.5)) +#define fsh(x) ((x) << 12) + +// derived from jidctint -- DCT_ISLOW +#define IDCT_1D(s0,s1,s2,s3,s4,s5,s6,s7) \ + int t0,t1,t2,t3,p1,p2,p3,p4,p5,x0,x1,x2,x3; \ + p2 = s2; \ + p3 = s6; \ + p1 = (p2+p3) * f2f(0.5411961f); \ + t2 = p1 + p3*f2f(-1.847759065f); \ + t3 = p1 + p2*f2f( 0.765366865f); \ + p2 = s0; \ + p3 = s4; \ + t0 = fsh(p2+p3); \ + t1 = fsh(p2-p3); \ + x0 = t0+t3; \ + x3 = t0-t3; \ + x1 = t1+t2; \ + x2 = t1-t2; \ + t0 = s7; \ + t1 = s5; \ + t2 = s3; \ + t3 = s1; \ + p3 = t0+t2; \ + p4 = t1+t3; \ + p1 = t0+t3; \ + p2 = t1+t2; \ + p5 = (p3+p4)*f2f( 1.175875602f); \ + t0 = t0*f2f( 0.298631336f); \ + t1 = t1*f2f( 2.053119869f); \ + t2 = t2*f2f( 3.072711026f); \ + t3 = t3*f2f( 1.501321110f); \ + p1 = p5 + p1*f2f(-0.899976223f); \ + p2 = p5 + p2*f2f(-2.562915447f); \ + p3 = p3*f2f(-1.961570560f); \ + p4 = p4*f2f(-0.390180644f); \ + t3 += p1+p4; \ + t2 += p2+p3; \ + t1 += p2+p4; \ + t0 += p1+p3; + +#ifdef STBI_SIMD +typedef unsigned short stbi_dequantize_t; +#else +typedef uint8 stbi_dequantize_t; +#endif + +// .344 seconds on 3*anemones.jpg +static void idct_block(uint8 *out, int out_stride, short data[64], stbi_dequantize_t *dequantize) +{ + int i,val[64],*v=val; + stbi_dequantize_t *dq = dequantize; + uint8 *o; + short *d = data; + + // columns + for (i=0; i < 8; ++i,++d,++dq, ++v) { + // if all zeroes, shortcut -- this avoids dequantizing 0s and IDCTing + if (d[ 8]==0 && d[16]==0 && d[24]==0 && d[32]==0 + && d[40]==0 && d[48]==0 && d[56]==0) { + // no shortcut 0 seconds + // (1|2|3|4|5|6|7)==0 0 seconds + // all separate -0.047 seconds + // 1 && 2|3 && 4|5 && 6|7: -0.047 seconds + int dcterm = d[0] * dq[0] << 2; + v[0] = v[8] = v[16] = v[24] = v[32] = v[40] = v[48] = v[56] = dcterm; + } else { + IDCT_1D(d[ 0]*dq[ 0],d[ 8]*dq[ 8],d[16]*dq[16],d[24]*dq[24], + d[32]*dq[32],d[40]*dq[40],d[48]*dq[48],d[56]*dq[56]) + // constants scaled things up by 1<<12; let's bring them back + // down, but keep 2 extra bits of precision + x0 += 512; x1 += 512; x2 += 512; x3 += 512; + v[ 0] = (x0+t3) >> 10; + v[56] = (x0-t3) >> 10; + v[ 8] = (x1+t2) >> 10; + v[48] = (x1-t2) >> 10; + v[16] = (x2+t1) >> 10; + v[40] = (x2-t1) >> 10; + v[24] = (x3+t0) >> 10; + v[32] = (x3-t0) >> 10; + } + } + + for (i=0, v=val, o=out; i < 8; ++i,v+=8,o+=out_stride) { + // no fast case since the first 1D IDCT spread components out + IDCT_1D(v[0],v[1],v[2],v[3],v[4],v[5],v[6],v[7]) + // constants scaled things up by 1<<12, plus we had 1<<2 from first + // loop, plus horizontal and vertical each scale by sqrt(8) so together + // we've got an extra 1<<3, so 1<<17 total we need to remove. + // so we want to round that, which means adding 0.5 * 1<<17, + // aka 65536. Also, we'll end up with -128 to 127 that we want + // to encode as 0..255 by adding 128, so we'll add that before the shift + x0 += 65536 + (128<<17); + x1 += 65536 + (128<<17); + x2 += 65536 + (128<<17); + x3 += 65536 + (128<<17); + // tried computing the shifts into temps, or'ing the temps to see + // if any were out of range, but that was slower + o[0] = clamp((x0+t3) >> 17); + o[7] = clamp((x0-t3) >> 17); + o[1] = clamp((x1+t2) >> 17); + o[6] = clamp((x1-t2) >> 17); + o[2] = clamp((x2+t1) >> 17); + o[5] = clamp((x2-t1) >> 17); + o[3] = clamp((x3+t0) >> 17); + o[4] = clamp((x3-t0) >> 17); + } +} + +#ifdef STBI_SIMD +static stbi_idct_8x8 stbi_idct_installed = idct_block; + +void stbi_install_idct(stbi_idct_8x8 func) +{ + stbi_idct_installed = func; +} +#endif + +#define MARKER_none 0xff +// if there's a pending marker from the entropy stream, return that +// otherwise, fetch from the stream and get a marker. if there's no +// marker, return 0xff, which is never a valid marker value +static uint8 get_marker(jpeg *j) +{ + uint8 x; + if (j->marker != MARKER_none) { x = j->marker; j->marker = MARKER_none; return x; } + x = get8u(j->s); + if (x != 0xff) return MARKER_none; + while (x == 0xff) + x = get8u(j->s); + return x; +} + +// in each scan, we'll have scan_n components, and the order +// of the components is specified by order[] +#define RESTART(x) ((x) >= 0xd0 && (x) <= 0xd7) + +// after a restart interval, reset the entropy decoder and +// the dc prediction +static void reset(jpeg *j) +{ + j->code_bits = 0; + j->code_buffer = 0; + j->nomore = 0; + j->img_comp[0].dc_pred = j->img_comp[1].dc_pred = j->img_comp[2].dc_pred = 0; + j->marker = MARKER_none; + j->todo = j->restart_interval ? j->restart_interval : 0x7fffffff; + // no more than 1<<31 MCUs if no restart_interal? that's plenty safe, + // since we don't even allow 1<<30 pixels +} + +static int parse_entropy_coded_data(jpeg *z) +{ + reset(z); + if (z->scan_n == 1) { + int i,j; + #ifdef STBI_SIMD + __declspec(align(16)) + #endif + short data[64]; + int n = z->order[0]; + // non-interleaved data, we just need to process one block at a time, + // in trivial scanline order + // number of blocks to do just depends on how many actual "pixels" this + // component has, independent of interleaved MCU blocking and such + int w = (z->img_comp[n].x+7) >> 3; + int h = (z->img_comp[n].y+7) >> 3; + for (j=0; j < h; ++j) { + for (i=0; i < w; ++i) { + if (!decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; + #ifdef STBI_SIMD + stbi_idct_installed(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); + #else + idct_block(z->img_comp[n].data+z->img_comp[n].w2*j*8+i*8, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); + #endif + // every data block is an MCU, so countdown the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(z->marker)) return 1; + reset(z); + } + } + } + } else { // interleaved! + int i,j,k,x,y; + short data[64]; + for (j=0; j < z->img_mcu_y; ++j) { + for (i=0; i < z->img_mcu_x; ++i) { + // scan an interleaved mcu... process scan_n components in order + for (k=0; k < z->scan_n; ++k) { + int n = z->order[k]; + // scan out an mcu's worth of this component; that's just determined + // by the basic H and V specified for the component + for (y=0; y < z->img_comp[n].v; ++y) { + for (x=0; x < z->img_comp[n].h; ++x) { + int x2 = (i*z->img_comp[n].h + x)*8; + int y2 = (j*z->img_comp[n].v + y)*8; + if (!decode_block(z, data, z->huff_dc+z->img_comp[n].hd, z->huff_ac+z->img_comp[n].ha, n)) return 0; + #ifdef STBI_SIMD + stbi_idct_installed(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant2[z->img_comp[n].tq]); + #else + idct_block(z->img_comp[n].data+z->img_comp[n].w2*y2+x2, z->img_comp[n].w2, data, z->dequant[z->img_comp[n].tq]); + #endif + } + } + } + // after all interleaved components, that's an interleaved MCU, + // so now count down the restart interval + if (--z->todo <= 0) { + if (z->code_bits < 24) grow_buffer_unsafe(z); + // if it's NOT a restart, then just bail, so we get corrupt data + // rather than no data + if (!RESTART(z->marker)) return 1; + reset(z); + } + } + } + } + return 1; +} + +static int process_marker(jpeg *z, int m) +{ + int L; + switch (m) { + case MARKER_none: // no marker found + return e("expected marker","Corrupt JPEG"); + + case 0xC2: // SOF - progressive + return e("progressive jpeg","JPEG format not supported (progressive)"); + + case 0xDD: // DRI - specify restart interval + if (get16(z->s) != 4) return e("bad DRI len","Corrupt JPEG"); + z->restart_interval = get16(z->s); + return 1; + + case 0xDB: // DQT - define quantization table + L = get16(z->s)-2; + while (L > 0) { + int q = get8(z->s); + int p = q >> 4; + int t = q & 15,i; + if (p != 0) return e("bad DQT type","Corrupt JPEG"); + if (t > 3) return e("bad DQT table","Corrupt JPEG"); + for (i=0; i < 64; ++i) + z->dequant[t][dezigzag[i]] = get8u(z->s); + #ifdef STBI_SIMD + for (i=0; i < 64; ++i) + z->dequant2[t][i] = z->dequant[t][i]; + #endif + L -= 65; + } + return L==0; + + case 0xC4: // DHT - define huffman table + L = get16(z->s)-2; + while (L > 0) { + uint8 *v; + int sizes[16],i,m=0; + int q = get8(z->s); + int tc = q >> 4; + int th = q & 15; + if (tc > 1 || th > 3) return e("bad DHT header","Corrupt JPEG"); + for (i=0; i < 16; ++i) { + sizes[i] = get8(z->s); + m += sizes[i]; + } + L -= 17; + if (tc == 0) { + if (!build_huffman(z->huff_dc+th, sizes)) return 0; + v = z->huff_dc[th].values; + } else { + if (!build_huffman(z->huff_ac+th, sizes)) return 0; + v = z->huff_ac[th].values; + } + for (i=0; i < m; ++i) + v[i] = get8u(z->s); + L -= m; + } + return L==0; + } + // check for comment block or APP blocks + if ((m >= 0xE0 && m <= 0xEF) || m == 0xFE) { + skip(z->s, get16(z->s)-2); + return 1; + } + return 0; +} + +// after we see SOS +static int process_scan_header(jpeg *z) +{ + int i; + int Ls = get16(z->s); + z->scan_n = get8(z->s); + if (z->scan_n < 1 || z->scan_n > 4 || z->scan_n > (int) z->s->img_n) return e("bad SOS component count","Corrupt JPEG"); + if (Ls != 6+2*z->scan_n) return e("bad SOS len","Corrupt JPEG"); + for (i=0; i < z->scan_n; ++i) { + int id = get8(z->s), which; + int q = get8(z->s); + for (which = 0; which < z->s->img_n; ++which) + if (z->img_comp[which].id == id) + break; + if (which == z->s->img_n) return 0; + z->img_comp[which].hd = q >> 4; if (z->img_comp[which].hd > 3) return e("bad DC huff","Corrupt JPEG"); + z->img_comp[which].ha = q & 15; if (z->img_comp[which].ha > 3) return e("bad AC huff","Corrupt JPEG"); + z->order[i] = which; + } + if (get8(z->s) != 0) return e("bad SOS","Corrupt JPEG"); + get8(z->s); // should be 63, but might be 0 + if (get8(z->s) != 0) return e("bad SOS","Corrupt JPEG"); + + return 1; +} + +static int process_frame_header(jpeg *z, int scan) +{ + stbi *s = z->s; + int Lf,p,i,q, h_max=1,v_max=1,c; + Lf = get16(s); if (Lf < 11) return e("bad SOF len","Corrupt JPEG"); // JPEG + p = get8(s); if (p != 8) return e("only 8-bit","JPEG format not supported: 8-bit only"); // JPEG baseline + s->img_y = get16(s); if (s->img_y == 0) return e("no header height", "JPEG format not supported: delayed height"); // Legal, but we don't handle it--but neither does IJG + s->img_x = get16(s); if (s->img_x == 0) return e("0 width","Corrupt JPEG"); // JPEG requires + c = get8(s); + if (c != 3 && c != 1) return e("bad component count","Corrupt JPEG"); // JFIF requires + s->img_n = c; + for (i=0; i < c; ++i) { + z->img_comp[i].data = NULL; + z->img_comp[i].linebuf = NULL; + } + + if (Lf != 8+3*s->img_n) return e("bad SOF len","Corrupt JPEG"); + + for (i=0; i < s->img_n; ++i) { + z->img_comp[i].id = get8(s); + if (z->img_comp[i].id != i+1) // JFIF requires + if (z->img_comp[i].id != i) // some version of jpegtran outputs non-JFIF-compliant files! + return e("bad component ID","Corrupt JPEG"); + q = get8(s); + z->img_comp[i].h = (q >> 4); if (!z->img_comp[i].h || z->img_comp[i].h > 4) return e("bad H","Corrupt JPEG"); + z->img_comp[i].v = q & 15; if (!z->img_comp[i].v || z->img_comp[i].v > 4) return e("bad V","Corrupt JPEG"); + z->img_comp[i].tq = get8(s); if (z->img_comp[i].tq > 3) return e("bad TQ","Corrupt JPEG"); + } + + if (scan != SCAN_load) return 1; + + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return e("too large", "Image too large to decode"); + + for (i=0; i < s->img_n; ++i) { + if (z->img_comp[i].h > h_max) h_max = z->img_comp[i].h; + if (z->img_comp[i].v > v_max) v_max = z->img_comp[i].v; + } + + // compute interleaved mcu info + z->img_h_max = h_max; + z->img_v_max = v_max; + z->img_mcu_w = h_max * 8; + z->img_mcu_h = v_max * 8; + z->img_mcu_x = (s->img_x + z->img_mcu_w-1) / z->img_mcu_w; + z->img_mcu_y = (s->img_y + z->img_mcu_h-1) / z->img_mcu_h; + + for (i=0; i < s->img_n; ++i) { + // number of effective pixels (e.g. for non-interleaved MCU) + z->img_comp[i].x = (s->img_x * z->img_comp[i].h + h_max-1) / h_max; + z->img_comp[i].y = (s->img_y * z->img_comp[i].v + v_max-1) / v_max; + // to simplify generation, we'll allocate enough memory to decode + // the bogus oversized data from using interleaved MCUs and their + // big blocks (e.g. a 16x16 iMCU on an image of width 33); we won't + // discard the extra data until colorspace conversion + z->img_comp[i].w2 = z->img_mcu_x * z->img_comp[i].h * 8; + z->img_comp[i].h2 = z->img_mcu_y * z->img_comp[i].v * 8; + z->img_comp[i].raw_data = malloc(z->img_comp[i].w2 * z->img_comp[i].h2+15); + if (z->img_comp[i].raw_data == NULL) { + for(--i; i >= 0; --i) { + free(z->img_comp[i].raw_data); + z->img_comp[i].data = NULL; + } + return e("outofmem", "Out of memory"); + } + // align blocks for installable-idct using mmx/sse + z->img_comp[i].data = (uint8*) (((size_t) z->img_comp[i].raw_data + 15) & ~15); + z->img_comp[i].linebuf = NULL; + } + + return 1; +} + +// use comparisons since in some cases we handle more than one case (e.g. SOF) +#define DNL(x) ((x) == 0xdc) +#define SOI(x) ((x) == 0xd8) +#define EOI(x) ((x) == 0xd9) +#define SOF(x) ((x) == 0xc0 || (x) == 0xc1) +#define SOS(x) ((x) == 0xda) + +static int decode_jpeg_header(jpeg *z, int scan) +{ + int m; + z->marker = MARKER_none; // initialize cached marker to empty + m = get_marker(z); + if (!SOI(m)) return e("no SOI","Corrupt JPEG"); + if (scan == SCAN_type) return 1; + m = get_marker(z); + while (!SOF(m)) { + if (!process_marker(z,m)) return 0; + m = get_marker(z); + while (m == MARKER_none) { + // some files have extra padding after their blocks, so ok, we'll scan + if (at_eof(z->s)) return e("no SOF", "Corrupt JPEG"); + m = get_marker(z); + } + } + if (!process_frame_header(z, scan)) return 0; + return 1; +} + +static int decode_jpeg_image(jpeg *j) +{ + int m; + j->restart_interval = 0; + if (!decode_jpeg_header(j, SCAN_load)) return 0; + m = get_marker(j); + while (!EOI(m)) { + if (SOS(m)) { + if (!process_scan_header(j)) return 0; + if (!parse_entropy_coded_data(j)) return 0; + if (j->marker == MARKER_none ) { + // handle 0s at the end of image data from IP Kamera 9060 + while (!at_eof(j->s)) { + int x = get8(j->s); + if (x == 255) { + j->marker = get8u(j->s); + break; + } else if (x != 0) { + return 0; + } + } + // if we reach eof without hitting a marker, get_marker() below will fail and we'll eventually return 0 + } + } else { + if (!process_marker(j, m)) return 0; + } + m = get_marker(j); + } + return 1; +} + +// static jfif-centered resampling (across block boundaries) + +typedef uint8 *(*resample_row_func)(uint8 *out, uint8 *in0, uint8 *in1, + int w, int hs); + +#define div4(x) ((uint8) ((x) >> 2)) + +static uint8 *resample_row_1(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + STBI_NOTUSED(out); + STBI_NOTUSED(in_far); + STBI_NOTUSED(w); + STBI_NOTUSED(hs); + return in_near; +} + +static uint8* resample_row_v_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate two samples vertically for every one in input + int i; + STBI_NOTUSED(hs); + for (i=0; i < w; ++i) + out[i] = div4(3*in_near[i] + in_far[i] + 2); + return out; +} + +static uint8* resample_row_h_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate two samples horizontally for every one in input + int i; + uint8 *input = in_near; + + if (w == 1) { + // if only one sample, can't do any interpolation + out[0] = out[1] = input[0]; + return out; + } + + out[0] = input[0]; + out[1] = div4(input[0]*3 + input[1] + 2); + for (i=1; i < w-1; ++i) { + int n = 3*input[i]+2; + out[i*2+0] = div4(n+input[i-1]); + out[i*2+1] = div4(n+input[i+1]); + } + out[i*2+0] = div4(input[w-2]*3 + input[w-1] + 2); + out[i*2+1] = input[w-1]; + + STBI_NOTUSED(in_far); + STBI_NOTUSED(hs); + + return out; +} + +#define div16(x) ((uint8) ((x) >> 4)) + +static uint8 *resample_row_hv_2(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // need to generate 2x2 samples for every one in input + int i,t0,t1; + if (w == 1) { + out[0] = out[1] = div4(3*in_near[0] + in_far[0] + 2); + return out; + } + + t1 = 3*in_near[0] + in_far[0]; + out[0] = div4(t1+2); + for (i=1; i < w; ++i) { + t0 = t1; + t1 = 3*in_near[i]+in_far[i]; + out[i*2-1] = div16(3*t0 + t1 + 8); + out[i*2 ] = div16(3*t1 + t0 + 8); + } + out[w*2-1] = div4(t1+2); + + STBI_NOTUSED(hs); + + return out; +} + +static uint8 *resample_row_generic(uint8 *out, uint8 *in_near, uint8 *in_far, int w, int hs) +{ + // resample with nearest-neighbor + int i,j; + STBI_NOTUSED(in_far); + for (i=0; i < w; ++i) + for (j=0; j < hs; ++j) + out[i*hs+j] = in_near[i]; + return out; +} + +#define float2fixed(x) ((int) ((x) * 65536 + 0.5)) + +// 0.38 seconds on 3*anemones.jpg (0.25 with processor = Pro) +// VC6 without processor=Pro is generating multiple LEAs per multiply! +static void YCbCr_to_RGB_row(uint8 *out, const uint8 *y, const uint8 *pcb, const uint8 *pcr, int count, int step) +{ + int i; + for (i=0; i < count; ++i) { + int y_fixed = (y[i] << 16) + 32768; // rounding + int r,g,b; + int cr = pcr[i] - 128; + int cb = pcb[i] - 128; + r = y_fixed + cr*float2fixed(1.40200f); + g = y_fixed - cr*float2fixed(0.71414f) - cb*float2fixed(0.34414f); + b = y_fixed + cb*float2fixed(1.77200f); + r >>= 16; + g >>= 16; + b >>= 16; + if ((unsigned) r > 255) { if (r < 0) r = 0; else r = 255; } + if ((unsigned) g > 255) { if (g < 0) g = 0; else g = 255; } + if ((unsigned) b > 255) { if (b < 0) b = 0; else b = 255; } + out[0] = (uint8)r; + out[1] = (uint8)g; + out[2] = (uint8)b; + out[3] = 255; + out += step; + } +} + +#ifdef STBI_SIMD +static stbi_YCbCr_to_RGB_run stbi_YCbCr_installed = YCbCr_to_RGB_row; + +void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func) +{ + stbi_YCbCr_installed = func; +} +#endif + + +// clean up the temporary component buffers +static void cleanup_jpeg(jpeg *j) +{ + int i; + for (i=0; i < j->s->img_n; ++i) { + if (j->img_comp[i].data) { + free(j->img_comp[i].raw_data); + j->img_comp[i].data = NULL; + } + if (j->img_comp[i].linebuf) { + free(j->img_comp[i].linebuf); + j->img_comp[i].linebuf = NULL; + } + } +} + +typedef struct +{ + resample_row_func resample; + uint8 *line0,*line1; + int hs,vs; // expansion factor in each axis + int w_lores; // horizontal pixels pre-expansion + int ystep; // how far through vertical expansion we are + int ypos; // which pre-expansion row we're on +} stbi_resample; + +static uint8 *load_jpeg_image(jpeg *z, int *out_x, int *out_y, int *comp, int req_comp) +{ + int n, decode_n; + // validate req_comp + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + z->s->img_n = 0; + + // load a jpeg image from whichever source + if (!decode_jpeg_image(z)) { cleanup_jpeg(z); return NULL; } + + // determine actual number of components to generate + n = req_comp ? req_comp : z->s->img_n; + + if (z->s->img_n == 3 && n < 3) + decode_n = 1; + else + decode_n = z->s->img_n; + + // resample and color-convert + { + int k; + uint i,j; + uint8 *output; + uint8 *coutput[4]; + + stbi_resample res_comp[4]; + + for (k=0; k < decode_n; ++k) { + stbi_resample *r = &res_comp[k]; + + // allocate line buffer big enough for upsampling off the edges + // with upsample factor of 4 + z->img_comp[k].linebuf = (uint8 *) malloc(z->s->img_x + 3); + if (!z->img_comp[k].linebuf) { cleanup_jpeg(z); return epuc("outofmem", "Out of memory"); } + + r->hs = z->img_h_max / z->img_comp[k].h; + r->vs = z->img_v_max / z->img_comp[k].v; + r->ystep = r->vs >> 1; + r->w_lores = (z->s->img_x + r->hs-1) / r->hs; + r->ypos = 0; + r->line0 = r->line1 = z->img_comp[k].data; + + if (r->hs == 1 && r->vs == 1) r->resample = resample_row_1; + else if (r->hs == 1 && r->vs == 2) r->resample = resample_row_v_2; + else if (r->hs == 2 && r->vs == 1) r->resample = resample_row_h_2; + else if (r->hs == 2 && r->vs == 2) r->resample = resample_row_hv_2; + else r->resample = resample_row_generic; + } + + // can't error after this so, this is safe + output = (uint8 *) malloc(n * z->s->img_x * z->s->img_y + 1); + if (!output) { cleanup_jpeg(z); return epuc("outofmem", "Out of memory"); } + + // now go ahead and resample + for (j=0; j < z->s->img_y; ++j) { + uint8 *out = output + n * z->s->img_x * j; + for (k=0; k < decode_n; ++k) { + stbi_resample *r = &res_comp[k]; + int y_bot = r->ystep >= (r->vs >> 1); + coutput[k] = r->resample(z->img_comp[k].linebuf, + y_bot ? r->line1 : r->line0, + y_bot ? r->line0 : r->line1, + r->w_lores, r->hs); + if (++r->ystep >= r->vs) { + r->ystep = 0; + r->line0 = r->line1; + if (++r->ypos < z->img_comp[k].y) + r->line1 += z->img_comp[k].w2; + } + } + if (n >= 3) { + uint8 *y = coutput[0]; + if (z->s->img_n == 3) { + #ifdef STBI_SIMD + stbi_YCbCr_installed(out, y, coutput[1], coutput[2], z->s.img_x, n); + #else + YCbCr_to_RGB_row(out, y, coutput[1], coutput[2], z->s->img_x, n); + #endif + } else + for (i=0; i < z->s->img_x; ++i) { + out[0] = out[1] = out[2] = y[i]; + out[3] = 255; // not used if n==3 + out += n; + } + } else { + uint8 *y = coutput[0]; + if (n == 1) + for (i=0; i < z->s->img_x; ++i) out[i] = y[i]; + else + for (i=0; i < z->s->img_x; ++i) *out++ = y[i], *out++ = 255; + } + } + cleanup_jpeg(z); + *out_x = z->s->img_x; + *out_y = z->s->img_y; + if (comp) *comp = z->s->img_n; // report original components, not output + return output; + } +} + +static unsigned char *stbi_jpeg_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + jpeg j; + j.s = s; + return load_jpeg_image(&j, x,y,comp,req_comp); +} + +static int stbi_jpeg_test(stbi *s) +{ + int r; + jpeg j; + j.s = s; + r = decode_jpeg_header(&j, SCAN_type); + stbi_rewind(s); + return r; +} + +static int stbi_jpeg_info_raw(jpeg *j, int *x, int *y, int *comp) +{ + if (!decode_jpeg_header(j, SCAN_header)) { + stbi_rewind( j->s ); + return 0; + } + if (x) *x = j->s->img_x; + if (y) *y = j->s->img_y; + if (comp) *comp = j->s->img_n; + return 1; +} + +static int stbi_jpeg_info(stbi *s, int *x, int *y, int *comp) +{ + jpeg j; + j.s = s; + return stbi_jpeg_info_raw(&j, x, y, comp); +} + +// public domain zlib decode v0.2 Sean Barrett 2006-11-18 +// simple implementation +// - all input must be provided in an upfront buffer +// - all output is written to a single output buffer (can malloc/realloc) +// performance +// - fast huffman + +// fast-way is faster to check than jpeg huffman, but slow way is slower +#define ZFAST_BITS 9 // accelerate all cases in default tables +#define ZFAST_MASK ((1 << ZFAST_BITS) - 1) + +// zlib-style huffman encoding +// (jpegs packs from left, zlib from right, so can't share code) +typedef struct +{ + uint16 fast[1 << ZFAST_BITS]; + uint16 firstcode[16]; + int maxcode[17]; + uint16 firstsymbol[16]; + uint8 size[288]; + uint16 value[288]; +} zhuffman; + +stbi_inline static int bitreverse16(int n) +{ + n = ((n & 0xAAAA) >> 1) | ((n & 0x5555) << 1); + n = ((n & 0xCCCC) >> 2) | ((n & 0x3333) << 2); + n = ((n & 0xF0F0) >> 4) | ((n & 0x0F0F) << 4); + n = ((n & 0xFF00) >> 8) | ((n & 0x00FF) << 8); + return n; +} + +stbi_inline static int bit_reverse(int v, int bits) +{ + assert(bits <= 16); + // to bit reverse n bits, reverse 16 and shift + // e.g. 11 bits, bit reverse and shift away 5 + return bitreverse16(v) >> (16-bits); +} + +static int zbuild_huffman(zhuffman *z, uint8 *sizelist, int num) +{ + int i,k=0; + int code, next_code[16], sizes[17]; + + // DEFLATE spec for generating codes + memset(sizes, 0, sizeof(sizes)); + memset(z->fast, 255, sizeof(z->fast)); + for (i=0; i < num; ++i) + ++sizes[sizelist[i]]; + sizes[0] = 0; + for (i=1; i < 16; ++i) + assert(sizes[i] <= (1 << i)); + code = 0; + for (i=1; i < 16; ++i) { + next_code[i] = code; + z->firstcode[i] = (uint16) code; + z->firstsymbol[i] = (uint16) k; + code = (code + sizes[i]); + if (sizes[i]) + if (code-1 >= (1 << i)) return e("bad codelengths","Corrupt JPEG"); + z->maxcode[i] = code << (16-i); // preshift for inner loop + code <<= 1; + k += sizes[i]; + } + z->maxcode[16] = 0x10000; // sentinel + for (i=0; i < num; ++i) { + int s = sizelist[i]; + if (s) { + int c = next_code[s] - z->firstcode[s] + z->firstsymbol[s]; + z->size[c] = (uint8)s; + z->value[c] = (uint16)i; + if (s <= ZFAST_BITS) { + int k = bit_reverse(next_code[s],s); + while (k < (1 << ZFAST_BITS)) { + z->fast[k] = (uint16) c; + k += (1 << s); + } + } + ++next_code[s]; + } + } + return 1; +} + +// zlib-from-memory implementation for PNG reading +// because PNG allows splitting the zlib stream arbitrarily, +// and it's annoying structurally to have PNG call ZLIB call PNG, +// we require PNG read all the IDATs and combine them into a single +// memory buffer + +typedef struct +{ + uint8 *zbuffer, *zbuffer_end; + int num_bits; + uint32 code_buffer; + + char *zout; + char *zout_start; + char *zout_end; + int z_expandable; + + zhuffman z_length, z_distance; +} zbuf; + +stbi_inline static int zget8(zbuf *z) +{ + if (z->zbuffer >= z->zbuffer_end) return 0; + return *z->zbuffer++; +} + +static void fill_bits(zbuf *z) +{ + do { + assert(z->code_buffer < (1U << z->num_bits)); + z->code_buffer |= zget8(z) << z->num_bits; + z->num_bits += 8; + } while (z->num_bits <= 24); +} + +stbi_inline static unsigned int zreceive(zbuf *z, int n) +{ + unsigned int k; + if (z->num_bits < n) fill_bits(z); + k = z->code_buffer & ((1 << n) - 1); + z->code_buffer >>= n; + z->num_bits -= n; + return k; +} + +stbi_inline static int zhuffman_decode(zbuf *a, zhuffman *z) +{ + int b,s,k; + if (a->num_bits < 16) fill_bits(a); + b = z->fast[a->code_buffer & ZFAST_MASK]; + if (b < 0xffff) { + s = z->size[b]; + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; + } + + // not resolved by fast table, so compute it the slow way + // use jpeg approach, which requires MSbits at top + k = bit_reverse(a->code_buffer, 16); + for (s=ZFAST_BITS+1; ; ++s) + if (k < z->maxcode[s]) + break; + if (s == 16) return -1; // invalid code! + // code size is s, so: + b = (k >> (16-s)) - z->firstcode[s] + z->firstsymbol[s]; + assert(z->size[b] == s); + a->code_buffer >>= s; + a->num_bits -= s; + return z->value[b]; +} + +static int expand(zbuf *z, int n) // need to make room for n bytes +{ + char *q; + int cur, limit; + if (!z->z_expandable) return e("output buffer limit","Corrupt PNG"); + cur = (int) (z->zout - z->zout_start); + limit = (int) (z->zout_end - z->zout_start); + while (cur + n > limit) + limit *= 2; + q = (char *) realloc(z->zout_start, limit); + if (q == NULL) return e("outofmem", "Out of memory"); + z->zout_start = q; + z->zout = q + cur; + z->zout_end = q + limit; + return 1; +} + +static int length_base[31] = { + 3,4,5,6,7,8,9,10,11,13, + 15,17,19,23,27,31,35,43,51,59, + 67,83,99,115,131,163,195,227,258,0,0 }; + +static int length_extra[31]= +{ 0,0,0,0,0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3,4,4,4,4,5,5,5,5,0,0,0 }; + +static int dist_base[32] = { 1,2,3,4,5,7,9,13,17,25,33,49,65,97,129,193, +257,385,513,769,1025,1537,2049,3073,4097,6145,8193,12289,16385,24577,0,0}; + +static int dist_extra[32] = +{ 0,0,0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,13,13}; + +static int parse_huffman_block(zbuf *a) +{ + for(;;) { + int z = zhuffman_decode(a, &a->z_length); + if (z < 256) { + if (z < 0) return e("bad huffman code","Corrupt PNG"); // error in huffman codes + if (a->zout >= a->zout_end) if (!expand(a, 1)) return 0; + *a->zout++ = (char) z; + } else { + uint8 *p; + int len,dist; + if (z == 256) return 1; + z -= 257; + len = length_base[z]; + if (length_extra[z]) len += zreceive(a, length_extra[z]); + z = zhuffman_decode(a, &a->z_distance); + if (z < 0) return e("bad huffman code","Corrupt PNG"); + dist = dist_base[z]; + if (dist_extra[z]) dist += zreceive(a, dist_extra[z]); + if (a->zout - a->zout_start < dist) return e("bad dist","Corrupt PNG"); + if (a->zout + len > a->zout_end) if (!expand(a, len)) return 0; + p = (uint8 *) (a->zout - dist); + while (len--) + *a->zout++ = *p++; + } + } +} + +static int compute_huffman_codes(zbuf *a) +{ + static uint8 length_dezigzag[19] = { 16,17,18,0,8,7,9,6,10,5,11,4,12,3,13,2,14,1,15 }; + zhuffman z_codelength; + uint8 lencodes[286+32+137];//padding for maximum single op + uint8 codelength_sizes[19]; + int i,n; + + int hlit = zreceive(a,5) + 257; + int hdist = zreceive(a,5) + 1; + int hclen = zreceive(a,4) + 4; + + memset(codelength_sizes, 0, sizeof(codelength_sizes)); + for (i=0; i < hclen; ++i) { + int s = zreceive(a,3); + codelength_sizes[length_dezigzag[i]] = (uint8) s; + } + if (!zbuild_huffman(&z_codelength, codelength_sizes, 19)) return 0; + + n = 0; + while (n < hlit + hdist) { + int c = zhuffman_decode(a, &z_codelength); + assert(c >= 0 && c < 19); + if (c < 16) + lencodes[n++] = (uint8) c; + else if (c == 16) { + c = zreceive(a,2)+3; + memset(lencodes+n, lencodes[n-1], c); + n += c; + } else if (c == 17) { + c = zreceive(a,3)+3; + memset(lencodes+n, 0, c); + n += c; + } else { + assert(c == 18); + c = zreceive(a,7)+11; + memset(lencodes+n, 0, c); + n += c; + } + } + if (n != hlit+hdist) return e("bad codelengths","Corrupt PNG"); + if (!zbuild_huffman(&a->z_length, lencodes, hlit)) return 0; + if (!zbuild_huffman(&a->z_distance, lencodes+hlit, hdist)) return 0; + return 1; +} + +static int parse_uncompressed_block(zbuf *a) +{ + uint8 header[4]; + int len,nlen,k; + if (a->num_bits & 7) + zreceive(a, a->num_bits & 7); // discard + // drain the bit-packed data into header + k = 0; + while (a->num_bits > 0) { + header[k++] = (uint8) (a->code_buffer & 255); // wtf this warns? + a->code_buffer >>= 8; + a->num_bits -= 8; + } + assert(a->num_bits == 0); + // now fill header the normal way + while (k < 4) + header[k++] = (uint8) zget8(a); + len = header[1] * 256 + header[0]; + nlen = header[3] * 256 + header[2]; + if (nlen != (len ^ 0xffff)) return e("zlib corrupt","Corrupt PNG"); + if (a->zbuffer + len > a->zbuffer_end) return e("read past buffer","Corrupt PNG"); + if (a->zout + len > a->zout_end) + if (!expand(a, len)) return 0; + memcpy(a->zout, a->zbuffer, len); + a->zbuffer += len; + a->zout += len; + return 1; +} + +static int parse_zlib_header(zbuf *a) +{ + int cmf = zget8(a); + int cm = cmf & 15; + /* int cinfo = cmf >> 4; */ + int flg = zget8(a); + if ((cmf*256+flg) % 31 != 0) return e("bad zlib header","Corrupt PNG"); // zlib spec + if (flg & 32) return e("no preset dict","Corrupt PNG"); // preset dictionary not allowed in png + if (cm != 8) return e("bad compression","Corrupt PNG"); // DEFLATE required for png + // window = 1 << (8 + cinfo)... but who cares, we fully buffer output + return 1; +} + +// @TODO: should statically initialize these for optimal thread safety +static uint8 default_length[288], default_distance[32]; +static void init_defaults(void) +{ + int i; // use <= to match clearly with spec + for (i=0; i <= 143; ++i) default_length[i] = 8; + for ( ; i <= 255; ++i) default_length[i] = 9; + for ( ; i <= 279; ++i) default_length[i] = 7; + for ( ; i <= 287; ++i) default_length[i] = 8; + + for (i=0; i <= 31; ++i) default_distance[i] = 5; +} + +int stbi_png_partial; // a quick hack to only allow decoding some of a PNG... I should implement real streaming support instead +static int parse_zlib(zbuf *a, int parse_header) +{ + int final, type; + if (parse_header) + if (!parse_zlib_header(a)) return 0; + a->num_bits = 0; + a->code_buffer = 0; + do { + final = zreceive(a,1); + type = zreceive(a,2); + if (type == 0) { + if (!parse_uncompressed_block(a)) return 0; + } else if (type == 3) { + return 0; + } else { + if (type == 1) { + // use fixed code lengths + if (!default_distance[31]) init_defaults(); + if (!zbuild_huffman(&a->z_length , default_length , 288)) return 0; + if (!zbuild_huffman(&a->z_distance, default_distance, 32)) return 0; + } else { + if (!compute_huffman_codes(a)) return 0; + } + if (!parse_huffman_block(a)) return 0; + } + if (stbi_png_partial && a->zout - a->zout_start > 65536) + break; + } while (!final); + return 1; +} + +static int do_zlib(zbuf *a, char *obuf, int olen, int exp, int parse_header) +{ + a->zout_start = obuf; + a->zout = obuf; + a->zout_end = obuf + olen; + a->z_expandable = exp; + + return parse_zlib(a, parse_header); +} + +char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen) +{ + zbuf a; + char *p = (char *) malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer + len; + if (do_zlib(&a, p, initial_size, 1, 1)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +char *stbi_zlib_decode_malloc(char const *buffer, int len, int *outlen) +{ + return stbi_zlib_decode_malloc_guesssize(buffer, len, 16384, outlen); +} + +char *stbi_zlib_decode_malloc_guesssize_headerflag(const char *buffer, int len, int initial_size, int *outlen, int parse_header) +{ + zbuf a; + char *p = (char *) malloc(initial_size); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer + len; + if (do_zlib(&a, p, initial_size, 1, parse_header)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +int stbi_zlib_decode_buffer(char *obuffer, int olen, char const *ibuffer, int ilen) +{ + zbuf a; + a.zbuffer = (uint8 *) ibuffer; + a.zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(&a, obuffer, olen, 0, 1)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +char *stbi_zlib_decode_noheader_malloc(char const *buffer, int len, int *outlen) +{ + zbuf a; + char *p = (char *) malloc(16384); + if (p == NULL) return NULL; + a.zbuffer = (uint8 *) buffer; + a.zbuffer_end = (uint8 *) buffer+len; + if (do_zlib(&a, p, 16384, 1, 0)) { + if (outlen) *outlen = (int) (a.zout - a.zout_start); + return a.zout_start; + } else { + free(a.zout_start); + return NULL; + } +} + +int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen) +{ + zbuf a; + a.zbuffer = (uint8 *) ibuffer; + a.zbuffer_end = (uint8 *) ibuffer + ilen; + if (do_zlib(&a, obuffer, olen, 0, 0)) + return (int) (a.zout - a.zout_start); + else + return -1; +} + +// public domain "baseline" PNG decoder v0.10 Sean Barrett 2006-11-18 +// simple implementation +// - only 8-bit samples +// - no CRC checking +// - allocates lots of intermediate memory +// - avoids problem of streaming data between subsystems +// - avoids explicit window management +// performance +// - uses stb_zlib, a PD zlib implementation with fast huffman decoding + + +typedef struct +{ + uint32 length; + uint32 type; +} chunk; + +#define PNG_TYPE(a,b,c,d) (((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) + +static chunk get_chunk_header(stbi *s) +{ + chunk c; + c.length = get32(s); + c.type = get32(s); + return c; +} + +static int check_png_header(stbi *s) +{ + static uint8 png_sig[8] = { 137,80,78,71,13,10,26,10 }; + int i; + for (i=0; i < 8; ++i) + if (get8u(s) != png_sig[i]) return e("bad png sig","Not a PNG"); + return 1; +} + +typedef struct +{ + stbi *s; + uint8 *idata, *expanded, *out; +} png; + + +enum { + F_none=0, F_sub=1, F_up=2, F_avg=3, F_paeth=4, + F_avg_first, F_paeth_first +}; + +static uint8 first_row_filter[5] = +{ + F_none, F_sub, F_none, F_avg_first, F_paeth_first +}; + +static int paeth(int a, int b, int c) +{ + int p = a + b - c; + int pa = abs(p-a); + int pb = abs(p-b); + int pc = abs(p-c); + if (pa <= pb && pa <= pc) return a; + if (pb <= pc) return b; + return c; +} + +// create the png data from post-deflated data +static int create_png_image_raw(png *a, uint8 *raw, uint32 raw_len, int out_n, uint32 x, uint32 y) +{ + stbi *s = a->s; + uint32 i,j,stride = x*out_n; + int k; + int img_n = s->img_n; // copy it into a local for later + assert(out_n == s->img_n || out_n == s->img_n+1); + if (stbi_png_partial) y = 1; + a->out = (uint8 *) malloc(x * y * out_n); + if (!a->out) return e("outofmem", "Out of memory"); + if (!stbi_png_partial) { + if (s->img_x == x && s->img_y == y) { + if (raw_len != (img_n * x + 1) * y) return e("not enough pixels","Corrupt PNG"); + } else { // interlaced: + if (raw_len < (img_n * x + 1) * y) return e("not enough pixels","Corrupt PNG"); + } + } + for (j=0; j < y; ++j) { + uint8 *cur = a->out + stride*j; + uint8 *prior = cur - stride; + int filter = *raw++; + if (filter > 4) return e("invalid filter","Corrupt PNG"); + // if first row, use special filter that doesn't sample previous row + if (j == 0) filter = first_row_filter[filter]; + // handle first pixel explicitly + for (k=0; k < img_n; ++k) { + switch (filter) { + case F_none : cur[k] = raw[k]; break; + case F_sub : cur[k] = raw[k]; break; + case F_up : cur[k] = raw[k] + prior[k]; break; + case F_avg : cur[k] = raw[k] + (prior[k]>>1); break; + case F_paeth : cur[k] = (uint8) (raw[k] + paeth(0,prior[k],0)); break; + case F_avg_first : cur[k] = raw[k]; break; + case F_paeth_first: cur[k] = raw[k]; break; + } + } + if (img_n != out_n) cur[img_n] = 255; + raw += img_n; + cur += out_n; + prior += out_n; + // this is a little gross, so that we don't switch per-pixel or per-component + if (img_n == out_n) { + #define CASE(f) \ + case f: \ + for (i=x-1; i >= 1; --i, raw+=img_n,cur+=img_n,prior+=img_n) \ + for (k=0; k < img_n; ++k) + switch (filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-img_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-img_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],prior[k],prior[k-img_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-img_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-img_n],0,0)); break; + } + #undef CASE + } else { + assert(img_n+1 == out_n); + #define CASE(f) \ + case f: \ + for (i=x-1; i >= 1; --i, cur[img_n]=255,raw+=img_n,cur+=out_n,prior+=out_n) \ + for (k=0; k < img_n; ++k) + switch (filter) { + CASE(F_none) cur[k] = raw[k]; break; + CASE(F_sub) cur[k] = raw[k] + cur[k-out_n]; break; + CASE(F_up) cur[k] = raw[k] + prior[k]; break; + CASE(F_avg) cur[k] = raw[k] + ((prior[k] + cur[k-out_n])>>1); break; + CASE(F_paeth) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],prior[k],prior[k-out_n])); break; + CASE(F_avg_first) cur[k] = raw[k] + (cur[k-out_n] >> 1); break; + CASE(F_paeth_first) cur[k] = (uint8) (raw[k] + paeth(cur[k-out_n],0,0)); break; + } + #undef CASE + } + } + return 1; +} + +static int create_png_image(png *a, uint8 *raw, uint32 raw_len, int out_n, int interlaced) +{ + uint8 *final; + int p; + int save; + if (!interlaced) + return create_png_image_raw(a, raw, raw_len, out_n, a->s->img_x, a->s->img_y); + save = stbi_png_partial; + stbi_png_partial = 0; + + // de-interlacing + final = (uint8 *) malloc(a->s->img_x * a->s->img_y * out_n); + for (p=0; p < 7; ++p) { + int xorig[] = { 0,4,0,2,0,1,0 }; + int yorig[] = { 0,0,4,0,2,0,1 }; + int xspc[] = { 8,8,4,4,2,2,1 }; + int yspc[] = { 8,8,8,4,4,2,2 }; + int i,j,x,y; + // pass1_x[4] = 0, pass1_x[5] = 1, pass1_x[12] = 1 + x = (a->s->img_x - xorig[p] + xspc[p]-1) / xspc[p]; + y = (a->s->img_y - yorig[p] + yspc[p]-1) / yspc[p]; + if (x && y) { + if (!create_png_image_raw(a, raw, raw_len, out_n, x, y)) { + free(final); + return 0; + } + for (j=0; j < y; ++j) + for (i=0; i < x; ++i) + memcpy(final + (j*yspc[p]+yorig[p])*a->s->img_x*out_n + (i*xspc[p]+xorig[p])*out_n, + a->out + (j*x+i)*out_n, out_n); + free(a->out); + raw += (x*out_n+1)*y; + raw_len -= (x*out_n+1)*y; + } + } + a->out = final; + + stbi_png_partial = save; + return 1; +} + +static int compute_transparency(png *z, uint8 tc[3], int out_n) +{ + stbi *s = z->s; + uint32 i, pixel_count = s->img_x * s->img_y; + uint8 *p = z->out; + + // compute color-based transparency, assuming we've + // already got 255 as the alpha value in the output + assert(out_n == 2 || out_n == 4); + + if (out_n == 2) { + for (i=0; i < pixel_count; ++i) { + p[1] = (p[0] == tc[0] ? 0 : 255); + p += 2; + } + } else { + for (i=0; i < pixel_count; ++i) { + if (p[0] == tc[0] && p[1] == tc[1] && p[2] == tc[2]) + p[3] = 0; + p += 4; + } + } + return 1; +} + +static int expand_palette(png *a, uint8 *palette, int len, int pal_img_n) +{ + uint32 i, pixel_count = a->s->img_x * a->s->img_y; + uint8 *p, *temp_out, *orig = a->out; + + p = (uint8 *) malloc(pixel_count * pal_img_n); + if (p == NULL) return e("outofmem", "Out of memory"); + + // between here and free(out) below, exitting would leak + temp_out = p; + + if (pal_img_n == 3) { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p += 3; + } + } else { + for (i=0; i < pixel_count; ++i) { + int n = orig[i]*4; + p[0] = palette[n ]; + p[1] = palette[n+1]; + p[2] = palette[n+2]; + p[3] = palette[n+3]; + p += 4; + } + } + free(a->out); + a->out = temp_out; + + STBI_NOTUSED(len); + + return 1; +} + +static int stbi_unpremultiply_on_load = 0; +static int stbi_de_iphone_flag = 0; + +void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply) +{ + stbi_unpremultiply_on_load = flag_true_if_should_unpremultiply; +} +void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert) +{ + stbi_de_iphone_flag = flag_true_if_should_convert; +} + +static void stbi_de_iphone(png *z) +{ + stbi *s = z->s; + uint32 i, pixel_count = s->img_x * s->img_y; + uint8 *p = z->out; + + if (s->img_out_n == 3) { // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + uint8 t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 3; + } + } else { + assert(s->img_out_n == 4); + if (stbi_unpremultiply_on_load) { + // convert bgr to rgb and unpremultiply + for (i=0; i < pixel_count; ++i) { + uint8 a = p[3]; + uint8 t = p[0]; + if (a) { + p[0] = p[2] * 255 / a; + p[1] = p[1] * 255 / a; + p[2] = t * 255 / a; + } else { + p[0] = p[2]; + p[2] = t; + } + p += 4; + } + } else { + // convert bgr to rgb + for (i=0; i < pixel_count; ++i) { + uint8 t = p[0]; + p[0] = p[2]; + p[2] = t; + p += 4; + } + } + } +} + +static int parse_png_file(png *z, int scan, int req_comp) +{ + uint8 palette[1024], pal_img_n=0; + uint8 has_trans=0, tc[3]; + uint32 ioff=0, idata_limit=0, i, pal_len=0; + int first=1,k,interlace=0, iphone=0; + stbi *s = z->s; + + z->expanded = NULL; + z->idata = NULL; + z->out = NULL; + + if (!check_png_header(s)) return 0; + + if (scan == SCAN_type) return 1; + + for (;;) { + chunk c = get_chunk_header(s); + switch (c.type) { + case PNG_TYPE('C','g','B','I'): + iphone = stbi_de_iphone_flag; + skip(s, c.length); + break; + case PNG_TYPE('I','H','D','R'): { + int depth,color,comp,filter; + if (!first) return e("multiple IHDR","Corrupt PNG"); + first = 0; + if (c.length != 13) return e("bad IHDR len","Corrupt PNG"); + s->img_x = get32(s); if (s->img_x > (1 << 24)) return e("too large","Very large image (corrupt?)"); + s->img_y = get32(s); if (s->img_y > (1 << 24)) return e("too large","Very large image (corrupt?)"); + depth = get8(s); if (depth != 8) return e("8bit only","PNG not supported: 8-bit only"); + color = get8(s); if (color > 6) return e("bad ctype","Corrupt PNG"); + if (color == 3) pal_img_n = 3; else if (color & 1) return e("bad ctype","Corrupt PNG"); + comp = get8(s); if (comp) return e("bad comp method","Corrupt PNG"); + filter= get8(s); if (filter) return e("bad filter method","Corrupt PNG"); + interlace = get8(s); if (interlace>1) return e("bad interlace method","Corrupt PNG"); + if (!s->img_x || !s->img_y) return e("0-pixel image","Corrupt PNG"); + if (!pal_img_n) { + s->img_n = (color & 2 ? 3 : 1) + (color & 4 ? 1 : 0); + if ((1 << 30) / s->img_x / s->img_n < s->img_y) return e("too large", "Image too large to decode"); + if (scan == SCAN_header) return 1; + } else { + // if paletted, then pal_n is our final components, and + // img_n is # components to decompress/filter. + s->img_n = 1; + if ((1 << 30) / s->img_x / 4 < s->img_y) return e("too large","Corrupt PNG"); + // if SCAN_header, have to scan to see if we have a tRNS + } + break; + } + + case PNG_TYPE('P','L','T','E'): { + if (first) return e("first not IHDR", "Corrupt PNG"); + if (c.length > 256*3) return e("invalid PLTE","Corrupt PNG"); + pal_len = c.length / 3; + if (pal_len * 3 != c.length) return e("invalid PLTE","Corrupt PNG"); + for (i=0; i < pal_len; ++i) { + palette[i*4+0] = get8u(s); + palette[i*4+1] = get8u(s); + palette[i*4+2] = get8u(s); + palette[i*4+3] = 255; + } + break; + } + + case PNG_TYPE('t','R','N','S'): { + if (first) return e("first not IHDR", "Corrupt PNG"); + if (z->idata) return e("tRNS after IDAT","Corrupt PNG"); + if (pal_img_n) { + if (scan == SCAN_header) { s->img_n = 4; return 1; } + if (pal_len == 0) return e("tRNS before PLTE","Corrupt PNG"); + if (c.length > pal_len) return e("bad tRNS len","Corrupt PNG"); + pal_img_n = 4; + for (i=0; i < c.length; ++i) + palette[i*4+3] = get8u(s); + } else { + if (!(s->img_n & 1)) return e("tRNS with alpha","Corrupt PNG"); + if (c.length != (uint32) s->img_n*2) return e("bad tRNS len","Corrupt PNG"); + has_trans = 1; + for (k=0; k < s->img_n; ++k) + tc[k] = (uint8) get16(s); // non 8-bit images will be larger + } + break; + } + + case PNG_TYPE('I','D','A','T'): { + if (first) return e("first not IHDR", "Corrupt PNG"); + if (pal_img_n && !pal_len) return e("no PLTE","Corrupt PNG"); + if (scan == SCAN_header) { s->img_n = pal_img_n; return 1; } + if (ioff + c.length > idata_limit) { + uint8 *p; + if (idata_limit == 0) idata_limit = c.length > 4096 ? c.length : 4096; + while (ioff + c.length > idata_limit) + idata_limit *= 2; + p = (uint8 *) realloc(z->idata, idata_limit); if (p == NULL) return e("outofmem", "Out of memory"); + z->idata = p; + } + if (!getn(s, z->idata+ioff,c.length)) return e("outofdata","Corrupt PNG"); + ioff += c.length; + break; + } + + case PNG_TYPE('I','E','N','D'): { + uint32 raw_len; + if (first) return e("first not IHDR", "Corrupt PNG"); + if (scan != SCAN_load) return 1; + if (z->idata == NULL) return e("no IDAT","Corrupt PNG"); + z->expanded = (uint8 *) stbi_zlib_decode_malloc_guesssize_headerflag((char *) z->idata, ioff, 16384, (int *) &raw_len, !iphone); + if (z->expanded == NULL) return 0; // zlib should set error + free(z->idata); z->idata = NULL; + if ((req_comp == s->img_n+1 && req_comp != 3 && !pal_img_n) || has_trans) + s->img_out_n = s->img_n+1; + else + s->img_out_n = s->img_n; + if (!create_png_image(z, z->expanded, raw_len, s->img_out_n, interlace)) return 0; + if (has_trans) + if (!compute_transparency(z, tc, s->img_out_n)) return 0; + if (iphone && s->img_out_n > 2) + stbi_de_iphone(z); + if (pal_img_n) { + // pal_img_n == 3 or 4 + s->img_n = pal_img_n; // record the actual colors we had + s->img_out_n = pal_img_n; + if (req_comp >= 3) s->img_out_n = req_comp; + if (!expand_palette(z, palette, pal_len, s->img_out_n)) + return 0; + } + free(z->expanded); z->expanded = NULL; + return 1; + } + + default: + // if critical, fail + if (first) return e("first not IHDR", "Corrupt PNG"); + if ((c.type & (1 << 29)) == 0) { + #ifndef STBI_NO_FAILURE_STRINGS + // not threadsafe + static char invalid_chunk[] = "XXXX chunk not known"; + invalid_chunk[0] = (uint8) (c.type >> 24); + invalid_chunk[1] = (uint8) (c.type >> 16); + invalid_chunk[2] = (uint8) (c.type >> 8); + invalid_chunk[3] = (uint8) (c.type >> 0); + #endif + return e(invalid_chunk, "PNG not supported: unknown chunk type"); + } + skip(s, c.length); + break; + } + // end of chunk, read and skip CRC + get32(s); + } +} + +static unsigned char *do_png(png *p, int *x, int *y, int *n, int req_comp) +{ + unsigned char *result=NULL; + if (req_comp < 0 || req_comp > 4) return epuc("bad req_comp", "Internal error"); + if (parse_png_file(p, SCAN_load, req_comp)) { + result = p->out; + p->out = NULL; + if (req_comp && req_comp != p->s->img_out_n) { + result = convert_format(result, p->s->img_out_n, req_comp, p->s->img_x, p->s->img_y); + p->s->img_out_n = req_comp; + if (result == NULL) return result; + } + *x = p->s->img_x; + *y = p->s->img_y; + if (n) *n = p->s->img_n; + } + free(p->out); p->out = NULL; + free(p->expanded); p->expanded = NULL; + free(p->idata); p->idata = NULL; + + return result; +} + +static unsigned char *stbi_png_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + png p; + p.s = s; + return do_png(&p, x,y,comp,req_comp); +} + +static int stbi_png_test(stbi *s) +{ + int r; + r = check_png_header(s); + stbi_rewind(s); + return r; +} + +static int stbi_png_info_raw(png *p, int *x, int *y, int *comp) +{ + if (!parse_png_file(p, SCAN_header, 0)) { + stbi_rewind( p->s ); + return 0; + } + if (x) *x = p->s->img_x; + if (y) *y = p->s->img_y; + if (comp) *comp = p->s->img_n; + return 1; +} + +static int stbi_png_info(stbi *s, int *x, int *y, int *comp) +{ + png p; + p.s = s; + return stbi_png_info_raw(&p, x, y, comp); +} + +// Microsoft/Windows BMP image + +static int bmp_test(stbi *s) +{ + int sz; + if (get8(s) != 'B') return 0; + if (get8(s) != 'M') return 0; + get32le(s); // discard filesize + get16le(s); // discard reserved + get16le(s); // discard reserved + get32le(s); // discard data offset + sz = get32le(s); + if (sz == 12 || sz == 40 || sz == 56 || sz == 108) return 1; + return 0; +} + +static int stbi_bmp_test(stbi *s) +{ + int r = bmp_test(s); + stbi_rewind(s); + return r; +} + + +// returns 0..31 for the highest set bit +static int high_bit(unsigned int z) +{ + int n=0; + if (z == 0) return -1; + if (z >= 0x10000) n += 16, z >>= 16; + if (z >= 0x00100) n += 8, z >>= 8; + if (z >= 0x00010) n += 4, z >>= 4; + if (z >= 0x00004) n += 2, z >>= 2; + if (z >= 0x00002) n += 1, z >>= 1; + return n; +} + +static int bitcount(unsigned int a) +{ + a = (a & 0x55555555) + ((a >> 1) & 0x55555555); // max 2 + a = (a & 0x33333333) + ((a >> 2) & 0x33333333); // max 4 + a = (a + (a >> 4)) & 0x0f0f0f0f; // max 8 per 4, now 8 bits + a = (a + (a >> 8)); // max 16 per 8 bits + a = (a + (a >> 16)); // max 32 per 8 bits + return a & 0xff; +} + +static int shiftsigned(int v, int shift, int bits) +{ + int result; + int z=0; + + if (shift < 0) v <<= -shift; + else v >>= shift; + result = v; + + z = bits; + while (z < 8) { + result += v >> z; + z += bits; + } + return result; +} + +static stbi_uc *bmp_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + uint8 *out; + unsigned int mr=0,mg=0,mb=0,ma=0, fake_a=0; + stbi_uc pal[256][4]; + int psize=0,i,j,compress=0,width; + int bpp, flip_vertically, pad, target, offset, hsz; + if (get8(s) != 'B' || get8(s) != 'M') return epuc("not BMP", "Corrupt BMP"); + get32le(s); // discard filesize + get16le(s); // discard reserved + get16le(s); // discard reserved + offset = get32le(s); + hsz = get32le(s); + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108) return epuc("unknown BMP", "BMP type not supported: unknown"); + if (hsz == 12) { + s->img_x = get16le(s); + s->img_y = get16le(s); + } else { + s->img_x = get32le(s); + s->img_y = get32le(s); + } + if (get16le(s) != 1) return epuc("bad BMP", "bad BMP"); + bpp = get16le(s); + if (bpp == 1) return epuc("monochrome", "BMP type not supported: 1-bit"); + flip_vertically = ((int) s->img_y) > 0; + s->img_y = abs((int) s->img_y); + if (hsz == 12) { + if (bpp < 24) + psize = (offset - 14 - 24) / 3; + } else { + compress = get32le(s); + if (compress == 1 || compress == 2) return epuc("BMP RLE", "BMP type not supported: RLE"); + get32le(s); // discard sizeof + get32le(s); // discard hres + get32le(s); // discard vres + get32le(s); // discard colorsused + get32le(s); // discard max important + if (hsz == 40 || hsz == 56) { + if (hsz == 56) { + get32le(s); + get32le(s); + get32le(s); + get32le(s); + } + if (bpp == 16 || bpp == 32) { + mr = mg = mb = 0; + if (compress == 0) { + if (bpp == 32) { + mr = 0xffu << 16; + mg = 0xffu << 8; + mb = 0xffu << 0; + ma = 0xffu << 24; + fake_a = 1; // @TODO: check for cases like alpha value is all 0 and switch it to 255 + } else { + mr = 31u << 10; + mg = 31u << 5; + mb = 31u << 0; + } + } else if (compress == 3) { + mr = get32le(s); + mg = get32le(s); + mb = get32le(s); + // not documented, but generated by photoshop and handled by mspaint + if (mr == mg && mg == mb) { + // ?!?!? + return epuc("bad BMP", "bad BMP"); + } + } else + return epuc("bad BMP", "bad BMP"); + } + } else { + assert(hsz == 108); + mr = get32le(s); + mg = get32le(s); + mb = get32le(s); + ma = get32le(s); + get32le(s); // discard color space + for (i=0; i < 12; ++i) + get32le(s); // discard color space parameters + } + if (bpp < 16) + psize = (offset - 14 - hsz) >> 2; + } + s->img_n = ma ? 4 : 3; + if (req_comp && req_comp >= 3) // we can directly decode 3 or 4 + target = req_comp; + else + target = s->img_n; // if they want monochrome, we'll post-convert + out = (stbi_uc *) malloc(target * s->img_x * s->img_y); + if (!out) return epuc("outofmem", "Out of memory"); + if (bpp < 16) { + int z=0; + if (psize == 0 || psize > 256) { free(out); return epuc("invalid", "Corrupt BMP"); } + for (i=0; i < psize; ++i) { + pal[i][2] = get8u(s); + pal[i][1] = get8u(s); + pal[i][0] = get8u(s); + if (hsz != 12) get8(s); + pal[i][3] = 255; + } + skip(s, offset - 14 - hsz - psize * (hsz == 12 ? 3 : 4)); + if (bpp == 4) width = (s->img_x + 1) >> 1; + else if (bpp == 8) width = s->img_x; + else { free(out); return epuc("bad bpp", "Corrupt BMP"); } + pad = (-width)&3; + for (j=0; j < (int) s->img_y; ++j) { + for (i=0; i < (int) s->img_x; i += 2) { + int v=get8(s),v2=0; + if (bpp == 4) { + v2 = v & 15; + v >>= 4; + } + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + if (i+1 == (int) s->img_x) break; + v = (bpp == 8) ? get8(s) : v2; + out[z++] = pal[v][0]; + out[z++] = pal[v][1]; + out[z++] = pal[v][2]; + if (target == 4) out[z++] = 255; + } + skip(s, pad); + } + } else { + int rshift=0,gshift=0,bshift=0,ashift=0,rcount=0,gcount=0,bcount=0,acount=0; + int z = 0; + int easy=0; + skip(s, offset - 14 - hsz); + if (bpp == 24) width = 3 * s->img_x; + else if (bpp == 16) width = 2*s->img_x; + else /* bpp = 32 and pad = 0 */ width=0; + pad = (-width) & 3; + if (bpp == 24) { + easy = 1; + } else if (bpp == 32) { + if (mb == 0xff && mg == 0xff00 && mr == 0x00ff0000 && ma == 0xff000000) + easy = 2; + } + if (!easy) { + if (!mr || !mg || !mb) { free(out); return epuc("bad masks", "Corrupt BMP"); } + // right shift amt to put high bit in position #7 + rshift = high_bit(mr)-7; rcount = bitcount(mr); + gshift = high_bit(mg)-7; gcount = bitcount(mr); + bshift = high_bit(mb)-7; bcount = bitcount(mr); + ashift = high_bit(ma)-7; acount = bitcount(mr); + } + for (j=0; j < (int) s->img_y; ++j) { + if (easy) { + for (i=0; i < (int) s->img_x; ++i) { + int a; + out[z+2] = get8u(s); + out[z+1] = get8u(s); + out[z+0] = get8u(s); + z += 3; + a = (easy == 2 ? get8(s) : 255); + if (target == 4) out[z++] = (uint8) a; + } + } else { + for (i=0; i < (int) s->img_x; ++i) { + uint32 v = (bpp == 16 ? get16le(s) : get32le(s)); + int a; + out[z++] = (uint8) shiftsigned(v & mr, rshift, rcount); + out[z++] = (uint8) shiftsigned(v & mg, gshift, gcount); + out[z++] = (uint8) shiftsigned(v & mb, bshift, bcount); + a = (ma ? shiftsigned(v & ma, ashift, acount) : 255); + if (target == 4) out[z++] = (uint8) a; + } + } + skip(s, pad); + } + } + if (flip_vertically) { + stbi_uc t; + for (j=0; j < (int) s->img_y>>1; ++j) { + stbi_uc *p1 = out + j *s->img_x*target; + stbi_uc *p2 = out + (s->img_y-1-j)*s->img_x*target; + for (i=0; i < (int) s->img_x*target; ++i) { + t = p1[i], p1[i] = p2[i], p2[i] = t; + } + } + } + + if (req_comp && req_comp != target) { + out = convert_format(out, target, req_comp, s->img_x, s->img_y); + if (out == NULL) return out; // convert_format frees input on failure + } + + *x = s->img_x; + *y = s->img_y; + if (comp) *comp = s->img_n; + return out; +} + +static stbi_uc *stbi_bmp_load(stbi *s,int *x, int *y, int *comp, int req_comp) +{ + return bmp_load(s, x,y,comp,req_comp); +} + + +// Targa Truevision - TGA +// by Jonathan Dummer + +static int tga_info(stbi *s, int *x, int *y, int *comp) +{ + int tga_w, tga_h, tga_comp; + int sz; + get8u(s); // discard Offset + sz = get8u(s); // color type + if( sz > 1 ) { + stbi_rewind(s); + return 0; // only RGB or indexed allowed + } + sz = get8u(s); // image type + // only RGB or grey allowed, +/- RLE + if ((sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11)) return 0; + skip(s,9); + tga_w = get16le(s); + if( tga_w < 1 ) { + stbi_rewind(s); + return 0; // test width + } + tga_h = get16le(s); + if( tga_h < 1 ) { + stbi_rewind(s); + return 0; // test height + } + sz = get8(s); // bits per pixel + // only RGB or RGBA or grey allowed + if ((sz != 8) && (sz != 16) && (sz != 24) && (sz != 32)) { + stbi_rewind(s); + return 0; + } + tga_comp = sz; + if (x) *x = tga_w; + if (y) *y = tga_h; + if (comp) *comp = tga_comp / 8; + return 1; // seems to have passed everything +} + +int stbi_tga_info(stbi *s, int *x, int *y, int *comp) +{ + return tga_info(s, x, y, comp); +} + +static int tga_test(stbi *s) +{ + int sz; + get8u(s); // discard Offset + sz = get8u(s); // color type + if ( sz > 1 ) return 0; // only RGB or indexed allowed + sz = get8u(s); // image type + if ( (sz != 1) && (sz != 2) && (sz != 3) && (sz != 9) && (sz != 10) && (sz != 11) ) return 0; // only RGB or grey allowed, +/- RLE + get16(s); // discard palette start + get16(s); // discard palette length + get8(s); // discard bits per palette color entry + get16(s); // discard x origin + get16(s); // discard y origin + if ( get16(s) < 1 ) return 0; // test width + if ( get16(s) < 1 ) return 0; // test height + sz = get8(s); // bits per pixel + if ( (sz != 8) && (sz != 16) && (sz != 24) && (sz != 32) ) return 0; // only RGB or RGBA or grey allowed + return 1; // seems to have passed everything +} + +static int stbi_tga_test(stbi *s) +{ + int res = tga_test(s); + stbi_rewind(s); + return res; +} + +static stbi_uc *tga_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + // read in the TGA header stuff + int tga_offset = get8u(s); + int tga_indexed = get8u(s); + int tga_image_type = get8u(s); + int tga_is_RLE = 0; + int tga_palette_start = get16le(s); + int tga_palette_len = get16le(s); + int tga_palette_bits = get8u(s); + int tga_x_origin = get16le(s); + int tga_y_origin = get16le(s); + int tga_width = get16le(s); + int tga_height = get16le(s); + int tga_bits_per_pixel = get8u(s); + int tga_inverted = get8u(s); + // image data + unsigned char *tga_data; + unsigned char *tga_palette = NULL; + int i, j; + unsigned char raw_data[4]; + unsigned char trans_data[4]; + int RLE_count = 0; + int RLE_repeating = 0; + int read_next_pixel = 1; + + // do a tiny bit of precessing + if ( tga_image_type >= 8 ) + { + tga_image_type -= 8; + tga_is_RLE = 1; + } + /* int tga_alpha_bits = tga_inverted & 15; */ + tga_inverted = 1 - ((tga_inverted >> 5) & 1); + + // error check + if ( //(tga_indexed) || + (tga_width < 1) || (tga_height < 1) || + (tga_image_type < 1) || (tga_image_type > 3) || + ((tga_bits_per_pixel != 8) && (tga_bits_per_pixel != 16) && + (tga_bits_per_pixel != 24) && (tga_bits_per_pixel != 32)) + ) + { + return NULL; // we don't report this as a bad TGA because we don't even know if it's TGA + } + + // If I'm paletted, then I'll use the number of bits from the palette + if ( tga_indexed ) + { + tga_bits_per_pixel = tga_palette_bits; + } + + // tga info + *x = tga_width; + *y = tga_height; + if ( (req_comp < 1) || (req_comp > 4) ) + { + // just use whatever the file was + req_comp = tga_bits_per_pixel / 8; + *comp = req_comp; + } else + { + // force a new number of components + *comp = tga_bits_per_pixel/8; + } + tga_data = (unsigned char*)malloc( tga_width * tga_height * req_comp ); + if (!tga_data) return epuc("outofmem", "Out of memory"); + + // skip to the data's starting position (offset usually = 0) + skip(s, tga_offset ); + // do I need to load a palette? + if ( tga_indexed ) + { + // any data to skip? (offset usually = 0) + skip(s, tga_palette_start ); + // load the palette + tga_palette = (unsigned char*)malloc( tga_palette_len * tga_palette_bits / 8 ); + if (!tga_palette) return epuc("outofmem", "Out of memory"); + if (!getn(s, tga_palette, tga_palette_len * tga_palette_bits / 8 )) { + free(tga_data); + free(tga_palette); + return epuc("bad palette", "Corrupt TGA"); + } + } + // load the data + trans_data[0] = trans_data[1] = trans_data[2] = trans_data[3] = 0; + for (i=0; i < tga_width * tga_height; ++i) + { + // if I'm in RLE mode, do I need to get a RLE chunk? + if ( tga_is_RLE ) + { + if ( RLE_count == 0 ) + { + // yep, get the next byte as a RLE command + int RLE_cmd = get8u(s); + RLE_count = 1 + (RLE_cmd & 127); + RLE_repeating = RLE_cmd >> 7; + read_next_pixel = 1; + } else if ( !RLE_repeating ) + { + read_next_pixel = 1; + } + } else + { + read_next_pixel = 1; + } + // OK, if I need to read a pixel, do it now + if ( read_next_pixel ) + { + // load however much data we did have + if ( tga_indexed ) + { + // read in 1 byte, then perform the lookup + int pal_idx = get8u(s); + if ( pal_idx >= tga_palette_len ) + { + // invalid index + pal_idx = 0; + } + pal_idx *= tga_bits_per_pixel / 8; + for (j = 0; j*8 < tga_bits_per_pixel; ++j) + { + raw_data[j] = tga_palette[pal_idx+j]; + } + } else + { + // read in the data raw + for (j = 0; j*8 < tga_bits_per_pixel; ++j) + { + raw_data[j] = get8u(s); + } + } + // convert raw to the intermediate format + switch (tga_bits_per_pixel) + { + case 8: + // Luminous => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 16: + // Luminous,Alpha => RGBA + trans_data[0] = raw_data[0]; + trans_data[1] = raw_data[0]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[1]; + break; + case 24: + // BGR => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = 255; + break; + case 32: + // BGRA => RGBA + trans_data[0] = raw_data[2]; + trans_data[1] = raw_data[1]; + trans_data[2] = raw_data[0]; + trans_data[3] = raw_data[3]; + break; + } + // clear the reading flag for the next pixel + read_next_pixel = 0; + } // end of reading a pixel + // convert to final format + switch (req_comp) + { + case 1: + // RGBA => Luminance + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + break; + case 2: + // RGBA => Luminance,Alpha + tga_data[i*req_comp+0] = compute_y(trans_data[0],trans_data[1],trans_data[2]); + tga_data[i*req_comp+1] = trans_data[3]; + break; + case 3: + // RGBA => RGB + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + break; + case 4: + // RGBA => RGBA + tga_data[i*req_comp+0] = trans_data[0]; + tga_data[i*req_comp+1] = trans_data[1]; + tga_data[i*req_comp+2] = trans_data[2]; + tga_data[i*req_comp+3] = trans_data[3]; + break; + } + // in case we're in RLE mode, keep counting down + --RLE_count; + } + // do I need to invert the image? + if ( tga_inverted ) + { + for (j = 0; j*2 < tga_height; ++j) + { + int index1 = j * tga_width * req_comp; + int index2 = (tga_height - 1 - j) * tga_width * req_comp; + for (i = tga_width * req_comp; i > 0; --i) + { + unsigned char temp = tga_data[index1]; + tga_data[index1] = tga_data[index2]; + tga_data[index2] = temp; + ++index1; + ++index2; + } + } + } + // clear my palette, if I had one + if ( tga_palette != NULL ) + { + free( tga_palette ); + } + // the things I do to get rid of an error message, and yet keep + // Microsoft's C compilers happy... [8^( + tga_palette_start = tga_palette_len = tga_palette_bits = + tga_x_origin = tga_y_origin = 0; + // OK, done + return tga_data; +} + +static stbi_uc *stbi_tga_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + return tga_load(s,x,y,comp,req_comp); +} + + +// ************************************************************************************************* +// Photoshop PSD loader -- PD by Thatcher Ulrich, integration by Nicolas Schulz, tweaked by STB + +static int psd_test(stbi *s) +{ + if (get32(s) != 0x38425053) return 0; // "8BPS" + else return 1; +} + +static int stbi_psd_test(stbi *s) +{ + int r = psd_test(s); + stbi_rewind(s); + return r; +} + +static stbi_uc *psd_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + int pixelCount; + int channelCount, compression; + int channel, i, count, len; + int w,h; + uint8 *out; + + // Check identifier + if (get32(s) != 0x38425053) // "8BPS" + return epuc("not PSD", "Corrupt PSD image"); + + // Check file type version. + if (get16(s) != 1) + return epuc("wrong version", "Unsupported version of PSD image"); + + // Skip 6 reserved bytes. + skip(s, 6 ); + + // Read the number of channels (R, G, B, A, etc). + channelCount = get16(s); + if (channelCount < 0 || channelCount > 16) + return epuc("wrong channel count", "Unsupported number of channels in PSD image"); + + // Read the rows and columns of the image. + h = get32(s); + w = get32(s); + + // Make sure the depth is 8 bits. + if (get16(s) != 8) + return epuc("unsupported bit depth", "PSD bit depth is not 8 bit"); + + // Make sure the color mode is RGB. + // Valid options are: + // 0: Bitmap + // 1: Grayscale + // 2: Indexed color + // 3: RGB color + // 4: CMYK color + // 7: Multichannel + // 8: Duotone + // 9: Lab color + if (get16(s) != 3) + return epuc("wrong color format", "PSD is not in RGB color format"); + + // Skip the Mode Data. (It's the palette for indexed color; other info for other modes.) + skip(s,get32(s) ); + + // Skip the image resources. (resolution, pen tool paths, etc) + skip(s, get32(s) ); + + // Skip the reserved data. + skip(s, get32(s) ); + + // Find out if the data is compressed. + // Known values: + // 0: no compression + // 1: RLE compressed + compression = get16(s); + if (compression > 1) + return epuc("bad compression", "PSD has an unknown compression format"); + + // Create the destination image. + out = (stbi_uc *) malloc(4 * w*h); + if (!out) return epuc("outofmem", "Out of memory"); + pixelCount = w*h; + + // Initialize the data to zero. + //memset( out, 0, pixelCount * 4 ); + + // Finally, the image data. + if (compression) { + // RLE as used by .PSD and .TIFF + // Loop until you get the number of unpacked bytes you are expecting: + // Read the next source byte into n. + // If n is between 0 and 127 inclusive, copy the next n+1 bytes literally. + // Else if n is between -127 and -1 inclusive, copy the next byte -n+1 times. + // Else if n is 128, noop. + // Endloop + + // The RLE-compressed data is preceeded by a 2-byte data count for each row in the data, + // which we're going to just skip. + skip(s, h * channelCount * 2 ); + + // Read the RLE data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out+channel; + if (channel >= channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = (channel == 3 ? 255 : 0), p += 4; + } else { + // Read the RLE data. + count = 0; + while (count < pixelCount) { + len = get8(s); + if (len == 128) { + // No-op. + } else if (len < 128) { + // Copy next len+1 bytes literally. + len++; + count += len; + while (len) { + *p = get8u(s); + p += 4; + len--; + } + } else if (len > 128) { + uint8 val; + // Next -len+1 bytes in the dest are replicated from next source byte. + // (Interpret len as a negative 8-bit int.) + len ^= 0x0FF; + len += 2; + val = get8u(s); + count += len; + while (len) { + *p = val; + p += 4; + len--; + } + } + } + } + } + + } else { + // We're at the raw image data. It's each channel in order (Red, Green, Blue, Alpha, ...) + // where each channel consists of an 8-bit value for each pixel in the image. + + // Read the data by channel. + for (channel = 0; channel < 4; channel++) { + uint8 *p; + + p = out + channel; + if (channel > channelCount) { + // Fill this channel with default data. + for (i = 0; i < pixelCount; i++) *p = channel == 3 ? 255 : 0, p += 4; + } else { + // Read the data. + for (i = 0; i < pixelCount; i++) + *p = get8u(s), p += 4; + } + } + } + + if (req_comp && req_comp != 4) { + out = convert_format(out, 4, req_comp, w, h); + if (out == NULL) return out; // convert_format frees input on failure + } + + if (comp) *comp = channelCount; + *y = h; + *x = w; + + return out; +} + +static stbi_uc *stbi_psd_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + return psd_load(s,x,y,comp,req_comp); +} + +// ************************************************************************************************* +// Softimage PIC loader +// by Tom Seddon +// +// See http://softimage.wiki.softimage.com/index.php/INFO:_PIC_file_format +// See http://ozviz.wasp.uwa.edu.au/~pbourke/dataformats/softimagepic/ + +static int pic_is4(stbi *s,const char *str) +{ + int i; + for (i=0; i<4; ++i) + if (get8(s) != (stbi_uc)str[i]) + return 0; + + return 1; +} + +static int pic_test(stbi *s) +{ + int i; + + if (!pic_is4(s,"\x53\x80\xF6\x34")) + return 0; + + for(i=0;i<84;++i) + get8(s); + + if (!pic_is4(s,"PICT")) + return 0; + + return 1; +} + +typedef struct +{ + stbi_uc size,type,channel; +} pic_packet_t; + +static stbi_uc *pic_readval(stbi *s, int channel, stbi_uc *dest) +{ + int mask=0x80, i; + + for (i=0; i<4; ++i, mask>>=1) { + if (channel & mask) { + if (at_eof(s)) return epuc("bad file","PIC file too short"); + dest[i]=get8u(s); + } + } + + return dest; +} + +static void pic_copyval(int channel,stbi_uc *dest,const stbi_uc *src) +{ + int mask=0x80,i; + + for (i=0;i<4; ++i, mask>>=1) + if (channel&mask) + dest[i]=src[i]; +} + +static stbi_uc *pic_load2(stbi *s,int width,int height,int *comp, stbi_uc *result) +{ + int act_comp=0,num_packets=0,y,chained; + pic_packet_t packets[10]; + + // this will (should...) cater for even some bizarre stuff like having data + // for the same channel in multiple packets. + do { + pic_packet_t *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return epuc("bad format","too many packets"); + + packet = &packets[num_packets++]; + + chained = get8(s); + packet->size = get8u(s); + packet->type = get8u(s); + packet->channel = get8u(s); + + act_comp |= packet->channel; + + if (at_eof(s)) return epuc("bad file","file too short (reading packets)"); + if (packet->size != 8) return epuc("bad format","packet isn't 8bpp"); + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); // has alpha channel? + + for(y=0; ytype) { + default: + return epuc("bad format","packet has bad compression type"); + + case 0: {//uncompressed + int x; + + for(x=0;xchannel,dest)) + return 0; + break; + } + + case 1://Pure RLE + { + int left=width, i; + + while (left>0) { + stbi_uc count,value[4]; + + count=get8u(s); + if (at_eof(s)) return epuc("bad file","file too short (pure read count)"); + + if (count > left) + count = (uint8) left; + + if (!pic_readval(s,packet->channel,value)) return 0; + + for(i=0; ichannel,dest,value); + left -= count; + } + } + break; + + case 2: {//Mixed RLE + int left=width; + while (left>0) { + int count = get8(s), i; + if (at_eof(s)) return epuc("bad file","file too short (mixed read count)"); + + if (count >= 128) { // Repeated + stbi_uc value[4]; + int i; + + if (count==128) + count = get16(s); + else + count -= 127; + if (count > left) + return epuc("bad file","scanline overrun"); + + if (!pic_readval(s,packet->channel,value)) + return 0; + + for(i=0;ichannel,dest,value); + } else { // Raw + ++count; + if (count>left) return epuc("bad file","scanline overrun"); + + for(i=0;ichannel,dest)) + return 0; + } + left-=count; + } + break; + } + } + } + } + + return result; +} + +static stbi_uc *pic_load(stbi *s,int *px,int *py,int *comp,int req_comp) +{ + stbi_uc *result; + int i, x,y; + + for (i=0; i<92; ++i) + get8(s); + + x = get16(s); + y = get16(s); + if (at_eof(s)) return epuc("bad file","file too short (pic header)"); + if ((1 << 28) / x < y) return epuc("too large", "Image too large to decode"); + + get32(s); //skip `ratio' + get16(s); //skip `fields' + get16(s); //skip `pad' + + // intermediate buffer is RGBA + result = (stbi_uc *) malloc(x*y*4); + memset(result, 0xff, x*y*4); + + if (!pic_load2(s,x,y,comp, result)) { + free(result); + result=0; + } + *px = x; + *py = y; + if (req_comp == 0) req_comp = *comp; + result=convert_format(result,4,req_comp,x,y); + + return result; +} + +static int stbi_pic_test(stbi *s) +{ + int r = pic_test(s); + stbi_rewind(s); + return r; +} + +static stbi_uc *stbi_pic_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + return pic_load(s,x,y,comp,req_comp); +} + +// ************************************************************************************************* +// GIF loader -- public domain by Jean-Marc Lienher -- simplified/shrunk by stb +typedef struct stbi_gif_lzw_struct { + int16 prefix; + uint8 first; + uint8 suffix; +} stbi_gif_lzw; + +typedef struct stbi_gif_struct +{ + int w,h; + stbi_uc *out; // output buffer (always 4 components) + int flags, bgindex, ratio, transparent, eflags; + uint8 pal[256][4]; + uint8 lpal[256][4]; + stbi_gif_lzw codes[4096]; + uint8 *color_table; + int parse, step; + int lflags; + int start_x, start_y; + int max_x, max_y; + int cur_x, cur_y; + int line_size; +} stbi_gif; + +static int gif_test(stbi *s) +{ + int sz; + if (get8(s) != 'G' || get8(s) != 'I' || get8(s) != 'F' || get8(s) != '8') return 0; + sz = get8(s); + if (sz != '9' && sz != '7') return 0; + if (get8(s) != 'a') return 0; + return 1; +} + +static int stbi_gif_test(stbi *s) +{ + int r = gif_test(s); + stbi_rewind(s); + return r; +} + +static void stbi_gif_parse_colortable(stbi *s, uint8 pal[256][4], int num_entries, int transp) +{ + int i; + for (i=0; i < num_entries; ++i) { + pal[i][2] = get8u(s); + pal[i][1] = get8u(s); + pal[i][0] = get8u(s); + pal[i][3] = transp ? 0 : 255; + } +} + +static int stbi_gif_header(stbi *s, stbi_gif *g, int *comp, int is_info) +{ + uint8 version; + if (get8(s) != 'G' || get8(s) != 'I' || get8(s) != 'F' || get8(s) != '8') + return e("not GIF", "Corrupt GIF"); + + version = get8u(s); + if (version != '7' && version != '9') return e("not GIF", "Corrupt GIF"); + if (get8(s) != 'a') return e("not GIF", "Corrupt GIF"); + + failure_reason = ""; + g->w = get16le(s); + g->h = get16le(s); + g->flags = get8(s); + g->bgindex = get8(s); + g->ratio = get8(s); + g->transparent = -1; + + if (comp != 0) *comp = 4; // can't actually tell whether it's 3 or 4 until we parse the comments + + if (is_info) return 1; + + if (g->flags & 0x80) + stbi_gif_parse_colortable(s,g->pal, 2 << (g->flags & 7), -1); + + return 1; +} + +static int stbi_gif_info_raw(stbi *s, int *x, int *y, int *comp) +{ + stbi_gif g; + if (!stbi_gif_header(s, &g, comp, 1)) { + stbi_rewind( s ); + return 0; + } + if (x) *x = g.w; + if (y) *y = g.h; + return 1; +} + +static void stbi_out_gif_code(stbi_gif *g, uint16 code) +{ + uint8 *p, *c; + + // recurse to decode the prefixes, since the linked-list is backwards, + // and working backwards through an interleaved image would be nasty + if (g->codes[code].prefix >= 0) + stbi_out_gif_code(g, g->codes[code].prefix); + + if (g->cur_y >= g->max_y) return; + + p = &g->out[g->cur_x + g->cur_y]; + c = &g->color_table[g->codes[code].suffix * 4]; + + if (c[3] >= 128) { + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } + g->cur_x += 4; + + if (g->cur_x >= g->max_x) { + g->cur_x = g->start_x; + g->cur_y += g->step; + + while (g->cur_y >= g->max_y && g->parse > 0) { + g->step = (1 << g->parse) * g->line_size; + g->cur_y = g->start_y + (g->step >> 1); + --g->parse; + } + } +} + +static uint8 *stbi_process_gif_raster(stbi *s, stbi_gif *g) +{ + uint8 lzw_cs; + int32 len, code; + uint32 first; + int32 codesize, codemask, avail, oldcode, bits, valid_bits, clear; + stbi_gif_lzw *p; + + lzw_cs = get8u(s); + clear = 1 << lzw_cs; + first = 1; + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + bits = 0; + valid_bits = 0; + for (code = 0; code < clear; code++) { + g->codes[code].prefix = -1; + g->codes[code].first = (uint8) code; + g->codes[code].suffix = (uint8) code; + } + + // support no starting clear code + avail = clear+2; + oldcode = -1; + + len = 0; + for(;;) { + if (valid_bits < codesize) { + if (len == 0) { + len = get8(s); // start new block + if (len == 0) + return g->out; + } + --len; + bits |= (int32) get8(s) << valid_bits; + valid_bits += 8; + } else { + int32 code = bits & codemask; + bits >>= codesize; + valid_bits -= codesize; + // @OPTIMIZE: is there some way we can accelerate the non-clear path? + if (code == clear) { // clear code + codesize = lzw_cs + 1; + codemask = (1 << codesize) - 1; + avail = clear + 2; + oldcode = -1; + first = 0; + } else if (code == clear + 1) { // end of stream code + skip(s, len); + while ((len = get8(s)) > 0) + skip(s,len); + return g->out; + } else if (code <= avail) { + if (first) return epuc("no clear code", "Corrupt GIF"); + + if (oldcode >= 0) { + p = &g->codes[avail++]; + if (avail > 4096) return epuc("too many codes", "Corrupt GIF"); + p->prefix = (int16) oldcode; + p->first = g->codes[oldcode].first; + p->suffix = (code == avail) ? p->first : g->codes[code].first; + } else if (code == avail) + return epuc("illegal code in raster", "Corrupt GIF"); + + stbi_out_gif_code(g, (uint16) code); + + if ((avail & codemask) == 0 && avail <= 0x0FFF) { + codesize++; + codemask = (1 << codesize) - 1; + } + + oldcode = code; + } else { + return epuc("illegal code in raster", "Corrupt GIF"); + } + } + } +} + +static void stbi_fill_gif_background(stbi_gif *g) +{ + int i; + uint8 *c = g->pal[g->bgindex]; + // @OPTIMIZE: write a dword at a time + for (i = 0; i < g->w * g->h * 4; i += 4) { + uint8 *p = &g->out[i]; + p[0] = c[2]; + p[1] = c[1]; + p[2] = c[0]; + p[3] = c[3]; + } +} + +// this function is designed to support animated gifs, although stb_image doesn't support it +static uint8 *stbi_gif_load_next(stbi *s, stbi_gif *g, int *comp, int req_comp) +{ + int i; + uint8 *old_out = 0; + + if (g->out == 0) { + if (!stbi_gif_header(s, g, comp,0)) return 0; // failure_reason set by stbi_gif_header + g->out = (uint8 *) malloc(4 * g->w * g->h); + if (g->out == 0) return epuc("outofmem", "Out of memory"); + stbi_fill_gif_background(g); + } else { + // animated-gif-only path + if (((g->eflags & 0x1C) >> 2) == 3) { + old_out = g->out; + g->out = (uint8 *) malloc(4 * g->w * g->h); + if (g->out == 0) return epuc("outofmem", "Out of memory"); + memcpy(g->out, old_out, g->w*g->h*4); + } + } + + for (;;) { + switch (get8(s)) { + case 0x2C: /* Image Descriptor */ + { + int32 x, y, w, h; + uint8 *o; + + x = get16le(s); + y = get16le(s); + w = get16le(s); + h = get16le(s); + if (((x + w) > (g->w)) || ((y + h) > (g->h))) + return epuc("bad Image Descriptor", "Corrupt GIF"); + + g->line_size = g->w * 4; + g->start_x = x * 4; + g->start_y = y * g->line_size; + g->max_x = g->start_x + w * 4; + g->max_y = g->start_y + h * g->line_size; + g->cur_x = g->start_x; + g->cur_y = g->start_y; + + g->lflags = get8(s); + + if (g->lflags & 0x40) { + g->step = 8 * g->line_size; // first interlaced spacing + g->parse = 3; + } else { + g->step = g->line_size; + g->parse = 0; + } + + if (g->lflags & 0x80) { + stbi_gif_parse_colortable(s,g->lpal, 2 << (g->lflags & 7), g->eflags & 0x01 ? g->transparent : -1); + g->color_table = (uint8 *) g->lpal; + } else if (g->flags & 0x80) { + for (i=0; i < 256; ++i) // @OPTIMIZE: reset only the previous transparent + g->pal[i][3] = 255; + if (g->transparent >= 0 && (g->eflags & 0x01)) + g->pal[g->transparent][3] = 0; + g->color_table = (uint8 *) g->pal; + } else + return epuc("missing color table", "Corrupt GIF"); + + o = stbi_process_gif_raster(s, g); + if (o == NULL) return NULL; + + if (req_comp && req_comp != 4) + o = convert_format(o, 4, req_comp, g->w, g->h); + return o; + } + + case 0x21: // Comment Extension. + { + int len; + if (get8(s) == 0xF9) { // Graphic Control Extension. + len = get8(s); + if (len == 4) { + g->eflags = get8(s); + get16le(s); // delay + g->transparent = get8(s); + } else { + skip(s, len); + break; + } + } + while ((len = get8(s)) != 0) + skip(s, len); + break; + } + + case 0x3B: // gif stream termination code + return (uint8 *) 1; + + default: + return epuc("unknown code", "Corrupt GIF"); + } + } +} + +static stbi_uc *stbi_gif_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + uint8 *u = 0; + stbi_gif g={0}; + + u = stbi_gif_load_next(s, &g, comp, req_comp); + if (u == (void *) 1) u = 0; // end of animated gif marker + if (u) { + *x = g.w; + *y = g.h; + } + + return u; +} + +static int stbi_gif_info(stbi *s, int *x, int *y, int *comp) +{ + return stbi_gif_info_raw(s,x,y,comp); +} + + +// ************************************************************************************************* +// Radiance RGBE HDR loader +// originally by Nicolas Schulz +#ifndef STBI_NO_HDR +static int hdr_test(stbi *s) +{ + const char *signature = "#?RADIANCE\n"; + int i; + for (i=0; signature[i]; ++i) + if (get8(s) != signature[i]) + return 0; + return 1; +} + +static int stbi_hdr_test(stbi* s) +{ + int r = hdr_test(s); + stbi_rewind(s); + return r; +} + +#define HDR_BUFLEN 1024 +static char *hdr_gettoken(stbi *z, char *buffer) +{ + int len=0; + char c = '\0'; + + c = (char) get8(z); + + while (!at_eof(z) && c != '\n') { + buffer[len++] = c; + if (len == HDR_BUFLEN-1) { + // flush to end of line + while (!at_eof(z) && get8(z) != '\n') + ; + break; + } + c = (char) get8(z); + } + + buffer[len] = 0; + return buffer; +} + +static void hdr_convert(float *output, stbi_uc *input, int req_comp) +{ + if ( input[3] != 0 ) { + float f1; + // Exponent + f1 = (float) ldexp(1.0f, input[3] - (int)(128 + 8)); + if (req_comp <= 2) + output[0] = (input[0] + input[1] + input[2]) * f1 / 3; + else { + output[0] = input[0] * f1; + output[1] = input[1] * f1; + output[2] = input[2] * f1; + } + if (req_comp == 2) output[1] = 1; + if (req_comp == 4) output[3] = 1; + } else { + switch (req_comp) { + case 4: output[3] = 1; /* fallthrough */ + case 3: output[0] = output[1] = output[2] = 0; + break; + case 2: output[1] = 1; /* fallthrough */ + case 1: output[0] = 0; + break; + } + } +} + +static float *hdr_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + char buffer[HDR_BUFLEN]; + char *token; + int valid = 0; + int width, height; + stbi_uc *scanline; + float *hdr_data; + int len; + unsigned char count, value; + int i, j, k, c1,c2, z; + + + // Check identifier + if (strcmp(hdr_gettoken(s,buffer), "#?RADIANCE") != 0) + return epf("not HDR", "Corrupt HDR image"); + + // Parse header + for(;;) { + token = hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) return epf("unsupported format", "Unsupported HDR format"); + + // Parse width and height + // can't use sscanf() if we're not using stdio! + token = hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + height = strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) return epf("unsupported data layout", "Unsupported HDR format"); + token += 3; + width = strtol(token, NULL, 10); + + *x = width; + *y = height; + + *comp = 3; + if (req_comp == 0) req_comp = 3; + + // Read data + hdr_data = (float *) malloc(height * width * req_comp * sizeof(float)); + + // Load image data + // image data is stored as some number of sca + if ( width < 8 || width >= 32768) { + // Read flat data + for (j=0; j < height; ++j) { + for (i=0; i < width; ++i) { + stbi_uc rgbe[4]; + main_decode_loop: + getn(s, rgbe, 4); + hdr_convert(hdr_data + j * width * req_comp + i * req_comp, rgbe, req_comp); + } + } + } else { + // Read RLE-encoded data + scanline = NULL; + + for (j = 0; j < height; ++j) { + c1 = get8(s); + c2 = get8(s); + len = get8(s); + if (c1 != 2 || c2 != 2 || (len & 0x80)) { + // not run-length encoded, so we have to actually use THIS data as a decoded + // pixel (note this can't be a valid pixel--one of RGB must be >= 128) + uint8 rgbe[4]; + rgbe[0] = (uint8) c1; + rgbe[1] = (uint8) c2; + rgbe[2] = (uint8) len; + rgbe[3] = (uint8) get8u(s); + hdr_convert(hdr_data, rgbe, req_comp); + i = 1; + j = 0; + free(scanline); + goto main_decode_loop; // yes, this makes no sense + } + len <<= 8; + len |= get8(s); + if (len != width) { free(hdr_data); free(scanline); return epf("invalid decoded scanline length", "corrupt HDR"); } + if (scanline == NULL) scanline = (stbi_uc *) malloc(width * 4); + + for (k = 0; k < 4; ++k) { + i = 0; + while (i < width) { + count = get8u(s); + if (count > 128) { + // Run + value = get8u(s); + count -= 128; + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = value; + } else { + // Dump + for (z = 0; z < count; ++z) + scanline[i++ * 4 + k] = get8u(s); + } + } + } + for (i=0; i < width; ++i) + hdr_convert(hdr_data+(j*width + i)*req_comp, scanline + i*4, req_comp); + } + free(scanline); + } + + return hdr_data; +} + +static float *stbi_hdr_load(stbi *s, int *x, int *y, int *comp, int req_comp) +{ + return hdr_load(s,x,y,comp,req_comp); +} + +static int stbi_hdr_info(stbi *s, int *x, int *y, int *comp) +{ + char buffer[HDR_BUFLEN]; + char *token; + int valid = 0; + + if (strcmp(hdr_gettoken(s,buffer), "#?RADIANCE") != 0) { + stbi_rewind( s ); + return 0; + } + + for(;;) { + token = hdr_gettoken(s,buffer); + if (token[0] == 0) break; + if (strcmp(token, "FORMAT=32-bit_rle_rgbe") == 0) valid = 1; + } + + if (!valid) { + stbi_rewind( s ); + return 0; + } + token = hdr_gettoken(s,buffer); + if (strncmp(token, "-Y ", 3)) { + stbi_rewind( s ); + return 0; + } + token += 3; + *y = strtol(token, &token, 10); + while (*token == ' ') ++token; + if (strncmp(token, "+X ", 3)) { + stbi_rewind( s ); + return 0; + } + token += 3; + *x = strtol(token, NULL, 10); + *comp = 3; + return 1; +} +#endif // STBI_NO_HDR + +static int stbi_bmp_info(stbi *s, int *x, int *y, int *comp) +{ + int hsz; + if (get8(s) != 'B' || get8(s) != 'M') { + stbi_rewind( s ); + return 0; + } + skip(s,12); + hsz = get32le(s); + if (hsz != 12 && hsz != 40 && hsz != 56 && hsz != 108) { + stbi_rewind( s ); + return 0; + } + if (hsz == 12) { + *x = get16le(s); + *y = get16le(s); + } else { + *x = get32le(s); + *y = get32le(s); + } + if (get16le(s) != 1) { + stbi_rewind( s ); + return 0; + } + *comp = get16le(s) / 8; + return 1; +} + +static int stbi_psd_info(stbi *s, int *x, int *y, int *comp) +{ + int channelCount; + if (get32(s) != 0x38425053) { + stbi_rewind( s ); + return 0; + } + if (get16(s) != 1) { + stbi_rewind( s ); + return 0; + } + skip(s, 6); + channelCount = get16(s); + if (channelCount < 0 || channelCount > 16) { + stbi_rewind( s ); + return 0; + } + *y = get32(s); + *x = get32(s); + if (get16(s) != 8) { + stbi_rewind( s ); + return 0; + } + if (get16(s) != 3) { + stbi_rewind( s ); + return 0; + } + *comp = 4; + return 1; +} + +static int stbi_pic_info(stbi *s, int *x, int *y, int *comp) +{ + int act_comp=0,num_packets=0,chained; + pic_packet_t packets[10]; + + skip(s, 92); + + *x = get16(s); + *y = get16(s); + if (at_eof(s)) return 0; + if ( (*x) != 0 && (1 << 28) / (*x) < (*y)) { + stbi_rewind( s ); + return 0; + } + + skip(s, 8); + + do { + pic_packet_t *packet; + + if (num_packets==sizeof(packets)/sizeof(packets[0])) + return 0; + + packet = &packets[num_packets++]; + chained = get8(s); + packet->size = get8u(s); + packet->type = get8u(s); + packet->channel = get8u(s); + act_comp |= packet->channel; + + if (at_eof(s)) { + stbi_rewind( s ); + return 0; + } + if (packet->size != 8) { + stbi_rewind( s ); + return 0; + } + } while (chained); + + *comp = (act_comp & 0x10 ? 4 : 3); + + return 1; +} + +static int stbi_info_main(stbi *s, int *x, int *y, int *comp) +{ + if (stbi_jpeg_info(s, x, y, comp)) + return 1; + if (stbi_png_info(s, x, y, comp)) + return 1; + if (stbi_gif_info(s, x, y, comp)) + return 1; + if (stbi_bmp_info(s, x, y, comp)) + return 1; + if (stbi_psd_info(s, x, y, comp)) + return 1; + if (stbi_pic_info(s, x, y, comp)) + return 1; + #ifndef STBI_NO_HDR + if (stbi_hdr_info(s, x, y, comp)) + return 1; + #endif + // test tga last because it's a crappy test! + if (stbi_tga_info(s, x, y, comp)) + return 1; + return e("unknown image type", "Image not of any known type, or corrupt"); +} + +#ifndef STBI_NO_STDIO +int stbi_info(char const *filename, int *x, int *y, int *comp) +{ + FILE *f = fopen(filename, "rb"); + int result; + if (!f) return e("can't fopen", "Unable to open file"); + result = stbi_info_from_file(f, x, y, comp); + fclose(f); + return result; +} + +int stbi_info_from_file(FILE *f, int *x, int *y, int *comp) +{ + int r; + stbi s; + long pos = ftell(f); + start_file(&s, f); + r = stbi_info_main(&s,x,y,comp); + fseek(f,pos,SEEK_SET); + return r; +} +#endif // !STBI_NO_STDIO + +int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp) +{ + stbi s; + start_mem(&s,buffer,len); + return stbi_info_main(&s,x,y,comp); +} + +int stbi_info_from_callbacks(stbi_io_callbacks const *c, void *user, int *x, int *y, int *comp) +{ + stbi s; + start_callbacks(&s, (stbi_io_callbacks *) c, user); + return stbi_info_main(&s,x,y,comp); +} + +#endif // STBI_HEADER_FILE_ONLY + +/* + revision history: + 1.34 + use STBI_NOTUSED in resample_row_generic() + 1.33 (2011-07-14) + make stbi_is_hdr work in STBI_NO_HDR (as specified), minor compiler-friendly improvements + 1.32 (2011-07-13) + support for "info" function for all supported filetypes (SpartanJ) + 1.31 (2011-06-20) + a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) + added ability to load files via callbacks to accomidate custom input streams (Ben Wenger) + removed deprecated format-specific test/load functions + removed support for installable file formats (stbi_loader) -- would have been broken for IO callbacks anyway + error cases in bmp and tga give messages and don't leak (Raymond Barbiero, grisha) + fix inefficiency in decoding 32-bit BMP (David Woo) + 1.29 (2010-08-16) + various warning fixes from Aurelien Pocheville + 1.28 (2010-08-01) + fix bug in GIF palette transparency (SpartanJ) + 1.27 (2010-08-01) + cast-to-uint8 to fix warnings + 1.26 (2010-07-24) + fix bug in file buffering for PNG reported by SpartanJ + 1.25 (2010-07-17) + refix trans_data warning (Won Chun) + 1.24 (2010-07-12) + perf improvements reading from files on platforms with lock-heavy fgetc() + minor perf improvements for jpeg + deprecated type-specific functions so we'll get feedback if they're needed + attempt to fix trans_data warning (Won Chun) + 1.23 fixed bug in iPhone support + 1.22 (2010-07-10) + removed image *writing* support + stbi_info support from Jetro Lauha + GIF support from Jean-Marc Lienher + iPhone PNG-extensions from James Brown + warning-fixes from Nicolas Schulz and Janez Zemva (i.e. Janez (U+017D)emva) + 1.21 fix use of 'uint8' in header (reported by jon blow) + 1.20 added support for Softimage PIC, by Tom Seddon + 1.19 bug in interlaced PNG corruption check (found by ryg) + 1.18 2008-08-02 + fix a threading bug (local mutable static) + 1.17 support interlaced PNG + 1.16 major bugfix - convert_format converted one too many pixels + 1.15 initialize some fields for thread safety + 1.14 fix threadsafe conversion bug + header-file-only version (#define STBI_HEADER_FILE_ONLY before including) + 1.13 threadsafe + 1.12 const qualifiers in the API + 1.11 Support installable IDCT, colorspace conversion routines + 1.10 Fixes for 64-bit (don't use "unsigned long") + optimized upsampling by Fabian "ryg" Giesen + 1.09 Fix format-conversion for PSD code (bad global variables!) + 1.08 Thatcher Ulrich's PSD code integrated by Nicolas Schulz + 1.07 attempt to fix C++ warning/errors again + 1.06 attempt to fix C++ warning/errors again + 1.05 fix TGA loading to return correct *comp and use good luminance calc + 1.04 default float alpha is 1, not 255; use 'void *' for stbi_image_free + 1.03 bugfixes to STBI_NO_STDIO, STBI_NO_HDR + 1.02 support for (subset of) HDR files, float interface for preferred access to them + 1.01 fix bug: possible bug in handling right-side up bmps... not sure + fix bug: the stbi_bmp_load() and stbi_tga_load() functions didn't work at all + 1.00 interface to zlib that skips zlib header + 0.99 correct handling of alpha in palette + 0.98 TGA loader by lonesock; dynamically add loaders (untested) + 0.97 jpeg errors on too large a file; also catch another malloc failure + 0.96 fix detection of invalid v value - particleman@mollyrocket forum + 0.95 during header scan, seek to markers in case of padding + 0.94 STBI_NO_STDIO to disable stdio usage; rename all #defines the same + 0.93 handle jpegtran output; verbose errors + 0.92 read 4,8,16,24,32-bit BMP files of several formats + 0.91 output 24-bit Windows 3.0 BMP files + 0.90 fix a few more warnings; bump version number to approach 1.0 + 0.61 bugfixes due to Marc LeBlanc, Christopher Lloyd + 0.60 fix compiling as c++ + 0.59 fix warnings: merge Dave Moore's -Wall fixes + 0.58 fix bug: zlib uncompressed mode len/nlen was wrong endian + 0.57 fix bug: jpg last huffman symbol before marker was >9 bits but less than 16 available + 0.56 fix bug: zlib uncompressed mode len vs. nlen + 0.55 fix bug: restart_interval not initialized to 0 + 0.54 allow NULL for 'int *comp' + 0.53 fix bug in png 3->4; speedup png decoding + 0.52 png handles req_comp=3,4 directly; minor cleanup; jpeg comments + 0.51 obey req_comp requests, 1-component jpegs return as 1-component, + on 'test' only check type, not whether we support this variant + 0.50 first released version +*/ diff --git a/stblib/stb_image.h b/stblib/stb_image.h new file mode 100644 index 0000000000..70c043ecf5 --- /dev/null +++ b/stblib/stb_image.h @@ -0,0 +1,329 @@ +/* stbi-1.33 - public domain JPEG/PNG reader - http://nothings.org/stb_image.c + when you control the images you're loading + no warranty implied; use at your own risk + + QUICK NOTES: + Primarily of interest to game developers and other people who can + avoid problematic images and only need the trivial interface + + JPEG baseline (no JPEG progressive) + PNG 8-bit only + + TGA (not sure what subset, if a subset) + BMP non-1bpp, non-RLE + PSD (composited view only, no extra channels) + + GIF (*comp always reports as 4-channel) + HDR (radiance rgbE format) + PIC (Softimage PIC) + + - decode from memory or through FILE (define STBI_NO_STDIO to remove code) + - decode from arbitrary I/O callbacks + - overridable dequantizing-IDCT, YCbCr-to-RGB conversion (define STBI_SIMD) + + Latest revisions: + 1.33 (2011-07-14) minor fixes suggested by Dave Moore + 1.32 (2011-07-13) info support for all filetypes (SpartanJ) + 1.31 (2011-06-19) a few more leak fixes, bug in PNG handling (SpartanJ) + 1.30 (2011-06-11) added ability to load files via io callbacks (Ben Wenger) + 1.29 (2010-08-16) various warning fixes from Aurelien Pocheville + + See end of file for full revision history. + + TODO: + stbi_info support for BMP,PSD,HDR,PIC + + + ============================ Contributors ========================= + + Image formats Optimizations & bugfixes + Sean Barrett (jpeg, png, bmp) Fabian "ryg" Giesen + Nicolas Schulz (hdr, psd) + Jonathan Dummer (tga) Bug fixes & warning fixes + Jean-Marc Lienher (gif) Marc LeBlanc + Tom Seddon (pic) Christpher Lloyd + Thatcher Ulrich (psd) Dave Moore + Won Chun + the Horde3D community + Extensions, features Janez Zemva + Jetro Lauha (stbi_info) Jonathan Blow + James "moose2000" Brown (iPhone PNG) Laurent Gomila + Ben "Disch" Wenger (io callbacks) Aruelien Pocheville + Martin "SpartanJ" Golini Ryamond Barbiero + David Woo + + + If your name should be here but isn't, let Sean know. + +*/ + +#ifndef STBI_INCLUDE_STB_IMAGE_H +#define STBI_INCLUDE_STB_IMAGE_H + +// To get a header file for this, either cut and paste the header, +// or create stb_image.h, #define STBI_HEADER_FILE_ONLY, and +// then include stb_image.c from it. + +//// begin header file //////////////////////////////////////////////////// +// +// Limitations: +// - no jpeg progressive support +// - non-HDR formats support 8-bit samples only (jpeg, png) +// - no delayed line count (jpeg) -- IJG doesn't support either +// - no 1-bit BMP +// - GIF always returns *comp=4 +// +// Basic usage (see HDR discussion below): +// int x,y,n; +// unsigned char *data = stbi_load(filename, &x, &y, &n, 0); +// // ... process data if not NULL ... +// // ... x = width, y = height, n = # 8-bit components per pixel ... +// // ... replace '0' with '1'..'4' to force that many components per pixel +// // ... but 'n' will always be the number that it would have been if you said 0 +// stbi_image_free(data) +// +// Standard parameters: +// int *x -- outputs image width in pixels +// int *y -- outputs image height in pixels +// int *comp -- outputs # of image components in image file +// int req_comp -- if non-zero, # of image components requested in result +// +// The return value from an image loader is an 'unsigned char *' which points +// to the pixel data. The pixel data consists of *y scanlines of *x pixels, +// with each pixel consisting of N interleaved 8-bit components; the first +// pixel pointed to is top-left-most in the image. There is no padding between +// image scanlines or between pixels, regardless of format. The number of +// components N is 'req_comp' if req_comp is non-zero, or *comp otherwise. +// If req_comp is non-zero, *comp has the number of components that _would_ +// have been output otherwise. E.g. if you set req_comp to 4, you will always +// get RGBA output, but you can check *comp to easily see if it's opaque. +// +// An output image with N components has the following components interleaved +// in this order in each pixel: +// +// N=#comp components +// 1 grey +// 2 grey, alpha +// 3 red, green, blue +// 4 red, green, blue, alpha +// +// If image loading fails for any reason, the return value will be NULL, +// and *x, *y, *comp will be unchanged. The function stbi_failure_reason() +// can be queried for an extremely brief, end-user unfriendly explanation +// of why the load failed. Define STBI_NO_FAILURE_STRINGS to avoid +// compiling these strings at all, and STBI_FAILURE_USERMSG to get slightly +// more user-friendly ones. +// +// Paletted PNG, BMP, GIF, and PIC images are automatically depalettized. +// +// =========================================================================== +// +// iPhone PNG support: +// +// By default we convert iphone-formatted PNGs back to RGB; nominally they +// would silently load as BGR, except the existing code should have just +// failed on such iPhone PNGs. But you can disable this conversion by +// by calling stbi_convert_iphone_png_to_rgb(0), in which case +// you will always just get the native iphone "format" through. +// +// Call stbi_set_unpremultiply_on_load(1) as well to force a divide per +// pixel to remove any premultiplied alpha *only* if the image file explicitly +// says there's premultiplied data (currently only happens in iPhone images, +// and only if iPhone convert-to-rgb processing is on). +// +// =========================================================================== +// +// HDR image support (disable by defining STBI_NO_HDR) +// +// stb_image now supports loading HDR images in general, and currently +// the Radiance .HDR file format, although the support is provided +// generically. You can still load any file through the existing interface; +// if you attempt to load an HDR file, it will be automatically remapped to +// LDR, assuming gamma 2.2 and an arbitrary scale factor defaulting to 1; +// both of these constants can be reconfigured through this interface: +// +// stbi_hdr_to_ldr_gamma(2.2f); +// stbi_hdr_to_ldr_scale(1.0f); +// +// (note, do not use _inverse_ constants; stbi_image will invert them +// appropriately). +// +// Additionally, there is a new, parallel interface for loading files as +// (linear) floats to preserve the full dynamic range: +// +// float *data = stbi_loadf(filename, &x, &y, &n, 0); +// +// If you load LDR images through this interface, those images will +// be promoted to floating point values, run through the inverse of +// constants corresponding to the above: +// +// stbi_ldr_to_hdr_scale(1.0f); +// stbi_ldr_to_hdr_gamma(2.2f); +// +// Finally, given a filename (or an open file or memory block--see header +// file for details) containing image data, you can query for the "most +// appropriate" interface to use (that is, whether the image is HDR or +// not), using: +// +// stbi_is_hdr(char *filename); +// +// =========================================================================== +// +// I/O callbacks +// +// I/O callbacks allow you to read from arbitrary sources, like packaged +// files or some other source. Data read from callbacks are processed +// through a small internal buffer (currently 128 bytes) to try to reduce +// overhead. +// +// The three functions you must define are "read" (reads some bytes of data), +// "skip" (skips some bytes of data), "eof" (reports if the stream is at the end). + + +#ifndef STBI_NO_STDIO + +#if defined(_MSC_VER) && _MSC_VER >= 0x1400 +#define _CRT_SECURE_NO_WARNINGS // suppress bogus warnings about fopen() +#endif + +#include +#endif + +#define STBI_VERSION 1 + +enum +{ + STBI_default = 0, // only used for req_comp + + STBI_grey = 1, + STBI_grey_alpha = 2, + STBI_rgb = 3, + STBI_rgb_alpha = 4 +}; + +typedef unsigned char stbi_uc; + +#ifdef __cplusplus +extern "C" { +#endif + +////////////////////////////////////////////////////////////////////////////// +// +// PRIMARY API - works on images of any type +// + +// +// load image by filename, open file, or memory buffer +// + +extern stbi_uc *stbi_load_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + +#ifndef STBI_NO_STDIO +extern stbi_uc *stbi_load (char const *filename, int *x, int *y, int *comp, int req_comp); +extern stbi_uc *stbi_load_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); +// for stbi_load_from_file, file pointer is left pointing immediately after image +#endif + +typedef struct +{ + int (*read) (void *user,char *data,int size); // fill 'data' with 'size' bytes. return number of bytes actually read + void (*skip) (void *user,unsigned n); // skip the next 'n' bytes + int (*eof) (void *user); // returns nonzero if we are at end of file/data +} stbi_io_callbacks; + +extern stbi_uc *stbi_load_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp); + +#ifndef STBI_NO_HDR + extern float *stbi_loadf_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp, int req_comp); + + #ifndef STBI_NO_STDIO + extern float *stbi_loadf (char const *filename, int *x, int *y, int *comp, int req_comp); + extern float *stbi_loadf_from_file (FILE *f, int *x, int *y, int *comp, int req_comp); + #endif + + extern float *stbi_loadf_from_callbacks (stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp, int req_comp); + + extern void stbi_hdr_to_ldr_gamma(float gamma); + extern void stbi_hdr_to_ldr_scale(float scale); + + extern void stbi_ldr_to_hdr_gamma(float gamma); + extern void stbi_ldr_to_hdr_scale(float scale); +#endif // STBI_NO_HDR + +// stbi_is_hdr is always defined +extern int stbi_is_hdr_from_callbacks(stbi_io_callbacks const *clbk, void *user); +extern int stbi_is_hdr_from_memory(stbi_uc const *buffer, int len); +#ifndef STBI_NO_STDIO +extern int stbi_is_hdr (char const *filename); +extern int stbi_is_hdr_from_file(FILE *f); +#endif // STBI_NO_STDIO + + +// get a VERY brief reason for failure +// NOT THREADSAFE +extern const char *stbi_failure_reason (void); + +// free the loaded image -- this is just free() +extern void stbi_image_free (void *retval_from_stbi_load); + +// get image dimensions & components without fully decoding +extern int stbi_info_from_memory(stbi_uc const *buffer, int len, int *x, int *y, int *comp); +extern int stbi_info_from_callbacks(stbi_io_callbacks const *clbk, void *user, int *x, int *y, int *comp); + +#ifndef STBI_NO_STDIO +extern int stbi_info (char const *filename, int *x, int *y, int *comp); +extern int stbi_info_from_file (FILE *f, int *x, int *y, int *comp); + +#endif + + + +// for image formats that explicitly notate that they have premultiplied alpha, +// we just return the colors as stored in the file. set this flag to force +// unpremultiplication. results are undefined if the unpremultiply overflow. +extern void stbi_set_unpremultiply_on_load(int flag_true_if_should_unpremultiply); + +// indicate whether we should process iphone images back to canonical format, +// or just pass them through "as-is" +extern void stbi_convert_iphone_png_to_rgb(int flag_true_if_should_convert); + + +// ZLIB client - used by PNG, available for other purposes + +extern char *stbi_zlib_decode_malloc_guesssize(const char *buffer, int len, int initial_size, int *outlen); +extern char *stbi_zlib_decode_malloc(const char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + +extern char *stbi_zlib_decode_noheader_malloc(const char *buffer, int len, int *outlen); +extern int stbi_zlib_decode_noheader_buffer(char *obuffer, int olen, const char *ibuffer, int ilen); + + +// define faster low-level operations (typically SIMD support) +#ifdef STBI_SIMD +typedef void (*stbi_idct_8x8)(stbi_uc *out, int out_stride, short data[64], unsigned short *dequantize); +// compute an integer IDCT on "input" +// input[x] = data[x] * dequantize[x] +// write results to 'out': 64 samples, each run of 8 spaced by 'out_stride' +// CLAMP results to 0..255 +typedef void (*stbi_YCbCr_to_RGB_run)(stbi_uc *output, stbi_uc const *y, stbi_uc const *cb, stbi_uc const *cr, int count, int step); +// compute a conversion from YCbCr to RGB +// 'count' pixels +// write pixels to 'output'; each pixel is 'step' bytes (either 3 or 4; if 4, write '255' as 4th), order R,G,B +// y: Y input channel +// cb: Cb input channel; scale/biased to be 0..255 +// cr: Cr input channel; scale/biased to be 0..255 + +extern void stbi_install_idct(stbi_idct_8x8 func); +extern void stbi_install_YCbCr_to_RGB(stbi_YCbCr_to_RGB_run func); +#endif // STBI_SIMD + + +#ifdef __cplusplus +} +#endif + +// +// +//// end header file ///////////////////////////////////////////////////// +#endif // STBI_INCLUDE_STB_IMAGE_H + From 4835ae35afafa88aac69b8db70743defc4c4ac4b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Wed, 11 Sep 2013 22:49:49 +0200 Subject: [PATCH 4/7] Keybord support Added * Implemented 'cellKb*' functions from 'sys_io' module, which are part of the libkb library. * Added corresponding entries in the 'Config > Settings' menu to change the handler of the keyboard. Supported handlers: Windows, Null. INFO: This keyboard library is *very* experimental and I am aware of some bugs. There will be improvements soon. --- rpcs3/Emu/Io/Keyboard.cpp | 38 +++ rpcs3/Emu/Io/Keyboard.h | 27 ++ rpcs3/Emu/Io/KeyboardHandler.h | 301 ++++++++++++++++++ rpcs3/Emu/Io/Null/NullKeyboardHandler.h | 24 ++ rpcs3/Emu/Io/Windows/WindowsKeyboardHandler.h | 162 ++++++++++ rpcs3/Emu/SysCalls/Modules/cellSysutil.cpp | 28 -- rpcs3/Emu/SysCalls/Modules/sys_io.cpp | 11 + rpcs3/Emu/SysCalls/SysCalls.h | 12 + rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp | 166 ++++++++++ rpcs3/Emu/System.cpp | 1 + rpcs3/Emu/System.h | 3 + rpcs3/Gui/MainFrame.cpp | 14 +- rpcs3/Ini.h | 8 +- rpcs3/rpcs3.vcxproj | 2 + rpcs3/rpcs3.vcxproj.filters | 6 + 15 files changed, 771 insertions(+), 32 deletions(-) create mode 100644 rpcs3/Emu/Io/Keyboard.cpp create mode 100644 rpcs3/Emu/Io/Keyboard.h create mode 100644 rpcs3/Emu/Io/KeyboardHandler.h create mode 100644 rpcs3/Emu/Io/Null/NullKeyboardHandler.h create mode 100644 rpcs3/Emu/Io/Windows/WindowsKeyboardHandler.h create mode 100644 rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp diff --git a/rpcs3/Emu/Io/Keyboard.cpp b/rpcs3/Emu/Io/Keyboard.cpp new file mode 100644 index 0000000000..e20e90e313 --- /dev/null +++ b/rpcs3/Emu/Io/Keyboard.cpp @@ -0,0 +1,38 @@ +#include "stdafx.h" +#include "Keyboard.h" +#include "Null/NullKeyboardHandler.h" +#include "Windows/WindowsKeyboardHandler.h" + +KeyboardManager::KeyboardManager() + : m_keyboard_handler(nullptr) + , m_inited(false) +{ +} + +KeyboardManager::~KeyboardManager() +{ +} + +void KeyboardManager::Init(const u32 max_connect) +{ + if(m_inited) return; + + switch(Ini.KeyboardHandlerMode.GetValue()) + { + case 1: m_keyboard_handler = new WindowsKeyboardHandler(); break; + + default: + case 0: m_keyboard_handler = new NullKeyboardHandler(); break; + } + + m_keyboard_handler->Init(max_connect); + m_inited = true; +} + +void KeyboardManager::Close() +{ + if(m_keyboard_handler) m_keyboard_handler->Close(); + m_keyboard_handler = nullptr; + + m_inited = false; +} \ No newline at end of file diff --git a/rpcs3/Emu/Io/Keyboard.h b/rpcs3/Emu/Io/Keyboard.h new file mode 100644 index 0000000000..5a44408fde --- /dev/null +++ b/rpcs3/Emu/Io/Keyboard.h @@ -0,0 +1,27 @@ +#pragma once + +#include "KeyboardHandler.h" + +class KeyboardManager //: public wxWindow +{ + bool m_inited; + KeyboardHandlerBase* m_keyboard_handler; + +public: + KeyboardManager(); + ~KeyboardManager(); + + void Init(const u32 max_connect); + void Close(); + + Array& GetKeyboards() { return m_keyboard_handler->GetKeyboards(); } + KbInfo& GetInfo() { return m_keyboard_handler->GetInfo(); } + Array& GetButtons(const u32 keyboard) { return m_keyboard_handler->GetButtons(keyboard); } + CellKbData& GetData(const u32 keyboard) { return m_keyboard_handler->GetData(keyboard); } + CellKbConfig& GetConfig(const u32 keyboard) { return m_keyboard_handler->GetConfig(keyboard); } + + bool IsInited() { return m_inited; } + +//private: + //DECLARE_EVENT_TABLE(); +}; \ No newline at end of file diff --git a/rpcs3/Emu/Io/KeyboardHandler.h b/rpcs3/Emu/Io/KeyboardHandler.h new file mode 100644 index 0000000000..44bd2bdfcf --- /dev/null +++ b/rpcs3/Emu/Io/KeyboardHandler.h @@ -0,0 +1,301 @@ +#pragma once + +extern u16 cellKbCnvRawCode(u32 arrange, u32 mkey, u32 led, u16 rawcode); // (TODO: Can it be problematic to place SysCalls in middle of nowhere?) + +enum KbPortStatus +{ + CELL_KB_STATUS_DISCONNECTED = 0x00000000, + CELL_KB_STATUS_CONNECTED = 0x00000001, +}; + +enum CellKbReadMode +{ + CELL_KB_RMODE_INPUTCHAR = 0, + CELL_KB_RMODE_PACKET = 1, +}; + +enum CellKbCodeType +{ + CELL_KB_CODETYPE_RAW = 0, + CELL_KB_CODETYPE_ASCII = 1, +}; + +enum KbMetaKeys +{ + CELL_KB_MKEY_L_CTRL = 0x00000001, + CELL_KB_MKEY_L_SHIFT = 0x00000002, + CELL_KB_MKEY_L_ALT = 0x00000004, + CELL_KB_MKEY_L_WIN = 0x00000008, + CELL_KB_MKEY_R_CTRL = 0x00000010, + CELL_KB_MKEY_R_SHIFT = 0x00000020, + CELL_KB_MKEY_R_ALT = 0x00000040, + CELL_KB_MKEY_R_WIN = 0x00000080, +}; + +enum Keys +{ + // Non-ASCII Raw data + CELL_KEYC_NO_EVENT = 0x00, + CELL_KEYC_E_ROLLOVER = 0x01, + CELL_KEYC_E_POSTFAIL = 0x02, + CELL_KEYC_E_UNDEF = 0x03, + CELL_KEYC_ESCAPE = 0x29, + CELL_KEYC_106_KANJI = 0x35, + CELL_KEYC_CAPS_LOCK = 0x39, + CELL_KEYC_F1 = 0x3a, + CELL_KEYC_F2 = 0x3b, + CELL_KEYC_F3 = 0x3c, + CELL_KEYC_F4 = 0x3d, + CELL_KEYC_F5 = 0x3e, + CELL_KEYC_F6 = 0x3f, + CELL_KEYC_F7 = 0x40, + CELL_KEYC_F8 = 0x41, + CELL_KEYC_F9 = 0x42, + CELL_KEYC_F10 = 0x43, + CELL_KEYC_F11 = 0x44, + CELL_KEYC_F12 = 0x45, + CELL_KEYC_PRINTSCREEN = 0x46, + CELL_KEYC_SCROLL_LOCK = 0x47, + CELL_KEYC_PAUSE = 0x48, + CELL_KEYC_INSERT = 0x49, + CELL_KEYC_HOME = 0x4a, + CELL_KEYC_PAGE_UP = 0x4b, + CELL_KEYC_DELETE = 0x4c, + CELL_KEYC_END = 0x4d, + CELL_KEYC_PAGE_DOWN = 0x4e, + CELL_KEYC_RIGHT_ARROW = 0x4f, + CELL_KEYC_LEFT_ARROW = 0x50, + CELL_KEYC_DOWN_ARROW = 0x51, + CELL_KEYC_UP_ARROW = 0x52, + CELL_KEYC_NUM_LOCK = 0x53, + CELL_KEYC_APPLICATION = 0x65, + CELL_KEYC_KANA = 0x88, + CELL_KEYC_HENKAN = 0x8a, + CELL_KEYC_MUHENKAN = 0x8b, + + // Raw keycodes for ASCII keys + CELL_KEYC_A = 0x04, + CELL_KEYC_B = 0x05, + CELL_KEYC_C = 0x06, + CELL_KEYC_D = 0x07, + CELL_KEYC_E = 0x08, + CELL_KEYC_F = 0x09, + CELL_KEYC_G = 0x0A, + CELL_KEYC_H = 0x0B, + CELL_KEYC_I = 0x0C, + CELL_KEYC_J = 0x0D, + CELL_KEYC_K = 0x0E, + CELL_KEYC_L = 0x0F, + CELL_KEYC_M = 0x10, + CELL_KEYC_N = 0x11, + CELL_KEYC_O = 0x12, + CELL_KEYC_P = 0x13, + CELL_KEYC_Q = 0x14, + CELL_KEYC_R = 0x15, + CELL_KEYC_S = 0x16, + CELL_KEYC_T = 0x17, + CELL_KEYC_U = 0x18, + CELL_KEYC_V = 0x19, + CELL_KEYC_W = 0x1A, + CELL_KEYC_X = 0x1B, + CELL_KEYC_Y = 0x1C, + CELL_KEYC_Z = 0x1D, + CELL_KEYC_1 = 0x1E, + CELL_KEYC_2 = 0x1F, + CELL_KEYC_3 = 0x20, + CELL_KEYC_4 = 0x21, + CELL_KEYC_5 = 0x22, + CELL_KEYC_6 = 0x23, + CELL_KEYC_7 = 0x24, + CELL_KEYC_8 = 0x25, + CELL_KEYC_9 = 0x26, + CELL_KEYC_0 = 0x27, + CELL_KEYC_ENTER = 0x28, + CELL_KEYC_ESC = 0x29, + CELL_KEYC_BS = 0x2A, + CELL_KEYC_TAB = 0x2B, + CELL_KEYC_SPACE = 0x2C, + CELL_KEYC_MINUS = 0x2D, + CELL_KEYC_EQUAL_101 = 0x2E, + CELL_KEYC_ACCENT_CIRCONFLEX_106 = 0x2E, + CELL_KEYC_LEFT_BRACKET_101 = 0x2F, + CELL_KEYC_ATMARK_106 = 0x2F, + CELL_KEYC_RIGHT_BRACKET_101 = 0x30, + CELL_KEYC_LEFT_BRACKET_106 = 0x30, + CELL_KEYC_BACKSLASH_101 = 0x31, + CELL_KEYC_RIGHT_BRACKET_106 = 0x32, + CELL_KEYC_SEMICOLON = 0x33, + CELL_KEYC_QUOTATION_101 = 0x34, + CELL_KEYC_COLON_106 = 0x34, + CELL_KEYC_COMMA = 0x36, + CELL_KEYC_PERIOD = 0x37, + CELL_KEYC_SLASH = 0x38, + //CELL_KEYC_CAPS_LOCK = 0x39, + CELL_KEYC_KPAD_NUMLOCK = 0x53, + CELL_KEYC_KPAD_SLASH = 0x54, + CELL_KEYC_KPAD_ASTERISK = 0x55, + CELL_KEYC_KPAD_MINUS = 0x56, + CELL_KEYC_KPAD_PLUS = 0x57, + CELL_KEYC_KPAD_ENTER = 0x58, + CELL_KEYC_KPAD_1 = 0x59, + CELL_KEYC_KPAD_2 = 0x5A, + CELL_KEYC_KPAD_3 = 0x5B, + CELL_KEYC_KPAD_4 = 0x5C, + CELL_KEYC_KPAD_5 = 0x5D, + CELL_KEYC_KPAD_6 = 0x5E, + CELL_KEYC_KPAD_7 = 0x5F, + CELL_KEYC_KPAD_8 = 0x60, + CELL_KEYC_KPAD_9 = 0x61, + CELL_KEYC_KPAD_0 = 0x62, + CELL_KEYC_KPAD_PERIOD = 0x63, + CELL_KEYC_BACKSLASH_106 = 0x87, + CELL_KEYC_YEN_106 = 0x89, +}; + +enum CellKbMappingType +{ + CELL_KB_MAPPING_101, + CELL_KB_MAPPING_106, + CELL_KB_MAPPING_106_KANA, + CELL_KB_MAPPING_GERMAN_GERMANY, + CELL_KB_MAPPING_SPANISH_SPAIN, + CELL_KB_MAPPING_FRENCH_FRANCE, + CELL_KB_MAPPING_ITALIAN_ITALY, + CELL_KB_MAPPING_DUTCH_NETHERLANDS, + CELL_KB_MAPPING_PORTUGUESE_PORTUGAL, + CELL_KB_MAPPING_RUSSIAN_RUSSIA, + CELL_KB_MAPPING_ENGLISH_UK, + CELL_KB_MAPPING_KOREAN_KOREA, + CELL_KB_MAPPING_NORWEGIAN_NORWAY, + CELL_KB_MAPPING_FINNISH_FINLAND, + CELL_KB_MAPPING_DANISH_DENMARK, + CELL_KB_MAPPING_SWEDISH_SWEDEN, + CELL_KB_MAPPING_CHINESE_TRADITIONAL, + CELL_KB_MAPPING_CHINESE_SIMPLIFIED, + CELL_KB_MAPPING_SWISS_FRENCH_SWITZERLAND, + CELL_KB_MAPPING_SWISS_GERMAN_SWITZERLAND, + CELL_KB_MAPPING_CANADIAN_FRENCH_CANADA, + CELL_KB_MAPPING_BELGIAN_BELGIUM, + CELL_KB_MAPPING_POLISH_POLAND, + CELL_KB_MAPPING_PORTUGUESE_BRAZIL, +}; + +static const u32 CELL_KB_MAX_KEYBOARDS = 127; +static const u32 CELL_KB_MAX_PORT_NUM = 7; +static const u32 CELL_KB_MAX_KEYCODES = 62; + +struct KbInfo +{ + u32 max_connect; + u32 now_connect; + u32 info; + u8 status[CELL_KB_MAX_KEYBOARDS]; +}; + +struct CellKbData +{ + u32 led; + u32 mkey; + s32 len; + u16 keycode[CELL_KB_MAX_KEYCODES]; + + CellKbData() + : led(0) + , mkey(0) + , len(0) + { + } +}; + +struct CellKbConfig +{ + u32 arrange; + u32 read_mode; + u32 code_type; + + CellKbConfig() + : arrange(CELL_KB_MAPPING_106) + , read_mode(CELL_KB_RMODE_INPUTCHAR) + , code_type(CELL_KB_CODETYPE_ASCII) + { + } +}; + +struct KbButton +{ + u32 m_keyCode; + u32 m_outKeyCode; + bool m_pressed; + //bool m_flush; + + KbButton(u32 keyCode, u32 outKeyCode) + : m_pressed(false) + //, m_flush(false) + , m_keyCode(keyCode) + , m_outKeyCode(outKeyCode) + { + } +}; + +struct Keyboard +{ + CellKbData m_data; + CellKbConfig m_config; + Array m_buttons; + + Keyboard() + : m_data() + , m_config() + { + } + + ~Keyboard() { m_buttons.Clear(); } +}; + +class KeyboardHandlerBase +{ +protected: + KbInfo m_info; + Array m_keyboards; + +public: + virtual void Init(const u32 max_connect)=0; + virtual void Close()=0; + + void Key(const u32 code, bool pressed) + { + for(u64 p=0; p& GetKeyboards() { return m_keyboards; } + Array& GetButtons(const u32 keyboard) { return GetKeyboards()[keyboard].m_buttons; } + CellKbData& GetData(const u32 keyboard) { return GetKeyboards()[keyboard].m_data; } + CellKbConfig& GetConfig(const u32 keyboard) { return GetKeyboards()[keyboard].m_config; } +}; \ No newline at end of file diff --git a/rpcs3/Emu/Io/Null/NullKeyboardHandler.h b/rpcs3/Emu/Io/Null/NullKeyboardHandler.h new file mode 100644 index 0000000000..d5693a78cc --- /dev/null +++ b/rpcs3/Emu/Io/Null/NullKeyboardHandler.h @@ -0,0 +1,24 @@ +#pragma once + +#include "Emu/Io/KeyboardHandler.h" + +class NullKeyboardHandler : public KeyboardHandlerBase +{ +public: + NullKeyboardHandler() + { + } + + virtual void Init(const u32 max_connect) + { + memset(&m_info, 0, sizeof(KbInfo)); + m_info.max_connect = max_connect; + m_keyboards.Clear(); + } + + virtual void Close() + { + memset(&m_info, 0, sizeof(KbInfo)); + m_keyboards.Clear(); + } +}; \ No newline at end of file diff --git a/rpcs3/Emu/Io/Windows/WindowsKeyboardHandler.h b/rpcs3/Emu/Io/Windows/WindowsKeyboardHandler.h new file mode 100644 index 0000000000..020dcd96fc --- /dev/null +++ b/rpcs3/Emu/Io/Windows/WindowsKeyboardHandler.h @@ -0,0 +1,162 @@ +#pragma once + +#include "Emu/Io/KeyboardHandler.h" + +class WindowsKeyboardHandler + : public wxWindow + , public KeyboardHandlerBase +{ + AppConnector m_app_connector; + +public: + WindowsKeyboardHandler() : wxWindow() + { + m_app_connector.Connect(wxEVT_KEY_DOWN, wxKeyEventHandler(WindowsKeyboardHandler::KeyDown), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_KEY_UP, wxKeyEventHandler(WindowsKeyboardHandler::KeyUp), (wxObject*)0, this); + } + + virtual void KeyDown(wxKeyEvent& event) { Key(event.GetKeyCode(), 1); event.Skip(); } + virtual void KeyUp(wxKeyEvent& event) { Key(event.GetKeyCode(), 0); event.Skip(); } + + virtual void Init(const u32 max_connect) + { + for(u32 i=0; i(GetKeyboards().GetCount(), max_connect); + m_info.info = 0; // Ownership of keyboard data: 0=Application, 1=System + m_info.status[0] = CELL_KB_STATUS_CONNECTED; // (TODO: Support for more keyboards) + } + + virtual void Close() + { + memset(&m_info, 0, sizeof(KbInfo)); + m_keyboards.Clear(); + } + + void LoadSettings() + { + // CELL_KB_RAWDAT + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_NO_EVENT)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_E_ROLLOVER)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_E_POSTFAIL)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_E_UNDEF)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_ESCAPE, CELL_KEYC_ESCAPE)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_106_KANJI)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_CAPS_LOCK)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F1, CELL_KEYC_F1)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F2, CELL_KEYC_F2)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F3, CELL_KEYC_F3)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F4, CELL_KEYC_F4)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F5, CELL_KEYC_F5)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F6, CELL_KEYC_F6)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F7, CELL_KEYC_F7)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F8, CELL_KEYC_F8)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F9, CELL_KEYC_F9)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F10, CELL_KEYC_F10)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F11, CELL_KEYC_F11)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_F12, CELL_KEYC_F12)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_PRINT, CELL_KEYC_PRINTSCREEN)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_SCROLL, CELL_KEYC_SCROLL_LOCK)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_PAUSE, CELL_KEYC_PAUSE)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_INSERT, CELL_KEYC_INSERT)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_HOME)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_PAGEUP, CELL_KEYC_PAGE_UP)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_DELETE, CELL_KEYC_DELETE)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_END, CELL_KEYC_END)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_PAGEDOWN, CELL_KEYC_PAGE_DOWN)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_RIGHT, CELL_KEYC_RIGHT_ARROW)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_LEFT, CELL_KEYC_LEFT_ARROW)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_DOWN, CELL_KEYC_DOWN_ARROW)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_UP, CELL_KEYC_UP_ARROW)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMLOCK, CELL_KEYC_NUM_LOCK)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_APPLICATION)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_KANA)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_HENKAN)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_MUHENKAN)); + + // CELL_KB_KEYPAD + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMLOCK, CELL_KEYC_KPAD_NUMLOCK)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD_DIVIDE, CELL_KEYC_KPAD_SLASH)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD_MULTIPLY, CELL_KEYC_KPAD_ASTERISK)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD_SUBTRACT, CELL_KEYC_KPAD_MINUS)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD_ADD, CELL_KEYC_KPAD_PLUS)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD_ENTER, CELL_KEYC_KPAD_ENTER)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD1, CELL_KEYC_KPAD_1)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD2, CELL_KEYC_KPAD_2)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD3, CELL_KEYC_KPAD_3)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD4, CELL_KEYC_KPAD_4)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD5, CELL_KEYC_KPAD_5)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD6, CELL_KEYC_KPAD_6)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD7, CELL_KEYC_KPAD_7)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD8, CELL_KEYC_KPAD_8)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD9, CELL_KEYC_KPAD_9)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD0, CELL_KEYC_KPAD_0)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_NUMPAD_DELETE, CELL_KEYC_KPAD_PERIOD)); + + // ASCII Printable characters + m_keyboards[0].m_buttons.Move(new KbButton('A', CELL_KEYC_A)); + m_keyboards[0].m_buttons.Move(new KbButton('B', CELL_KEYC_B)); + m_keyboards[0].m_buttons.Move(new KbButton('C', CELL_KEYC_C)); + m_keyboards[0].m_buttons.Move(new KbButton('D', CELL_KEYC_D)); + m_keyboards[0].m_buttons.Move(new KbButton('E', CELL_KEYC_E)); + m_keyboards[0].m_buttons.Move(new KbButton('F', CELL_KEYC_F)); + m_keyboards[0].m_buttons.Move(new KbButton('G', CELL_KEYC_G)); + m_keyboards[0].m_buttons.Move(new KbButton('H', CELL_KEYC_H)); + m_keyboards[0].m_buttons.Move(new KbButton('I', CELL_KEYC_I)); + m_keyboards[0].m_buttons.Move(new KbButton('J', CELL_KEYC_J)); + m_keyboards[0].m_buttons.Move(new KbButton('K', CELL_KEYC_K)); + m_keyboards[0].m_buttons.Move(new KbButton('L', CELL_KEYC_L)); + m_keyboards[0].m_buttons.Move(new KbButton('M', CELL_KEYC_M)); + m_keyboards[0].m_buttons.Move(new KbButton('N', CELL_KEYC_N)); + m_keyboards[0].m_buttons.Move(new KbButton('O', CELL_KEYC_O)); + m_keyboards[0].m_buttons.Move(new KbButton('P', CELL_KEYC_P)); + m_keyboards[0].m_buttons.Move(new KbButton('Q', CELL_KEYC_Q)); + m_keyboards[0].m_buttons.Move(new KbButton('R', CELL_KEYC_R)); + m_keyboards[0].m_buttons.Move(new KbButton('S', CELL_KEYC_S)); + m_keyboards[0].m_buttons.Move(new KbButton('T', CELL_KEYC_T)); + m_keyboards[0].m_buttons.Move(new KbButton('U', CELL_KEYC_U)); + m_keyboards[0].m_buttons.Move(new KbButton('V', CELL_KEYC_V)); + m_keyboards[0].m_buttons.Move(new KbButton('W', CELL_KEYC_W)); + m_keyboards[0].m_buttons.Move(new KbButton('X', CELL_KEYC_X)); + m_keyboards[0].m_buttons.Move(new KbButton('Y', CELL_KEYC_Y)); + m_keyboards[0].m_buttons.Move(new KbButton('Z', CELL_KEYC_Z)); + + m_keyboards[0].m_buttons.Move(new KbButton('1', CELL_KEYC_1)); + m_keyboards[0].m_buttons.Move(new KbButton('2', CELL_KEYC_2)); + m_keyboards[0].m_buttons.Move(new KbButton('3', CELL_KEYC_3)); + m_keyboards[0].m_buttons.Move(new KbButton('4', CELL_KEYC_4)); + m_keyboards[0].m_buttons.Move(new KbButton('5', CELL_KEYC_5)); + m_keyboards[0].m_buttons.Move(new KbButton('6', CELL_KEYC_6)); + m_keyboards[0].m_buttons.Move(new KbButton('7', CELL_KEYC_7)); + m_keyboards[0].m_buttons.Move(new KbButton('8', CELL_KEYC_8)); + m_keyboards[0].m_buttons.Move(new KbButton('9', CELL_KEYC_9)); + m_keyboards[0].m_buttons.Move(new KbButton('0', CELL_KEYC_0)); + + m_keyboards[0].m_buttons.Move(new KbButton(WXK_RETURN, CELL_KEYC_ENTER)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_ESC)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_TAB, CELL_KEYC_TAB)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_SPACE, CELL_KEYC_SPACE)); + m_keyboards[0].m_buttons.Move(new KbButton(WXK_SUBTRACT, CELL_KEYC_MINUS)); + m_keyboards[0].m_buttons.Move(new KbButton('=', CELL_KEYC_EQUAL_101)); + m_keyboards[0].m_buttons.Move(new KbButton('^', CELL_KEYC_ACCENT_CIRCONFLEX_106)); + //m_keyboards[0].m_buttons.Move(new KbButton('(', CELL_KEYC_LEFT_BRACKET_101)); + m_keyboards[0].m_buttons.Move(new KbButton('@', CELL_KEYC_ATMARK_106)); + //m_keyboards[0].m_buttons.Move(new KbButton(')', CELL_KEYC_RIGHT_BRACKET_101)); + m_keyboards[0].m_buttons.Move(new KbButton('(', CELL_KEYC_LEFT_BRACKET_106)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_BACKSLASH_101)); + m_keyboards[0].m_buttons.Move(new KbButton('(', CELL_KEYC_RIGHT_BRACKET_106)); + m_keyboards[0].m_buttons.Move(new KbButton(';', CELL_KEYC_SEMICOLON)); + m_keyboards[0].m_buttons.Move(new KbButton('"', CELL_KEYC_QUOTATION_101)); + m_keyboards[0].m_buttons.Move(new KbButton(':', CELL_KEYC_COLON_106)); + m_keyboards[0].m_buttons.Move(new KbButton(',', CELL_KEYC_COMMA)); + m_keyboards[0].m_buttons.Move(new KbButton('.', CELL_KEYC_PERIOD)); + m_keyboards[0].m_buttons.Move(new KbButton('/', CELL_KEYC_SLASH)); + m_keyboards[0].m_buttons.Move(new KbButton('\\', CELL_KEYC_BACKSLASH_106)); + //m_keyboards[0].m_buttons.Move(new KbButton(, CELL_KEYC_YEN_106)); + } +}; \ No newline at end of file diff --git a/rpcs3/Emu/SysCalls/Modules/cellSysutil.cpp b/rpcs3/Emu/SysCalls/Modules/cellSysutil.cpp index 1602f4fc52..643879ccdd 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellSysutil.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellSysutil.cpp @@ -99,34 +99,6 @@ enum CELL_SYSUTIL_PAD_RUMBLE_ON = 1, }; -enum -{ - CELL_KB_MAPPING_101, - CELL_KB_MAPPING_106, - CELL_KB_MAPPING_106_KANA, - CELL_KB_MAPPING_GERMAN_GERMANY, - CELL_KB_MAPPING_SPANISH_SPAIN, - CELL_KB_MAPPING_FRENCH_FRANCE, - CELL_KB_MAPPING_ITALIAN_ITALY, - CELL_KB_MAPPING_DUTCH_NETHERLANDS, - CELL_KB_MAPPING_PORTUGUESE_PORTUGAL, - CELL_KB_MAPPING_RUSSIAN_RUSSIA, - CELL_KB_MAPPING_ENGLISH_UK, - CELL_KB_MAPPING_KOREAN_KOREA, - CELL_KB_MAPPING_NORWEGIAN_NORWAY, - CELL_KB_MAPPING_FINNISH_FINLAND, - CELL_KB_MAPPING_DANISH_DENMARK, - CELL_KB_MAPPING_SWEDISH_SWEDEN, - CELL_KB_MAPPING_CHINESE_TRADITIONAL, - CELL_KB_MAPPING_CHINESE_SIMPLIFIED, - CELL_KB_MAPPING_SWISS_FRENCH_SWITZERLAND, - CELL_KB_MAPPING_SWISS_GERMAN_SWITZERLAND, - CELL_KB_MAPPING_CANADIAN_FRENCH_CANADA, - CELL_KB_MAPPING_BELGIAN_BELGIUM, - CELL_KB_MAPPING_POLISH_POLAND, - CELL_KB_MAPPING_PORTUGUESE_BRAZIL, -}; - void cellSysutil_init(); Module cellSysutil(0x0015, cellSysutil_init); diff --git a/rpcs3/Emu/SysCalls/Modules/sys_io.cpp b/rpcs3/Emu/SysCalls/Modules/sys_io.cpp index 385f5c99d1..4339052ecf 100644 --- a/rpcs3/Emu/SysCalls/Modules/sys_io.cpp +++ b/rpcs3/Emu/SysCalls/Modules/sys_io.cpp @@ -15,4 +15,15 @@ void sys_io_init() sys_io.AddFunc(0xf65544ee, cellPadSetActDirect); sys_io.AddFunc(0xa703a51d, cellPadGetInfo2); sys_io.AddFunc(0x578e3c98, cellPadSetPortSetting); + + sys_io.AddFunc(0x433f6ec0, cellKbInit); + sys_io.AddFunc(0xbfce3285, cellKbEnd); + sys_io.AddFunc(0x2073b7f6, cellKbClearBuf); + sys_io.AddFunc(0x4ab1fa77, cellKbCnvRawCode); + sys_io.AddFunc(0x2f1774d5, cellKbGetInfo); + sys_io.AddFunc(0xff0a21b7, cellKbRead); + sys_io.AddFunc(0xa5f85e4d, cellKbSetCodeType); + sys_io.AddFunc(0x3f72c56e, cellKbSetLEDStatus); + sys_io.AddFunc(0xdeefdfa7, cellKbSetReadMode); + sys_io.AddFunc(0x1f71ecbe, cellKbGetConfiguration); } diff --git a/rpcs3/Emu/SysCalls/SysCalls.h b/rpcs3/Emu/SysCalls/SysCalls.h index edbb9a6dae..68e12d882a 100644 --- a/rpcs3/Emu/SysCalls/SysCalls.h +++ b/rpcs3/Emu/SysCalls/SysCalls.h @@ -216,6 +216,18 @@ extern int cellPadSetActDirect(u32 port_no, u32 param_addr); extern int cellPadGetInfo2(u32 info_addr); extern int cellPadSetPortSetting(u32 port_no, u32 port_setting); +//cellKb +extern int cellKbInit(u32 max_connect); +extern int cellKbEnd(); +extern int cellKbClearBuf(u32 port_no); +extern u16 cellKbCnvRawCode(u32 arrange, u32 mkey, u32 led, u16 rawcode); +extern int cellKbGetInfo(u32 info_addr); +extern int cellKbRead(u32 port_no, u32 data_addr); +extern int cellKbSetCodeType(u32 port_no, u32 type); +extern int cellKbSetLEDStatus(u32 port_no, u8 led); +extern int cellKbSetReadMode(u32 port_no, u32 rmode); +extern int cellKbGetConfiguration(u32 port_no, u32 config_addr); + //cellGcm extern int cellGcmCallback(u32 context_addr, u32 count); diff --git a/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp b/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp new file mode 100644 index 0000000000..90073fe946 --- /dev/null +++ b/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp @@ -0,0 +1,166 @@ +#include "stdafx.h" +#include "Emu/Io/Keyboard.h" +#include "Emu/SysCalls/SysCalls.h" + +extern Module sys_io; + +enum CELL_KB_ERROR_CODE +{ + CELL_KB_ERROR_FATAL = 0x80121001, + CELL_KB_ERROR_INVALID_PARAMETER = 0x80121002, + CELL_KB_ERROR_ALREADY_INITIALIZED = 0x80121003, + CELL_KB_ERROR_UNINITIALIZED = 0x80121004, + CELL_KB_ERROR_RESOURCE_ALLOCATION_FAILED = 0x80121005, + CELL_KB_ERROR_READ_FAILED = 0x80121006, + CELL_KB_ERROR_NO_DEVICE = 0x80121007, + CELL_KB_ERROR_SYS_SETTING_FAILED = 0x80121008, +}; + +int cellKbInit(u32 max_connect) +{ + sys_io.Log("cellKbInit(max_connect=%d)", max_connect); + if(Emu.GetKeyboardManager().IsInited()) return CELL_KB_ERROR_ALREADY_INITIALIZED; + if(max_connect > 7) return CELL_KB_ERROR_INVALID_PARAMETER; + + Emu.GetKeyboardManager().Init(max_connect); + return CELL_OK; +} + +int cellKbEnd() +{ + sys_io.Log("cellKbEnd()"); + if(!Emu.GetKeyboardManager().IsInited()) return CELL_KB_ERROR_UNINITIALIZED; + Emu.GetKeyboardManager().Close(); + return CELL_OK; +} + +int cellKbClearBuf(u32 port_no) +{ + sys_io.Log("cellKbClearBuf(port_no=%d)", port_no); + if(!Emu.GetKeyboardManager().IsInited()) return CELL_KB_ERROR_UNINITIALIZED; + if(port_no >= Emu.GetKeyboardManager().GetKeyboards().GetCount()) return CELL_KB_ERROR_INVALID_PARAMETER; + + //? + + return CELL_OK; +} + +u16 cellKbCnvRawCode(u32 arrange, u32 mkey, u32 led, u16 rawcode) +{ + sys_io.Log("cellKbCnvRawCode(arrange=%d,mkey=%d,led=%d,rawcode=%d)", arrange, mkey, led, rawcode); + + // CELL_KB_RAWDAT + if ((rawcode >= 0x00 && rawcode <= 0x03) || rawcode == 0x29 || rawcode == 0x35 || + (rawcode >= 0x39 && rawcode <= 0x53) || rawcode == 0x65 || rawcode == 0x88 || + rawcode == 0x8A || rawcode == 0x8B) + { + return rawcode | 0x8000; + } + + // CELL_KB_NUMPAD + if (rawcode >= 0x59 && rawcode <= 0x61) return (rawcode - 0x28) | 0x4000; // '1' - '9' + if (rawcode == 0x62) return 0x30 | 0x4000; // '0' + if (rawcode == 0x53) return 0x00 | 0x4000; // 'Num Lock' + if (rawcode == 0x54) return 0x2F | 0x4000; // '/' + if (rawcode == 0x55) return 0x2A | 0x4000; // '*' + if (rawcode == 0x56) return 0x2D | 0x4000; // '-' + if (rawcode == 0x57) return 0x2B | 0x4000; // '+' + if (rawcode == 0x58) return 0x0A | 0x4000; // '\n' + + // ASCII + if (rawcode >= 0x04 && rawcode <= 0x1D) return rawcode + 0x3D; // 'A' - 'Z' + if (rawcode >= 0x1E && rawcode <= 0x26) return rawcode + 0x13; // '1' - '9' + if (rawcode == 0x27) return 0x30; // '0' + if (rawcode == 0x28) return 0x0A; // '\n' + if (rawcode == 0x2B) return 0x09; // '\t' + if (rawcode == 0x2C) return 0x20; // ' ' + if (rawcode == 0x2D) return 0x2D; // '-' + if (rawcode == 0x2E) return 0x3D; // '=' + + // (TODO: Add more cases) + + return 0x0000; +} + +int cellKbGetInfo(u32 info_addr) +{ + sys_io.Log("cellKbGetInfo(info_addr=0x%x)", info_addr); + if(!Emu.GetKeyboardManager().IsInited()) return CELL_KB_ERROR_UNINITIALIZED; + + const KbInfo& current_info = Emu.GetKeyboardManager().GetInfo(); + mem_class_t info(info_addr); + info += current_info.max_connect; + info += current_info.now_connect; + info += current_info.info; + for(u32 i=0; i& keyboards = Emu.GetKeyboardManager().GetKeyboards(); + if(!Emu.GetKeyboardManager().IsInited()) return CELL_KB_ERROR_UNINITIALIZED; + if(port_no >= keyboards.GetCount()) return CELL_KB_ERROR_INVALID_PARAMETER; + + CellKbData& current_data = Emu.GetKeyboardManager().GetData(port_no); + + mem_class_t data(data_addr); + data += current_data.led; + data += current_data.mkey; + data += min((u32)current_data.len, CELL_KB_MAX_KEYCODES); + for(s32 i=0; iAppend("Windows"); //cbox_pad_handler->Append("DirectInput"); + cbox_keyboard_handler->Append("Null"); + cbox_keyboard_handler->Append("Windows"); + //cbox_pad_handler->Append("DirectInput"); + chbox_gs_vsync->SetValue(Ini.GSVSyncEnable.GetValue()); cbox_cpu_decoder->SetSelection(Ini.CPUDecoderMode.GetValue() ? Ini.CPUDecoderMode.GetValue() - 1 : 0); @@ -322,6 +328,7 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) cbox_gs_resolution->SetSelection(ResolutionIdToNum(Ini.GSResolution.GetValue()) - 1); cbox_gs_aspect->SetSelection(Ini.GSAspectRatio.GetValue() - 1); cbox_pad_handler->SetSelection(Ini.PadHandlerMode.GetValue()); + cbox_keyboard_handler->SetSelection(Ini.KeyboardHandlerMode.GetValue()); s_round_cpu_decoder->Add(cbox_cpu_decoder, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_cpu->Add(s_round_cpu_decoder, wxSizerFlags().Border(wxALL, 5).Expand()); @@ -335,7 +342,9 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) s_round_gs->Add(chbox_gs_vsync, wxSizerFlags().Border(wxALL, 5)); s_round_pad_handler->Add(cbox_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); + s_round_keyboard_handler->Add(cbox_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_pad->Add(s_round_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); + s_round_pad->Add(s_round_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); wxBoxSizer* s_b_panel(new wxBoxSizer(wxHORIZONTAL)); @@ -359,6 +368,7 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) Ini.GSAspectRatio.SetValue(cbox_gs_aspect->GetSelection() + 1); Ini.GSVSyncEnable.SetValue(chbox_gs_vsync->GetValue()); Ini.PadHandlerMode.SetValue(cbox_pad_handler->GetSelection()); + Ini.KeyboardHandlerMode.SetValue(cbox_keyboard_handler->GetSelection()); Ini.Save(); } diff --git a/rpcs3/Ini.h b/rpcs3/Ini.h index 829db83c46..e9b2c92652 100644 --- a/rpcs3/Ini.h +++ b/rpcs3/Ini.h @@ -98,6 +98,7 @@ public: IniEntry GSAspectRatio; IniEntry GSVSyncEnable; IniEntry PadHandlerMode; + IniEntry KeyboardHandlerMode; public: Inis() : DefPath("EmuSettings") @@ -113,8 +114,9 @@ public: GSAspectRatio.Init("AspectRatio", path); GSVSyncEnable.Init("VSyncEnable", path); - path = DefPath + "\\" + "Pad"; - PadHandlerMode.Init("HandlerMode", path); + path = DefPath + "\\" + "IO"; + PadHandlerMode.Init("PadHandlerMode", path); + KeyboardHandlerMode.Init("KeyboardHandlerMode", path); } void Load() @@ -125,6 +127,7 @@ public: GSAspectRatio.Load(1); GSVSyncEnable.Load(false); PadHandlerMode.Load(0); + KeyboardHandlerMode.Load(0); } void Save() @@ -135,6 +138,7 @@ public: GSAspectRatio.Save(); GSVSyncEnable.Save(); PadHandlerMode.Save(); + KeyboardHandlerMode.Save(); } }; diff --git a/rpcs3/rpcs3.vcxproj b/rpcs3/rpcs3.vcxproj index fe1af9b6ff..637e88a80a 100644 --- a/rpcs3/rpcs3.vcxproj +++ b/rpcs3/rpcs3.vcxproj @@ -221,6 +221,7 @@ + @@ -230,6 +231,7 @@ + diff --git a/rpcs3/rpcs3.vcxproj.filters b/rpcs3/rpcs3.vcxproj.filters index 09c648cf32..7a15323230 100644 --- a/rpcs3/rpcs3.vcxproj.filters +++ b/rpcs3/rpcs3.vcxproj.filters @@ -295,6 +295,12 @@ Emu\SysCalls\Modules + + Emu\Io + + + Emu\SysCalls\lv2 + From 1b7302c0bafc1b6bd87fb602edb93374efe77b53 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Fri, 13 Sep 2013 19:29:54 +0200 Subject: [PATCH 5/7] Keybord support improved *Improved Keyboard: - Fixed crash when using Null keyboard handler - Added support for meta keys - Added support for {caps, num, scroll} lock. *Small issues fixed in previous module: - cellGifDec: Fixed wrong information of packet field in cellGifDecReadHeader. - cellGifDec: Set DataOutInfo.recordType as 1 since the output is always an image. - cellJpgDec: Set colorSpace as 3, until a better function is implemented. - cellJpgDec, cellPngDec: Added RGBA support. *Changed layout of "Config > Settings" menu. --- rpcs3/Emu/Io/KeyboardHandler.h | 56 +++++++-- rpcs3/Emu/Io/Null/NullKeyboardHandler.h | 4 + rpcs3/Emu/Io/Windows/WindowsKeyboardHandler.h | 15 ++- rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp | 7 +- rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp | 112 ++++++++++++++---- rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp | 107 +++++++++++++---- rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp | 13 +- rpcs3/Gui/MainFrame.cpp | 22 ++-- 8 files changed, 262 insertions(+), 74 deletions(-) diff --git a/rpcs3/Emu/Io/KeyboardHandler.h b/rpcs3/Emu/Io/KeyboardHandler.h index 44bd2bdfcf..463ae9516a 100644 --- a/rpcs3/Emu/Io/KeyboardHandler.h +++ b/rpcs3/Emu/Io/KeyboardHandler.h @@ -16,8 +16,17 @@ enum CellKbReadMode enum CellKbCodeType { - CELL_KB_CODETYPE_RAW = 0, - CELL_KB_CODETYPE_ASCII = 1, + CELL_KB_CODETYPE_RAW = 0, + CELL_KB_CODETYPE_ASCII = 1, +}; + +enum KbLedCodes +{ + CELL_KB_LED_NUM_LOCK = 0x00000001, + CELL_KB_LED_CAPS_LOCK = 0x00000002, + CELL_KB_LED_SCROLL_LOCK = 0x00000004, + CELL_KB_LED_COMPOSE = 0x00000008, + CELL_KB_LED_KANA = 0x00000016, }; enum KbMetaKeys @@ -203,7 +212,7 @@ struct CellKbData : led(0) , mkey(0) , len(0) - { + { // (TODO: Set initial state of led) } }; @@ -226,11 +235,9 @@ struct KbButton u32 m_keyCode; u32 m_outKeyCode; bool m_pressed; - //bool m_flush; KbButton(u32 keyCode, u32 outKeyCode) : m_pressed(false) - //, m_flush(false) , m_keyCode(keyCode) , m_outKeyCode(outKeyCode) { @@ -276,17 +283,40 @@ public: if (pressed) { - u16 kcode; - if (config.code_type == CELL_KB_CODETYPE_RAW) - { - kcode = button.m_outKeyCode; + + if (code == 308 || code == 307 || code == 306 || + code == 393 || code == 396 || code == 394) + { // Meta Keys + data.mkey |= button.m_outKeyCode; } - else //config.code_type == CELL_KB_CODETYPE_ASCII + else { - kcode = cellKbCnvRawCode(config.arrange, data.mkey, data.led, button.m_outKeyCode); + // Led Keys + if (code == 364) data.led ^= CELL_KB_LED_NUM_LOCK; + if (code == 311) data.led ^= CELL_KB_LED_CAPS_LOCK; + if (code == 365) data.led ^= CELL_KB_LED_SCROLL_LOCK; + + u16 kcode; + if (config.code_type == CELL_KB_CODETYPE_RAW) + { + kcode = button.m_outKeyCode; + } + else //config.code_type == CELL_KB_CODETYPE_ASCII + { + kcode = cellKbCnvRawCode(config.arrange, data.mkey, data.led, button.m_outKeyCode); + } + data.keycode[data.len % CELL_KB_MAX_KEYCODES] = kcode; + data.len++; + } + } + + if (!pressed) + { + if (code == 308 || code == 307 || code == 306 || + code == 393 || code == 396 || code == 394) + { // Meta Keys + data.mkey &= ~button.m_outKeyCode; } - data.keycode[data.len % CELL_KB_MAX_KEYCODES] = kcode; - data.len++; } } diff --git a/rpcs3/Emu/Io/Null/NullKeyboardHandler.h b/rpcs3/Emu/Io/Null/NullKeyboardHandler.h index d5693a78cc..4a9598a71e 100644 --- a/rpcs3/Emu/Io/Null/NullKeyboardHandler.h +++ b/rpcs3/Emu/Io/Null/NullKeyboardHandler.h @@ -14,6 +14,10 @@ public: memset(&m_info, 0, sizeof(KbInfo)); m_info.max_connect = max_connect; m_keyboards.Clear(); + for(u32 i=0; i> 7; - current_info.SColorResolution = (packedField >> 4) & 7; + current_info.SColorResolution = ((packedField >> 4) & 7)+1; current_info.SSortFlag = (packedField >> 3) & 1; - current_info.SSizeOfGlobalColorTable = packedField & 7; + current_info.SSizeOfGlobalColorTable = (packedField & 7)+1; current_info.SBackGroundColor = Memory.Read8(buffer+11); current_info.SPixelAspectRatio = Memory.Read8(buffer+12); @@ -171,6 +171,9 @@ int cellGifDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC } Memory.Free(buffer); + //The output data is an image (dataOutInfo.recordType = 1) + Memory.Write32(dataOutInfo_addr, 1); + return CELL_OK; } diff --git a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp index d8bac54859..05eeaf7f04 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp @@ -21,6 +21,19 @@ enum CELL_JPGDEC_ERROR_CB_PARAM = 0x80611109, }; +enum CellJpgDecColorSpace +{ + CELL_JPG_UNKNOWN = 0, + CELL_JPG_GRAYSCALE = 1, + CELL_JPG_RGB = 2, + CELL_JPG_YCbCr = 3, + CELL_JPG_RGBA = 10, + CELL_JPG_UPSAMPLE_ONLY = 11, + CELL_JPG_ARGB = 20, + CELL_JPG_GRAYSCALE_TO_ALPHA_RGBA = 40, + CELL_JPG_GRAYSCALE_TO_ALPHA_ARGB = 41, +}; + struct CellJpgDecInfo { u32 imageWidth; @@ -40,6 +53,36 @@ struct CellJpgDecSrc u32 spuThreadEnable; // CellJpgDecSpuThreadEna }; +struct CellJpgDecInParam +{ + u32 *commandPtr; + u32 downScale; + u32 method; // CellJpgDecMethod + u32 outputMode; // CellJpgDecOutputMode + u32 outputColorSpace; // CellJpgDecColorSpace + u8 outputColorAlpha; + u8 reserved[3]; +}; + +struct CellJpgDecOutParam +{ + u64 outputWidthByte; + u32 outputWidth; + u32 outputHeight; + u32 outputComponents; + u32 outputMode; // CellJpgDecOutputMode + u32 outputColorSpace; // CellJpgDecColorSpace + u32 downScale; + u32 useMemorySpace; +}; + +struct CellJpgDecSubHandle //Custom struct +{ + u32 fd; + u64 fileSize; + CellJpgDecInParam inParam; +}; + CellJpgDecInfo current_info; CellJpgDecSrc current_src; @@ -72,30 +115,39 @@ int cellJpgDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInf //current_src.streamSize = Memory.Read32(src_addr+20); //current_src.spuThreadEnable = Memory.Read32(src_addr+24); - u32& fd_addr = subHandle_addr; // Set file descriptor as sub handler of the decoder + CellJpgDecSubHandle *subHandle = new CellJpgDecSubHandle; + + // Get file descriptor + u32 fd_addr = Memory.Alloc(sizeof(u32), 1); int ret = cellFsOpen(current_src.fileName, 0, fd_addr, NULL, 0); + subHandle->fd = Memory.Read32(fd_addr); + Memory.Free(fd_addr); if(ret != 0) return CELL_JPGDEC_ERROR_OPEN_FILE; + // Get size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(subHandle->fd, sb_addr); + subHandle->fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); + + // From now, every u32 subHandle argument is a pointer to a CellPngDecSubHandle struct. + Memory.Write32(subHandle_addr, (u32)subHandle); + return CELL_OK; } int cellJpgDecClose(u32 mainHandle, u32 subHandle) { - u32& fd = subHandle; - cellFsClose(fd); + cellFsClose( ((CellJpgDecSubHandle*)subHandle)->fd ); + delete (CellJpgDecSubHandle*)subHandle; return CELL_OK; } int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) { - u32& fd = subHandle; - - //Get size of file - u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct - cellFsFstat(fd, sb_addr); - u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - Memory.Free(sb_addr); + const u32& fd = ((CellJpgDecSubHandle*)subHandle)->fd; + const u64& fileSize = ((CellJpgDecSubHandle*)subHandle)->fileSize; //Copy the JPG file to a buffer u32 buffer = Memory.Alloc(fileSize,1); @@ -134,7 +186,7 @@ int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) current_info.imageWidth = Memory.Read8(buffer+i+7)*256 + Memory.Read8(buffer+i+8); current_info.imageHeight = Memory.Read8(buffer+i+5)*256 + Memory.Read8(buffer+i+6); current_info.numComponents = 0; // Unimplemented - current_info.colorSpace = 0; // Unimplemented + current_info.colorSpace = 3; // Unimplemented mem_class_t info(info_addr); info += current_info.imageWidth; @@ -148,13 +200,9 @@ int cellJpgDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataCtrlParam_addr, u32 dataOutInfo_addr) { - u32& fd = subHandle; - - //Get size of file - u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct - cellFsFstat(fd, sb_addr); - u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - Memory.Free(sb_addr); + const u32& fd = ((CellJpgDecSubHandle*)subHandle)->fd; + const u64& fileSize = ((CellJpgDecSubHandle*)subHandle)->fileSize; + const CellJpgDecInParam& inParam = ((CellJpgDecSubHandle*)subHandle)->inParam; // (TODO: We should use the outParam) //Copy the JPG file to a buffer u32 buffer = Memory.Alloc(fileSize,1); @@ -177,20 +225,34 @@ int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC return CELL_JPGDEC_ERROR_STREAM_FORMAT; } u32 image_size = width * height * 4; - for(u32 i = 0; i < image_size; i+=4){ - Memory.Write8(data_addr+i+0, image[i+3]); - Memory.Write8(data_addr+i+1, image[i+0]); - Memory.Write8(data_addr+i+2, image[i+1]); - Memory.Write8(data_addr+i+3, image[i+2]); + if (inParam.outputColorSpace == CELL_JPG_RGBA){ + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+0]); + Memory.Write8(data_addr+i+1, image[i+1]); + Memory.Write8(data_addr+i+2, image[i+2]); + Memory.Write8(data_addr+i+3, image[i+3]); + } + } + else if (inParam.outputColorSpace == CELL_JPG_ARGB){ + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+3]); + Memory.Write8(data_addr+i+1, image[i+0]); + Memory.Write8(data_addr+i+2, image[i+1]); + Memory.Write8(data_addr+i+3, image[i+2]); + } } Memory.Free(buffer); return CELL_OK; } -int cellJpgDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outParam) +int cellJpgDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam_addr, u32 outParam_addr) { - UNIMPLEMENTED_FUNC(cellJpgDec); + CellJpgDecInParam& inParam = ((CellJpgDecSubHandle*)subHandle)->inParam; + inParam.outputColorSpace = Memory.Read32(inParam_addr+16); + + // (TODO) + return CELL_OK; } diff --git a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp index 70659c9fed..d260312d8f 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp @@ -22,6 +22,16 @@ enum CELL_PNGDEC_ERROR_CB_PARAM = 0x8061120a, }; +enum CellPngDecColorSpace +{ + CELL_PNGDEC_GRAYSCALE = 1, + CELL_PNGDEC_RGB = 2, + CELL_PNGDEC_PALETTE = 4, + CELL_PNGDEC_GRAYSCALE_ALPHA = 9, + CELL_PNGDEC_RGBA = 10, + CELL_PNGDEC_ARGB = 20, +}; + struct CellPngDecInfo { u32 imageWidth; @@ -44,6 +54,36 @@ struct CellPngDecSrc u32 spuThreadEnable; // CellPngDecSpuThreadEna }; +struct CellPngDecInParam +{ + u32 *commandPtr; + u32 outputMode; // CellPngDecOutputMode + u32 outputColorSpace; // CellPngDecColorSpace + u32 outputBitDepth; + u32 outputPackFlag; // CellPngDecPackFlag + u32 outputAlphaSelect; // CellPngDecAlphaSelect + u32 outputColorAlpha; +}; + +struct CellPngDecOutParam +{ + u64 outputWidthByte; + u32 outputWidth; + u32 outputHeight; + u32 outputComponents; + u32 outputBitDepth; + u32 outputMode; // CellPngDecOutputMode + u32 outputColorSpace; // CellPngDecColorSpace + u32 useMemorySpace; +}; + +struct CellPngDecSubHandle //Custom struct +{ + u32 fd; + u64 fileSize; + CellPngDecInParam inParam; +}; + CellPngDecInfo current_info; CellPngDecSrc current_src; @@ -70,30 +110,41 @@ int cellPngDecOpen(u32 mainHandle, u32 subHandle_addr, u32 src_addr, u32 openInf //current_src.streamSize = Memory.Read32(src_addr+20); //current_src.spuThreadEnable = Memory.Read32(src_addr+24); - u32& fd_addr = subHandle_addr; // Set file descriptor as sub handler of the decoder + CellPngDecSubHandle *subHandle = new CellPngDecSubHandle; + + // Get file descriptor + u32 fd_addr = Memory.Alloc(sizeof(u32), 1); int ret = cellFsOpen(current_src.fileName, 0, fd_addr, NULL, 0); + subHandle->fd = Memory.Read32(fd_addr); + Memory.Free(fd_addr); if(ret != 0) return CELL_PNGDEC_ERROR_OPEN_FILE; + // Get size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(subHandle->fd, sb_addr); + subHandle->fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); + + // From now, every u32 subHandle argument is a pointer to a CellPngDecSubHandle struct. + Memory.Write32(subHandle_addr, (u32)subHandle); + return CELL_OK; } int cellPngDecClose(u32 mainHandle, u32 subHandle) { - u32& fd = subHandle; - cellFsClose(fd); + cellFsClose( ((CellPngDecSubHandle*)subHandle)->fd ); + delete (CellPngDecSubHandle*)subHandle; return CELL_OK; } int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) { - u32& fd = subHandle; + const u32& fd = ((CellPngDecSubHandle*)subHandle)->fd; + const u64& fileSize = ((CellPngDecSubHandle*)subHandle)->fileSize; //Check size of file - u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct - cellFsFstat(fd, sb_addr); - u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - Memory.Free(sb_addr); if(fileSize < 29) return CELL_PNGDEC_ERROR_HEADER; // Error: The file is smaller than the length of a PNG header //Write the header to buffer @@ -134,17 +185,13 @@ int cellPngDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataCtrlParam_addr, u32 dataOutInfo_addr) { - u32& fd = subHandle; - - //Get size of file - u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct - cellFsFstat(fd, sb_addr); - u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - Memory.Free(sb_addr); + const u32& fd = ((CellPngDecSubHandle*)subHandle)->fd; + const u64& fileSize = ((CellPngDecSubHandle*)subHandle)->fileSize; + const CellPngDecInParam& inParam = ((CellPngDecSubHandle*)subHandle)->inParam; // (TODO: We should use the outParam) //Copy the PNG file to a buffer u32 buffer = Memory.Alloc(fileSize,1); - u32 pos_addr = Memory.Alloc(8,1); + u32 pos_addr = Memory.Alloc(sizeof(u64),1); cellFsLseek(fd, 0, 0, pos_addr); cellFsRead(fd, buffer, fileSize, NULL); Memory.Free(pos_addr); @@ -163,20 +210,34 @@ int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC return CELL_PNGDEC_ERROR_STREAM_FORMAT; } u32 image_size = width * height * 4; - for(u32 i = 0; i < image_size; i+=4){ - Memory.Write8(data_addr+i+0, image[i+3]); - Memory.Write8(data_addr+i+1, image[i+0]); - Memory.Write8(data_addr+i+2, image[i+1]); - Memory.Write8(data_addr+i+3, image[i+2]); + if (inParam.outputColorSpace == CELL_PNGDEC_RGBA){ + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+0]); + Memory.Write8(data_addr+i+1, image[i+1]); + Memory.Write8(data_addr+i+2, image[i+2]); + Memory.Write8(data_addr+i+3, image[i+3]); + } + } + else if (inParam.outputColorSpace == CELL_PNGDEC_ARGB){ + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+3]); + Memory.Write8(data_addr+i+1, image[i+0]); + Memory.Write8(data_addr+i+2, image[i+1]); + Memory.Write8(data_addr+i+3, image[i+2]); + } } Memory.Free(buffer); return CELL_OK; } -int cellPngDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outParam) +int cellPngDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam_addr, u32 outParam_addr) { - UNIMPLEMENTED_FUNC(cellPngDec); + CellPngDecInParam& inParam = ((CellPngDecSubHandle*)subHandle)->inParam; + inParam.outputColorSpace = Memory.Read32(inParam_addr+8); + + // (TODO) + return CELL_OK; } diff --git a/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp b/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp index 90073fe946..f01e368e91 100644 --- a/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp +++ b/rpcs3/Emu/SysCalls/lv2/SC_Keyboard.cpp @@ -68,7 +68,14 @@ u16 cellKbCnvRawCode(u32 arrange, u32 mkey, u32 led, u16 rawcode) if (rawcode == 0x58) return 0x0A | 0x4000; // '\n' // ASCII - if (rawcode >= 0x04 && rawcode <= 0x1D) return rawcode + 0x3D; // 'A' - 'Z' + if (rawcode >= 0x04 && rawcode <= 0x1D) // 'A' - 'Z' + { + rawcode -= + (mkey&(CELL_KB_MKEY_L_SHIFT|CELL_KB_MKEY_R_SHIFT)) ? + ((led&(CELL_KB_LED_CAPS_LOCK)) ? 0 : 0x20) : + ((led&(CELL_KB_LED_CAPS_LOCK)) ? 0x20 : 0); + return rawcode + 0x5D; + } if (rawcode >= 0x1E && rawcode <= 0x26) return rawcode + 0x13; // '1' - '9' if (rawcode == 0x27) return 0x30; // '0' if (rawcode == 0x28) return 0x0A; // '\n' @@ -76,6 +83,10 @@ u16 cellKbCnvRawCode(u32 arrange, u32 mkey, u32 led, u16 rawcode) if (rawcode == 0x2C) return 0x20; // ' ' if (rawcode == 0x2D) return 0x2D; // '-' if (rawcode == 0x2E) return 0x3D; // '=' + if (rawcode == 0x36) return 0x2C; // ',' + if (rawcode == 0x37) return 0x2E; // '.' + if (rawcode == 0x38) return 0x2F; // '/' + if (rawcode == 0x87) return 0x5C; // '\' // (TODO: Add more cases) diff --git a/rpcs3/Gui/MainFrame.cpp b/rpcs3/Gui/MainFrame.cpp index 063681289a..ad555891f6 100644 --- a/rpcs3/Gui/MainFrame.cpp +++ b/rpcs3/Gui/MainFrame.cpp @@ -274,7 +274,9 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) wxDialog diag(this, wxID_ANY, "Settings", wxDefaultPosition); - wxBoxSizer* s_panel(new wxBoxSizer(wxVERTICAL)); + wxBoxSizer* s_panel(new wxBoxSizer(wxHORIZONTAL)); + wxBoxSizer* s_subpanel1(new wxBoxSizer(wxVERTICAL)); + wxBoxSizer* s_subpanel2(new wxBoxSizer(wxVERTICAL)); wxStaticBoxSizer* s_round_cpu( new wxStaticBoxSizer( wxVERTICAL, &diag, _("CPU") ) ); wxStaticBoxSizer* s_round_cpu_decoder( new wxStaticBoxSizer( wxVERTICAL, &diag, _("Decoder") ) ); @@ -284,7 +286,7 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) wxStaticBoxSizer* s_round_gs_res( new wxStaticBoxSizer( wxVERTICAL, &diag, _("Default resolution") ) ); wxStaticBoxSizer* s_round_gs_aspect( new wxStaticBoxSizer( wxVERTICAL, &diag, _("Default aspect ratio") ) ); - wxStaticBoxSizer* s_round_pad( new wxStaticBoxSizer( wxVERTICAL, &diag, _("IO") ) ); + wxStaticBoxSizer* s_round_io( new wxStaticBoxSizer( wxVERTICAL, &diag, _("IO") ) ); wxStaticBoxSizer* s_round_pad_handler( new wxStaticBoxSizer( wxVERTICAL, &diag, _("Pad Handler") ) ); wxStaticBoxSizer* s_round_keyboard_handler( new wxStaticBoxSizer( wxVERTICAL, &diag, _("Keyboard Handler") ) ); @@ -343,8 +345,8 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) s_round_pad_handler->Add(cbox_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_keyboard_handler->Add(cbox_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); - s_round_pad->Add(s_round_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); - s_round_pad->Add(s_round_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); + s_round_io->Add(s_round_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); + s_round_io->Add(s_round_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); wxBoxSizer* s_b_panel(new wxBoxSizer(wxHORIZONTAL)); @@ -353,10 +355,14 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) //wxBoxSizer* s_conf_panel(new wxBoxSizer(wxHORIZONTAL)); - s_panel->Add(s_round_cpu, wxSizerFlags().Border(wxALL, 5).Expand()); - s_panel->Add(s_round_gs, wxSizerFlags().Border(wxALL, 5).Expand()); - s_panel->Add(s_round_pad, wxSizerFlags().Border(wxALL, 5).Expand()); - s_panel->Add(s_b_panel, wxSizerFlags().Border(wxALL, 8).Expand()); + s_subpanel1->Add(s_round_cpu, wxSizerFlags().Border(wxALL, 5).Expand()); + s_subpanel1->Add(s_round_gs, wxSizerFlags().Border(wxALL, 5).Expand()); + s_subpanel2->Add(s_round_io, wxSizerFlags().Border(wxALL, 5).Expand()); + s_subpanel1->Add(s_b_panel, wxSizerFlags().Border(wxALL, 8).Expand()); + + s_subpanel2->AddSpacer(200); + s_panel->Add(s_subpanel1, wxSizerFlags().Border(wxALL, 5).Expand()); + s_panel->Add(s_subpanel2, wxSizerFlags().Border(wxALL, 5).Expand()); diag.SetSizerAndFit( s_panel ); From 1024a7c7c41f2c747a875e891ad7bed79730b6a8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Sat, 14 Sep 2013 20:20:57 +0200 Subject: [PATCH 6/7] Mouse support added *Implemented 'cellMouse*' functions from 'sys_io' module, which are part of the libmouse library. * Added corresponding entries in the 'Config > Settings' menu to change the handler of the mouse. Supported handlers: Windows, Null. * cellGifDec: Fixed some errors and added support for RGBA color. (Remember: The mouse support is very experimental) --- rpcs3/Emu/Io/Mouse.cpp | 38 +++++ rpcs3/Emu/Io/Mouse.h | 26 ++++ rpcs3/Emu/Io/MouseHandler.h | 160 +++++++++++++++++++++ rpcs3/Emu/Io/Null/NullMouseHandler.h | 24 ++++ rpcs3/Emu/Io/Windows/WindowsMouseHandler.h | 59 ++++++++ rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp | 103 +++++++++---- rpcs3/Emu/SysCalls/Modules/sys_io.cpp | 11 ++ rpcs3/Emu/SysCalls/SysCalls.h | 12 ++ rpcs3/Emu/SysCalls/lv2/SC_Mouse.cpp | 142 ++++++++++++++++++ rpcs3/Emu/System.cpp | 1 + rpcs3/Emu/System.h | 3 + rpcs3/Gui/MainFrame.cpp | 12 +- rpcs3/Ini.h | 4 + rpcs3/rpcs3.vcxproj | 2 + rpcs3/rpcs3.vcxproj.filters | 6 + 15 files changed, 578 insertions(+), 25 deletions(-) create mode 100644 rpcs3/Emu/Io/Mouse.cpp create mode 100644 rpcs3/Emu/Io/Mouse.h create mode 100644 rpcs3/Emu/Io/MouseHandler.h create mode 100644 rpcs3/Emu/Io/Null/NullMouseHandler.h create mode 100644 rpcs3/Emu/Io/Windows/WindowsMouseHandler.h create mode 100644 rpcs3/Emu/SysCalls/lv2/SC_Mouse.cpp diff --git a/rpcs3/Emu/Io/Mouse.cpp b/rpcs3/Emu/Io/Mouse.cpp new file mode 100644 index 0000000000..7ad1dd0631 --- /dev/null +++ b/rpcs3/Emu/Io/Mouse.cpp @@ -0,0 +1,38 @@ +#include "stdafx.h" +#include "Mouse.h" +#include "Null/NullMouseHandler.h" +#include "Windows/WindowsMouseHandler.h" + +MouseManager::MouseManager() + : m_mouse_handler(nullptr) + , m_inited(false) +{ +} + +MouseManager::~MouseManager() +{ +} + +void MouseManager::Init(const u32 max_connect) +{ + if(m_inited) return; + + switch(Ini.MouseHandlerMode.GetValue()) + { + case 1: m_mouse_handler = new WindowsMouseHandler(); break; + + default: + case 0: m_mouse_handler = new NullMouseHandler(); break; + } + + m_mouse_handler->Init(max_connect); + m_inited = true; +} + +void MouseManager::Close() +{ + if(m_mouse_handler) m_mouse_handler->Close(); + m_mouse_handler = nullptr; + + m_inited = false; +} \ No newline at end of file diff --git a/rpcs3/Emu/Io/Mouse.h b/rpcs3/Emu/Io/Mouse.h new file mode 100644 index 0000000000..90f2336fbc --- /dev/null +++ b/rpcs3/Emu/Io/Mouse.h @@ -0,0 +1,26 @@ +#pragma once + +#include "MouseHandler.h" + +class MouseManager //: public wxWindow +{ + bool m_inited; + MouseHandlerBase* m_mouse_handler; + +public: + MouseManager(); + ~MouseManager(); + + void Init(const u32 max_connect); + void Close(); + + Array& GetMice() { return m_mouse_handler->GetMice(); } + MouseInfo& GetInfo() { return m_mouse_handler->GetInfo(); } + CellMouseData& GetData(const u32 mouse) { return m_mouse_handler->GetData(mouse); } + CellMouseRawData& GetRawData(const u32 mouse) { return m_mouse_handler->GetRawData(mouse); } + + bool IsInited() { return m_inited; } + +//private: + //DECLARE_EVENT_TABLE(); +}; \ No newline at end of file diff --git a/rpcs3/Emu/Io/MouseHandler.h b/rpcs3/Emu/Io/MouseHandler.h new file mode 100644 index 0000000000..b23a404633 --- /dev/null +++ b/rpcs3/Emu/Io/MouseHandler.h @@ -0,0 +1,160 @@ +#pragma once + +enum MousePortStatus +{ + CELL_MOUSE_STATUS_DISCONNECTED = 0x00000000, + CELL_MOUSE_STATUS_CONNECTED = 0x00000001, +}; + +enum MouseDataUpdate +{ + CELL_MOUSE_DATA_UPDATE = 1, + CELL_MOUSE_DATA_NON = 0, +}; + +enum MouseButtonCodes +{ + CELL_MOUSE_BUTTON_1 = 0x00000001, + CELL_MOUSE_BUTTON_2 = 0x00000002, + CELL_MOUSE_BUTTON_3 = 0x00000004, + CELL_MOUSE_BUTTON_4 = 0x00000008, + CELL_MOUSE_BUTTON_5 = 0x00000010, + CELL_MOUSE_BUTTON_6 = 0x00000020, + CELL_MOUSE_BUTTON_7 = 0x00000040, + CELL_MOUSE_BUTTON_8 = 0x00000080, +}; + +static const u32 CELL_MAX_MICE = 127; +static const u32 CELL_MOUSE_MAX_DATA_LIST_NUM = 8; +static const u32 CELL_MOUSE_MAX_CODES = 64; + +struct MouseInfo +{ + u32 max_connect; + u32 now_connect; + u32 info; + u16 vendor_id[CELL_MAX_MICE]; + u16 product_id[CELL_MAX_MICE]; + u8 status[CELL_MAX_MICE]; +}; + +struct CellMouseRawData +{ + s32 len; + u8 data[CELL_MOUSE_MAX_CODES]; + + CellMouseRawData() + : len(0) + { + } +}; + +struct CellMouseData +{ + u8 update; + u8 buttons; + s8 x_axis; + s8 y_axis; + s8 wheel; + s8 tilt; // (TODO) + + CellMouseData() + : update(0) + , buttons(0) + , x_axis(0) + , y_axis(0) + , wheel(0) + , tilt(0) + { + } +}; + +struct CellMouseDataList +{ + u32 list_num; + CellMouseData list[CELL_MOUSE_MAX_DATA_LIST_NUM]; + + CellMouseDataList() + : list_num(0) + { + } +}; + +struct Mouse +{ + s16 x_pos; + s16 y_pos; + + CellMouseData m_data; + CellMouseRawData m_rawdata; + + Mouse() + : m_data() + , m_rawdata() + { + } +}; + +class MouseHandlerBase +{ +protected: + MouseInfo m_info; + Array m_mice; + +public: + virtual void Init(const u32 max_connect)=0; + virtual void Close()=0; + + void Button(u8 button, bool pressed) + { + for(u64 p=0; p& GetMice() { return m_mice; } + CellMouseData& GetData(const u32 mouse) { return GetMice()[mouse].m_data; } + CellMouseRawData& GetRawData(const u32 mouse) { return GetMice()[mouse].m_rawdata; } +}; \ No newline at end of file diff --git a/rpcs3/Emu/Io/Null/NullMouseHandler.h b/rpcs3/Emu/Io/Null/NullMouseHandler.h new file mode 100644 index 0000000000..1abe6fe2c8 --- /dev/null +++ b/rpcs3/Emu/Io/Null/NullMouseHandler.h @@ -0,0 +1,24 @@ +#pragma once + +#include "Emu/Io/MouseHandler.h" + +class NullMouseHandler : public MouseHandlerBase +{ +public: + NullMouseHandler() + { + } + + virtual void Init(const u32 max_connect) + { + memset(&m_info, 0, sizeof(MouseInfo)); + m_info.max_connect = max_connect; + m_mice.Clear(); + } + + virtual void Close() + { + memset(&m_info, 0, sizeof(MouseInfo)); + m_mice.Clear(); + } +}; \ No newline at end of file diff --git a/rpcs3/Emu/Io/Windows/WindowsMouseHandler.h b/rpcs3/Emu/Io/Windows/WindowsMouseHandler.h new file mode 100644 index 0000000000..dce6832dec --- /dev/null +++ b/rpcs3/Emu/Io/Windows/WindowsMouseHandler.h @@ -0,0 +1,59 @@ +#pragma once + +#include "Emu/Io/MouseHandler.h" + +class WindowsMouseHandler + : public wxWindow + , public MouseHandlerBase +{ + AppConnector m_app_connector; + +public: + WindowsMouseHandler() : wxWindow() + { + m_app_connector.Connect(wxEVT_LEFT_DOWN, wxMouseEventHandler(WindowsMouseHandler::MouseButtonDown), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_RIGHT_DOWN, wxMouseEventHandler(WindowsMouseHandler::MouseButtonDown), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_MIDDLE_DOWN, wxMouseEventHandler(WindowsMouseHandler::MouseButtonDown), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_LEFT_UP, wxMouseEventHandler(WindowsMouseHandler::MouseButtonUp), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_RIGHT_UP, wxMouseEventHandler(WindowsMouseHandler::MouseButtonUp), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_MIDDLE_UP, wxMouseEventHandler(WindowsMouseHandler::MouseButtonUp), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_MOUSEWHEEL, wxMouseEventHandler(WindowsMouseHandler::MouseScroll), (wxObject*)0, this); + m_app_connector.Connect(wxEVT_MOTION, wxMouseEventHandler(WindowsMouseHandler::MouseMove), (wxObject*)0, this); + } + + virtual void MouseButtonDown(wxMouseEvent& event) + { + if (event.LeftDown()) MouseHandlerBase::Button(CELL_MOUSE_BUTTON_1, 1); + else if (event.RightDown()) MouseHandlerBase::Button(CELL_MOUSE_BUTTON_2, 1); + else if (event.MiddleDown())MouseHandlerBase::Button(CELL_MOUSE_BUTTON_3, 1); + event.Skip(); + } + virtual void MouseButtonUp(wxMouseEvent& event) + { + if (event.LeftUp()) MouseHandlerBase::Button(CELL_MOUSE_BUTTON_1, 0); + else if (event.RightUp()) MouseHandlerBase::Button(CELL_MOUSE_BUTTON_2, 0); + else if (event.MiddleUp()) MouseHandlerBase::Button(CELL_MOUSE_BUTTON_3, 0); + event.Skip(); + } + virtual void MouseScroll(wxMouseEvent& event) { MouseHandlerBase::Scroll(event.GetWheelRotation()); event.Skip(); } + virtual void MouseMove(wxMouseEvent& event) { MouseHandlerBase::Move(event.m_x, event.m_y); event.Skip(); } + + virtual void Init(const u32 max_connect) + { + m_mice.Move(new Mouse()); + memset(&m_info, 0, sizeof(MouseInfo)); + m_info.max_connect = max_connect; + m_info.now_connect = min(GetMice().GetCount(), max_connect); + m_info.info = 0; // Ownership of mouse data: 0=Application, 1=System + m_info.status[0] = CELL_MOUSE_STATUS_CONNECTED; // (TODO: Support for more mice) + for(u32 i=1; ifd = Memory.Read32(fd_addr); + Memory.Free(fd_addr); if(ret != 0) return CELL_GIFDEC_ERROR_OPEN_FILE; + // Get size of file + u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct + cellFsFstat(subHandle->fd, sb_addr); + subHandle->fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size + Memory.Free(sb_addr); + + // From now, every u32 subHandle argument is a pointer to a CellPngDecSubHandle struct. + Memory.Write32(subHandle_addr, (u32)subHandle); + return CELL_OK; } int cellGifDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) { - u32& fd = subHandle; - - //Check size of file - u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct - cellFsFstat(fd, sb_addr); - u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - Memory.Free(sb_addr); + const u32& fd = ((CellGifDecSubHandle*)subHandle)->fd; + const u64& fileSize = ((CellGifDecSubHandle*)subHandle)->fileSize; //Write the header to buffer u32 buffer = Memory.Alloc(13,1); // Alloc buffer for GIF header @@ -126,21 +168,21 @@ int cellGifDecReadHeader(u32 mainHandle, u32 subHandle, u32 info_addr) return CELL_OK; } -int cellGifDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam, u32 outParam) +int cellGifDecSetParameter(u32 mainHandle, u32 subHandle, u32 inParam_addr, u32 outParam_addr) { - UNIMPLEMENTED_FUNC(cellGifDec); + CellGifDecInParam& inParam = ((CellGifDecSubHandle*)subHandle)->inParam; + inParam.colorSpace = Memory.Read32(inParam_addr+4); + + // (TODO) + return CELL_OK; } int cellGifDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataCtrlParam_addr, u32 dataOutInfo_addr) { - u32& fd = subHandle; - - //Get size of file - u32 sb_addr = Memory.Alloc(52,1); // Alloc a CellFsStat struct - cellFsFstat(fd, sb_addr); - u64 fileSize = Memory.Read64(sb_addr+36); // Get CellFsStat.st_size - Memory.Free(sb_addr); + const u32& fd = ((CellGifDecSubHandle*)subHandle)->fd; + const u64& fileSize = ((CellGifDecSubHandle*)subHandle)->fileSize; + const CellGifDecInParam& inParam = ((CellGifDecSubHandle*)subHandle)->inParam; // (TODO: We should use the outParam) //Copy the GIF file to a buffer u32 buffer = Memory.Alloc(fileSize,1); @@ -163,24 +205,37 @@ int cellGifDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC return CELL_GIFDEC_ERROR_STREAM_FORMAT; } u32 image_size = width * height * 4; - for(u32 i = 0; i < image_size; i+=4){ - Memory.Write8(data_addr+i+0, image[i+3]); - Memory.Write8(data_addr+i+1, image[i+0]); - Memory.Write8(data_addr+i+2, image[i+1]); - Memory.Write8(data_addr+i+3, image[i+2]); + if (inParam.colorSpace == CELL_GIFDEC_RGBA){ + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+0]); + Memory.Write8(data_addr+i+1, image[i+1]); + Memory.Write8(data_addr+i+2, image[i+2]); + Memory.Write8(data_addr+i+3, image[i+3]); // (This can be optimized by using Write32) + } + } + if (inParam.colorSpace == CELL_GIFDEC_ARGB){ + for(u32 i = 0; i < image_size; i+=4){ + Memory.Write8(data_addr+i+0, image[i+3]); + Memory.Write8(data_addr+i+1, image[i+0]); + Memory.Write8(data_addr+i+2, image[i+1]); + Memory.Write8(data_addr+i+3, image[i+2]); + } } Memory.Free(buffer); //The output data is an image (dataOutInfo.recordType = 1) Memory.Write32(dataOutInfo_addr, 1); + u32 outExtensionData_addr = Memory.Alloc(20,1); // (TODO: Is this the best way to avoid exceptions when trying to access this data? Will this produce a memory leak?) + Memory.Write32(dataOutInfo_addr+8, outExtensionData_addr); + return CELL_OK; } int cellGifDecClose(u32 mainHandle, u32 subHandle) { - u32& fd = subHandle; - cellFsClose(fd); + cellFsClose( ((CellGifDecSubHandle*)subHandle)->fd ); + delete (CellGifDecSubHandle*)subHandle; return CELL_OK; } diff --git a/rpcs3/Emu/SysCalls/Modules/sys_io.cpp b/rpcs3/Emu/SysCalls/Modules/sys_io.cpp index 4339052ecf..db6e727c0e 100644 --- a/rpcs3/Emu/SysCalls/Modules/sys_io.cpp +++ b/rpcs3/Emu/SysCalls/Modules/sys_io.cpp @@ -26,4 +26,15 @@ void sys_io_init() sys_io.AddFunc(0x3f72c56e, cellKbSetLEDStatus); sys_io.AddFunc(0xdeefdfa7, cellKbSetReadMode); sys_io.AddFunc(0x1f71ecbe, cellKbGetConfiguration); + + sys_io.AddFunc(0xc9030138, cellMouseInit); + sys_io.AddFunc(0x3ef66b95, cellMouseClearBuf); + sys_io.AddFunc(0xe10183ce, cellMouseEnd); + sys_io.AddFunc(0x5baf30fb, cellMouseGetInfo); + sys_io.AddFunc(0x4d0b3b1f, cellMouseInfoTabletMode); + sys_io.AddFunc(0x3138e632, cellMouseGetData); + sys_io.AddFunc(0x6bd131f0, cellMouseGetDataList); + sys_io.AddFunc(0x2d16da4f, cellMouseSetTabletMode); + sys_io.AddFunc(0x21a62e9b, cellMouseGetTabletDataList); + sys_io.AddFunc(0xa328cc35, cellMouseGetRawData); } diff --git a/rpcs3/Emu/SysCalls/SysCalls.h b/rpcs3/Emu/SysCalls/SysCalls.h index 68e12d882a..159f168d57 100644 --- a/rpcs3/Emu/SysCalls/SysCalls.h +++ b/rpcs3/Emu/SysCalls/SysCalls.h @@ -228,6 +228,18 @@ extern int cellKbSetLEDStatus(u32 port_no, u8 led); extern int cellKbSetReadMode(u32 port_no, u32 rmode); extern int cellKbGetConfiguration(u32 port_no, u32 config_addr); +//cellMouse +extern int cellMouseInit(u32 max_connect); +extern int cellMouseClearBuf(u32 port_no); +extern int cellMouseEnd(); +extern int cellMouseGetInfo(u32 info_addr); +extern int cellMouseInfoTabletMode(u32 port_no, u32 info_addr); +extern int cellMouseGetData(u32 port_no, u32 data_addr); +extern int cellMouseGetDataList(u32 port_no, u32 data_addr); +extern int cellMouseSetTabletMode(u32 port_no, u32 mode); +extern int cellMouseGetTabletDataList(u32 port_no, u32 data_addr); +extern int cellMouseGetRawData(u32 port_no, u32 data_addr); + //cellGcm extern int cellGcmCallback(u32 context_addr, u32 count); diff --git a/rpcs3/Emu/SysCalls/lv2/SC_Mouse.cpp b/rpcs3/Emu/SysCalls/lv2/SC_Mouse.cpp new file mode 100644 index 0000000000..9161fba651 --- /dev/null +++ b/rpcs3/Emu/SysCalls/lv2/SC_Mouse.cpp @@ -0,0 +1,142 @@ +#include "stdafx.h" +#include "Emu/Io/Mouse.h" +#include "Emu/SysCalls/SysCalls.h" + +extern Module sys_io; + +enum CELL_MOUSE_ERROR_CODE +{ + CELL_MOUSE_ERROR_FATAL = 0x80121201, + CELL_MOUSE_ERROR_INVALID_PARAMETER = 0x80121202, + CELL_MOUSE_ERROR_ALREADY_INITIALIZED = 0x80121203, + CELL_MOUSE_ERROR_UNINITIALIZED = 0x80121204, + CELL_MOUSE_ERROR_RESOURCE_ALLOCATION_FAILED = 0x80121205, + CELL_MOUSE_ERROR_DATA_READ_FAILED = 0x80121206, + CELL_MOUSE_ERROR_NO_DEVICE = 0x80121207, + CELL_MOUSE_ERROR_SYS_SETTING_FAILED = 0x80121208, +}; + +int cellMouseInit(u32 max_connect) +{ + sys_io.Log("cellMouseInit(max_connect=%d)", max_connect); + if(Emu.GetMouseManager().IsInited()) return CELL_MOUSE_ERROR_ALREADY_INITIALIZED; + if(max_connect > 7) return CELL_MOUSE_ERROR_INVALID_PARAMETER; + + Emu.GetMouseManager().Init(max_connect); + return CELL_OK; +} + + +int cellMouseClearBuf(u32 port_no) +{ + sys_io.Log("cellMouseClearBuf(port_no=%d)", port_no); + if(!Emu.GetMouseManager().IsInited()) return CELL_MOUSE_ERROR_UNINITIALIZED; + if(port_no >= Emu.GetMouseManager().GetMice().GetCount()) return CELL_MOUSE_ERROR_INVALID_PARAMETER; + + //? + + return CELL_OK; +} + +int cellMouseEnd() +{ + sys_io.Log("cellMouseEnd()"); + if(!Emu.GetMouseManager().IsInited()) return CELL_MOUSE_ERROR_UNINITIALIZED; + Emu.GetMouseManager().Close(); + return CELL_OK; +} + +int cellMouseGetInfo(u32 info_addr) +{ + sys_io.Log("cellMouseGetInfo(info_addr=0x%x)", info_addr); + if(!Emu.GetMouseManager().IsInited()) return CELL_MOUSE_ERROR_UNINITIALIZED; + + const MouseInfo& current_info = Emu.GetMouseManager().GetInfo(); + mem_class_t info(info_addr); + info += current_info.max_connect; + info += current_info.now_connect; + info += current_info.info; + for(u32 i=0; i= Emu.GetMouseManager().GetMice().GetCount()) return CELL_MOUSE_ERROR_INVALID_PARAMETER; + + mem_class_t info(info_addr); + info += 0; // Unimplemented: (0=Tablet mode is not supported) + info += 1; // Unimplemented: (1=Mouse mode) + + return CELL_OK; +} + +int cellMouseGetData(u32 port_no, u32 data_addr) +{ + sys_io.Log("cellMouseGetData(port_no=%d,data_addr=0x%x)", port_no,data_addr); + if(!Emu.GetMouseManager().IsInited()) return CELL_MOUSE_ERROR_UNINITIALIZED; + if(port_no >= Emu.GetMouseManager().GetMice().GetCount()) return CELL_MOUSE_ERROR_NO_DEVICE; + + CellMouseData& current_data = Emu.GetMouseManager().GetData(port_no); + mem_class_t data(data_addr); + data += current_data.update; + data += current_data.buttons; + data += current_data.x_axis; + data += current_data.y_axis; + data += current_data.wheel; + data += current_data.tilt; + + current_data.update = CELL_MOUSE_DATA_NON; + current_data.x_axis = 0; + current_data.y_axis = 0; + current_data.wheel = 0; + + return CELL_OK; +} + +int cellMouseGetDataList(u32 port_no, u32 data_addr) +{ + UNIMPLEMENTED_FUNC(sys_io); + + return CELL_OK; +} + +int cellMouseSetTabletMode(u32 port_no, u32 mode) +{ + UNIMPLEMENTED_FUNC(sys_io); + + return CELL_OK; +} + +int cellMouseGetTabletDataList(u32 port_no, u32 data_addr) +{ + UNIMPLEMENTED_FUNC(sys_io); + + return CELL_OK; +} + +int cellMouseGetRawData(u32 port_no, u32 data_addr) +{ + UNIMPLEMENTED_FUNC(sys_io); + + /*sys_io.Log("cellMouseGetRawData(port_no=%d,data_addr=0x%x)", port_no,data_addr); + if(!Emu.GetMouseManager().IsInited()) return CELL_MOUSE_ERROR_UNINITIALIZED; + if(port_no >= Emu.GetMouseManager().GetMice().GetCount()) return CELL_MOUSE_ERROR_NO_DEVICE; + + CellMouseRawData& current_rawdata = Emu.GetMouseManager().GetRawData(port_no); + mem_class_t data(data_addr); + data += current_rawdata.len; + for(s32 i=0; iAppend("Windows"); //cbox_pad_handler->Append("DirectInput"); + cbox_mouse_handler->Append("Null"); + cbox_mouse_handler->Append("Windows"); + //cbox_pad_handler->Append("DirectInput"); + chbox_gs_vsync->SetValue(Ini.GSVSyncEnable.GetValue()); cbox_cpu_decoder->SetSelection(Ini.CPUDecoderMode.GetValue() ? Ini.CPUDecoderMode.GetValue() - 1 : 0); @@ -331,6 +337,7 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) cbox_gs_aspect->SetSelection(Ini.GSAspectRatio.GetValue() - 1); cbox_pad_handler->SetSelection(Ini.PadHandlerMode.GetValue()); cbox_keyboard_handler->SetSelection(Ini.KeyboardHandlerMode.GetValue()); + cbox_mouse_handler->SetSelection(Ini.MouseHandlerMode.GetValue()); s_round_cpu_decoder->Add(cbox_cpu_decoder, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_cpu->Add(s_round_cpu_decoder, wxSizerFlags().Border(wxALL, 5).Expand()); @@ -345,8 +352,10 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) s_round_pad_handler->Add(cbox_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_keyboard_handler->Add(cbox_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); + s_round_mouse_handler->Add(cbox_mouse_handler, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_io->Add(s_round_pad_handler, wxSizerFlags().Border(wxALL, 5).Expand()); s_round_io->Add(s_round_keyboard_handler, wxSizerFlags().Border(wxALL, 5).Expand()); + s_round_io->Add(s_round_mouse_handler, wxSizerFlags().Border(wxALL, 5).Expand()); wxBoxSizer* s_b_panel(new wxBoxSizer(wxHORIZONTAL)); @@ -360,7 +369,7 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) s_subpanel2->Add(s_round_io, wxSizerFlags().Border(wxALL, 5).Expand()); s_subpanel1->Add(s_b_panel, wxSizerFlags().Border(wxALL, 8).Expand()); - s_subpanel2->AddSpacer(200); + s_subpanel2->AddSpacer(180); s_panel->Add(s_subpanel1, wxSizerFlags().Border(wxALL, 5).Expand()); s_panel->Add(s_subpanel2, wxSizerFlags().Border(wxALL, 5).Expand()); @@ -375,6 +384,7 @@ void MainFrame::Config(wxCommandEvent& WXUNUSED(event)) Ini.GSVSyncEnable.SetValue(chbox_gs_vsync->GetValue()); Ini.PadHandlerMode.SetValue(cbox_pad_handler->GetSelection()); Ini.KeyboardHandlerMode.SetValue(cbox_keyboard_handler->GetSelection()); + Ini.MouseHandlerMode.SetValue(cbox_mouse_handler->GetSelection()); Ini.Save(); } diff --git a/rpcs3/Ini.h b/rpcs3/Ini.h index e9b2c92652..dff605bfeb 100644 --- a/rpcs3/Ini.h +++ b/rpcs3/Ini.h @@ -99,6 +99,7 @@ public: IniEntry GSVSyncEnable; IniEntry PadHandlerMode; IniEntry KeyboardHandlerMode; + IniEntry MouseHandlerMode; public: Inis() : DefPath("EmuSettings") @@ -117,6 +118,7 @@ public: path = DefPath + "\\" + "IO"; PadHandlerMode.Init("PadHandlerMode", path); KeyboardHandlerMode.Init("KeyboardHandlerMode", path); + MouseHandlerMode.Init("MouseHandlerMode", path); } void Load() @@ -128,6 +130,7 @@ public: GSVSyncEnable.Load(false); PadHandlerMode.Load(0); KeyboardHandlerMode.Load(0); + MouseHandlerMode.Load(0); } void Save() @@ -139,6 +142,7 @@ public: GSVSyncEnable.Save(); PadHandlerMode.Save(); KeyboardHandlerMode.Save(); + MouseHandlerMode.Save(); } }; diff --git a/rpcs3/rpcs3.vcxproj b/rpcs3/rpcs3.vcxproj index 637e88a80a..893c671409 100644 --- a/rpcs3/rpcs3.vcxproj +++ b/rpcs3/rpcs3.vcxproj @@ -222,6 +222,7 @@ + @@ -234,6 +235,7 @@ + diff --git a/rpcs3/rpcs3.vcxproj.filters b/rpcs3/rpcs3.vcxproj.filters index 7a15323230..1606f87a8a 100644 --- a/rpcs3/rpcs3.vcxproj.filters +++ b/rpcs3/rpcs3.vcxproj.filters @@ -301,6 +301,12 @@ Emu\SysCalls\lv2 + + Emu\Io + + + Emu\SysCalls\lv2 + From b85d39bfc34d4522f4877839ad63ef4405c02faa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Alexandro=20S=C3=A1nchez=20Bach?= Date: Sat, 14 Sep 2013 22:47:59 +0200 Subject: [PATCH 7/7] Minor issues fixed *Memory leak in image decodification modules. *Changed mouse wheel delta value. --- rpcs3/Emu/Io/MouseHandler.h | 2 +- rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp | 13 ++++++------- rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp | 11 +++++------ rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp | 11 +++++------ 4 files changed, 17 insertions(+), 20 deletions(-) diff --git a/rpcs3/Emu/Io/MouseHandler.h b/rpcs3/Emu/Io/MouseHandler.h index b23a404633..d937bf25c9 100644 --- a/rpcs3/Emu/Io/MouseHandler.h +++ b/rpcs3/Emu/Io/MouseHandler.h @@ -127,7 +127,7 @@ public: { CellMouseData& data = GetData(p); data.update = CELL_MOUSE_DATA_UPDATE; - data.wheel = rotation; + data.wheel = rotation/120; //120=event.GetWheelDelta() } } } diff --git a/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp index 1049103aeb..e35da3dd45 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellGifDec.cpp @@ -197,20 +197,19 @@ int cellGifDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC for(u32 i = 0; i < fileSize; i++){ gif[i] = Memory.Read8(buffer+i); } + Memory.Free(buffer); unsigned char *image = stbi_load_from_memory((const unsigned char*)gif, fileSize, &width, &height, &actual_components, 4); - if (!image) - { - Memory.Free(buffer); - return CELL_GIFDEC_ERROR_STREAM_FORMAT; - } + delete[] gif; + if (!image) return CELL_GIFDEC_ERROR_STREAM_FORMAT; + u32 image_size = width * height * 4; if (inParam.colorSpace == CELL_GIFDEC_RGBA){ for(u32 i = 0; i < image_size; i+=4){ Memory.Write8(data_addr+i+0, image[i+0]); Memory.Write8(data_addr+i+1, image[i+1]); Memory.Write8(data_addr+i+2, image[i+2]); - Memory.Write8(data_addr+i+3, image[i+3]); // (This can be optimized by using Write32) + Memory.Write8(data_addr+i+3, image[i+3]); } } if (inParam.colorSpace == CELL_GIFDEC_ARGB){ @@ -221,7 +220,7 @@ int cellGifDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC Memory.Write8(data_addr+i+3, image[i+2]); } } - Memory.Free(buffer); + delete[] image; //The output data is an image (dataOutInfo.recordType = 1) Memory.Write32(dataOutInfo_addr, 1); diff --git a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp index 05eeaf7f04..c9cbaba239 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellJpgDec.cpp @@ -217,13 +217,12 @@ int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC for(u32 i = 0; i < fileSize; i++){ jpg[i] = Memory.Read8(buffer+i); } + Memory.Free(buffer); unsigned char *image = stbi_load_from_memory((const unsigned char*)jpg, fileSize, &width, &height, &actual_components, 4); - if (!image) - { - Memory.Free(buffer); - return CELL_JPGDEC_ERROR_STREAM_FORMAT; - } + delete[] jpg; + if (!image) return CELL_JPGDEC_ERROR_STREAM_FORMAT; + u32 image_size = width * height * 4; if (inParam.outputColorSpace == CELL_JPG_RGBA){ for(u32 i = 0; i < image_size; i+=4){ @@ -241,7 +240,7 @@ int cellJpgDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC Memory.Write8(data_addr+i+3, image[i+2]); } } - Memory.Free(buffer); + delete[] image; return CELL_OK; } diff --git a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp index d260312d8f..1a93351259 100644 --- a/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp +++ b/rpcs3/Emu/SysCalls/Modules/cellPngDec.cpp @@ -202,13 +202,12 @@ int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC for(u32 i = 0; i < fileSize; i++){ png[i] = Memory.Read8(buffer+i); } + Memory.Free(buffer); unsigned char *image = stbi_load_from_memory((const unsigned char*)png, fileSize, &width, &height, &actual_components, 4); - if (!image) - { - Memory.Free(buffer); - return CELL_PNGDEC_ERROR_STREAM_FORMAT; - } + delete[] png; + if (!image) return CELL_PNGDEC_ERROR_STREAM_FORMAT; + u32 image_size = width * height * 4; if (inParam.outputColorSpace == CELL_PNGDEC_RGBA){ for(u32 i = 0; i < image_size; i+=4){ @@ -226,7 +225,7 @@ int cellPngDecDecodeData(u32 mainHandle, u32 subHandle, u32 data_addr, u32 dataC Memory.Write8(data_addr+i+3, image[i+2]); } } - Memory.Free(buffer); + delete[] image; return CELL_OK; }