1# scripts/pnglibconf.dfa - library build configuration control 2# 3@/*- pnglibconf.dfn intermediate file 4@ * generated from scripts/pnglibconf.dfa 5@ */ 6# 7com pnglibconf.h - library build configuration 8com 9version 10com 11com Copyright (c) 1998-2017 Glenn Randers-Pehrson 12com 13com This code is released under the libpng license. 14com For conditions of distribution and use, see the disclaimer 15com and license in png.h 16com 17 18file pnglibconf.h scripts/pnglibconf.dfa PNGLCONF_H 19 20# This file is preprocessed by scripts/options.awk and the 21# C compiler to generate 'pnglibconf.h' - a list of all the 22# configuration options. The file lists the various options 23# that can *only* be specified during the libpng build; 24# pnglibconf.h freezes the definitions selected for the specific 25# build. 26# 27# The syntax is detailed in scripts/options.awk; this is a summary 28# only: 29# 30# setting <name> [requires ...] [default] 31# #define PNG_<name> <value> /* value comes from current setting */ 32# option <name> [requires ...] [if ...] [enables ...] [disabled] 33# #define PNG_<name>_SUPPORTED if the requirements are met and 34# enable the other options listed 35# chunk <name> [requires ...] [enables ...] [disabled] 36# Enable chunk processing for the given ancillary chunk; any 37# 'requires something' expands to READ_something for read and 38# WRITE_something for write, but the enables list members are 39# used as given (e.g. enables GAMMA just expands to that on the 40# correspond READ_name and WRITE_name lines.) 41# 42# "," may be used to separate options on an 'option' line and is ignored; it 43# doesn't change the meaning of the line. (NOT setting, where "," becomes 44# part of the setting!) A comma at the end of an option line causes a 45# continuation (the next line is included in the option too.) 46# 47# Note that the 'on' and 'off' keywords, while valid on both option 48# and chunk, should not be used in this file because they force the 49# relevant options on or off. 50 51#---------------------------------------------------------------------- 52 53# The following setting, option and chunk values can all be changed 54# while building libpng: 55# 56# setting: change 'setting' lines to fine tune library performance; 57# changes to the settings don't affect the libpng API functionally 58# 59# option: change 'option' lines to remove or add capabilities from 60# or to the library; options change the library API 61# 62# chunk: change 'chunk' lines to remove capabilities to process 63# optional ('ancillary') chunks. This does not prevent PNG 64# decoding but does change the libpng API because some chunks 65# will be ignored. 66# 67# There are three ways of disabling features, in no particular order: 68# 69# 1) Create 'pngusr.h', enter the required private build information 70# detailed below and #define PNG_NO_<option> for each option you 71# don't want in that file in that file. You can also turn on options 72# using PNG_<option>_SUPPORTED. When you have finished rerun 73# configure and rebuild pnglibconf.h file with -DPNG_USER_CONFIG: 74# 75# make clean 76# CPPFLAGS='-DPNG_USER_CONFIG' ./configure 77# make pnglibconf.h 78# 79# pngusr.h is only used during the creation of pnglibconf.h, but it 80# is safer to ensure that -DPNG_USER_CONFIG is specified throughout 81# the build by changing the CPPFLAGS passed to the initial ./configure 82# 83# 2) Add definitions of the settings you want to change to 84# CPPFLAGS; for example: 85# 86# -DPNG_DEFAULT_READ_MACROS=0 87# 88# (This would change the default to *not* use read macros.) Be 89# very careful to change only settings that don't alter the API 90# because this approach bypasses the private build checking. You 91# can also change settings from pngpriv.h (read pngpriv.h) safely 92# without API changes. Do that in the same way. 93# 94# 3) Write a new '.dfa' file (say 'pngusr.dfa') and in this file 95# provide override values for setting entries and turn option or 96# chunk values explicitly 'on' or 'off': 97# 98# setting FOO default VALUE 99# option BAR [on|off] 100# 101# Then add this file to the options.awk command line (the *first* 102# one) after this file. The make macro DFA_XTRA is provided to make 103# this easier (set it like CPPFLAGS prior to running ./configure). 104# Look at the builds below contrib/pngminim for some extreme examples 105# of how this can be used. 106# 107# Don't edit this file unless you are contributing a patch to 108# libpng and need new or modified options/settings. 109#---------------------------------------------------------------------- 110 111# The following causes commented out #undef lines to be written to 112# pnglibconf.h; this can be stopped by logunsupported=0 in a later 113# file or on the command line (after pnglibconf.dfa) 114 115logunsupported = 1 116 117# The following allows the output from configure to modify the contents of 118# pnglibconf.h 119 120@#ifdef HAVE_CONFIG_H 121@# include "config.h" 122@#endif 123 124# PNG_USER_CONFIG has to be defined on the compiler command line 125# to cause pngusr.h to be read while constructing pnglibconf.h 126# 127# If you create a private DLL you need to define the following 128# macros in the file 'pngusr.h' and set -DPNG_USER_CONFIG for 129# compilation (i.e. in CPPFLAGS.) 130# #define PNG_USER_PRIVATEBUILD \ 131# <Describes by whom and why this version of the DLL was built> 132# e.g. #define PNG_USER_PRIVATEBUILD "Build by MyCompany for xyz reasons." 133# #define PNG_USER_DLLFNAME_POSTFIX <two-letter postfix that serve to 134# distinguish your DLL from those of the official release. These 135# correspond to the trailing letters that come after the version 136# number and must match your private DLL name> 137# e.g. // private DLL "libpng13gx.dll" 138# #define PNG_USER_DLLFNAME_POSTFIX "gx" 139# 140# The following macros are also at your disposal if you want to complete the 141# DLL VERSIONINFO structure. 142# - PNG_USER_VERSIONINFO_COMMENTS 143# - PNG_USER_VERSIONINFO_COMPANYNAME 144# - PNG_USER_VERSIONINFO_LEGALTRADEMARKS 145 146# It is necessary to include configures definitions here so that AC_DEFINE 147# in configure.ac works in a comprehensible way 148@#if defined(HAVE_CONFIG_H) && !defined(PNG_NO_CONFIG_H) 149@# include "config.h" 150@#endif 151 152@#ifdef PNG_USER_CONFIG 153@# include "pngusr.h" 154@#endif 155 156# This is a special fixup for the Watcom C compiler on Windows, which has 157# multiple procedure call standards. Unless PNG_API_RULE is set explicitly 158# (i.e. if it is not defined at this point) it will be forced to '2' here when 159# using Watcom. This indicates to the other header files that Watcom behaviour 160# is required where appropriate. 161 162@#ifdef __WATCOMC__ 163@# ifndef PNG_API_RULE 164@# define PNG_API_RULE 2 /* Use Watcom calling conventions */ 165@# endif 166@#endif 167 168# IN DEVELOPMENT 169# These are currently experimental features; define them if you want (NOTE: 170# experimental options must be disabled before they are defined in this file!) 171 172# NONE 173 174# Note that PNG_USER_CONFIG only has an effect when building 175# pnglibconf.h 176 177setting USER_CONFIG 178setting USER_PRIVATEBUILD 179setting USER_DLLFNAME_POSTFIX 180setting USER_VERSIONINFO_COMMENTS 181setting USER_VERSIONINFO_COMPANYNAME 182setting USER_VERSIONINFO_LEGALTRADEMARKS 183 184# Record the 'API rule' used to select calling conventions on 185# those systems that support such things (see all the comments in 186# pngconf.h) 187# Changing this setting has a fundamental affect on the PNG ABI, 188# do not release shared libraries with this changed. 189 190setting API_RULE default 0 191 192# This allows a prefix to be added to the front of every API functon name (and 193# therefore every symbol) by redefining all the function names with the prefix 194# at the end of pnglibconf.h. It also turns on similar internal symbol renaming 195# by causing a similar build-time only file, pngprefix.h, to be generated. 196 197setting PREFIX 198 199# Implementation specific control of the optimizations, enabled by those 200# hardware or software options that need it (typically when run-time choices 201# must be made by the user) 202option SET_OPTION disabled 203 204# These options are specific to the ARM NEON hardware optimizations. At present 205# these optimizations depend on GCC specific pre-processing of an assembler (.S) 206# file so they probably won't work with other compilers. 207# 208# ARM_NEON_OPT: unset: check at compile time (__ARM_NEON__ must be defined by 209# the compiler, typically as a result of specifying 210# CC="gcc -mfpu=neon".) 211# 0: disable (even if the CPU has a NEON FPU.) 212# 1: check at run time (via ARM_NEON_{API,CHECK}) 213# 2: switch on unconditionally (inadvisable - instead pass 214# -mfpu=neon to GCC in CC) 215# When building libpng avoid using any setting other than '0'; '1' is 216# set automatically when either 'API' or 'CHECK' are configured in, 217# '2' should not be necessary as -mfpu=neon will achieve the same 218# effect as well as applying NEON optimizations to the rest of the 219# libpng code. 220# NOTE: any setting other than '0' requires ALIGNED_MEMORY 221# ARM_NEON_API: (PNG_ARM_NEON == 1) allow the optimization to be switched on 222# with png_set_option 223# ARM_NEON_CHECK: (PNG_ARM_NEON == 1) compile a run-time check to see if Neon 224# extensions are supported. This is poorly supported and 225# deprecated - use the png_set_option API. 226setting ARM_NEON_OPT 227option ARM_NEON_API disabled requires ALIGNED_MEMORY enables SET_OPTION, 228 sets ARM_NEON_OPT 1 229option ARM_NEON_CHECK disabled requires ALIGNED_MEMORY, 230 sets ARM_NEON_OPT 1 231 232# These options are specific to the PowerPC VSX hardware optimizations. 233# 234# POWERPC_VSX_OPT: unset: check at compile time (__PPC64__,__ALTIVEC__,__VSX__ 235# must be defined by the compiler, typically as a result 236# of specifying 237# "-mvsx -maltivec" compiler flags) 238# 0: disable (even if the CPU supports VSX.) 239# 1: check at run time (via POWERPC_VSX_{API,CHECK}) 240# 2: switch on unconditionally (inadvisable - instead pass 241# -mvsx -maltivec to compiler options) 242# When building libpng avoid using any setting other than '0'; '1' is 243# set automatically when either 'API' or 'CHECK' are configured in, 244# '2' should not be necessary as "-mvsx -maltivec" will achieve the same 245# effect as well as applying VSX optimizations to the rest of the 246# libpng code. 247# POWERPC_VSX_API: (PNG_POWERPC_VSX == 1) allow the optimization to be switched on 248# with png_set_option 249# POWERPC_VSX_CHECK: (PNG_POWERPC_VSX == 1) compile a run-time check to see if VSX 250# extensions are supported. This is supported not for all OSes 251# (see contrib/powerpc/README) 252setting POWERPC_VSX_OPT 253option POWERPC_VSX_API disabled enables SET_OPTION, 254 sets POWERPC_VSX_OPT 1 255option POWERPC_VSX_CHECK disabled, 256 sets POWERPC_VSX_OPT 1 257 258 259# These settings configure the default compression level (0-9) and 'strategy'; 260# strategy is as defined by the implementors of zlib. It describes the input 261# data and modifies the zlib parameters in an attempt to optimize the balance 262# between search and huffman encoding in the zlib algorithms. The defaults are 263# the zlib.h defaults - the apparently recursive definition does not arise 264# because the name of the setting is prefixed by PNG_ 265# 266# The TEXT values are the defaults when writing compressed text (all forms) 267 268# Include the zlib header so that the defaults below are known 269@# include <zlib.h> 270 271# The '@' here means to substitute the value when pnglibconf.h is built 272setting Z_DEFAULT_COMPRESSION default @Z_DEFAULT_COMPRESSION 273# TODO: why aren't these Z_RLE; zlib.h says that Z_RLE, specifically, is 274# appropriate for PNG images, maybe it doesn't exist in all versions? 275setting Z_DEFAULT_STRATEGY default @Z_FILTERED 276setting Z_DEFAULT_NOFILTER_STRATEGY default @Z_DEFAULT_STRATEGY 277setting ZLIB_VERNUM default @ZLIB_VERNUM 278 279# Linkage of: 280# 281# API: libpng API functions 282# CALLBACK: internal non-file-local callbacks 283# FUNCTION: internal non-file-local functions 284# DATA: internal non-file-local (const) data 285setting LINKAGE_API default extern 286setting LINKAGE_CALLBACK default extern 287setting LINKAGE_FUNCTION default extern 288setting LINKAGE_DATA default extern 289 290setting TEXT_Z_DEFAULT_COMPRESSION default @Z_DEFAULT_COMPRESSION 291setting TEXT_Z_DEFAULT_STRATEGY default @Z_DEFAULT_STRATEGY 292 293# Default to using the read macros 294 295setting DEFAULT_READ_MACROS default 1 296 297# The alternative is to call functions to read PNG values, if 298# the functions are turned *off* the read macros must always 299# be enabled, so turning this off will actually force the 300# USE_READ_MACROS option on (see pngconf.h) 301 302option READ_INT_FUNCTIONS requires READ 303 304# The same for write but these can only be switched off if no writing 305# is required at all - hence the use of a 'disabled', not a 'requires'. 306# If these are needed, they are enabled in the 'WRITE options' section 307# below. 308 309option WRITE_INT_FUNCTIONS disabled 310 311# Error controls 312# 313# WARNINGS: normally on, if off no warnings are generated 314# ERROR_TEXT: normally on, if off errors happen but there is no message 315# ERROR_NUMBERS: unimplemented feature, therefore disabled 316# BENIGN_ERRORS: support for just issuing warnings for recoverable errors 317# 318# BENIGN_READ_ERRORS: 319# By default recoverable errors on read should just generate warnings, 320# generally safe but PNG files that don't conform to the specification will 321# be accepted if a meaningful result can be produced. 322# 323# BENIGN_WRITE_ERRORS: 324# By default recoverable errors on write should just generate warnings, 325# not generally safe because this allows the application to write invalid 326# PNG files. Applications should enable this themselves; it's useful 327# because it means that a failure to write an ancilliary chunk can often be 328# ignored. 329 330option WARNINGS 331option ERROR_TEXT 332option ERROR_NUMBERS disabled 333 334option BENIGN_ERRORS 335option BENIGN_WRITE_ERRORS requires BENIGN_ERRORS disabled 336option BENIGN_READ_ERRORS requires BENIGN_ERRORS 337 338 339# Generic options - affect both read and write. 340 341option MNG_FEATURES 342 343# Arithmetic options, the first is the big switch that chooses between internal 344# floating and fixed point arithmetic implementations - it does not affect any 345# APIs. The second two (the _POINT settings) switch off individual APIs. 346# 347# Prior to libpng 1.6.8 one of the API (_POINT) variants had to be selected. At 348# 1.6.8 this restriction has been removed; the simplified API can be used 349# without enabling any of the low level fixed/floating APIs. 350 351option FLOATING_ARITHMETIC 352option FLOATING_POINT 353option FIXED_POINT 354 355# This protects us against compilers that run on a windowing system 356# and thus don't have or would rather us not use the stdio types: 357# stdin, stdout, and stderr. The only one currently used is stderr 358# in png_error() and png_warning(). #defining PNG_NO_CONSOLE_IO will 359# prevent these from being compiled and used. #defining PNG_NO_STDIO 360# will also prevent these, plus will prevent the entire set of stdio 361# macros and functions (FILE *, printf, etc.) from being compiled and used, 362# unless (PNG_DEBUG > 0) has been #defined. 363 364option STDIO 365option CONSOLE_IO requires STDIO 366 367# Note: prior to 1.5.0 this option could not be disabled if STDIO 368# was enabled. Prior to 1.5.3 this option required STDIO 369 370option TIME_RFC1123 371 372# PNG_SETJMP_NOT_SUPPORTED is an old equivalent for NO_SETJMP 373 374option SETJMP 375= NO_SETJMP SETJMP_NOT_SUPPORTED 376 377# If this is disabled it is not possible for apps to get the 378# values from the 'info' structure, this effectively removes 379# quite a lot of the READ API. 380 381option EASY_ACCESS 382 383# Added at libpng-1.2.0 384 385option USER_MEM 386 387# Added at libpng-1.4.0 388 389option IO_STATE 390 391# Libpng limits: limit the size of images and data on read. 392# 393# If this option is disabled all the limit checking code will be disabled: 394 395option USER_LIMITS requires READ 396 397# The default settings given below for the limits mean that libpng will 398# limit the size of images or the size of data in ancilliary chunks to less 399# than the specification or implementation limits. Settings have the 400# following interpretations: 401# 402# USER_WIDTH_MAX: maximum width of an image that will be read 403# USER_HEIGHT_MAX: maximum height 404# USER_CHUNK_MALLOC_MAX: maximum in-memory (decompressed) size of a single chunk 405# USER_CHUNK_CACHE_MAX: maximum number of chunks to be cached 406# 407# Only chunks that are variable in number are counted towards the 408 409# Use 0x7fffffff for unlimited 410setting USER_WIDTH_MAX default 1000000 411setting USER_HEIGHT_MAX default 1000000 412 413# Use 0 for unlimited 414setting USER_CHUNK_CACHE_MAX default 1000 415setting USER_CHUNK_MALLOC_MAX default 8000000 416 417# If this option is enabled APIs to set the above limits at run time are added; 418# without this the hardwired (compile time) limits will be used. 419option SET_USER_LIMITS requires USER_LIMITS 420 421# All of the following options relate to code capabilities for 422# processing image data before creating a PNG or after reading one. 423# You can remove these capabilities safely and still be PNG 424# conformant, however the library that results is still non-standard. 425# See the comments above about how to change options and settings. 426 427# READ options 428# 429# WARNING: in libpng 1.5 maintained configuration compatibility with earlier 430# versions. In some cases turning off an option turned off other options, in 431# others it was ineffective unless dependent options were also turned off. 432# Libpng 1.6 changes this: in general if you turn off an option that affects 433# APIs it stays off and simply disables APIs that depend on it. 434# 435# As a result if you simply port the libpng 1.5 configuration to libpng 1.6 you 436# will probably see build failures due to missing APIs. Fixing these failures 437# requires some, perhaps considerable, knowledge of what your libpng using 438# applications are doing, fortunately there is no great reason for you to move 439# to libpng 1.6; the new interfaces in 1.6 will take several years to become 440# popular. 441 442option READ enables READ_INTERLACING SET_OPTION 443 444# Disabling READ_16BIT does not disable reading 16-bit PNG files, but it 445# forces them to be chopped down to 8-bit, and disables any 16-bit 446# processing after that has happened. You need to be sure to enable 447# READ_SCALE_16_TO_8 or READ_STRIP_16_TO_8 when you disable READ_16BIT for 448# this to work properly. You should disable the other option if you need to 449# ensure a particular conversion (otherwise the app can chose.) 450 451option READ_16BIT requires READ enables 16BIT 452 453option READ_QUANTIZE requires READ 454 455option READ_TRANSFORMS requires READ 456= NO_READ_TRANSFORMS READ_TRANSFORMS_NOT_SUPPORTED 457 458# Read gamma handling. Gamma processing is a core part of libpng and many of 459# the capabilities are dependent on libpng performing gamma correction. 460# 461# In libpng 1.6 disabling gamma processing (setting PNG_NO_READ_GAMMA) 462# consistently disables those parts of the API that depend on it. Prior to 463# 1.6.0 this was not true; the results were unpredictable and varied between 464# releases. 465# 466# If you disable gamma processing and your program no longer compiles you need 467# to ask whether you really need the APIs that are missing. If you do then you 468# almost certainly need the gamma processing. 469# 470# If you handle gamma issues outside libpng then you do not need the libpng 471# gamma processing; and it is an enormous waste of space. You just need to 472# remove the use of libpng APIs that depend on it. 473option READ_GAMMA requires READ_TRANSFORMS, READ_gAMA, READ_sRGB 474 475option READ_ALPHA_MODE requires READ_TRANSFORMS, READ_GAMMA 476option READ_BACKGROUND requires READ_TRANSFORMS, READ_STRIP_ALPHA, READ_GAMMA 477option READ_BGR requires READ_TRANSFORMS 478option READ_EXPAND_16 requires READ_TRANSFORMS, READ_16BIT, READ_EXPAND 479option READ_EXPAND requires READ_TRANSFORMS 480option READ_FILLER requires READ_TRANSFORMS 481option READ_GRAY_TO_RGB requires READ_TRANSFORMS 482option READ_INVERT_ALPHA requires READ_TRANSFORMS 483option READ_INVERT requires READ_TRANSFORMS 484option READ_PACK requires READ_TRANSFORMS 485option READ_PACKSWAP requires READ_TRANSFORMS 486option READ_RGB_TO_GRAY requires READ_TRANSFORMS, READ_GAMMA enables COLORSPACE 487option READ_SCALE_16_TO_8 requires READ_TRANSFORMS 488option READ_SHIFT requires READ_TRANSFORMS 489option READ_STRIP_16_TO_8 requires READ_TRANSFORMS 490option READ_STRIP_ALPHA requires READ_TRANSFORMS 491option READ_SWAP_ALPHA requires READ_TRANSFORMS 492option READ_SWAP requires READ_TRANSFORMS, READ_16BIT 493option READ_USER_TRANSFORM requires READ_TRANSFORMS 494 495option PROGRESSIVE_READ requires READ 496option SEQUENTIAL_READ requires READ 497 498# You can define PNG_NO_PROGRESSIVE_READ if you don't do progressive reading. 499# This is not talking about interlacing capability! You'll still have 500# interlacing unless you turn off the following which is required 501# for PNG-compliant decoders. (In other words, do not do this - in 502# fact it can't be disabled from the command line!) 503#option READ_INTERLACING requires READ 504 505option READ_COMPOSITE_NODIV requires READ 506= NO_READ_COMPOSITE_NODIV NO_READ_COMPOSITED_NODIV 507 508# Inch conversions 509 510option INCH_CONVERSIONS 511= INCH_CONVERSIONS INCH_CONVERSIONS 512 513# API to build a grayscale palette 514# NOTE: this is not used internally by libpng at present. 515 516option BUILD_GRAYSCALE_PALETTE 517 518# WRITE options 519 520option WRITE enables WRITE_INT_FUNCTIONS 521 522# Disabling WRITE_16BIT prevents 16-bit PNG files from being 523# generated. 524option WRITE_16BIT requires WRITE enables 16BIT 525 526option WRITE_TRANSFORMS requires WRITE 527= NO_WRITE_TRANSFORMS WRITE_TRANSFORMS_NOT_SUPPORTED 528 529option WRITE_SHIFT requires WRITE_TRANSFORMS 530option WRITE_PACK requires WRITE_TRANSFORMS 531option WRITE_BGR requires WRITE_TRANSFORMS 532option WRITE_SWAP requires WRITE_TRANSFORMS, WRITE_16BIT 533option WRITE_PACKSWAP requires WRITE_TRANSFORMS 534option WRITE_INVERT requires WRITE_TRANSFORMS 535option WRITE_FILLER requires WRITE_TRANSFORMS 536option WRITE_SWAP_ALPHA requires WRITE_TRANSFORMS 537option WRITE_INVERT_ALPHA requires WRITE_TRANSFORMS 538option WRITE_USER_TRANSFORM requires WRITE_TRANSFORMS 539 540# This is not required for PNG-compliant encoders, but can cause 541# trouble if left undefined 542 543option WRITE_INTERLACING requires WRITE 544 545# Deprecated, will be removed. 546option WRITE_WEIGHTED_FILTER requires WRITE 547 548option WRITE_FLUSH requires WRITE 549 550# Note: these can be turned off explicitly if not required by the 551# apps implementing the user transforms 552option USER_TRANSFORM_PTR if READ_USER_TRANSFORM, WRITE_USER_TRANSFORM 553option USER_TRANSFORM_INFO if READ_USER_TRANSFORM, WRITE_USER_TRANSFORM 554 555# This enables API to set compression parameters for compressing 556# non-IDAT chunks (zTXt, iTXt, iCCP, and unknown chunks). This feature 557# was added at libpng-1.5.3. 558option WRITE_CUSTOMIZE_ZTXT_COMPRESSION requires WRITE 559option WRITE_CUSTOMIZE_COMPRESSION requires WRITE 560 561# Any chunks you are not interested in, you can undef here. The 562# ones that allocate memory may be expecially important (hIST, 563# tEXt, zTXt, tRNS, pCAL). Others will just save time and make png_info 564# a bit smaller. 565 566# The size of the png_text structure changed in libpng-1.0.6 when 567# iTXt support was added. iTXt support was turned off by default through 568# libpng-1.2.x, to support old apps that malloc the png_text structure 569# instead of calling png_set_text() and letting libpng malloc it. It 570# was turned on by default in libpng-1.4.0. 571 572option READ_ANCILLARY_CHUNKS requires READ 573# PNG_READ_ANCILLARY_CHUNKS_NOT_SUPPORTED is deprecated. 574= NO_READ_ANCILLARY_CHUNKS READ_ANCILLARY_CHUNKS_NOT_SUPPORTED 575 576option WRITE_ANCILLARY_CHUNKS requires WRITE 577# PNG_WRITE_ANCILLARY_CHUNKS_NOT_SUPPORTED is deprecated. 578= NO_WRITE_ANCILLARY_CHUNKS WRITE_ANCILLARY_CHUNKS_NOT_SUPPORTED 579 580# These options disable *all* the text chunks if turned off 581 582option TEXT disabled 583option READ_TEXT requires READ_ANCILLARY_CHUNKS enables TEXT 584option WRITE_TEXT requires WRITE_ANCILLARY_CHUNKS enables TEXT 585 586# Moved to pnglibconf.h at libpng-1.5.0 587# Feature support: in 1.4 this was in pngconf.h, but the following 588# features have no affect on the libpng API. Add library 589# only features to the end of this list. Add features that 590# affect the API above. (Note: the list of chunks follows 591# the library-only settings.) 592# 593# BUILD TIME ONLY OPTIONS 594# These options do not affect the API but rather alter how the 595# API is implemented, they get recorded in pnglibconf.h, but 596# can't be changed by the application. 597 598# Colorspace support (enabled as required); just the support for colorant 599# information. Gamma support, likewise, is just support for the gamma 600# information, READ_GAMMA is required for gamma transformations (so it 601# is possible to read PNG gamma without enabling all the libpng transform 602# code - do this for applications that do their own gamma processing) 603# 604# As of 1.6.0 COLORSPACE is only useful if the application processes the 605# information; this is because the library does not do any colorspace 606# processing, it just validates the data in the PNG file. 607 608option GAMMA disabled 609option COLORSPACE enables GAMMA disabled 610 611# When an ICC profile is read, or png_set, it will be checked for a match 612# against known sRGB profiles if the sRGB handling is enabled. The 613# PNG_sRGB_PROFILE_CHECKS setting controls how much work is done during the 614# check: 615# 616# -1: Don't do any sRGB profile checking. 617# 618# 0: Just validate the profile MD5 signature if present, otherwise use 619# the checks in option 1. 620# 621# 1: Additionally check the length, intent and adler32 checksum of the 622# actual data. If enabled this will reject known profiles that have 623# had the rendering intent in the header changed as well as other edits 624# done without updating the checksum. See the discussion below. 625# 626# 2: Additionally checksum all the data using the ethernet CRC32 algorithm. 627# This makes it more difficult to fake profiles and makes it less likely 628# to get a false positive on profiles with no signature, but is probably 629# just a waste of time since all currently approved ICC sRGB profiles have 630# a secure MD5 signature. 631# 632# The rendering intent. An ICC profile stores an intended rendering intent, 633# but does not include the value in the signature. The intent is documented 634# as the intent that should be used when combining two profiles. The sRGB 635# profile is intended, however, to be used with any of the four defined intents. 636# For this reason the sRGB chunk includes an 'intent' to be used when displaying 637# the image (intent is really a property of the image not the profile.) 638# 639# Unfortunately the iCCP chunk does not. It may therefore be that some 640# applications modify the intent in profiles (including sRGB profiles) to work 641# round this problem. Selecting an option other than option '0' will cause such 642# modified profiles to be rejected. 643# 644# Security. The use of Adler32 and CRC32 checksums does not help significantly 645# with any security issues. It is relatively easy to produce arbitrary profiles 646# with the required checksums on current computer systems. Nevertheless 647# security does not seem to be an issue because the only consequence of a false 648# positive is a false assertion that the profile is an sRGB profile. This might 649# be used to hide data from libpng using applications, but it doesn't seem 650# possible to damage them. 651 652setting sRGB_PROFILE_CHECKS default 2 653 654# Artificially align memory - the code typically aligns to 8 byte 655# boundaries if this is switched on, it's a small waste of space 656# but can help (in theory) on some architectures. Only affects 657# internal structures. Added at libpng 1.4.0 658 659option ALIGNED_MEMORY 660 661# Buggy compilers (e.g., gcc 2.7.2.2) need PNG_NO_POINTER_INDEXING 662# See png[wr]util.c, normally this should always be *on* 663 664option POINTER_INDEXING 665 666# Other defines for things like memory and the like can go here. 667 668# BUILD TIME SETTINGS 669# Like build time options these do not affect the API, but they 670# may be useful to applications because they record details of 671# how the API will behave particularly with regard to overall 672# accuracy. 673 674# This controls how fine the quantizing gets. As this allocates 675# a largish chunk of memory (32K), those who are not as concerned 676# with quantizing quality can decrease some or all of these. 677 678setting QUANTIZE_RED_BITS default 5 679setting QUANTIZE_GREEN_BITS default 5 680setting QUANTIZE_BLUE_BITS default 5 681 682# This controls how fine the gamma correction becomes when you 683# are only interested in 8 bits anyway. Increasing this value 684# results in more memory being used, and more pow() functions 685# being called to fill in the gamma tables. Don't set this value 686# less than 8, and even that may not work (I haven't tested it). 687 688setting MAX_GAMMA_8 default 11 689 690# This controls how much a difference in gamma we can tolerate before 691# we actually start doing gamma conversion, it's a fixed point value, 692# so the default below is 0.05, meaning libpng ignores corrections in 693# the range 0.95 to 1.05 694 695setting GAMMA_THRESHOLD_FIXED default 5000 696 697# Precision to use when converting a floating point value to a PNG 698# extension format string in an sCAL chunk (only relevant if the 699# floating point API is enabled) 700 701setting sCAL_PRECISION default 5 702 703# This is the size of the compression buffer, and thus the size of 704# an IDAT chunk. Make this whatever size you feel is best for your 705# machine. One of these will be allocated per png_struct. When this 706# is full, it writes the data to the disk, and does some other 707# calculations. Making this an extremely small size may slow 708# the library down, but you may want to experiment to determine 709# where it becomes significant, if you are concerned with memory 710# usage. Note that zlib allocates at least 32Kb also. For readers, 711# this describes the size of the buffer available to read the data in. 712# Unless this gets smaller than the size of a row (compressed), 713# it should not make much difference how big this is. 714 715setting ZBUF_SIZE default 8192 716 717# This is the size of the decompression buffer used when counting or checking 718# the decompressed size of an LZ stream from a compressed ancilliary chunk; the 719# decompressed data is never used so a different size may be optimal. This size 720# was determined using contrib/libtests/timepng.c with compressed zTXt data 721# around 11MByte in size. Slight speed improvements (up to about 14% in 722# timepng) can be achieved by very large increases (to 32kbyte) on regular data, 723# but highly compressible data shows only around 2% improvement. The size is 724# chosen to minimize the effects of DoS attacks based on using very large 725# amounts of highly compressible data. 726 727setting INFLATE_BUF_SIZE default 1024 728 729# This is the maximum amount of IDAT data that the sequential reader will 730# process at one time. The setting does not affect the size of IDAT chunks 731# read, just the amount read at once. Neither does it affect the progressive 732# reader, which processes just the amount of data the application gives it. 733# The sequential reader is currently unable to process more than one IDAT at 734# once - it has to read and process each one in turn. There is no point setting 735# this to a value larger than the IDAT chunks typically encountered (it would 736# just waste memory) but there may be some point in reducing it below the value 737# of ZBUF_SIZE (the size of IDAT chunks written by libpng.) 738 739setting IDAT_READ_SIZE default PNG_ZBUF_SIZE 740 741# Ancillary chunks 742chunk bKGD 743chunk cHRM enables COLORSPACE 744chunk eXIf 745chunk gAMA enables GAMMA 746chunk hIST 747chunk iCCP enables COLORSPACE, GAMMA 748chunk iTXt enables TEXT 749chunk oFFs 750chunk pCAL 751chunk pHYs 752chunk sBIT 753chunk sCAL 754chunk sPLT 755chunk sRGB enables COLORSPACE, GAMMA, SET_OPTION 756chunk tEXt requires TEXT 757chunk tIME 758chunk tRNS 759chunk zTXt enables TEXT 760 761# This only affects support of the optional PLTE chunk in RGB and RGBA 762# images. Notice that READ_ANCILLARY_CHUNKS therefore disables part 763# of the regular chunk reading too. 764 765option READ_OPT_PLTE requires READ_ANCILLARY_CHUNKS 766 767# Unknown chunk handling 768# 769# 'UNKNOWN_CHUNKS' is a global option to disable all unknown chunk handling on 770# read or write; everything else below requires it (directly or indirectly). 771option UNKNOWN_CHUNKS 772 773# There are three main options to control the ability to read and write unknown 774# chunks. If either read option is turned on then unknown chunks will be read, 775# otherwise they are skipped. If the write option is turned on unknown chunks 776# set by png_set_unknown_chunks will be written otherwise it is an error to call 777# that API on a write struct. 778option WRITE_UNKNOWN_CHUNKS requires WRITE requires UNKNOWN_CHUNKS 779option WRITE_UNKNOWN_CHUNKS enables STORE_UNKNOWN_CHUNKS 780 781# The first way to read user chunks is to have libpng save them for a later call 782# to png_get_unknown_chunks, the application must call 783# png_set_keep_unknown_chunks to cause this to actually happen (see png.h) 784option SAVE_UNKNOWN_CHUNKS requires READ requires SET_UNKNOWN_CHUNKS 785option SAVE_UNKNOWN_CHUNKS enables READ_UNKNOWN_CHUNKS, STORE_UNKNOWN_CHUNKS 786 787# The second approach is to use an application provided callback to process the 788# chunks, the callback can either handle the chunk entirely itself or request 789# that libpng store the chunk for later retrieval via png_get_unknown_chunks. 790# 791# NOTE: If STORE_UNKNOWN_CHUNKS is not enabled (which is the default if 792# both SAVE_UNKNOWN_CHUNKS and WRITE_UNKNOWN_CHUNKS are disabled) then a 793# 0 result from the callback will be ignored because no support for saving 794# unknown chunks has been compiled in. The normal symptom is that your app 795# fails to compile because png_get_unknown_chunks is no longer defined in png.h. 796# If you encounter this issue simply enable STORE_UNKNOWN_CHUNKS in your build. 797# 798# Note that there is no 'WRITE_USER_CHUNKS' so the USER_CHUNKS option is always 799# the same as READ_USER_CHUNKS at present 800option READ_USER_CHUNKS requires READ, UNKNOWN_CHUNKS 801option READ_USER_CHUNKS enables READ_UNKNOWN_CHUNKS, USER_CHUNKS 802 803# Two further options are provided to allow detailed control of the handling. 804# The first enables png_set_keep_unknown_chunks; this allows the default to be 805# changed from discarding unknown chunks and allows per-chunk control. This is 806# required to use the SAVE_UNKNOWN_CHUNKS option. If enabled this option also 807# applies to write (see png.h), otherwise the write API simply writes all the 808# chunks it is given. 809# 810# The second option extends the unknown handling to allow known chunks to be 811# handled as though they were unknown. This option doesn't change any APIs, it 812# merely turns on the code to check known as well as unknown chunks. 813# 814# This option no longer affects the write code. It can be safely disabled and 815# will prevent applications stopping libpng reading known chunks. 816option SET_UNKNOWN_CHUNKS requires UNKNOWN_CHUNKS 817option HANDLE_AS_UNKNOWN requires SET_UNKNOWN_CHUNKS 818 819# The following options are derived from the above and should not be turned on 820# explicitly. 821option READ_UNKNOWN_CHUNKS requires UNKNOWN_CHUNKS disabled 822option STORE_UNKNOWN_CHUNKS requires UNKNOWN_CHUNKS disabled 823 824option CONVERT_tIME requires WRITE_ANCILLARY_CHUNKS 825# The "tm" structure is not supported on WindowsCE 826 827@#ifdef _WIN32_WCE 828@# define PNG_NO_CONVERT_tIME 829@#endif 830 831option WRITE_FILTER requires WRITE 832 833option SAVE_INT_32 disabled 834# png_save_int_32 is required internally for writing the ancillary chunks oFFs 835# and pCAL and for both reading and writing iCCP (for the generation/checking of 836# the corresponding cHRM/gAMA chunks) if full ICC is supported. 837 838# added at libpng-1.5.4 839 840option WRITE_OPTIMIZE_CMF requires WRITE 841 842option READ_COMPRESSED_TEXT disabled 843option READ_iCCP enables READ_COMPRESSED_TEXT 844option READ_iTXt enables READ_COMPRESSED_TEXT 845option READ_zTXt enables READ_COMPRESSED_TEXT 846 847option WRITE_oFFs enables SAVE_INT_32 848option WRITE_pCAL enables SAVE_INT_32 849option WRITE_cHRM enables SAVE_INT_32 850 851option WRITE_COMPRESSED_TEXT disabled 852option WRITE_iCCP enables WRITE_COMPRESSED_TEXT 853option WRITE_iTXt enables WRITE_COMPRESSED_TEXT 854option WRITE_zTXt enables WRITE_COMPRESSED_TEXT 855 856# Turn this off to disable png_read_png() and png_write_png() and 857# leave the row_pointers member out of the info structure. 858 859option INFO_IMAGE 860 861# added at libpng-1.5.10 862# Turn this off to disable warning about invalid palette index and 863# leave the num_palette_max member out of the png structure. 864 865option CHECK_FOR_INVALID_INDEX enables READ_CHECK_FOR_INVALID_INDEX 866option CHECK_FOR_INVALID_INDEX enables WRITE_CHECK_FOR_INVALID_INDEX 867option READ_CHECK_FOR_INVALID_INDEX requires READ, CHECK_FOR_INVALID_INDEX 868option WRITE_CHECK_FOR_INVALID_INDEX requires WRITE, CHECK_FOR_INVALID_INDEX 869 870# added at libpng-1.5.15 871option GET_PALETTE_MAX enables READ_GET_PALETTE_MAX WRITE_GET_PALETTE_MAX 872option READ_GET_PALETTE_MAX requires READ_CHECK_FOR_INVALID_INDEX disabled 873option WRITE_GET_PALETTE_MAX requires WRITE_CHECK_FOR_INVALID_INDEX disabled 874 875# Simplified API options (added at libpng-1.6.0) 876# In libpng 1.6.8 the handling of these options was changed to used 'requires' 877# throughout, so that disabling some of the low level support always disables 878# the base simplified read/write API. This much simplifies the handling and 879# makes 'everything = off' work in a more intuitive way. It eliminates a 880# previously reported feature that APIs previously enabled by the simplified 881# API couldn't be turned off without explicitly turning off the simplified 882# APIs. 883# 884# Read: 885option SIMPLIFIED_READ, 886 requires SEQUENTIAL_READ, READ_TRANSFORMS, SETJMP, BENIGN_ERRORS, 887 READ_EXPAND, READ_16BIT, READ_EXPAND_16, READ_SCALE_16_TO_8, 888 READ_RGB_TO_GRAY, READ_ALPHA_MODE, READ_BACKGROUND, READ_STRIP_ALPHA, 889 READ_FILLER, READ_SWAP, READ_PACK, READ_GRAY_TO_RGB, READ_GAMMA, 890 READ_tRNS, READ_bKGD, READ_gAMA, READ_cHRM, READ_sRGB, READ_sBIT 891 892# AFIRST and BGR read options: 893# Prior to libpng 1.6.8 these were disabled but switched on if the low level 894# libpng routines that do the swaps were enabled. This worked but was 895# confusing. In libpng 1.6.8 the options were changed to simple 'requires' 896# and are enabled by default. This should work the same way in practice. 897option SIMPLIFIED_READ_AFIRST enables FORMAT_AFIRST, 898 requires SIMPLIFIED_READ READ_SWAP_ALPHA 899 900option SIMPLIFIED_READ_BGR enables FORMAT_BGR, 901 requires SIMPLIFIED_READ READ_BGR 902 903# Write: 904option SIMPLIFIED_WRITE, 905 requires WRITE, SETJMP, WRITE_SWAP, WRITE_PACK, 906 WRITE_tRNS, WRITE_gAMA, WRITE_sRGB, WRITE_cHRM 907 908# 1.6.22: allow simplified write without stdio support: 909option SIMPLIFIED_WRITE_STDIO requires SIMPLIFIED_WRITE STDIO 910 911option SIMPLIFIED_WRITE_AFIRST enables FORMAT_AFIRST, 912 requires SIMPLIFIED_WRITE WRITE_SWAP_ALPHA 913 914option SIMPLIFIED_WRITE_BGR enables FORMAT_BGR, 915 requires SIMPLIFIED_WRITE WRITE_BGR 916 917# Formats: 918option FORMAT_AFIRST disabled 919option FORMAT_BGR disabled 920