Fix OOB write in BuildHuffmanTable.

First, BuildHuffmanTable is called to check if the data is valid.
If it is and the table is not big enough, more memory is allocated.

This will make sure that valid (but unoptimized because of unbalanced
codes) streams are still decodable.

Bug: chromium:1479274
Change-Id: I31c36dbf3aa78d35ecf38706b50464fd3d375741
(cherry picked from commit 902bc91903)
(cherry picked from commit 2af26267cd)

.gitattributes

@@ -1,5 +1,6 @@
 .gitattributes export-ignore
 .gitignore export-ignore
 .mailmap export-ignore
+*.bat text eol=crlf
 *.pdf -text -diff
 *.ppm -text -diff

.gitignore

@@ -1,6 +1,7 @@
 *.l[ao]
 *.[ao]
 *.pc
+.DS_Store
 .deps
 .libs
 /aclocal.m4
@@ -20,9 +21,11 @@
 Makefile
 Makefile.in
 examples/anim_diff
+examples/anim_dump
 examples/[cdv]webp
 examples/gif2webp
 examples/img2webp
+examples/webpinfo
 examples/webpmux
 src/webp/config.h*
 src/webp/stamp-h1
@@ -31,11 +34,21 @@ src/webp/stamp-h1
 *.idb
 *.pdb
 /iosbuild
-/WebP.framework
+/xcframeworkbuild
+/WebP*.*framework
 CMakeCache.txt
 CMakeFiles/
 cmake_install.cmake
 .gradle
 /build
 extras/get_disto
+extras/vwebp_sdl
 extras/webp_quality
+tests/fuzzer/advanced_api_fuzzer
+tests/fuzzer/animation_api_fuzzer
+tests/fuzzer/animdecoder_fuzzer
+tests/fuzzer/animencoder_fuzzer
+tests/fuzzer/demux_api_fuzzer
+tests/fuzzer/enc_dec_fuzzer
+tests/fuzzer/mux_demux_api_fuzzer
+tests/fuzzer/simple_api_fuzzer

AUTHORS

@@ -2,9 +2,11 @@ Contributors:
 - Aidan O'Loan (aidanol at gmail dot com)
 - Alan Browning (browning at google dot com)
 - Charles Munger (clm at google dot com)
+- Cheng Yi (cyi at google dot com)
 - Christian Duvivier (cduvivier at google dot com)
 - Djordje Pesut (djordje dot pesut at imgtec dot com)
 - Hui Su (huisu at google dot com)
+- Ingvar Stepanyan (rreverser at google dot com)
 - James Zern (jzern at google dot com)
 - Jan Engelhardt (jengelh at medozas dot de)
 - Jehan (jehan at girinstud dot io)
@@ -21,6 +23,7 @@ Contributors:
 - Mislav Bradac (mislavm at google dot com)
 - Nico Weber (thakis at chromium dot org)
 - Noel Chromium (noel at chromium dot org)
+- Oliver Wolff (oliver dot wolff at qt dot io)
 - Owen Rodley (orodley at google dot com)
 - Parag Salasakar (img dot mips1 at gmail dot com)
 - Pascal Massimino (pascal dot massimino at gmail dot com)
@@ -39,5 +42,6 @@ Contributors:
 - Vikas Arora (vikasa at google dot com)
 - Vincent Rabaud (vrabaud at google dot com)
 - Vlad Tsyrklevich (vtsyrklevich at chromium dot org)
+- Wan-Teh Chang (wtc at google dot com)
 - Yang Zhang (yang dot zhang at arm dot com)
 - Yannis Guyon (yguyon at google dot com)

CMakeLists.txt

@@ -1,9 +1,24 @@
+##
+##  Copyright (c) 2020 Google LLC.
+##
+##  Use of this source code is governed by a BSD-style license
+##  that can be found in the LICENSE file in the root of the source
+##  tree. An additional intellectual property rights grant can be found
+##  in the file PATENTS.  All contributing project authors may
+##  be found in the AUTHORS file in the root of the source tree.
+
 cmake_minimum_required(VERSION 3.5)
 
 project(WebP C)
 
 # Options for coder / decoder executables.
-option(WEBP_ENABLE_SIMD "Enable any SIMD optimization." ON)
+if(NOT EMSCRIPTEN)
+  # Disable SIMD on Emscripten by default, as it's a new unstable Wasm feature.
+  # Users can still explicitly opt-in to make a SIMD-enabled build.
+  set(WEBP_ENABLE_SIMD_DEFAULT ON)
+endif()
+option(WEBP_ENABLE_SIMD "Enable any SIMD optimization."
+       ${WEBP_ENABLE_SIMD_DEFAULT})
 option(WEBP_BUILD_ANIM_UTILS "Build animation utilities." ON)
 option(WEBP_BUILD_CWEBP "Build the cwebp command line tool." ON)
 option(WEBP_BUILD_DWEBP "Build the dwebp command line tool." ON)
@@ -26,7 +41,6 @@ if(WIN32)
 endif()
 
 if(WEBP_BUILD_WEBP_JS)
-  set(WEBP_ENABLE_SIMD OFF)
   set(WEBP_BUILD_ANIM_UTILS OFF)
   set(WEBP_BUILD_CWEBP OFF)
   set(WEBP_BUILD_DWEBP OFF)
@@ -36,6 +50,10 @@ if(WEBP_BUILD_WEBP_JS)
   set(WEBP_BUILD_WEBPINFO OFF)
   set(WEBP_BUILD_WEBPMUX OFF)
   set(WEBP_BUILD_EXTRAS OFF)
+
+  if(WEBP_ENABLE_SIMD)
+    message("wasm2js does not support SIMD, disabling webp.js generation.")
+  endif()
 endif()
 
 set(WEBP_DEP_LIBRARIES)
@@ -43,7 +61,7 @@ set(WEBP_DEP_INCLUDE_DIRS)
 
 if(NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release"
-      CACHE "Build type: Release, Debug, MinSizeRel or RelWithDebInfo" STRING
+      CACHE STRING "Build type: Release, Debug, MinSizeRel or RelWithDebInfo"
       FORCE)
 endif()
 
@@ -164,6 +182,29 @@ add_definitions(-DHAVE_CONFIG_H)
 
 # ##############################################################################
 # Build the webpdecoder library.
+
+# Creates a source file with an unused stub function in $CMAKE_BINARY_DIR and
+# adds it to the specified target. Currently used only with Xcode.
+#
+# See also:
+# https://cmake.org/cmake/help/v3.18/command/add_library.html#object-libraries
+# "Some native build systems (such as Xcode) may not like targets that have
+# only object files, so consider adding at least one real source file to any
+# target that references $<TARGET_OBJECTS:objlib>."
+function(libwebp_add_stub_file TARGET)
+  set(stub_source_dir "${CMAKE_BINARY_DIR}")
+  set(stub_source_file
+      "${stub_source_dir}/libwebp_${TARGET}_stub.c")
+  set(stub_source_code
+      "// Generated file. DO NOT EDIT!\n"
+      "// C source file created for target ${TARGET}.\n"
+      "void libwebp_${TARGET}_stub_function(void)\;\n"
+      "void libwebp_${TARGET}_stub_function(void) {}\n")
+  file(WRITE "${stub_source_file}" ${stub_source_code})
+
+  target_sources(${TARGET} PRIVATE ${stub_source_file})
+endfunction()
+
 if(MSVC)
   # avoid security warnings for e.g., fopen() used in the examples.
   add_definitions(-D_CRT_SECURE_NO_WARNINGS)
@@ -190,6 +231,9 @@ add_library(webpdecoder
             $<TARGET_OBJECTS:webpdecode>
             $<TARGET_OBJECTS:webpdspdecode>
             $<TARGET_OBJECTS:webputilsdecode>)
+if(XCODE)
+  libwebp_add_stub_file(webpdecoder)
+endif()
 target_link_libraries(webpdecoder ${WEBP_DEP_LIBRARIES})
 target_include_directories(
   webpdecoder
@@ -229,6 +273,9 @@ add_library(webp
             $<TARGET_OBJECTS:webpdsp>
             $<TARGET_OBJECTS:webpencode>
             $<TARGET_OBJECTS:webputils>)
+if(XCODE)
+  libwebp_add_stub_file(webp)
+endif()
 target_link_libraries(webp ${WEBP_DEP_LIBRARIES})
 target_include_directories(webp
                            PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
@@ -349,20 +396,21 @@ if(WEBP_BUILD_ANIM_UTILS
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "EXAMPLEUTIL_SRCS"
                     "example_util_[^ ]*")
   list(APPEND EXAMPLEUTIL_SRCS ${CMAKE_CURRENT_SOURCE_DIR}/examples/stopwatch.h)
-  add_library(exampleutil ${EXAMPLEUTIL_SRCS})
+  add_library(exampleutil STATIC ${EXAMPLEUTIL_SRCS})
+  target_link_libraries(exampleutil imageioutil)
   target_include_directories(
     exampleutil
     PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>)
 
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEIOUTILS_SRCS"
                     "imageio_util_[^ ]*")
-  add_library(imageioutil ${IMAGEIOUTILS_SRCS})
+  add_library(imageioutil STATIC ${IMAGEIOUTILS_SRCS})
   target_link_libraries(imageioutil webp)
 
   # Image-decoding utility library.
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEDEC_SRCS"
                     "imagedec_[^ ]*")
-  add_library(imagedec ${IMAGEDEC_SRCS})
+  add_library(imagedec STATIC ${IMAGEDEC_SRCS})
   target_link_libraries(imagedec
                         imageioutil
                         webpdemux
@@ -372,8 +420,8 @@ if(WEBP_BUILD_ANIM_UTILS
   # Image-encoding utility library.
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEENC_SRCS"
                     "imageenc_[^ ]*")
-  add_library(imageenc ${IMAGEENC_SRCS})
-  target_link_libraries(imageenc webp)
+  add_library(imageenc STATIC ${IMAGEENC_SRCS})
+  target_link_libraries(imageenc imageioutil webp)
 
   set_property(TARGET exampleutil
                       imageioutil
@@ -466,12 +514,12 @@ if(WEBP_BUILD_VWEBP)
     target_link_libraries(vwebp
                           ${OPENGL_LIBRARIES}
                           exampleutil
-                          GLUT::GLUT
+                          ${GLUT_glut_LIBRARY}
                           imageioutil
                           webp
                           webpdemux)
     target_include_directories(vwebp
-                               PRIVATE GLUT::GLUT
+                               PRIVATE ${GLUT_INCLUDE_DIR}
                                        ${CMAKE_CURRENT_BINARY_DIR}/src
                                        ${OPENGL_INCLUDE_DIR})
     install(TARGETS vwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
@@ -524,7 +572,9 @@ if(WEBP_BUILD_EXTRAS)
   # webp_quality
   add_executable(webp_quality ${WEBP_QUALITY_SRCS} ${WEBP_EXTRAS_SRCS})
   target_link_libraries(webp_quality exampleutil imagedec)
-  target_include_directories(webp_quality PRIVATE ${CMAKE_CURRENT_SOURCE_DIR})
+  target_include_directories(webp_quality
+                             PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                     ${CMAKE_CURRENT_BINARY_DIR})
   install(TARGETS webp_quality RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 
   # vwebp_sdl
@@ -545,18 +595,21 @@ if(WEBP_BUILD_EXTRAS)
 endif()
 
 if(WEBP_BUILD_WEBP_JS)
-  # JavaScript version
-  add_executable(webp_js ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
-  target_link_libraries(webp_js webpdecoder SDL)
-  target_include_directories(webp_js PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
-  set(WEBP_HAVE_SDL 1)
-  set_target_properties(
-    webp_js
-    PROPERTIES LINK_FLAGS
-               "-s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
-       -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
-  set_target_properties(webp_js PROPERTIES OUTPUT_NAME webp)
-  target_compile_definitions(webp_js PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
+  # wasm2js does not support SIMD.
+  if(NOT WEBP_ENABLE_SIMD)
+    # JavaScript version
+    add_executable(webp_js ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
+    target_link_libraries(webp_js webpdecoder SDL)
+    target_include_directories(webp_js PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
+    set(WEBP_HAVE_SDL 1)
+    set_target_properties(
+      webp_js
+      PROPERTIES LINK_FLAGS "-s WASM=0 \
+         -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+         -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
+    set_target_properties(webp_js PROPERTIES OUTPUT_NAME webp)
+    target_compile_definitions(webp_js PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
+  endif()
 
   # WASM version
   add_executable(webp_wasm ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
@@ -607,7 +660,6 @@ if(WEBP_BUILD_ANIM_UTILS)
 endif()
 
 # Install the different headers and libraries.
-include(GNUInstallDirs)
 install(TARGETS ${INSTALLED_LIBRARIES}
         EXPORT ${PROJECT_NAME}Targets
         PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/webp

ChangeLog

@@ -1,3 +1,109 @@
+170a8712 Fix check_c_source_compiles with pthread.
+85995719 disable CombinedShannonEntropy_SSE2 on x86
+8af7436f Merge "{ios,xcframework}build.sh: make min version(s) more visible" into 1.2.0
+e56c3c5b pngdec: raise memory limit if needed
+13b8e9fe {ios,xcframework}build.sh: make min version(s) more visible
+a9225410 animdecoder_fuzzer: fix memory leak
+d6c2285d update gradle to 6.1.1
+52ce6333 update NEWS
+28c49820 bump version to 1.2.0
+7363dff2 webp/encode.h: restore WEBP_ENCODER_ABI_VERSION to v1.1.0
+826aafa5 update AUTHORS
+63258823 animdecoder_fuzzer: validate canvas size
+9eb26381 CMake: remove duplicate "include(GNUInstallDirs)"
+2e7bed79 WebPPicture: clarify the ownership of user-owned data.
+cccf5e33 webpmux: add an '-set loop <value>' option
+c9a3f6a1 Merge changes Ie29f9867,I289c54c4
+319f56f1 iosbuild.sh: sync some aspects of xcframeworkbuild.sh
+e8e8db98 add xcframeworkbuild.sh
+ae545534 dsp.h: allow config.h to override MSVC SIMD autodetection
+fef789f3 Merge "cmake: fix per-file assembly flags"
+fc14fc03 Have C encoding predictors use decoding predictors.
+7656f0b3 README,cosmetics: fix a couple typos
+d2e245ea cmake: disable webp.js if WEBP_ENABLE_SIMD=1
+96099a79 cmake: fix per-file assembly flags
+5abb5582 Merge "cmake: fix compilation w/Xcode generator"
+8484a120 cmake: fix compilation w/Xcode generator
+d7bf01c9 Merge changes Ifcae0f38,Iee2d7401
+36c81ff6 WASM-SIMD: port 2 patches from rreverser@'s tree
+988b02ab Merge "Couple of fixes to allow SIMD on Emscripten"
+26faf770 wicdec: fail with animated images
+ab2d08a8 [cd]webp: document lack of animated webp support
+52273943 Couple of fixes to allow SIMD on Emscripten
+8870ba7f Fix skia bug #10952
+4b3c6953 Detect if StoreFrame read more than anmf_payload_size bytes
+17fd4ba8 webp/decode.h,cosmetics: normalize 'flip' comment
+411d3677 remove some unreachable break statements
+3700ffd7 WebPPictureHasTransparency: remove unreachable return
+83604bf3 {animencoder,enc_dec}_fuzzer: convert some abort()s to returns
+eb44119c Merge changes I8ae09473,I678c8b1e
+9f6055fc fuzz_utils.h: rename max() to Max()
+695788e7 fuzz_utils.h: make functions WEBP_INLINE
+906c1fcd make ImgIoUtilReadFile use WebPMalloc instead of malloc
+8cb7e536 rename demux_api_fuzzer.c -> mux_demux_api_fuzzer.c
+443db47d add animdecoder_fuzzer.cc
+36a6eea3 Merge "import fuzzers from oss-fuzz/chromium"
+ec5f12c1 Makefile.vc: remove deprecated /Gm option
+64425a08 picture_tools_enc: fix windows build warning
+bd94090a import fuzzers from oss-fuzz/chromium
+cf847cba use WEBP_DSP_INIT_FUNC for Init{GammaTables*,GetCoeffs}
+55a080e5 Add WebPReplaceTransparentPixels() in dsp
+84739717 GetBackgroundColorGIF: promote to uint32_t before << 24
+def64e92 cwebp: Fix -print_psnr for near_lossless
+cf2f88b3 Add palette and spatial for q >= 75 and -m 5
+f0110bae Add no-color cache configuration to the cruncher
+749a8b99 Better estimate of the cache cost.
+4f9f00cc Use spatial predictors on top of palette no matter what.
+7658c686 Add spatial prediction on top of palette in cruncher.
+133ff0e3 webp_js: force WASM=0 option explicitly
+e3c259a2 Fix integer overflow in EmitFancyRGB.
+b3ff0bde man/{gif2,img2}webp,webpmux: normalize some wording
+f9b30586 fix ABI breakage introduced by 6a0ff358
+1d58dcfc README.webp_js: update note about emscripten version
+44070266 README.webp_js: s/fastcomp/upstream/
+2565fa8f README.webp_js: update cmake command
+47309ef5 webp: WEBP_OFFSET_PTR()
+687ab00e DC{4,8,16}_NEON: replace vmovl w/vaddl
+1b92fe75 DC16_NEON,aarch64: use vaddlv
+53f3d8cf dec_neon,DC8_NEON: use vaddlv instead of movl+vaddv
+27d08240 Fix integer overflow in WebPAnimDecoderGetNext()
+69776e38 Merge "remove call to MBAnalyzeBestIntra4Mode for method >= 5"
+a99078c1 remove call to MBAnalyzeBestIntra4Mode for method >= 5
+22e404cc CMakeLists.txt: fix set(CACHE) argument order
+71690b52 fix MSVC warning
+6a0ff358 Enc: add a qmin / qmax range for quality factor
+0fa56f30 Merge tag 'v1.1.0'
+6cf504d0 PNM decoding: handle max_value != 255
+d7844e97 update ChangeLog (tag: v1.1.0-rc2, tag: v1.1.0, origin/1.1.0)
+7f006436 Makefile.vc: fix webp_quality.exe link
+cf047e83 Makefile.vc: fix webp_quality.exe link
+c074c653 update NEWS
+30f09551 bump version to 1.1.0
+a76694a1 update AUTHORS
+6e3ef7b3 extras: fix WEBP_SWAP_16BIT_CSP check
+47178dbd extras: add WebPUnmultiplyARGB() convenience function
+22cbae33 idec_dec: fix 0 offset of NULL pointer
+290dd0b4 muxread: fix 0 offset of NULL pointer
+0df474ac Merge "lossless_(enc_|)sse2: avoid offsetting a NULL pointer"
+c6b75a19 lossless_(enc_|)sse2: avoid offsetting a NULL pointer
+295e5e38 fix UBSAN warning
+e2575e05 DC8_NEON,aarch64: use vaddv
+b0e09e34 dec_neon: Fix build failure under some toolchains
+cf0e903c dsp/lossless: Fix non gcc ARM builds
+bb7bc40b Remove ubsan errors.
+78881b76 CMake: fix GLUT library link
+9f750f7a cmake: fix BUILD_SHARED_LIBS build on mac
+17850e74 libwebp: Remove char-subscripts warning in pnmdec.c
+2fa2552d Merge "Expose WebPMalloc() in addition to WebPFree()"
+a4df4aae Expose WebPMalloc() in addition to WebPFree()
+853ea3d8 imageio/tiff: Return error before allocating bad tile size
+af650c0b Fix a Wxor-used-as-pow false positive
+601ef17c libwebp.py: update to swig 3.0.12
+0e48d889 bugfix: last alpha rows were incorrectly decoded
+24d2ccb4 webp: Fix imageio ReadPNM() TUPLTYPE
+fab8f9cf cosmetics: normalize '*' association
+94138e0e update .gitignore
+0fe1a89d update ChangeLog (tag: v1.0.3-rc1, tag: v1.0.3)
 2ad0916d update NEWS
 1287362b bump version to 1.0.3
 7b968cc2 update AUTHORS

Makefile.vc

@@ -28,7 +28,7 @@ PLATFORM_LDFLAGS = /SAFESEH
 
 NOLOGO     = /nologo
 CCNODBG    = cl.exe $(NOLOGO) /O2 /DNDEBUG
-CCDEBUG    = cl.exe $(NOLOGO) /Od /Gm /Zi /D_DEBUG /RTC1
+CCDEBUG    = cl.exe $(NOLOGO) /Od /Zi /D_DEBUG /RTC1
 CFLAGS     = /I. /Isrc $(NOLOGO) /W3 /EHsc /c
 CFLAGS     = $(CFLAGS) /DWIN32 /D_CRT_SECURE_NO_WARNINGS /DWIN32_LEAN_AND_MEAN
 LDFLAGS    = /LARGEADDRESSAWARE /MANIFEST /NXCOMPAT /DYNAMICBASE
@@ -391,7 +391,13 @@ $(DIRBIN)\get_disto.exe: $(IMAGEIO_DEC_OBJS) $(IMAGEIO_UTIL_OBJS)
 $(DIRBIN)\get_disto.exe: $(LIBWEBPDEMUX) $(LIBWEBP)
 $(DIRBIN)\webp_quality.exe: $(DIROBJ)\extras\webp_quality.obj
 $(DIRBIN)\webp_quality.exe: $(IMAGEIO_UTIL_OBJS)
-$(DIRBIN)\webp_quality.exe: $(EXTRAS_OBJS) $(LIBWEBP)
+$(DIRBIN)\webp_quality.exe: $(EXTRAS_OBJS)
+# EXTRA_OBJS requires private symbols from dsp. Explicitly add those when
+# building libwebp as a dll.
+!IF "$(DLLBUILD)" == "TRUE"
+$(DIRBIN)\webp_quality.exe: $(DSP_DEC_OBJS)
+!ENDIF
+$(DIRBIN)\webp_quality.exe: $(LIBWEBP)
 $(DIRBIN)\webpinfo.exe: $(DIROBJ)\examples\webpinfo.obj
 $(DIRBIN)\webpinfo.exe: $(IMAGEIO_DEC_OBJS)
 $(DIRBIN)\webpinfo.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS)

NEWS

@@ -1,3 +1,26 @@
+- 12/23/2020: version 1.2.0
+  * API changes:
+    - libwebp:
+      encode.h: add a qmin / qmax range for quality factor (cwebp adds -qrange)
+  * lossless encoder improvements
+  * SIMD support for Wasm builds
+  * add xcframeworkbuild.sh, supports Mac Catalyst builds
+  * import fuzzers from oss-fuzz & chromium (#409)
+  * webpmux: add an '-set loop <value>' option (#494)
+  * toolchain updates and bug fixes (#449, #463, #470, #475, #477, #478, #479,
+    #488, #491)
+
+- 12/18/2019: version 1.1.0
+  * API changes:
+    - libwebp:
+      WebPMalloc (issue #442)
+    - extras:
+      WebPUnmultiplyARGB
+  * alpha decode fix (issue #439)
+  * toolchain updates and bug fixes
+    (chromium: #1026858, #1027136, #1027409, #1028620, #1028716, #995200)
+    (oss-fuzz: #19430, #19447)
+
 - 7/4/2019: version 1.0.3
   This is a binary compatible release.
   * resize fixes for Nx1 sizes and the addition of non-opaque alpha values for

README

@@ -4,7 +4,7 @@
           \__\__/\____/\_____/__/ ____  ___
                 / _/ /    \    \ /  _ \/ _/
                /  \_/   / /   \ \   __/  \__
-               \____/____/\_____/_____/____/v1.0.3
+               \____/____/\_____/_____/____/v1.2.0
 
 Description:
 ============
@@ -113,7 +113,7 @@ make install
 
 CMake:
 ------
-With CMake, you can compile libwebp, cwebp, dwebp, gif2web, img2webp, webpinfo
+With CMake, you can compile libwebp, cwebp, dwebp, gif2webp, img2webp, webpinfo
 and the JS bindings.
 
 Prerequisites:
@@ -225,6 +225,7 @@ Usage:
 
 If input size (-s) for an image is not specified, it is
 assumed to be a PNG, JPEG, TIFF or WebP file.
+Note: Animated PNG and WebP files are not supported.
 
 Options:
   -h / -help ............. short help
@@ -254,6 +255,8 @@ Options:
   -partition_limit <int> . limit quality to fit the 512k limit on
                            the first partition (0=no degradation ... 100=full)
   -pass <int> ............ analysis pass number (1..10)
+  -qrange <min> <max> .... specifies the permissible quality range
+                           (default: 0 100)
   -crop <x> <y> <w> <h> .. crop picture with the given rectangle
   -resize <w> <h> ........ resize picture (after any cropping)
   -mt .................... use multi-threading if available
@@ -294,6 +297,7 @@ Experimental Options:
   -af .................... auto-adjust filter strength
   -pre <int> ............. pre-processing filter
 
+
 The main options you might want to try in order to further tune the
 visual quality are:
  -preset
@@ -341,7 +345,9 @@ The full list of options is available using -h:
 > dwebp -h
 Usage: dwebp in_file [options] [-o out_file]
 
-Decodes the WebP image file to PNG format [Default]
+Decodes the WebP image file to PNG format [Default].
+Note: Animated WebP files are not supported.
+
 Use following options to convert into alternate image formats:
   -pam ......... save the raw RGBA samples as a color PAM
   -ppm ......... save the raw RGB samples as a color PPM
@@ -423,7 +429,7 @@ Prerequisites:
 1) OpenGL & OpenGL Utility Toolkit (GLUT)
   Linux:
     $ sudo apt-get install freeglut3-dev mesa-common-dev
-  Mac + XCode:
+  Mac + Xcode:
     - These libraries should be available in the OpenGL / GLUT frameworks.
   Windows:
     http://freeglut.sourceforge.net/index.php#download

README.mux

@@ -1,7 +1,7 @@
           __   __  ____  ____  ____  __ __  _     __ __
          /  \\/  \/  _ \/  _ \/  _ \/  \  \/ \___/_ / _\
          \       /   __/  _  \   __/      /  /  (_/  /__
-          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v1.0.3
+          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v1.2.0
 
 
 Description:
@@ -44,6 +44,7 @@ GET_OPTIONS:
 
 SET_OPTIONS:
  Set color profile/metadata:
+   loop LOOP_COUNT   set the loop count
    icc  file.icc     set ICC profile
    exif file.exif    set EXIF metadata
    xmp  file.xmp     set XMP metadata

README.webp_js

@@ -19,21 +19,20 @@ using Emscripten and CMake.
    accessible. This is the toolchain file used by CMake to invoke Emscripten.
    If $EMSCRIPTEN is unset search for Emscripten.cmake under $EMSDK and set
    $EMSCRIPTEN accordingly, for example:
-   unix-like environments: export EMSCRIPTEN=$EMSDK/fastcomp/emscripten
-   windows: set EMSCRIPTEN=%EMSDK%\fastcomp\emscripten
+   unix-like environments: export EMSCRIPTEN=$EMSDK/upstream/emscripten
+   windows: set EMSCRIPTEN=%EMSDK%\upstream\emscripten
 
  - configure the project 'WEBP_JS' with CMake using:
 
  cd webp_js && \
  cmake -DWEBP_BUILD_WEBP_JS=ON \
-       -DEMSCRIPTEN_GENERATE_BITCODE_STATIC_LIBRARIES=1 \
        -DCMAKE_TOOLCHAIN_FILE=$EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake \
        ../
 
  - compile webp.js using 'make'.
 
  - that's it! Upon completion, you should have the webp.js and
-   webp.js.mem files generated.
+   webp.wasm files generated.
 
 The callable JavaScript function is WebPToSDL(), which decodes a raw WebP
 bitstream into a canvas. See webp_js/index.html for a simple usage sample
@@ -59,8 +58,9 @@ Web-Assembly (WASM) version:
   See webp_js/index_wasm.html for a simple demo page using the WASM version
   of the library.
 
-  You will need a fairly recent version of Emscripten (at least 1.37.8) and of
-  your WASM-enabled browser to run this version. Consider it very experimental!
+  You will need a fairly recent version of Emscripten (at least 1.37.8,
+  latest-upstream is recommended) and of your WASM-enabled browser to run this
+  version.
 
 Caveat:
 =======
@@ -78,3 +78,6 @@ Caveat:
     https://github.com/kripken/emscripten/issues/3788
 
     Therefore, SSE2 optimization is currently disabled in CMakeLists.txt.
+
+  - If WEBP_ENABLE_SIMD is set to 1 the JavaScript version (webp.js) will be
+    disabled as wasm2js does not support SIMD.

build.gradle

@@ -432,8 +432,3 @@ model {
     }
   }
 }
-
-// Task to generate the wrapper.
-task wrapper(type: Wrapper) {
-  gradleVersion = '2.13'
-}

cmake/config.h.in

@@ -93,10 +93,6 @@
 /* Define to the version of this package. */
 #cmakedefine PACKAGE_VERSION "@PACKAGE_VERSION@"
 
-/* Define to necessary symbol if this constant uses a non-standard name on
-   your system. */
-#cmakedefine PTHREAD_CREATE_JOINABLE 1
-
 /* Define to 1 if you have the ANSI C header files. */
 #cmakedefine STDC_HEADERS 1
 

cmake/cpu.cmake

@@ -31,9 +31,17 @@ endfunction()
 set(WEBP_SIMD_FLAGS "SSE41;SSE2;MIPS32;MIPS_DSP_R2;NEON;MSA")
 set(WEBP_SIMD_FILE_EXTENSIONS
     "_sse41.c;_sse2.c;_mips32.c;_mips_dsp_r2.c;_neon.c;_msa.c")
-if(MSVC)
-  # MSVC does not have a SSE4 flag but AVX support implies SSE4 support.
-  set(SIMD_ENABLE_FLAGS "/arch:AVX;/arch:SSE2;;;;")
+if(MSVC AND CMAKE_C_COMPILER_ID STREQUAL "MSVC")
+  # With at least Visual Studio 12 (2013)+ /arch is not necessary to build SSE2
+  # or SSE4 code unless a lesser /arch is forced. MSVC does not have a SSE4
+  # flag, but an AVX one. Using that with SSE4 code risks generating illegal
+  # instructions when used on machines with SSE4 only. The flags are left for
+  # older (untested) versions to avoid any potential compatibility issues.
+  if(MSVC_VERSION GREATER_EQUAL 1800 AND NOT CMAKE_C_FLAGS MATCHES "/arch:")
+    set(SIMD_ENABLE_FLAGS)
+  else()
+    set(SIMD_ENABLE_FLAGS "/arch:AVX;/arch:SSE2;;;;")
+  endif()
   set(SIMD_DISABLE_FLAGS)
 else()
   set(SIMD_ENABLE_FLAGS
@@ -57,9 +65,14 @@ endif()
 
 list(LENGTH WEBP_SIMD_FLAGS WEBP_SIMD_FLAGS_LENGTH)
 math(EXPR WEBP_SIMD_FLAGS_RANGE "${WEBP_SIMD_FLAGS_LENGTH} - 1")
-unset(HIGHEST_SSE_FLAG)
 
 foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
+  # With Emscripten 2.0.9 -msimd128 -mfpu=neon will enable NEON, but the
+  # source will fail to compile.
+  if(EMSCRIPTEN AND ${I_SIMD} GREATER_EQUAL 2)
+    break()
+  endif()
+
   list(GET WEBP_SIMD_FLAGS ${I_SIMD} WEBP_SIMD_FLAG)
 
   # First try with no extra flag added as the compiler might have default flags
@@ -70,10 +83,15 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
   webp_check_compiler_flag(${WEBP_SIMD_FLAG} ${WEBP_ENABLE_SIMD})
   if(NOT WEBP_HAVE_${WEBP_SIMD_FLAG})
     list(GET SIMD_ENABLE_FLAGS ${I_SIMD} SIMD_COMPILE_FLAG)
+    if(EMSCRIPTEN)
+      set(SIMD_COMPILE_FLAG "-msimd128 ${SIMD_COMPILE_FLAG}")
+    endif()
     set(CMAKE_REQUIRED_FLAGS ${SIMD_COMPILE_FLAG})
     webp_check_compiler_flag(${WEBP_SIMD_FLAG} ${WEBP_ENABLE_SIMD})
   else()
-    if(MSVC)
+    if(MSVC AND SIMD_ENABLE_FLAGS)
+      # The detection for SSE2/SSE4 support under MSVC is based on the compiler
+      # version so e.g., clang-cl will require flags to enable the assembly.
       list(GET SIMD_ENABLE_FLAGS ${I_SIMD} SIMD_COMPILE_FLAG)
     else()
       set(SIMD_COMPILE_FLAG " ")
@@ -84,17 +102,10 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
   file(GLOB SIMD_FILES "${CMAKE_CURRENT_LIST_DIR}/../"
             "src/dsp/*${WEBP_SIMD_FILE_EXTENSION}")
   if(WEBP_HAVE_${WEBP_SIMD_FLAG})
-    if(${I_SIMD} LESS 2 AND NOT HIGHEST_SSE_FLAG)
-      set(HIGHEST_SSE_FLAG ${SIMD_COMPILE_FLAG})
-    endif()
     # Memorize the file and flags.
     foreach(FILE ${SIMD_FILES})
       list(APPEND WEBP_SIMD_FILES_TO_INCLUDE ${FILE})
-      if(${I_SIMD} LESS 2)
-        list(APPEND WEBP_SIMD_FLAGS_TO_INCLUDE ${HIGHEST_SSE_FLAG})
-      else()
-        list(APPEND WEBP_SIMD_FLAGS_TO_INCLUDE ${SIMD_COMPILE_FLAG})
-      endif()
+      list(APPEND WEBP_SIMD_FLAGS_TO_INCLUDE ${SIMD_COMPILE_FLAG})
     endforeach()
   else()
     # Remove the file from the list.

cmake/deps.cmake

@@ -22,20 +22,23 @@ check_c_source_compiles("
   " HAVE_BUILTIN_BSWAP64)
 
 # Check for libraries.
-find_package(Threads)
+if(NOT WEBP_BUILD_WEBP_JS)
+  # Disable pThreads for WASM.
+  find_package(Threads)
+endif()
 if(Threads_FOUND)
-  if(CMAKE_USE_PTHREADS_INIT)
+  # work around cmake bug on QNX (https://cmake.org/Bug/view.php?id=11333)
+  if(CMAKE_USE_PTHREADS_INIT AND NOT CMAKE_SYSTEM_NAME STREQUAL "QNX")
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -pthread")
   endif()
-  foreach(PTHREAD_TEST HAVE_PTHREAD_PRIO_INHERIT PTHREAD_CREATE_UNDETACHED)
-    check_c_source_compiles("
-        #include <pthread.h>
-        int main (void) {
-          int attr = ${PTHREAD_TEST};
-          return attr;
-        }
-      " ${PTHREAD_TEST})
-  endforeach()
+  check_c_source_compiles("
+      #include <pthread.h>
+      int main (void) {
+        int attr = PTHREAD_PRIO_INHERIT;
+        return attr;
+      }
+    " FLAG_HAVE_PTHREAD_PRIO_INHERIT)
+  set(HAVE_PTHREAD_PRIO_INHERIT ${FLAG_HAVE_PTHREAD_PRIO_INHERIT})
   list(APPEND WEBP_DEP_LIBRARIES ${CMAKE_THREAD_LIBS_INIT})
 endif()
 set(WEBP_USE_THREAD ${Threads_FOUND})

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([libwebp], [1.0.3],
+AC_INIT([libwebp], [1.2.0],
         [https://bugs.chromium.org/p/webp],,
         [http://developers.google.com/speed/webp])
 AC_CANONICAL_HOST
@@ -80,6 +80,7 @@ TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wparentheses-equality])
 TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wshadow])
 TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wshorten-64-to-32])
 TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wundef])
+TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wunreachable-code-aggressive])
 TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wunreachable-code])
 TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wunused-but-set-variable])
 TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wunused])

examples/anim_util.c

@@ -60,15 +60,15 @@ static int AllocateFrames(AnimatedImage* const image, uint32_t num_frames) {
       !CheckSizeForOverflow(total_frame_size)) {
     return 0;
   }
-  mem = (uint8_t*)malloc((size_t)total_size);
-  frames = (DecodedFrame*)malloc((size_t)total_frame_size);
+  mem = (uint8_t*)WebPMalloc((size_t)total_size);
+  frames = (DecodedFrame*)WebPMalloc((size_t)total_frame_size);
 
   if (mem == NULL || frames == NULL) {
-    free(mem);
-    free(frames);
+    WebPFree(mem);
+    WebPFree(frames);
     return 0;
   }
-  free(image->raw_mem);
+  WebPFree(image->raw_mem);
   image->num_frames = num_frames;
   image->frames = frames;
   for (i = 0; i < num_frames; ++i) {
@@ -82,8 +82,8 @@ static int AllocateFrames(AnimatedImage* const image, uint32_t num_frames) {
 
 void ClearAnimatedImage(AnimatedImage* const image) {
   if (image != NULL) {
-    free(image->raw_mem);
-    free(image->frames);
+    WebPFree(image->raw_mem);
+    WebPFree(image->frames);
     image->num_frames = 0;
     image->frames = NULL;
     image->raw_mem = NULL;
@@ -165,7 +165,7 @@ static int DumpFrame(const char filename[], const char dump_folder[],
   base_name = (base_name == NULL) ? (const W_CHAR*)filename : base_name + 1;
   max_len = WSTRLEN(dump_folder) + 1 + WSTRLEN(base_name)
           + strlen("_frame_") + strlen(".pam") + 8;
-  file_name = (W_CHAR*)malloc(max_len * sizeof(*file_name));
+  file_name = (W_CHAR*)WebPMalloc(max_len * sizeof(*file_name));
   if (file_name == NULL) goto End;
 
   if (WSNPRINTF(file_name, max_len, "%s/%s_frame_%d.pam",
@@ -197,7 +197,7 @@ static int DumpFrame(const char filename[], const char dump_folder[],
   ok = 1;
  End:
   if (f != NULL) fclose(f);
-  free(file_name);
+  WebPFree(file_name);
   return ok;
 }
 
@@ -405,7 +405,7 @@ static uint32_t GetBackgroundColorGIF(GifFileType* gif) {
     return 0xffffffff;  // Invalid: assume white.
   } else {
     const GifColorType color = color_map->Colors[gif->SBackGroundColor];
-    return (0xff << 24) |
+    return (0xffu << 24) |
            (color.Red << 16) |
            (color.Green << 8) |
            (color.Blue << 0);

examples/cwebp.c

@@ -12,6 +12,7 @@
 //
 // Author: Skal (pascal.massimino@gmail.com)
 
+#include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -23,6 +24,7 @@
 #include "../examples/example_util.h"
 #include "../imageio/image_dec.h"
 #include "../imageio/imageio_util.h"
+#include "../imageio/webpdec.h"
 #include "./stopwatch.h"
 #include "./unicode.h"
 #include "webp/encode.h"
@@ -92,7 +94,7 @@ static int ReadPicture(const char* const filename, WebPPicture* const pic,
     WFPRINTF(stderr, "Error! Could not process file %s\n",
              (const W_CHAR*)filename);
   }
-  free((void*)data);
+  WebPFree((void*)data);
   return ok;
 }
 
@@ -119,7 +121,7 @@ static int ReadPicture(const char* const filename, WebPPicture* const pic,
     WFPRINTF(stderr, "Error! Could not process file %s\n",
              (const W_CHAR*)filename);
   }
-  free((void*)data);
+  WebPFree((void*)data);
   return ok;
 }
 
@@ -128,7 +130,8 @@ static int ReadPicture(const char* const filename, WebPPicture* const pic,
 static void AllocExtraInfo(WebPPicture* const pic) {
   const int mb_w = (pic->width + 15) / 16;
   const int mb_h = (pic->height + 15) / 16;
-  pic->extra_info = (uint8_t*)malloc(mb_w * mb_h * sizeof(*pic->extra_info));
+  pic->extra_info =
+      (uint8_t*)WebPMalloc(mb_w * mb_h * sizeof(*pic->extra_info));
 }
 
 static void PrintByteCount(const int bytes[4], int total_size,
@@ -147,7 +150,7 @@ static void PrintPercents(const int counts[4]) {
   int s;
   const int total = counts[0] + counts[1] + counts[2] + counts[3];
   for (s = 0; s < 4; ++s) {
-    fprintf(stderr, "|      %2d%%", (int)(100. * counts[s] / total + .5));
+    fprintf(stderr, "|     %3d%%", (int)(100. * counts[s] / total + .5));
   }
   fprintf(stderr, "| %7d\n", total);
 }
@@ -524,6 +527,7 @@ static void HelpLong(void) {
   printf(" cwebp [-preset <...>] [options] in_file [-o out_file]\n\n");
   printf("If input size (-s) for an image is not specified, it is\n"
          "assumed to be a PNG, JPEG, TIFF or WebP file.\n");
+  printf("Note: Animated PNG and WebP files are not supported.\n");
 #ifdef HAVE_WINCODEC_H
   printf("Windows builds can take as input any of the files handled by WIC.\n");
 #endif
@@ -564,6 +568,8 @@ static void HelpLong(void) {
   printf("                           "
          "the first partition (0=no degradation ... 100=full)\n");
   printf("  -pass <int> ............ analysis pass number (1..10)\n");
+  printf("  -qrange <min> <max> .... specifies the permissible quality range\n"
+         "                           (default: 0 100)\n");
   printf("  -crop <x> <y> <w> <h> .. crop picture with the given rectangle\n");
   printf("  -resize <w> <h> ........ resize picture (after any cropping)\n");
   printf("  -mt .................... use multi-threading if available\n");
@@ -640,10 +646,10 @@ static const char* const kErrorMessages[VP8_ENC_ERROR_LAST] = {
 
 //------------------------------------------------------------------------------
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   int return_value = -1;
-  const char *in_file = NULL, *out_file = NULL, *dump_file = NULL;
-  FILE *out = NULL;
+  const char* in_file = NULL, *out_file = NULL, *dump_file = NULL;
+  FILE* out = NULL;
   int c;
   int short_output = 0;
   int quiet = 0;
@@ -663,6 +669,7 @@ int main(int argc, const char *argv[]) {
   WebPConfig config;
   WebPAuxStats stats;
   WebPMemoryWriter memory_writer;
+  int use_memory_writer;
   Metadata metadata;
   Stopwatch stop_watch;
 
@@ -690,9 +697,9 @@ int main(int argc, const char *argv[]) {
     } else if (!strcmp(argv[c], "-H") || !strcmp(argv[c], "-longhelp")) {
       HelpLong();
       FREE_WARGV_AND_RETURN(0);
-    } else if (!strcmp(argv[c], "-o") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-o") && c + 1 < argc) {
       out_file = (const char*)GET_WARGV(argv, ++c);
-    } else if (!strcmp(argv[c], "-d") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-d") && c + 1 < argc) {
       dump_file = (const char*)GET_WARGV(argv, ++c);
       config.show_compressed = 1;
     } else if (!strcmp(argv[c], "-print_psnr")) {
@@ -706,7 +713,7 @@ int main(int argc, const char *argv[]) {
       print_distortion = 2;
     } else if (!strcmp(argv[c], "-short")) {
       ++short_output;
-    } else if (!strcmp(argv[c], "-s") && c < argc - 2) {
+    } else if (!strcmp(argv[c], "-s") && c + 2 < argc) {
       picture.width = ExUtilGetInt(argv[++c], 0, &parse_error);
       picture.height = ExUtilGetInt(argv[++c], 0, &parse_error);
       if (picture.width > WEBP_MAX_DIMENSION || picture.width < 0 ||
@@ -716,30 +723,30 @@ int main(int argc, const char *argv[]) {
                 picture.width, picture.height);
         goto Error;
       }
-    } else if (!strcmp(argv[c], "-m") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-m") && c + 1 < argc) {
       config.method = ExUtilGetInt(argv[++c], 0, &parse_error);
       use_lossless_preset = 0;   // disable -z option
-    } else if (!strcmp(argv[c], "-q") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-q") && c + 1 < argc) {
       config.quality = ExUtilGetFloat(argv[++c], &parse_error);
       use_lossless_preset = 0;   // disable -z option
-    } else if (!strcmp(argv[c], "-z") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-z") && c + 1 < argc) {
       lossless_preset = ExUtilGetInt(argv[++c], 0, &parse_error);
       if (use_lossless_preset != 0) use_lossless_preset = 1;
-    } else if (!strcmp(argv[c], "-alpha_q") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-alpha_q") && c + 1 < argc) {
       config.alpha_quality = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-alpha_method") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-alpha_method") && c + 1 < argc) {
       config.alpha_compression = ExUtilGetInt(argv[++c], 0, &parse_error);
     } else if (!strcmp(argv[c], "-alpha_cleanup")) {
       // This flag is obsolete, does opposite of -exact.
       config.exact = 0;
     } else if (!strcmp(argv[c], "-exact")) {
       config.exact = 1;
-    } else if (!strcmp(argv[c], "-blend_alpha") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-blend_alpha") && c + 1 < argc) {
       blend_alpha = 1;
       // background color is given in hex with an optional '0x' prefix
       background_color = ExUtilGetInt(argv[++c], 16, &parse_error);
       background_color = background_color & 0x00ffffffu;
-    } else if (!strcmp(argv[c], "-alpha_filter") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-alpha_filter") && c + 1 < argc) {
       ++c;
       if (!strcmp(argv[c], "none")) {
         config.alpha_filtering = 0;
@@ -755,10 +762,10 @@ int main(int argc, const char *argv[]) {
       keep_alpha = 0;
     } else if (!strcmp(argv[c], "-lossless")) {
       config.lossless = 1;
-    } else if (!strcmp(argv[c], "-near_lossless") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-near_lossless") && c + 1 < argc) {
       config.near_lossless = ExUtilGetInt(argv[++c], 0, &parse_error);
       config.lossless = 1;  // use near-lossless only with lossless
-    } else if (!strcmp(argv[c], "-hint") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-hint") && c + 1 < argc) {
       ++c;
       if (!strcmp(argv[c], "photo")) {
         config.image_hint = WEBP_HINT_PHOTO;
@@ -770,13 +777,13 @@ int main(int argc, const char *argv[]) {
         fprintf(stderr, "Error! Unrecognized image hint: %s\n", argv[c]);
         goto Error;
       }
-    } else if (!strcmp(argv[c], "-size") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-size") && c + 1 < argc) {
       config.target_size = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-psnr") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-psnr") && c + 1 < argc) {
       config.target_PSNR = ExUtilGetFloat(argv[++c], &parse_error);
-    } else if (!strcmp(argv[c], "-sns") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-sns") && c + 1 < argc) {
       config.sns_strength = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-f") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-f") && c + 1 < argc) {
       config.filter_strength = ExUtilGetInt(argv[++c], 0, &parse_error);
     } else if (!strcmp(argv[c], "-af")) {
       config.autofilter = 1;
@@ -790,27 +797,32 @@ int main(int argc, const char *argv[]) {
       config.filter_type = 1;
     } else if (!strcmp(argv[c], "-nostrong")) {
       config.filter_type = 0;
-    } else if (!strcmp(argv[c], "-sharpness") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-sharpness") && c + 1 < argc) {
       config.filter_sharpness = ExUtilGetInt(argv[++c], 0, &parse_error);
     } else if (!strcmp(argv[c], "-sharp_yuv")) {
       config.use_sharp_yuv = 1;
-    } else if (!strcmp(argv[c], "-pass") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-pass") && c + 1 < argc) {
       config.pass = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-pre") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-qrange") && c + 2 < argc) {
+      config.qmin = ExUtilGetInt(argv[++c], 0, &parse_error);
+      config.qmax = ExUtilGetInt(argv[++c], 0, &parse_error);
+      if (config.qmin < 0) config.qmin = 0;
+      if (config.qmax > 100) config.qmax = 100;
+    } else if (!strcmp(argv[c], "-pre") && c + 1 < argc) {
       config.preprocessing = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-segments") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-segments") && c + 1 < argc) {
       config.segments = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-partition_limit") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-partition_limit") && c + 1 < argc) {
       config.partition_limit = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-map") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-map") && c + 1 < argc) {
       picture.extra_info_type = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-crop") && c < argc - 4) {
+    } else if (!strcmp(argv[c], "-crop") && c + 4 < argc) {
       crop = 1;
       crop_x = ExUtilGetInt(argv[++c], 0, &parse_error);
       crop_y = ExUtilGetInt(argv[++c], 0, &parse_error);
       crop_w = ExUtilGetInt(argv[++c], 0, &parse_error);
       crop_h = ExUtilGetInt(argv[++c], 0, &parse_error);
-    } else if (!strcmp(argv[c], "-resize") && c < argc - 2) {
+    } else if (!strcmp(argv[c], "-resize") && c + 2 < argc) {
       resize_w = ExUtilGetInt(argv[++c], 0, &parse_error);
       resize_h = ExUtilGetInt(argv[++c], 0, &parse_error);
 #ifndef WEBP_DLL
@@ -826,7 +838,7 @@ int main(int argc, const char *argv[]) {
       show_progress = 1;
     } else if (!strcmp(argv[c], "-quiet")) {
       quiet = 1;
-    } else if (!strcmp(argv[c], "-preset") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-preset") && c + 1 < argc) {
       WebPPreset preset;
       ++c;
       if (!strcmp(argv[c], "default")) {
@@ -849,7 +861,7 @@ int main(int argc, const char *argv[]) {
         fprintf(stderr, "Error! Could initialize configuration with preset.\n");
         goto Error;
       }
-    } else if (!strcmp(argv[c], "-metadata") && c < argc - 1) {
+    } else if (!strcmp(argv[c], "-metadata") && c + 1 < argc) {
       static const struct {
         const char* option;
         int flag;
@@ -897,7 +909,7 @@ int main(int argc, const char *argv[]) {
     } else if (!strcmp(argv[c], "-v")) {
       verbose = 1;
     } else if (!strcmp(argv[c], "--")) {
-      if (c < argc - 1) in_file = (const char*)GET_WARGV(argv, ++c);
+      if (c + 1 < argc) in_file = (const char*)GET_WARGV(argv, ++c);
       break;
     } else if (argv[c][0] == '-') {
       fprintf(stderr, "Error! Unknown option '%s'\n", argv[c]);
@@ -973,6 +985,14 @@ int main(int argc, const char *argv[]) {
     const double read_time = StopwatchReadAndReset(&stop_watch);
     fprintf(stderr, "Time to read input: %.3fs\n", read_time);
   }
+  // The bitstream should be kept in memory when metadata must be appended
+  // before writing it to a file/stream, and/or when the near-losslessly encoded
+  // bitstream must be decoded for distortion computation (lossy will modify the
+  // 'picture' but not the lossless pipeline).
+  // Otherwise directly write the bitstream to a file.
+  use_memory_writer = (out_file != NULL && keep_metadata) ||
+                      (!quiet && print_distortion >= 0 && config.lossless &&
+                       config.near_lossless < 100);
 
   // Open the output
   if (out_file != NULL) {
@@ -987,15 +1007,19 @@ int main(int argc, const char *argv[]) {
         WFPRINTF(stderr, "Saving file '%s'\n", (const W_CHAR*)out_file);
       }
     }
-    if (keep_metadata == 0) {
-      picture.writer = MyWriter;
-      picture.custom_ptr = (void*)out;
-    } else {
+    if (use_memory_writer) {
       picture.writer = WebPMemoryWrite;
       picture.custom_ptr = (void*)&memory_writer;
+    } else {
+      picture.writer = MyWriter;
+      picture.custom_ptr = (void*)out;
     }
   } else {
     out = NULL;
+    if (use_memory_writer) {
+      picture.writer = WebPMemoryWrite;
+      picture.custom_ptr = (void*)&memory_writer;
+    }
     if (!quiet && !short_output) {
       fprintf(stderr, "No output file specified (no -o flag). Encoding will\n");
       fprintf(stderr, "be performed, but its results discarded.\n\n");
@@ -1074,8 +1098,12 @@ int main(int argc, const char *argv[]) {
   if (picture.extra_info_type > 0) {
     AllocExtraInfo(&picture);
   }
-  if (print_distortion >= 0) {  // Save original picture for later comparison
-    WebPPictureCopy(&picture, &original_picture);
+  // Save original picture for later comparison. Only for lossy as lossless does
+  // not modify 'picture' (even near-lossless).
+  if (print_distortion >= 0 && !config.lossless &&
+      !WebPPictureCopy(&picture, &original_picture)) {
+    fprintf(stderr, "Error! Cannot copy temporary picture\n");
+    goto Error;
   }
 
   // Compress.
@@ -1093,7 +1121,38 @@ int main(int argc, const char *argv[]) {
     fprintf(stderr, "Time to encode picture: %.3fs\n", encode_time);
   }
 
-  // Write info
+  // Get the decompressed image for the lossless pipeline.
+  if (!quiet && print_distortion >= 0 && config.lossless) {
+    if (config.near_lossless == 100) {
+      // Pure lossless: image was not modified, make 'original_picture' a view
+      // of 'picture' by copying all members except the freeable pointers.
+      original_picture = picture;
+      original_picture.memory_ = original_picture.memory_argb_ = NULL;
+    } else {
+      // Decode the bitstream stored in 'memory_writer' to get the altered image
+      // to 'picture'; save the 'original_picture' beforehand.
+      assert(use_memory_writer);
+      original_picture = picture;
+      if (!WebPPictureInit(&picture)) {  // Do not free 'picture'.
+        fprintf(stderr, "Error! Version mismatch!\n");
+        goto Error;
+      }
+
+      picture.use_argb = 1;
+      if (!ReadWebP(memory_writer.mem, memory_writer.size, &picture,
+                    /*keep_alpha=*/WebPPictureHasTransparency(&picture),
+                    /*metadata=*/NULL)) {
+        fprintf(stderr, "Error! Cannot decode encoded WebP bitstream\n");
+        fprintf(stderr, "Error code: %d (%s)\n", picture.error_code,
+                kErrorMessages[picture.error_code]);
+        goto Error;
+      }
+      picture.stats = original_picture.stats;
+    }
+    original_picture.stats = NULL;
+  }
+
+  // Write the YUV planes to a PGM file. Only available for lossy.
   if (dump_file) {
     if (picture.use_argb) {
       fprintf(stderr, "Warning: can't dump file (-d option) "
@@ -1104,31 +1163,29 @@ int main(int argc, const char *argv[]) {
     }
   }
 
-  if (keep_metadata != 0) {
-    if (out != NULL) {
-      if (!WriteWebPWithMetadata(out, &picture, &memory_writer,
-                                 &metadata, keep_metadata, &metadata_written)) {
-        fprintf(stderr, "Error writing WebP file with metadata!\n");
-        goto Error;
-      }
-    } else {  // output is disabled, just display the metadata stats.
-      const struct {
-        const MetadataPayload* const payload;
-        int flag;
-      } *iter, info[] = {
-        { &metadata.exif, METADATA_EXIF },
-        { &metadata.iccp, METADATA_ICC },
-        { &metadata.xmp, METADATA_XMP },
-        { NULL, 0 }
-      };
-      uint32_t unused1 = 0;
-      uint64_t unused2 = 0;
+  if (use_memory_writer && out != NULL &&
+      !WriteWebPWithMetadata(out, &picture, &memory_writer, &metadata,
+                             keep_metadata, &metadata_written)) {
+    fprintf(stderr, "Error writing WebP file!\n");
+    goto Error;
+  }
 
-      for (iter = info; iter->payload != NULL; ++iter) {
-        if (UpdateFlagsAndSize(iter->payload, !!(keep_metadata & iter->flag),
-                               0, &unused1, &unused2)) {
-          metadata_written |= iter->flag;
-        }
+  if (out == NULL && keep_metadata) {
+    // output is disabled, just display the metadata stats.
+    const struct {
+      const MetadataPayload* const payload;
+      int flag;
+    } *iter, info[] = {{&metadata.exif, METADATA_EXIF},
+                       {&metadata.iccp, METADATA_ICC},
+                       {&metadata.xmp, METADATA_XMP},
+                       {NULL, 0}};
+    uint32_t unused1 = 0;
+    uint64_t unused2 = 0;
+
+    for (iter = info; iter->payload != NULL; ++iter) {
+      if (UpdateFlagsAndSize(iter->payload, !!(keep_metadata & iter->flag),
+                             /*flag=*/0, &unused1, &unused2)) {
+        metadata_written |= iter->flag;
       }
     }
   }
@@ -1168,7 +1225,7 @@ int main(int argc, const char *argv[]) {
 
  Error:
   WebPMemoryWriterClear(&memory_writer);
-  free(picture.extra_info);
+  WebPFree(picture.extra_info);
   MetadataFree(&metadata);
   WebPPictureFree(&picture);
   WebPPictureFree(&original_picture);

examples/dwebp.c

@@ -76,7 +76,8 @@ static int SaveOutput(const WebPDecBuffer* const buffer,
 
 static void Help(void) {
   printf("Usage: dwebp in_file [options] [-o out_file]\n\n"
-         "Decodes the WebP image file to PNG format [Default]\n"
+         "Decodes the WebP image file to PNG format [Default].\n"
+         "Note: Animated WebP files are not supported.\n\n"
          "Use following options to convert into alternate image formats:\n"
          "  -pam ......... save the raw RGBA samples as a color PAM\n"
          "  -ppm ......... save the raw RGB samples as a color PPM\n"
@@ -132,7 +133,7 @@ static uint8_t* AllocateExternalBuffer(WebPDecoderConfig* config,
                      format == RGB_565) ? 2
                   : 4;
     uint32_t stride = bpp * w + 7;   // <- just for exercising
-    external_buffer = (uint8_t*)malloc(stride * h);
+    external_buffer = (uint8_t*)WebPMalloc(stride * h);
     if (external_buffer == NULL) return NULL;
     output_buffer->u.RGBA.stride = stride;
     output_buffer->u.RGBA.size = stride * h;
@@ -145,7 +146,7 @@ static uint8_t* AllocateExternalBuffer(WebPDecoderConfig* config,
     uint32_t total_size = stride * h * (has_alpha ? 2 : 1)
                         + 2 * uv_stride * (h + 1) / 2;
     assert(format >= YUV && format <= YUVA);
-    external_buffer = (uint8_t*)malloc(total_size);
+    external_buffer = (uint8_t*)WebPMalloc(total_size);
     if (external_buffer == NULL) return NULL;
     tmp = external_buffer;
     output_buffer->u.YUVA.y = tmp;
@@ -176,10 +177,10 @@ static uint8_t* AllocateExternalBuffer(WebPDecoderConfig* config,
   return external_buffer;
 }
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   int ok = 0;
-  const char *in_file = NULL;
-  const char *out_file = NULL;
+  const char* in_file = NULL;
+  const char* out_file = NULL;
 
   WebPDecoderConfig config;
   WebPDecBuffer* const output_buffer = &config.output;
@@ -412,8 +413,8 @@ int main(int argc, const char *argv[]) {
   }
  Exit:
   WebPFreeDecBuffer(output_buffer);
-  free((void*)external_buffer);
-  free((void*)data);
+  WebPFree((void*)external_buffer);
+  WebPFree((void*)data);
   FREE_WARGV_AND_RETURN(ok ? 0 : -1);
 }
 

examples/example_util.c

@@ -75,7 +75,7 @@ static void ResetCommandLineArguments(int argc, const char* argv[],
 void ExUtilDeleteCommandLineArguments(CommandLineArguments* const args) {
   if (args != NULL) {
     if (args->own_argv_) {
-      free((void*)args->argv_);
+      WebPFree((void*)args->argv_);
       WebPDataClear(&args->argv_data_);
     }
     ResetCommandLineArguments(0, NULL, args);
@@ -102,7 +102,7 @@ int ExUtilInitCommandLineArguments(int argc, const char* argv[],
       return 0;
     }
     args->own_argv_ = 1;
-    args->argv_ = (const char**)malloc(MAX_ARGC * sizeof(*args->argv_));
+    args->argv_ = (const char**)WebPMalloc(MAX_ARGC * sizeof(*args->argv_));
     if (args->argv_ == NULL) return 0;
 
     argc = 0;

examples/gif2webp.c

@@ -96,12 +96,12 @@ static void Help(void) {
 
 //------------------------------------------------------------------------------
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   int verbose = 0;
   int gif_error = GIF_ERROR;
   WebPMuxError err = WEBP_MUX_OK;
   int ok = 0;
-  const W_CHAR *in_file = NULL, *out_file = NULL;
+  const W_CHAR* in_file = NULL, *out_file = NULL;
   GifFileType* gif = NULL;
   int frame_duration = 0;
   int frame_timestamp = 0;
@@ -379,7 +379,7 @@ int main(int argc, const char *argv[]) {
       }
       case EXTENSION_RECORD_TYPE: {
         int extension;
-        GifByteType *data = NULL;
+        GifByteType* data = NULL;
         if (DGifGetExtension(gif, &extension, &data) == GIF_ERROR) {
           goto End;
         }
@@ -593,7 +593,7 @@ int main(int argc, const char *argv[]) {
 
 #else  // !WEBP_HAVE_GIF
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   fprintf(stderr, "GIF support not enabled in %s.\n", argv[0]);
   (void)argc;
   return 0;

examples/gifdec.c

@@ -137,7 +137,7 @@ int GIFReadFrame(GifFileType* const gif, int transparent_index,
   }
   dst = sub_image.argb;
 
-  tmp = (uint8_t*)malloc(rect.width * sizeof(*tmp));
+  tmp = (uint8_t*)WebPMalloc(rect.width * sizeof(*tmp));
   if (tmp == NULL) goto End;
 
   if (image_desc->Interlace) {  // Interlaced image.
@@ -168,7 +168,7 @@ int GIFReadFrame(GifFileType* const gif, int transparent_index,
  End:
   if (!ok) picture->error_code = sub_image.error_code;
   WebPPictureFree(&sub_image);
-  free(tmp);
+  WebPFree(tmp);
   return ok;
 }
 

examples/img2webp.c

@@ -84,7 +84,7 @@ static int ReadImage(const char filename[], WebPPicture* const pic) {
   if (!ImgIoUtilReadFile(filename, &data, &data_size)) return 0;
   reader = WebPGuessImageReader(data, data_size);
   ok = reader(data, data_size, pic, 1, NULL);
-  free((void*)data);
+  WebPFree((void*)data);
   return ok;
 }
 

examples/vwebp.c

@@ -479,7 +479,7 @@ static void Help(void) {
       "  'q' / 'Q' / ESC .... quit\n");
 }
 
-int main(int argc, char *argv[]) {
+int main(int argc, char* argv[]) {
   int c;
   WebPDecoderConfig* const config = &kParams.config;
   WebPIterator* const curr = &kParams.curr_frame;
@@ -630,7 +630,7 @@ int main(int argc, char *argv[]) {
 
 #else   // !WEBP_HAVE_GL
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   fprintf(stderr, "OpenGL support not enabled in %s.\n", argv[0]);
   (void)argc;
   return 0;

examples/webpmux.c

@@ -26,6 +26,7 @@
     webpmux -set icc image_profile.icc in.webp -o out_icc_container.webp
     webpmux -set exif image_metadata.exif in.webp -o out_exif_container.webp
     webpmux -set xmp image_metadata.xmp in.webp -o out_xmp_container.webp
+    webpmux -set loop 1 in.webp -o out_looped.webp
 
   Extract relevant data from WebP container file:
     webpmux -get frame n in.webp -o out_frame.webp
@@ -97,6 +98,7 @@ typedef enum {
   FEATURE_ICCP,
   FEATURE_ANMF,
   FEATURE_DURATION,
+  FEATURE_LOOP,
   LAST_FEATURE
 } FeatureType;
 
@@ -314,6 +316,7 @@ static void PrintHelp(void) {
   printf("\n");
   printf("SET_OPTIONS:\n");
   printf(" Set color profile/metadata:\n");
+  printf("   loop LOOP_COUNT   set the loop count\n");
   printf("   icc  file.icc     set ICC profile\n");
   printf("   exif file.exif    set EXIF metadata\n");
   printf("   xmp  file.xmp     set XMP metadata\n");
@@ -768,6 +771,13 @@ static int ParseCommandLine(Config* config, const W_CHAR** const unicode_argv) {
         arg->params_ = argv[i + 1];
         ++feature_arg_index;
         i += 2;
+      } else if (!strcmp(argv[i], "loop") &&
+                 (config->action_type_ == ACTION_SET)) {
+        CHECK_NUM_ARGS_AT_LEAST(2, ErrParse);
+        config->type_ = FEATURE_LOOP;
+        arg->params_ = argv[i + 1];
+        ++feature_arg_index;
+        i += 2;
       } else {  // Assume input file.
         if (config->input_ == NULL) {
           config->input_ = wargv[i];
@@ -1011,13 +1021,38 @@ static int Process(const Config* config) {
           ok = ExUtilReadFileToWebPData(config->args_[0].filename_, &chunk);
           if (!ok) goto Err2;
           err = WebPMuxSetChunk(mux, kFourccList[config->type_], &chunk, 1);
-          free((void*)chunk.bytes);
+          WebPDataClear(&chunk);
           if (err != WEBP_MUX_OK) {
             ERROR_GOTO3("ERROR (%s): Could not set the %s.\n",
                         ErrorString(err), kDescriptions[config->type_], Err2);
           }
           break;
         }
+        case FEATURE_LOOP: {
+          WebPMuxAnimParams params = { 0xFFFFFFFF, 0 };
+          int parse_error = 0;
+          const int loop_count =
+              ExUtilGetInt(config->args_[0].params_, 10, &parse_error);
+          if (loop_count < 0 || loop_count > 65535 || parse_error) {
+            ERROR_GOTO1("ERROR: Loop count must be in the range 0 to 65535.\n",
+                        Err2);
+          }
+          ok = CreateMux(config->input_, &mux);
+          if (!ok) goto Err2;
+          ok = (WebPMuxGetAnimationParams(mux, &params) == WEBP_MUX_OK);
+          if (!ok) {
+            ERROR_GOTO1("ERROR: input file does not seem to be an animation.\n",
+                        Err2);
+          }
+          params.loop_count = loop_count;
+          err = WebPMuxSetAnimationParams(mux, &params);
+          ok = (err == WEBP_MUX_OK);
+          if (!ok) {
+            ERROR_GOTO2("ERROR (%s): Could not set animation parameters.\n",
+                        ErrorString(err), Err2);
+          }
+          break;
+        }
         default: {
           ERROR_GOTO1("ERROR: Invalid feature for action 'set'.\n", Err2);
           break;
@@ -1045,7 +1080,7 @@ static int Process(const Config* config) {
         int* durations = NULL;
         WebPMux* new_mux = DuplicateMuxHeader(mux);
         if (new_mux == NULL) goto Err2;
-        durations = (int*)malloc((size_t)num_frames * sizeof(*durations));
+        durations = (int*)WebPMalloc((size_t)num_frames * sizeof(*durations));
         if (durations == NULL) goto Err2;
         for (i = 0; i < num_frames; ++i) durations[i] = -1;
 
@@ -1103,7 +1138,7 @@ static int Process(const Config* config) {
         new_mux = NULL;
 
  Err3:
-        free(durations);
+        WebPFree(durations);
         WebPMuxDelete(new_mux);
         if (!ok) goto Err2;
       }

extras/extras.c

@@ -12,13 +12,14 @@
 
 #include "extras/extras.h"
 #include "webp/format_constants.h"
+#include "src/dsp/dsp.h"
 
 #include <assert.h>
 #include <string.h>
 
 #define XTRA_MAJ_VERSION 1
-#define XTRA_MIN_VERSION 0
-#define XTRA_REV_VERSION 3
+#define XTRA_MIN_VERSION 2
+#define XTRA_REV_VERSION 0
 
 //------------------------------------------------------------------------------
 
@@ -57,7 +58,7 @@ int WebPImportRGB565(const uint8_t* rgb565, WebPPicture* pic) {
   for (y = 0; y < pic->height; ++y) {
     const int width = pic->width;
     for (x = 0; x < width; ++x) {
-#ifdef WEBP_SWAP_16BIT_CSP
+#if defined(WEBP_SWAP_16BIT_CSP) && (WEBP_SWAP_16BIT_CSP == 1)
       const uint32_t rg = rgb565[2 * x + 1];
       const uint32_t gb = rgb565[2 * x + 0];
 #else
@@ -90,7 +91,7 @@ int WebPImportRGB4444(const uint8_t* rgb4444, WebPPicture* pic) {
   for (y = 0; y < pic->height; ++y) {
     const int width = pic->width;
     for (x = 0; x < width; ++x) {
-#ifdef WEBP_SWAP_16BIT_CSP
+#if defined(WEBP_SWAP_16BIT_CSP) && (WEBP_SWAP_16BIT_CSP == 1)
       const uint32_t rg = rgb4444[2 * x + 1];
       const uint32_t ba = rgb4444[2 * x + 0];
 #else
@@ -144,3 +145,18 @@ int WebPImportColorMappedARGB(const uint8_t* indexed, int indexed_stride,
 }
 
 //------------------------------------------------------------------------------
+
+int WebPUnmultiplyARGB(WebPPicture* pic) {
+  int y;
+  uint32_t* dst;
+  if (pic == NULL || pic->use_argb != 1 || pic->argb == NULL) return 0;
+  WebPInitAlphaProcessing();
+  dst = pic->argb;
+  for (y = 0; y < pic->height; ++y) {
+    WebPMultARGBRow(dst, pic->width, /*inverse=*/1);
+    dst += pic->argb_stride;
+  }
+  return 1;
+}
+
+//------------------------------------------------------------------------------

extras/extras.h

@@ -19,7 +19,7 @@ extern "C" {
 
 #include "webp/encode.h"
 
-#define WEBP_EXTRAS_ABI_VERSION 0x0001    // MAJOR(8b) + MINOR(8b)
+#define WEBP_EXTRAS_ABI_VERSION 0x0002    // MAJOR(8b) + MINOR(8b)
 
 //------------------------------------------------------------------------------
 
@@ -51,6 +51,13 @@ WebPImportColorMappedARGB(const uint8_t* indexed, int indexed_stride,
                           const uint32_t palette[], int palette_size,
                           WebPPicture* pic);
 
+// Convert the ARGB content of 'pic' from associated to unassociated.
+// 'pic' can be for instance the result of calling of some WebPPictureImportXXX
+// functions, with pic->use_argb set to 'true'. It is assumed (and not checked)
+// that the pre-multiplied r/g/b values as less or equal than the alpha value.
+// Return false in case of error (invalid parameter, ...).
+WEBP_EXTERN int WebPUnmultiplyARGB(WebPPicture* pic);
+
 //------------------------------------------------------------------------------
 
 // Parse a bitstream, search for VP8 (lossy) header and report a

extras/get_disto.c

@@ -226,7 +226,7 @@ static void Help(void) {
           " Also handles PNG, JPG and TIFF files, in addition to WebP.\n");
 }
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   WebPPicture pic1, pic2;
   size_t size1 = 0, size2 = 0;
   int ret = 1;

extras/vwebp_sdl.c

@@ -92,7 +92,7 @@ int main(int argc, char* argv[]) {
 
 #else  // !WEBP_HAVE_SDL
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   fprintf(stderr, "SDL support not enabled in %s.\n", argv[0]);
   (void)argc;
   return 0;

extras/webp_quality.c

@@ -15,7 +15,7 @@
 #include "imageio/imageio_util.h"
 #include "../examples/unicode.h"
 
-int main(int argc, const char *argv[]) {
+int main(int argc, const char* argv[]) {
   int c;
   int quiet = 0;
   int ok = 1;

gradle/wrapper/gradle-wrapper.properties

@@ -1,6 +1,5 @@
-#Thu May 12 17:06:25 CEST 2016
 distributionBase=GRADLE_USER_HOME
 distributionPath=wrapper/dists
+distributionUrl=https\://services.gradle.org/distributions/gradle-6.1.1-bin.zip
 zipStoreBase=GRADLE_USER_HOME
 zipStorePath=wrapper/dists
-distributionUrl=https\://services.gradle.org/distributions/gradle-2.13-bin.zip

gradlew

@@ -1,4 +1,20 @@
-#!/usr/bin/env bash
+#!/usr/bin/env sh
+
+#
+# Copyright 2015 the original author or authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#      https://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
 
 ##############################################################################
 ##
@@ -28,16 +44,16 @@ APP_NAME="Gradle"
 APP_BASE_NAME=`basename "$0"`
 
 # Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
-DEFAULT_JVM_OPTS=""
+DEFAULT_JVM_OPTS='"-Xmx64m" "-Xms64m"'
 
 # Use the maximum available, or set MAX_FD != -1 to use that value.
 MAX_FD="maximum"
 
-warn ( ) {
+warn () {
     echo "$*"
 }
 
-die ( ) {
+die () {
     echo
     echo "$*"
     echo
@@ -109,8 +125,8 @@ if $darwin; then
     GRADLE_OPTS="$GRADLE_OPTS \"-Xdock:name=$APP_NAME\" \"-Xdock:icon=$APP_HOME/media/gradle.icns\""
 fi
 
-# For Cygwin, switch paths to Windows format before running java
-if $cygwin ; then
+# For Cygwin or MSYS, switch paths to Windows format before running java
+if [ "$cygwin" = "true" -o "$msys" = "true" ] ; then
     APP_HOME=`cygpath --path --mixed "$APP_HOME"`
     CLASSPATH=`cygpath --path --mixed "$CLASSPATH"`
     JAVACMD=`cygpath --unix "$JAVACMD"`
@@ -138,27 +154,30 @@ if $cygwin ; then
         else
             eval `echo args$i`="\"$arg\""
         fi
-        i=$((i+1))
+        i=`expr $i + 1`
     done
     case $i in
-        (0) set -- ;;
-        (1) set -- "$args0" ;;
-        (2) set -- "$args0" "$args1" ;;
-        (3) set -- "$args0" "$args1" "$args2" ;;
-        (4) set -- "$args0" "$args1" "$args2" "$args3" ;;
-        (5) set -- "$args0" "$args1" "$args2" "$args3" "$args4" ;;
-        (6) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" ;;
-        (7) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" ;;
-        (8) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" "$args7" ;;
-        (9) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" "$args7" "$args8" ;;
+        0) set -- ;;
+        1) set -- "$args0" ;;
+        2) set -- "$args0" "$args1" ;;
+        3) set -- "$args0" "$args1" "$args2" ;;
+        4) set -- "$args0" "$args1" "$args2" "$args3" ;;
+        5) set -- "$args0" "$args1" "$args2" "$args3" "$args4" ;;
+        6) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" ;;
+        7) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" ;;
+        8) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" "$args7" ;;
+        9) set -- "$args0" "$args1" "$args2" "$args3" "$args4" "$args5" "$args6" "$args7" "$args8" ;;
     esac
 fi
 
-# Split up the JVM_OPTS And GRADLE_OPTS values into an array, following the shell quoting and substitution rules
-function splitJvmOpts() {
-    JVM_OPTS=("$@")
+# Escape application args
+save () {
+    for i do printf %s\\n "$i" | sed "s/'/'\\\\''/g;1s/^/'/;\$s/\$/' \\\\/" ; done
+    echo " "
 }
-eval splitJvmOpts $DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS
-JVM_OPTS[${#JVM_OPTS[*]}]="-Dorg.gradle.appname=$APP_BASE_NAME"
+APP_ARGS=`save "$@"`
 
-exec "$JAVACMD" "${JVM_OPTS[@]}" -classpath "$CLASSPATH" org.gradle.wrapper.GradleWrapperMain "$@"
+# Collect all arguments for the java command, following the shell quoting and substitution rules
+eval set -- $DEFAULT_JVM_OPTS $JAVA_OPTS $GRADLE_OPTS "\"-Dorg.gradle.appname=$APP_BASE_NAME\"" -classpath "\"$CLASSPATH\"" org.gradle.wrapper.GradleWrapperMain "$APP_ARGS"
+
+exec "$JAVACMD" "$@"

gradlew.bat

@@ -1,90 +1,100 @@
-@if "%DEBUG%" == "" @echo off
-@rem ##########################################################################
-@rem
-@rem  Gradle startup script for Windows
-@rem
-@rem ##########################################################################
-
-@rem Set local scope for the variables with windows NT shell
-if "%OS%"=="Windows_NT" setlocal
-
-set DIRNAME=%~dp0
-if "%DIRNAME%" == "" set DIRNAME=.
-set APP_BASE_NAME=%~n0
-set APP_HOME=%DIRNAME%
-
-@rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
-set DEFAULT_JVM_OPTS=
-
-@rem Find java.exe
-if defined JAVA_HOME goto findJavaFromJavaHome
-
-set JAVA_EXE=java.exe
-%JAVA_EXE% -version >NUL 2>&1
-if "%ERRORLEVEL%" == "0" goto init
-
-echo.
-echo ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
-echo.
-echo Please set the JAVA_HOME variable in your environment to match the
-echo location of your Java installation.
-
-goto fail
-
-:findJavaFromJavaHome
-set JAVA_HOME=%JAVA_HOME:"=%
-set JAVA_EXE=%JAVA_HOME%/bin/java.exe
-
-if exist "%JAVA_EXE%" goto init
-
-echo.
-echo ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME%
-echo.
-echo Please set the JAVA_HOME variable in your environment to match the
-echo location of your Java installation.
-
-goto fail
-
-:init
-@rem Get command-line arguments, handling Windows variants
-
-if not "%OS%" == "Windows_NT" goto win9xME_args
-if "%@eval[2+2]" == "4" goto 4NT_args
-
-:win9xME_args
-@rem Slurp the command line arguments.
-set CMD_LINE_ARGS=
-set _SKIP=2
-
-:win9xME_args_slurp
-if "x%~1" == "x" goto execute
-
-set CMD_LINE_ARGS=%*
-goto execute
-
-:4NT_args
-@rem Get arguments from the 4NT Shell from JP Software
-set CMD_LINE_ARGS=%$
-
-:execute
-@rem Setup the command line
-
-set CLASSPATH=%APP_HOME%\gradle\wrapper\gradle-wrapper.jar
-
-@rem Execute Gradle
-"%JAVA_EXE%" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% "-Dorg.gradle.appname=%APP_BASE_NAME%" -classpath "%CLASSPATH%" org.gradle.wrapper.GradleWrapperMain %CMD_LINE_ARGS%
-
-:end
-@rem End local scope for the variables with windows NT shell
-if "%ERRORLEVEL%"=="0" goto mainEnd
-
-:fail
-rem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of
-rem the _cmd.exe /c_ return code!
-if  not "" == "%GRADLE_EXIT_CONSOLE%" exit 1
-exit /b 1
-
-:mainEnd
-if "%OS%"=="Windows_NT" endlocal
-
-:omega
+@rem
+@rem Copyright 2015 the original author or authors.
+@rem
+@rem Licensed under the Apache License, Version 2.0 (the "License");
+@rem you may not use this file except in compliance with the License.
+@rem You may obtain a copy of the License at
+@rem
+@rem      https://www.apache.org/licenses/LICENSE-2.0
+@rem
+@rem Unless required by applicable law or agreed to in writing, software
+@rem distributed under the License is distributed on an "AS IS" BASIS,
+@rem WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+@rem See the License for the specific language governing permissions and
+@rem limitations under the License.
+@rem
+
+@if "%DEBUG%" == "" @echo off
+@rem ##########################################################################
+@rem
+@rem  Gradle startup script for Windows
+@rem
+@rem ##########################################################################
+
+@rem Set local scope for the variables with windows NT shell
+if "%OS%"=="Windows_NT" setlocal
+
+set DIRNAME=%~dp0
+if "%DIRNAME%" == "" set DIRNAME=.
+set APP_BASE_NAME=%~n0
+set APP_HOME=%DIRNAME%
+
+@rem Add default JVM options here. You can also use JAVA_OPTS and GRADLE_OPTS to pass JVM options to this script.
+set DEFAULT_JVM_OPTS="-Xmx64m" "-Xms64m"
+
+@rem Find java.exe
+if defined JAVA_HOME goto findJavaFromJavaHome
+
+set JAVA_EXE=java.exe
+%JAVA_EXE% -version >NUL 2>&1
+if "%ERRORLEVEL%" == "0" goto init
+
+echo.
+echo ERROR: JAVA_HOME is not set and no 'java' command could be found in your PATH.
+echo.
+echo Please set the JAVA_HOME variable in your environment to match the
+echo location of your Java installation.
+
+goto fail
+
+:findJavaFromJavaHome
+set JAVA_HOME=%JAVA_HOME:"=%
+set JAVA_EXE=%JAVA_HOME%/bin/java.exe
+
+if exist "%JAVA_EXE%" goto init
+
+echo.
+echo ERROR: JAVA_HOME is set to an invalid directory: %JAVA_HOME%
+echo.
+echo Please set the JAVA_HOME variable in your environment to match the
+echo location of your Java installation.
+
+goto fail
+
+:init
+@rem Get command-line arguments, handling Windows variants
+
+if not "%OS%" == "Windows_NT" goto win9xME_args
+
+:win9xME_args
+@rem Slurp the command line arguments.
+set CMD_LINE_ARGS=
+set _SKIP=2
+
+:win9xME_args_slurp
+if "x%~1" == "x" goto execute
+
+set CMD_LINE_ARGS=%*
+
+:execute
+@rem Setup the command line
+
+set CLASSPATH=%APP_HOME%\gradle\wrapper\gradle-wrapper.jar
+
+@rem Execute Gradle
+"%JAVA_EXE%" %DEFAULT_JVM_OPTS% %JAVA_OPTS% %GRADLE_OPTS% "-Dorg.gradle.appname=%APP_BASE_NAME%" -classpath "%CLASSPATH%" org.gradle.wrapper.GradleWrapperMain %CMD_LINE_ARGS%
+
+:end
+@rem End local scope for the variables with windows NT shell
+if "%ERRORLEVEL%"=="0" goto mainEnd
+
+:fail
+rem Set variable GRADLE_EXIT_CONSOLE if you need the _script_ return code instead of
+rem the _cmd.exe /c_ return code!
+if  not "" == "%GRADLE_EXIT_CONSOLE%" exit 1
+exit /b 1
+
+:mainEnd
+if "%OS%"=="Windows_NT" endlocal
+
+:omega

imageio/imageio_util.c

@@ -91,7 +91,7 @@ int ImgIoUtilReadFile(const char* const file_name,
   file_size = ftell(in);
   fseek(in, 0, SEEK_SET);
   // we allocate one extra byte for the \0 terminator
-  file_data = (uint8_t*)malloc(file_size + 1);
+  file_data = (uint8_t*)WebPMalloc(file_size + 1);
   if (file_data == NULL) {
     fclose(in);
     WFPRINTF(stderr, "memory allocation failure when reading file %s\n",
@@ -104,7 +104,7 @@ int ImgIoUtilReadFile(const char* const file_name,
   if (!ok) {
     WFPRINTF(stderr, "Could not read %d bytes of data from file %s\n",
              (int)file_size, (const W_CHAR*)file_name);
-    free(file_data);
+    WebPFree(file_data);
     return 0;
   }
   file_data[file_size] = '\0';  // convenient 0-terminator

imageio/imageio_util.h

@@ -29,7 +29,7 @@ FILE* ImgIoUtilSetBinaryMode(FILE* file);
 
 // Allocates storage for entire file 'file_name' and returns contents and size
 // in 'data' and 'data_size'. Returns 1 on success, 0 otherwise. '*data' should
-// be deleted using free().
+// be deleted using WebPFree().
 // Note: for convenience, the data will be null-terminated with an extra byte
 // (not accounted for in *data_size), in case the file is text and intended
 // to be used as a C-string.

imageio/pngdec.c

@@ -259,6 +259,15 @@ int ReadPNG(const uint8_t* const data, size_t data_size,
     goto End;
   }
 
+#if LOCAL_PNG_PREREQ(1,5) || \
+    (LOCAL_PNG_PREREQ(1,4) && PNG_LIBPNG_VER_RELEASE >= 1)
+  // If it looks like the bitstream is going to need more memory than libpng's
+  // internal limit (default: 8M), try to (reasonably) raise it.
+  if (data_size > png_get_chunk_malloc_max(png) && data_size < (1u << 24)) {
+    png_set_chunk_malloc_max(png, data_size);
+  }
+#endif
+
   info = png_create_info_struct(png);
   if (info == NULL) goto Error;
   end_info = png_create_info_struct(png);

imageio/pnmdec.c

@@ -26,15 +26,15 @@ typedef enum {
   DEPTH_FLAG      = 1 << 2,
   MAXVAL_FLAG     = 1 << 3,
   TUPLE_FLAG      = 1 << 4,
-  ALL_NEEDED_FLAGS = 0x1f
+  ALL_NEEDED_FLAGS = WIDTH_FLAG | HEIGHT_FLAG | DEPTH_FLAG | MAXVAL_FLAG
 } PNMFlags;
 
 typedef struct {
   const uint8_t* data;
   size_t data_size;
   int width, height;
-  int bytes_per_px;   // 1, 3, 4
-  int depth;
+  int bytes_per_px;
+  int depth;          // 1 (grayscale), 2 (grayscale + alpha), 3 (rgb), 4 (rgba)
   int max_value;
   int type;           // 5, 6 or 7
   int seen_flags;
@@ -74,6 +74,7 @@ static size_t ReadPAMFields(PNMInfo* const info, size_t off) {
   char out[MAX_LINE_SIZE + 1];
   size_t out_size;
   int tmp;
+  int expected_depth = -1;
   assert(info != NULL);
   while (1) {
     off = ReadLine(info->data, off, info->data_size, out, &out_size);
@@ -95,13 +96,16 @@ static size_t ReadPAMFields(PNMInfo* const info, size_t off) {
       info->seen_flags |= MAXVAL_FLAG;
       info->max_value = tmp;
     } else if (!strcmp(out, "TUPLTYPE RGB_ALPHA")) {
-      info->bytes_per_px = 4;
+      expected_depth = 4;
       info->seen_flags |= TUPLE_FLAG;
     } else if (!strcmp(out, "TUPLTYPE RGB")) {
-      info->bytes_per_px = 3;
+      expected_depth = 3;
+      info->seen_flags |= TUPLE_FLAG;
+    } else if (!strcmp(out, "TUPLTYPE GRAYSCALE_ALPHA")) {
+      expected_depth = 2;
       info->seen_flags |= TUPLE_FLAG;
     } else if (!strcmp(out, "TUPLTYPE GRAYSCALE")) {
-      info->bytes_per_px = 1;
+      expected_depth = 1;
       info->seen_flags |= TUPLE_FLAG;
     } else if (!strcmp(out, "ENDHDR")) {
       break;
@@ -110,23 +114,24 @@ static size_t ReadPAMFields(PNMInfo* const info, size_t off) {
       int i;
       if (out_size > 20) sprintf(out + 20 - strlen(kEllipsis), kEllipsis);
       for (i = 0; i < (int)strlen(out); ++i) {
-        if (!isprint(out[i])) out[i] = ' ';
+        // isprint() might trigger a "char-subscripts" warning if given a char.
+        if (!isprint((int)out[i])) out[i] = ' ';
       }
       fprintf(stderr, "PAM header error: unrecognized entry [%s]\n", out);
       return 0;
     }
   }
-  if (!(info->seen_flags & TUPLE_FLAG)) {
-    if (info->depth > 0 && info->depth <= 4 && info->depth != 2) {
-      info->seen_flags |= TUPLE_FLAG;
-      info->bytes_per_px = info->depth * (info->max_value > 255 ? 2 : 1);
-    } else {
-      fprintf(stderr, "PAM: invalid bitdepth (%d).\n", info->depth);
-      return 0;
-    }
+  if (!(info->seen_flags & ALL_NEEDED_FLAGS)) {
+    fprintf(stderr, "PAM header error: missing tags%s%s%s%s\n",
+            (info->seen_flags & WIDTH_FLAG) ? "" : " WIDTH",
+            (info->seen_flags & HEIGHT_FLAG) ? "" : " HEIGHT",
+            (info->seen_flags & DEPTH_FLAG) ? "" : " DEPTH",
+            (info->seen_flags & MAXVAL_FLAG) ? "" : " MAXVAL");
+    return 0;
   }
-  if (info->seen_flags != ALL_NEEDED_FLAGS) {
-    fprintf(stderr, "PAM: incomplete header.\n");
+  if (expected_depth != -1 && info->depth != expected_depth) {
+    fprintf(stderr, "PAM header error: expected DEPTH %d but got DEPTH %d\n",
+            expected_depth, info->depth);
     return 0;
   }
   return off;
@@ -160,16 +165,15 @@ static size_t ReadHeader(PNMInfo* const info) {
 
     // finish initializing missing fields
     info->depth = (info->type == 5) ? 1 : 3;
-    info->bytes_per_px = info->depth * (info->max_value > 255 ? 2 : 1);
   }
   // perform some basic numerical validation
   if (info->width <= 0 || info->height <= 0 ||
       info->type <= 0 || info->type >= 9 ||
-      info->depth <= 0 || info->depth == 2 || info->depth > 4 ||
-      info->bytes_per_px < info->depth ||
+      info->depth <= 0 || info->depth > 4 ||
       info->max_value <= 0 || info->max_value >= 65536) {
     return 0;
   }
+  info->bytes_per_px = info->depth * (info->max_value > 255 ? 2 : 1);
   return off;
 }
 
@@ -178,7 +182,7 @@ int ReadPNM(const uint8_t* const data, size_t data_size,
             struct Metadata* const metadata) {
   int ok = 0;
   int i, j;
-  uint64_t stride, pixel_bytes;
+  uint64_t stride, pixel_bytes, sample_size, depth;
   uint8_t* rgb = NULL, *tmp_rgb;
   size_t offset;
   PNMInfo info;
@@ -209,8 +213,10 @@ int ReadPNM(const uint8_t* const data, size_t data_size,
     fprintf(stderr, "Truncated PNM file (P%d).\n", info.type);
     goto End;
   }
-  stride =
-      (uint64_t)(info.bytes_per_px < 3 ? 3 : info.bytes_per_px) * info.width;
+  sample_size = (info.max_value > 255) ? 2 : 1;
+  // final depth
+  depth = (info.depth == 1 || info.depth == 3 || !keep_alpha) ? 3 : 4;
+  stride = depth * info.width;
   if (stride != (size_t)stride ||
       !ImgIoUtilCheckSizeArgumentsOverflow(stride, info.height)) {
     goto End;
@@ -219,30 +225,63 @@ int ReadPNM(const uint8_t* const data, size_t data_size,
   rgb = (uint8_t*)malloc((size_t)stride * info.height);
   if (rgb == NULL) goto End;
 
-  // Convert input
+  // Convert input.
+  // We only optimize for the sample_size=1, max_value=255, depth=1 case.
   tmp_rgb = rgb;
   for (j = 0; j < info.height; ++j) {
-    assert(offset + info.bytes_per_px * info.width <= data_size);
-    if (info.depth == 1) {
-      // convert grayscale -> RGB
-      for (i = 0; i < info.width; ++i) {
-        const uint8_t v = data[offset + i];
-        tmp_rgb[3 * i + 0] = tmp_rgb[3 * i + 1] = tmp_rgb[3 * i + 2] = v;
-      }
-    } else if (info.depth == 3) {   // RGB
-      memcpy(tmp_rgb, data + offset, 3 * info.width * sizeof(*data));
-    } else if (info.depth == 4) {   // RGBA
-      memcpy(tmp_rgb, data + offset, 4 * info.width * sizeof(*data));
-    }
+    const uint8_t* in = data + offset;
     offset += info.bytes_per_px * info.width;
+    assert(offset <= data_size);
+    if (info.max_value == 255 && info.depth >= 3) {
+      // RGB or RGBA
+      if (info.depth == 3 || keep_alpha) {
+        memcpy(tmp_rgb, in, info.depth * info.width * sizeof(*in));
+      } else {
+        assert(info.depth == 4 && !keep_alpha);
+        for (i = 0; i < info.width; ++i) {
+          tmp_rgb[3 * i + 0] = in[4 * i + 0];
+          tmp_rgb[3 * i + 1] = in[4 * i + 1];
+          tmp_rgb[3 * i + 2] = in[4 * i + 2];
+        }
+      }
+    } else {
+      // Unoptimized case, we need to handle non-trivial operations:
+      //   * convert 16b to 8b (if max_value > 255)
+      //   * rescale to [0..255] range (if max_value != 255)
+      //   * drop the alpha channel (if keep_alpha is false)
+      const uint32_t round = info.max_value / 2;
+      int k = 0;
+      for (i = 0; i < info.width * info.depth; ++i) {
+        uint32_t v = (sample_size == 2) ? 256u * in[2 * i + 0] + in[2 * i + 1]
+                   : in[i];
+        if (info.max_value != 255) v = (v * 255u + round) / info.max_value;
+        if (v > 255u) v = 255u;
+        if (info.depth > 2) {
+          if (!keep_alpha && info.depth == 4 && (i % 4) == 3) {
+            // skip alpha
+          } else {
+            tmp_rgb[k] = v;
+            k += 1;
+          }
+        } else if (info.depth == 1 || (i % 2) == 0) {
+          tmp_rgb[k + 0] = tmp_rgb[k + 1] = tmp_rgb[k + 2] = v;
+          k += 3;
+        } else if (keep_alpha && info.depth == 2) {
+          tmp_rgb[k] = v;
+          k += 1;
+        } else {
+          // skip alpha
+        }
+      }
+    }
     tmp_rgb += stride;
   }
 
   // WebP conversion.
   pic->width = info.width;
   pic->height = info.height;
-  ok = (info.depth == 4) ? WebPPictureImportRGBA(pic, rgb, (int)stride)
-                         : WebPPictureImportRGB(pic, rgb, (int)stride);
+  ok = (depth == 4) ? WebPPictureImportRGBA(pic, rgb, (int)stride)
+                    : WebPPictureImportRGB(pic, rgb, (int)stride);
   if (!ok) goto End;
 
   ok = 1;

imageio/tiffdec.c

@@ -156,7 +156,7 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
              Metadata* const metadata) {
   MyData my_data = { data, (toff_t)data_size, 0 };
   TIFF* tif;
-  uint32_t width, height;
+  uint32_t image_width, image_height, tile_width, tile_height;
   uint16_t samples_per_px = 0;
   uint16_t extra_samples = 0;
   uint16_t* extra_samples_ptr = NULL;
@@ -189,15 +189,25 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
   }
   if (samples_per_px < 3 || samples_per_px > 4) goto End;  // not supported
 
-  if (!(TIFFGetField(tif, TIFFTAG_IMAGEWIDTH, &width) &&
-        TIFFGetField(tif, TIFFTAG_IMAGELENGTH, &height))) {
+  if (!(TIFFGetField(tif, TIFFTAG_IMAGEWIDTH, &image_width) &&
+        TIFFGetField(tif, TIFFTAG_IMAGELENGTH, &image_height))) {
     fprintf(stderr, "Error! Cannot retrieve TIFF image dimensions.\n");
     goto End;
   }
-  if (!ImgIoUtilCheckSizeArgumentsOverflow((uint64_t)width * height,
+  if (!ImgIoUtilCheckSizeArgumentsOverflow((uint64_t)image_width * image_height,
                                            sizeof(*raster))) {
     goto End;
   }
+  // According to spec, a tile can be bigger than the image. However it should
+  // be a multiple of 16 and not way too large, so check that it's not more than
+  // twice the image size, for dimensions above some arbitrary minimum 32.
+  if ((TIFFGetField(tif, TIFFTAG_TILEWIDTH, &tile_width) &&
+       tile_width > 32 && tile_width / 2 > image_width) ||
+      (TIFFGetField(tif, TIFFTAG_TILELENGTH, &tile_height) &&
+       tile_height > 32 && tile_height / 2 > image_height)) {
+    fprintf(stderr, "Error! TIFF tile dimensions are too big.\n");
+    goto End;
+  }
   if (samples_per_px > 3 && !TIFFGetField(tif, TIFFTAG_EXTRASAMPLES,
                                           &extra_samples, &extra_samples_ptr)) {
     fprintf(stderr, "Error! Cannot retrieve TIFF ExtraSamples info.\n");
@@ -205,27 +215,28 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
   }
 
   // _Tiffmalloc uses a signed type for size.
-  alloc_size = (int64_t)((uint64_t)width * height * sizeof(*raster));
+  alloc_size =
+      (int64_t)((uint64_t)image_width * image_height * sizeof(*raster));
   if (alloc_size < 0 || alloc_size != (tsize_t)alloc_size) goto End;
 
   raster = (uint32*)_TIFFmalloc((tsize_t)alloc_size);
   if (raster != NULL) {
-    if (TIFFReadRGBAImageOriented(tif, width, height, raster,
+    if (TIFFReadRGBAImageOriented(tif, image_width, image_height, raster,
                                   ORIENTATION_TOPLEFT, 1)) {
-      const int stride = width * sizeof(*raster);
-      pic->width = width;
-      pic->height = height;
+      const int stride = image_width * sizeof(*raster);
+      pic->width = image_width;
+      pic->height = image_height;
       // TIFF data is ABGR
 #ifdef WORDS_BIGENDIAN
-      TIFFSwabArrayOfLong(raster, width * height);
+      TIFFSwabArrayOfLong(raster, image_width * image_height);
 #endif
       // if we have an alpha channel, we must un-multiply from rgbA to RGBA
       if (extra_samples == 1 && extra_samples_ptr != NULL &&
           extra_samples_ptr[0] == EXTRASAMPLE_ASSOCALPHA) {
         uint32_t y;
         uint8_t* tmp = (uint8_t*)raster;
-        for (y = 0; y < height; ++y) {
-          MultARGBRow(tmp, width);
+        for (y = 0; y < image_height; ++y) {
+          MultARGBRow(tmp, image_width);
           tmp += stride;
         }
       }

imageio/webpdec.c

@@ -65,7 +65,7 @@ int LoadWebP(const char* const in_file,
 
   status = WebPGetFeatures(*data, *data_size, bitstream);
   if (status != VP8_STATUS_OK) {
-    free((void*)*data);
+    WebPFree((void*)*data);
     *data = NULL;
     *data_size = 0;
     PrintWebPError(in_file, status);

imageio/wicdec.c

@@ -298,9 +298,15 @@ int ReadPictureWithWIC(const char* const filename,
           factory, stream, NULL,
           WICDecodeMetadataCacheOnDemand, &decoder));
   IFS(IWICBitmapDecoder_GetFrameCount(decoder, &frame_count));
-  if (SUCCEEDED(hr) && frame_count == 0) {
-    fprintf(stderr, "No frame found in input file.\n");
-    hr = E_FAIL;
+  if (SUCCEEDED(hr)) {
+    if (frame_count == 0) {
+      fprintf(stderr, "No frame found in input file.\n");
+      hr = E_FAIL;
+    } else if (frame_count > 1) {
+      // WIC will be tried before native WebP decoding so avoid duplicating the
+      // error message.
+      hr = E_FAIL;
+    }
   }
   IFS(IWICBitmapDecoder_GetFrame(decoder, 0, &frame));
   IFS(IWICBitmapFrameDecode_GetPixelFormat(frame, &src_pixel_format));

iosbuild.sh

@@ -1,8 +1,9 @@
 #!/bin/bash
 #
-# This script generates 'WebP.framework' and 'WebPDecoder.framework'. An iOS
-# app can decode WebP images by including 'WebPDecoder.framework' and both
-# encode and decode WebP images by including 'WebP.framework'.
+# This script generates 'WebP.framework' and 'WebPDecoder.framework',
+# 'WebPDemux.framework' and 'WebPMux.framework'.
+# An iOS app can decode WebP images by including 'WebPDecoder.framework' and
+# both encode and decode WebP images by including 'WebP.framework'.
 #
 # Run ./iosbuild.sh to generate the frameworks under the current directory
 # (the previous build will be erased if it exists).
@@ -12,6 +13,9 @@
 
 set -e
 
+# Set this variable based on the desired minimum deployment target.
+readonly IOS_MIN_VERSION=6.0
+
 # Extract the latest SDK version from the final field of the form: iphoneosX.Y
 readonly SDK=$(xcodebuild -showsdks \
   | grep iphoneos | sort | tail -n 1 | awk '{print substr($NF, 9)}'
@@ -50,13 +54,27 @@ if [[ -z "${SDK}" ]]; then
   exit 1
 elif [[ ${SDK%%.*} -gt 8 ]]; then
   EXTRA_CFLAGS="-fembed-bitcode"
-elif [[ ${SDK} < 6.0 ]]; then
+elif [[ ${SDK%%.*} -le 6 ]]; then
   echo "You need iOS SDK version 6.0 or above"
   exit 1
-else
-  echo "iOS SDK Version ${SDK}"
 fi
 
+echo "Xcode Version: ${XCODE}"
+echo "iOS SDK Version: ${SDK}"
+
+if [[ -e "${BUILDDIR}" || -e "${TARGETDIR}" || -e "${DECTARGETDIR}" \
+      || -e "${MUXTARGETDIR}" || -e "${DEMUXTARGETDIR}" ]]; then
+  cat << EOF
+WARNING: The following directories will be deleted:
+WARNING:   ${BUILDDIR}
+WARNING:   ${TARGETDIR}
+WARNING:   ${DECTARGETDIR}
+WARNING:   ${MUXTARGETDIR}
+WARNING:   ${DEMUXTARGETDIR}
+WARNING: The build will continue in 5 seconds...
+EOF
+  sleep 5
+fi
 rm -rf ${BUILDDIR} ${TARGETDIR} ${DECTARGETDIR} \
     ${MUXTARGETDIR} ${DEMUXTARGETDIR}
 mkdir -p ${BUILDDIR} ${TARGETDIR}/Headers/ ${DECTARGETDIR}/Headers/ \
@@ -64,12 +82,12 @@ mkdir -p ${BUILDDIR} ${TARGETDIR}/Headers/ ${DECTARGETDIR}/Headers/ \
 
 if [[ ! -e ${SRCDIR}/configure ]]; then
   if ! (cd ${SRCDIR} && sh autogen.sh); then
-    cat <<EOT
+    cat << EOF
 Error creating configure script!
 This script requires the autoconf/automake and libtool to build. MacPorts can
 be used to obtain these:
 http://www.macports.org/install.php
-EOT
+EOF
     exit 1
   fi
 fi
@@ -103,7 +121,7 @@ for PLATFORM in ${PLATFORMS}; do
   SDKROOT="${PLATFORMSROOT}/"
   SDKROOT+="${PLATFORM}.platform/Developer/SDKs/${PLATFORM}${SDK}.sdk/"
   CFLAGS="-arch ${ARCH2:-${ARCH}} -pipe -isysroot ${SDKROOT} -O3 -DNDEBUG"
-  CFLAGS+=" -miphoneos-version-min=6.0 ${EXTRA_CFLAGS}"
+  CFLAGS+=" -miphoneos-version-min=${IOS_MIN_VERSION} ${EXTRA_CFLAGS}"
 
   set -x
   export PATH="${DEVROOT}/usr/bin:${OLDPATH}"
@@ -148,3 +166,5 @@ echo "DEMUXLIBLIST = ${DEMUXLIBLIST}"
 cp -a ${SRCDIR}/src/webp/{decode,types,mux_types,demux}.h \
     ${DEMUXTARGETDIR}/Headers/
 ${LIPO} -create ${DEMUXLIBLIST} -output ${DEMUXTARGETDIR}/WebPDemux
+
+echo  "SUCCESS"

man/cwebp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH CWEBP 1 "January 11, 2019"
+.TH CWEBP 1 "November 19, 2020"
 .SH NAME
 cwebp \- compress an image file to a WebP file
 .SH SYNOPSIS
@@ -13,6 +13,7 @@ command.
 .PP
 \fBcwebp\fP compresses an image using the WebP format.
 Input format can be either PNG, JPEG, TIFF, WebP or raw Y'CbCr samples.
+Note: Animated PNG and WebP files are not supported.
 .SH OPTIONS
 The basic options are:
 .TP
@@ -134,6 +135,13 @@ options \fB\-size\fP or \fB\-psnr\fP. Maximum value is 10, default is 1.
 If options \fB\-size\fP or \fB\-psnr\fP were used, but \fB\-pass\fP wasn't
 specified, a default value of '6' passes will be used.
 .TP
+.BI \-qrange " int int
+Specifies the permissible interval for the quality factor. This is particularly
+useful when using multi-pass (\fB\-size\fP or \fB\-psnr\fP options).
+Default is 0 100.
+If the quality factor is outside this range, it will be clamped.
+If the minimum value must be less or equal to the maximum one.
+.TP
 .B \-af
 Turns auto\-filter on. This algorithm will spend additional time optimizing
 the filtering strength to reach a well\-balanced quality.

man/dwebp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH DWEBP 1 "June 23, 2016"
+.TH DWEBP 1 "November 19, 2020"
 .SH NAME
 dwebp \- decompress a WebP file to an image file
 .SH SYNOPSIS
@@ -12,6 +12,7 @@ This manual page documents the
 command.
 .PP
 \fBdwebp\fP decompresses WebP files into PNG, PAM, PPM or PGM images.
+Note: Animated WebP files are not supported.
 .SH OPTIONS
 The basic options are:
 .TP

man/gif2webp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH GIF2WEBP 1 "January 25, 2018"
+.TH GIF2WEBP 1 "May 1, 2020"
 .SH NAME
 gif2webp \- Convert a GIF image to WebP
 .SH SYNOPSIS
@@ -62,9 +62,9 @@ larger file size and lower compression quality.
 .TP
 .BI \-min_size
 Encode image to achieve smallest size. This disables key frame insertion and
-picks the dispose method resulting in smallest output for each frame. It uses
-lossless compression by default, but can be combined with \-q, \-m, \-lossy or
-\-mixed options.
+picks the dispose method resulting in the smallest output for each frame. It
+uses lossless compression by default, but can be combined with \-q, \-m,
+\-lossy or \-mixed options.
 .TP
 .BI \-kmin " int
 .TP

man/img2webp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH IMG2WEBP 1 "April 3, 2018"
+.TH IMG2WEBP 1 "May 1, 2020"
 .SH NAME
 img2webp \- create animated WebP file from a sequence of input images.
 .SH SYNOPSIS
@@ -16,8 +16,8 @@ command.
 \fBimg2webp\fP compresses a sequence of images using the animated WebP format.
 Input images can either be PNG, JPEG, TIFF or WebP.
 If a single file name (not starting with the character '\-') is supplied as
-the argument, the command line argument are actually tokenized from this file.
-This allows for easy scripting or using large number of arguments.
+the argument, the command line arguments are actually tokenized from this file.
+This allows for easy scripting or using a large number of arguments.
 .SH FILE-LEVEL OPTIONS
 The file-level options are applied at the beginning of the compression process,
 before the input frames are read.
@@ -27,7 +27,7 @@ Specify the name of the output WebP file.
 .TP
 .BI \-min_size
 Encode images to achieve smallest size. This disables key frame insertion and
-picks the parameters resulting in smallest output for each frame. It uses
+picks the parameters resulting in the smallest output for each frame. It uses
 lossless compression by default, but can be combined with \-q, \-m, \-lossy or
 \-mixed options.
 .TP

man/webpmux.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH WEBPMUX 1 "April 23, 2018"
+.TH WEBPMUX 1 "December 12, 2020"
 .SH NAME
 webpmux \- create animated WebP files from non\-animated WebP images, extract
 frames from animated WebP images, and manage XMP/EXIF metadata and ICC profile.
@@ -59,7 +59,7 @@ command.
 to add/extract/strip XMP/EXIF metadata and ICC profile.
 If a single file name (not starting with the character '\-') is supplied as
 the argument, the command line arguments are actually tokenized from this file.
-This allows for easy scripting or using large number of arguments.
+This allows for easy scripting or using a large number of arguments.
 .SH OPTIONS
 .SS GET_OPTIONS (\-get):
 .TP
@@ -77,6 +77,11 @@ Get nth frame from an animated image. (n = 0 has a special meaning: last frame).
 
 .SS SET_OPTIONS (\-set)
 .TP
+.BI loop " loop_count
+Set loop count on an animated file.
+.P
+Where: 'loop_count' must be in range [0, 65535].
+.TP
 .BI icc " file.icc
 Set ICC profile.
 .P

src/Makefile.am

@@ -36,7 +36,7 @@ libwebp_la_LIBADD += utils/libwebputils.la
 # other than the ones listed on the command line, i.e., after linking, it will
 # not have unresolved symbols. Some platforms (Windows among them) require all
 # symbols in shared libraries to be resolved at library creation.
-libwebp_la_LDFLAGS = -no-undefined -version-info 7:5:0
+libwebp_la_LDFLAGS = -no-undefined -version-info 8:1:1
 libwebpincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebp.pc
 
@@ -48,7 +48,7 @@ if BUILD_LIBWEBPDECODER
   libwebpdecoder_la_LIBADD += dsp/libwebpdspdecode.la
   libwebpdecoder_la_LIBADD += utils/libwebputilsdecode.la
 
-  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 3:5:0
+  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 4:1:1
   pkgconfig_DATA += libwebpdecoder.pc
 endif
 

src/dec/frame_dec.c

@@ -732,7 +732,7 @@ static int AllocateMemory(VP8Decoder* const dec) {
   mem += f_info_size;
   dec->thread_ctx_.id_ = 0;
   dec->thread_ctx_.f_info_ = dec->f_info_;
-  if (dec->mt_method_ > 0) {
+  if (dec->filter_type_ > 0 && dec->mt_method_ > 0) {
     // secondary cache line. The deblocking process need to make use of the
     // filtering strength from previous macroblock row, while the new ones
     // are being decoded in parallel. We'll just swap the pointers.

src/dec/idec_dec.c

@@ -166,9 +166,11 @@ static int AppendToMemBuffer(WebPIDecoder* const idec,
   VP8Decoder* const dec = (VP8Decoder*)idec->dec_;
   MemBuffer* const mem = &idec->mem_;
   const int need_compressed_alpha = NeedCompressedAlpha(idec);
-  const uint8_t* const old_start = mem->buf_ + mem->start_;
+  const uint8_t* const old_start =
+      (mem->buf_ == NULL) ? NULL : mem->buf_ + mem->start_;
   const uint8_t* const old_base =
       need_compressed_alpha ? dec->alpha_data_ : old_start;
+  assert(mem->buf_ != NULL || mem->start_ == 0);
   assert(mem->mode_ == MEM_MODE_APPEND);
   if (data_size > MAX_CHUNK_PAYLOAD) {
     // security safeguard: trying to allocate more than what the format
@@ -184,7 +186,7 @@ static int AppendToMemBuffer(WebPIDecoder* const idec,
     uint8_t* const new_buf =
         (uint8_t*)WebPSafeMalloc(extra_size, sizeof(*new_buf));
     if (new_buf == NULL) return 0;
-    memcpy(new_buf, old_base, current_size);
+    if (old_base != NULL) memcpy(new_buf, old_base, current_size);
     WebPSafeFree(mem->buf_);
     mem->buf_ = new_buf;
     mem->buf_size_ = (size_t)extra_size;
@@ -192,6 +194,7 @@ static int AppendToMemBuffer(WebPIDecoder* const idec,
     mem->end_ = current_size;
   }
 
+  assert(mem->buf_ != NULL);
   memcpy(mem->buf_ + mem->end_, data, data_size);
   mem->end_ += data_size;
   assert(mem->end_ <= mem->buf_size_);
@@ -204,7 +207,9 @@ static int RemapMemBuffer(WebPIDecoder* const idec,
                           const uint8_t* const data, size_t data_size) {
   MemBuffer* const mem = &idec->mem_;
   const uint8_t* const old_buf = mem->buf_;
-  const uint8_t* const old_start = old_buf + mem->start_;
+  const uint8_t* const old_start =
+      (old_buf == NULL) ? NULL : old_buf + mem->start_;
+  assert(old_buf != NULL || mem->start_ == 0);
   assert(mem->mode_ == MEM_MODE_MAP);
 
   if (data_size < mem->buf_size_) return 0;  // can't remap to a shorter buffer!

src/dec/io_dec.c

@@ -25,21 +25,16 @@
 static int EmitYUV(const VP8Io* const io, WebPDecParams* const p) {
   WebPDecBuffer* output = p->output;
   const WebPYUVABuffer* const buf = &output->u.YUVA;
-  uint8_t* const y_dst = buf->y + io->mb_y * buf->y_stride;
-  uint8_t* const u_dst = buf->u + (io->mb_y >> 1) * buf->u_stride;
-  uint8_t* const v_dst = buf->v + (io->mb_y >> 1) * buf->v_stride;
+  uint8_t* const y_dst = buf->y + (size_t)io->mb_y * buf->y_stride;
+  uint8_t* const u_dst = buf->u + (size_t)(io->mb_y >> 1) * buf->u_stride;
+  uint8_t* const v_dst = buf->v + (size_t)(io->mb_y >> 1) * buf->v_stride;
   const int mb_w = io->mb_w;
   const int mb_h = io->mb_h;
   const int uv_w = (mb_w + 1) / 2;
   const int uv_h = (mb_h + 1) / 2;
-  int j;
-  for (j = 0; j < mb_h; ++j) {
-    memcpy(y_dst + j * buf->y_stride, io->y + j * io->y_stride, mb_w);
-  }
-  for (j = 0; j < uv_h; ++j) {
-    memcpy(u_dst + j * buf->u_stride, io->u + j * io->uv_stride, uv_w);
-    memcpy(v_dst + j * buf->v_stride, io->v + j * io->uv_stride, uv_w);
-  }
+  WebPCopyPlane(io->y, io->y_stride, y_dst, buf->y_stride, mb_w, mb_h);
+  WebPCopyPlane(io->u, io->uv_stride, u_dst, buf->u_stride, uv_w, uv_h);
+  WebPCopyPlane(io->v, io->uv_stride, v_dst, buf->v_stride, uv_w, uv_h);
   return io->mb_h;
 }
 
@@ -47,7 +42,7 @@ static int EmitYUV(const VP8Io* const io, WebPDecParams* const p) {
 static int EmitSampledRGB(const VP8Io* const io, WebPDecParams* const p) {
   WebPDecBuffer* const output = p->output;
   WebPRGBABuffer* const buf = &output->u.RGBA;
-  uint8_t* const dst = buf->rgba + io->mb_y * buf->stride;
+  uint8_t* const dst = buf->rgba + (size_t)io->mb_y * buf->stride;
   WebPSamplerProcessPlane(io->y, io->y_stride,
                           io->u, io->v, io->uv_stride,
                           dst, buf->stride, io->mb_w, io->mb_h,
@@ -62,7 +57,7 @@ static int EmitSampledRGB(const VP8Io* const io, WebPDecParams* const p) {
 static int EmitFancyRGB(const VP8Io* const io, WebPDecParams* const p) {
   int num_lines_out = io->mb_h;   // a priori guess
   const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* dst = buf->rgba + io->mb_y * buf->stride;
+  uint8_t* dst = buf->rgba + (size_t)io->mb_y * buf->stride;
   WebPUpsampleLinePairFunc upsample = WebPUpsamplers[p->output->colorspace];
   const uint8_t* cur_y = io->y;
   const uint8_t* cur_u = io->u;
@@ -133,7 +128,7 @@ static int EmitAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
   const WebPYUVABuffer* const buf = &p->output->u.YUVA;
   const int mb_w = io->mb_w;
   const int mb_h = io->mb_h;
-  uint8_t* dst = buf->a + io->mb_y * buf->a_stride;
+  uint8_t* dst = buf->a + (size_t)io->mb_y * buf->a_stride;
   int j;
   (void)expected_num_lines_out;
   assert(expected_num_lines_out == mb_h);
@@ -186,7 +181,7 @@ static int EmitAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
         (colorspace == MODE_ARGB || colorspace == MODE_Argb);
     const WebPRGBABuffer* const buf = &p->output->u.RGBA;
     int num_rows;
-    const int start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
+    const size_t start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
     uint8_t* const base_rgba = buf->rgba + start_y * buf->stride;
     uint8_t* const dst = base_rgba + (alpha_first ? 0 : 3);
     const int has_alpha = WebPDispatchAlpha(alpha, io->width, mb_w,
@@ -210,7 +205,7 @@ static int EmitAlphaRGBA4444(const VP8Io* const io, WebPDecParams* const p,
     const WEBP_CSP_MODE colorspace = p->output->colorspace;
     const WebPRGBABuffer* const buf = &p->output->u.RGBA;
     int num_rows;
-    const int start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
+    const size_t start_y = GetAlphaSourceRow(io, &alpha, &num_rows);
     uint8_t* const base_rgba = buf->rgba + start_y * buf->stride;
 #if (WEBP_SWAP_16BIT_CSP == 1)
     uint8_t* alpha_dst = base_rgba;
@@ -276,9 +271,9 @@ static int EmitRescaledYUV(const VP8Io* const io, WebPDecParams* const p) {
 static int EmitRescaledAlphaYUV(const VP8Io* const io, WebPDecParams* const p,
                                 int expected_num_lines_out) {
   const WebPYUVABuffer* const buf = &p->output->u.YUVA;
-  uint8_t* const dst_a = buf->a + p->last_y * buf->a_stride;
+  uint8_t* const dst_a = buf->a + (size_t)p->last_y * buf->a_stride;
   if (io->a != NULL) {
-    uint8_t* const dst_y = buf->y + p->last_y * buf->y_stride;
+    uint8_t* const dst_y = buf->y + (size_t)p->last_y * buf->y_stride;
     const int num_lines_out = Rescale(io->a, io->width, io->mb_h, p->scaler_a);
     assert(expected_num_lines_out == num_lines_out);
     if (num_lines_out > 0) {   // unmultiply the Y
@@ -356,7 +351,7 @@ static int ExportRGB(WebPDecParams* const p, int y_pos) {
   const WebPYUV444Converter convert =
       WebPYUV444Converters[p->output->colorspace];
   const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* dst = buf->rgba + y_pos * buf->stride;
+  uint8_t* dst = buf->rgba + (size_t)y_pos * buf->stride;
   int num_lines_out = 0;
   // For RGB rescaling, because of the YUV420, current scan position
   // U/V can be +1/-1 line from the Y one.  Hence the double test.
@@ -383,15 +378,15 @@ static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {
   while (j < mb_h) {
     const int y_lines_in =
         WebPRescalerImport(p->scaler_y, mb_h - j,
-                           io->y + j * io->y_stride, io->y_stride);
+                           io->y + (size_t)j * io->y_stride, io->y_stride);
     j += y_lines_in;
     if (WebPRescaleNeededLines(p->scaler_u, uv_mb_h - uv_j)) {
-      const int u_lines_in =
-          WebPRescalerImport(p->scaler_u, uv_mb_h - uv_j,
-                             io->u + uv_j * io->uv_stride, io->uv_stride);
-      const int v_lines_in =
-          WebPRescalerImport(p->scaler_v, uv_mb_h - uv_j,
-                             io->v + uv_j * io->uv_stride, io->uv_stride);
+      const int u_lines_in = WebPRescalerImport(
+          p->scaler_u, uv_mb_h - uv_j, io->u + (size_t)uv_j * io->uv_stride,
+          io->uv_stride);
+      const int v_lines_in = WebPRescalerImport(
+          p->scaler_v, uv_mb_h - uv_j, io->v + (size_t)uv_j * io->uv_stride,
+          io->uv_stride);
       (void)v_lines_in;   // remove a gcc warning
       assert(u_lines_in == v_lines_in);
       uv_j += u_lines_in;
@@ -403,7 +398,7 @@ static int EmitRescaledRGB(const VP8Io* const io, WebPDecParams* const p) {
 
 static int ExportAlpha(WebPDecParams* const p, int y_pos, int max_lines_out) {
   const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* const base_rgba = buf->rgba + y_pos * buf->stride;
+  uint8_t* const base_rgba = buf->rgba + (size_t)y_pos * buf->stride;
   const WEBP_CSP_MODE colorspace = p->output->colorspace;
   const int alpha_first =
       (colorspace == MODE_ARGB || colorspace == MODE_Argb);
@@ -431,7 +426,7 @@ static int ExportAlpha(WebPDecParams* const p, int y_pos, int max_lines_out) {
 static int ExportAlphaRGBA4444(WebPDecParams* const p, int y_pos,
                                int max_lines_out) {
   const WebPRGBABuffer* const buf = &p->output->u.RGBA;
-  uint8_t* const base_rgba = buf->rgba + y_pos * buf->stride;
+  uint8_t* const base_rgba = buf->rgba + (size_t)y_pos * buf->stride;
 #if (WEBP_SWAP_16BIT_CSP == 1)
   uint8_t* alpha_dst = base_rgba;
 #else
@@ -470,7 +465,7 @@ static int EmitRescaledAlphaRGB(const VP8Io* const io, WebPDecParams* const p,
     int lines_left = expected_num_out_lines;
     const int y_end = p->last_y + lines_left;
     while (lines_left > 0) {
-      const int row_offset = scaler->src_y - io->mb_y;
+      const int64_t row_offset = (int64_t)scaler->src_y - io->mb_y;
       WebPRescalerImport(scaler, io->mb_h + io->mb_y - scaler->src_y,
                          io->a + row_offset * io->width, io->width);
       lines_left -= p->emit_alpha_row(p, y_end - lines_left, lines_left);

src/dec/vp8_dec.c

@@ -494,13 +494,11 @@ static int GetCoeffsAlt(VP8BitReader* const br,
   return 16;
 }
 
-static WEBP_TSAN_IGNORE_FUNCTION void InitGetCoeffs(void) {
-  if (GetCoeffs == NULL) {
-    if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
-      GetCoeffs = GetCoeffsAlt;
-    } else {
-      GetCoeffs = GetCoeffsFast;
-    }
+WEBP_DSP_INIT_FUNC(InitGetCoeffs) {
+  if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
+    GetCoeffs = GetCoeffsAlt;
+  } else {
+    GetCoeffs = GetCoeffsFast;
   }
 }
 

src/dec/vp8i_dec.h

@@ -31,8 +31,8 @@ extern "C" {
 
 // version numbers
 #define DEC_MAJ_VERSION 1
-#define DEC_MIN_VERSION 0
-#define DEC_REV_VERSION 3
+#define DEC_MIN_VERSION 2
+#define DEC_REV_VERSION 0
 
 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
 // Constraints are: We need to store one 16x16 block of luma samples (y),

src/dec/vp8l_dec.c

@@ -951,7 +951,6 @@ static WEBP_INLINE void CopyBlock8b(uint8_t* const dst, int dist, int length) {
         break;
       default:
         goto Copy;
-        break;
     }
     CopySmallPattern8b(src, dst, length, pattern);
     return;

src/dec/vp8li_dec.h

@@ -37,7 +37,7 @@ struct VP8LTransform {
   int                    bits_;   // subsampling bits defining transform window.
   int                    xsize_;  // transform window X index.
   int                    ysize_;  // transform window Y index.
-  uint32_t              *data_;   // transform data.
+  uint32_t*              data_;   // transform data.
 };
 
 typedef struct {
@@ -48,9 +48,9 @@ typedef struct {
   int             huffman_mask_;
   int             huffman_subsample_bits_;
   int             huffman_xsize_;
-  uint32_t       *huffman_image_;
+  uint32_t*       huffman_image_;
   int             num_htree_groups_;
-  HTreeGroup     *htree_groups_;
+  HTreeGroup*     htree_groups_;
   HuffmanTables   huffman_tables_;
 } VP8LMetadata;
 
@@ -58,13 +58,13 @@ typedef struct VP8LDecoder VP8LDecoder;
 struct VP8LDecoder {
   VP8StatusCode    status_;
   VP8LDecodeState  state_;
-  VP8Io           *io_;
+  VP8Io*           io_;
 
-  const WebPDecBuffer *output_;    // shortcut to io->opaque->output
+  const WebPDecBuffer* output_;    // shortcut to io->opaque->output
 
-  uint32_t        *pixels_;        // Internal data: either uint8_t* for alpha
+  uint32_t*        pixels_;        // Internal data: either uint8_t* for alpha
                                    // or uint32_t* for BGRA.
-  uint32_t        *argb_cache_;    // Scratch buffer for temporary BGRA storage.
+  uint32_t*        argb_cache_;    // Scratch buffer for temporary BGRA storage.
 
   VP8LBitReader    br_;
   int              incremental_;   // if true, incremental decoding is expected
@@ -86,8 +86,8 @@ struct VP8LDecoder {
   // or'd bitset storing the transforms types.
   uint32_t         transforms_seen_;
 
-  uint8_t         *rescaler_memory;  // Working memory for rescaling work.
-  WebPRescaler    *rescaler;         // Common rescaler for all channels.
+  uint8_t*         rescaler_memory;  // Working memory for rescaling work.
+  WebPRescaler*    rescaler;         // Common rescaler for all channels.
 };
 
 //------------------------------------------------------------------------------

src/demux/Makefile.am

@@ -13,6 +13,6 @@ noinst_HEADERS =
 noinst_HEADERS += ../webp/format_constants.h
 
 libwebpdemux_la_LIBADD = ../libwebp.la
-libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:6:0
+libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:7:0
 libwebpdemuxincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebpdemux.pc

src/demux/anim_decode.c

@@ -346,12 +346,15 @@ int WebPAnimDecoderGetNext(WebPAnimDecoder* dec,
   {
     const uint8_t* in = iter.fragment.bytes;
     const size_t in_size = iter.fragment.size;
-    const size_t out_offset =
-        (iter.y_offset * width + iter.x_offset) * NUM_CHANNELS;
+    const uint32_t stride = width * NUM_CHANNELS;  // at most 25 + 2 bits
+    const uint64_t out_offset = (uint64_t)iter.y_offset * stride +
+                                (uint64_t)iter.x_offset * NUM_CHANNELS;  // 53b
+    const uint64_t size = (uint64_t)iter.height * stride;  // at most 25 + 27b
     WebPDecoderConfig* const config = &dec->config_;
     WebPRGBABuffer* const buf = &config->output.u.RGBA;
-    buf->stride = NUM_CHANNELS * width;
-    buf->size = buf->stride * iter.height;
+    if ((size_t)size != size) goto Error;
+    buf->stride = (int)stride;
+    buf->size = (size_t)size;
     buf->rgba = dec->curr_frame_ + out_offset;
 
     if (WebPDecode(in, in_size, config) != VP8_STATUS_OK) {

src/demux/demux.c

@@ -24,8 +24,8 @@
 #include "src/webp/format_constants.h"
 
 #define DMUX_MAJ_VERSION 1
-#define DMUX_MIN_VERSION 0
-#define DMUX_REV_VERSION 3
+#define DMUX_MIN_VERSION 2
+#define DMUX_REV_VERSION 0
 
 typedef struct {
   size_t start_;        // start location of the data
@@ -312,6 +312,7 @@ static ParseStatus ParseAnimationFrame(
   int bits;
   MemBuffer* const mem = &dmux->mem_;
   Frame* frame;
+  size_t start_offset;
   ParseStatus status =
       NewFrame(mem, ANMF_CHUNK_SIZE, frame_chunk_size, &frame);
   if (status != PARSE_OK) return status;
@@ -332,7 +333,11 @@ static ParseStatus ParseAnimationFrame(
 
   // Store a frame only if the animation flag is set there is some data for
   // this frame is available.
+  start_offset = mem->start_;
   status = StoreFrame(dmux->num_frames_ + 1, anmf_payload_size, mem, frame);
+  if (status != PARSE_ERROR && mem->start_ - start_offset > anmf_payload_size) {
+    status = PARSE_ERROR;
+  }
   if (status != PARSE_ERROR && is_animation && frame->frame_num_ > 0) {
     added_frame = AddFrame(dmux, frame);
     if (added_frame) {

src/demux/libwebpdemux.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,0,3
- PRODUCTVERSION 1,0,0,3
+ FILEVERSION 1,0,2,0
+ PRODUCTVERSION 1,0,2,0
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpdemux DLL"
-            VALUE "FileVersion", "1.0.3"
+            VALUE "FileVersion", "1.2.0"
             VALUE "InternalName", "libwebpdemux.dll"
             VALUE "LegalCopyright", "Copyright (C) 2019"
             VALUE "OriginalFilename", "libwebpdemux.dll"
             VALUE "ProductName", "WebP Image Demuxer"
-            VALUE "ProductVersion", "1.0.3"
+            VALUE "ProductVersion", "1.2.0"
         END
     END
     BLOCK "VarFileInfo"

src/dsp/alpha_processing.c

@@ -359,6 +359,11 @@ static int HasAlpha32b_C(const uint8_t* src, int length) {
   return 0;
 }
 
+static void AlphaReplace_C(uint32_t* src, int length, uint32_t color) {
+  int x;
+  for (x = 0; x < length; ++x) if ((src[x] >> 24) == 0) src[x] = color;
+}
+
 //------------------------------------------------------------------------------
 // Simple channel manipulations.
 
@@ -400,6 +405,7 @@ void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
 
 int (*WebPHasAlpha8b)(const uint8_t* src, int length);
 int (*WebPHasAlpha32b)(const uint8_t* src, int length);
+void (*WebPAlphaReplace)(uint32_t* src, int length, uint32_t color);
 
 //------------------------------------------------------------------------------
 // Init function
@@ -428,6 +434,7 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
 
   WebPHasAlpha8b = HasAlpha8b_C;
   WebPHasAlpha32b = HasAlpha32b_C;
+  WebPAlphaReplace = AlphaReplace_C;
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
@@ -469,4 +476,5 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
   assert(WebPPackRGB != NULL);
   assert(WebPHasAlpha8b != NULL);
   assert(WebPHasAlpha32b != NULL);
+  assert(WebPAlphaReplace != NULL);
 }

src/dsp/alpha_processing_sse2.c

@@ -265,6 +265,27 @@ static int HasAlpha32b_SSE2(const uint8_t* src, int length) {
   return 0;
 }
 
+static void AlphaReplace_SSE2(uint32_t* src, int length, uint32_t color) {
+  const __m128i m_color = _mm_set1_epi32(color);
+  const __m128i zero = _mm_setzero_si128();
+  int i = 0;
+  for (; i + 8 <= length; i += 8) {
+    const __m128i a0 = _mm_loadu_si128((const __m128i*)(src + i + 0));
+    const __m128i a1 = _mm_loadu_si128((const __m128i*)(src + i + 4));
+    const __m128i b0 = _mm_srai_epi32(a0, 24);
+    const __m128i b1 = _mm_srai_epi32(a1, 24);
+    const __m128i c0 = _mm_cmpeq_epi32(b0, zero);
+    const __m128i c1 = _mm_cmpeq_epi32(b1, zero);
+    const __m128i d0 = _mm_and_si128(c0, m_color);
+    const __m128i d1 = _mm_and_si128(c1, m_color);
+    const __m128i e0 = _mm_andnot_si128(c0, a0);
+    const __m128i e1 = _mm_andnot_si128(c1, a1);
+    _mm_storeu_si128((__m128i*)(src + i + 0), _mm_or_si128(d0, e0));
+    _mm_storeu_si128((__m128i*)(src + i + 4), _mm_or_si128(d1, e1));
+  }
+  for (; i < length; ++i) if ((src[i] >> 24) == 0) src[i] = color;
+}
+
 // -----------------------------------------------------------------------------
 // Apply alpha value to rows
 
@@ -334,6 +355,7 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessingSSE2(void) {
 
   WebPHasAlpha8b = HasAlpha8b_SSE2;
   WebPHasAlpha32b = HasAlpha32b_SSE2;
+  WebPAlphaReplace = AlphaReplace_SSE2;
 }
 
 #else  // !WEBP_USE_SSE2

src/dsp/cpu.c

@@ -55,12 +55,18 @@ static WEBP_INLINE void GetCPUInfo(int cpu_info[4], int info_type) {
     : "=a"(cpu_info[0]), "=b"(cpu_info[1]), "=c"(cpu_info[2]), "=d"(cpu_info[3])
     : "a"(info_type), "c"(0));
 }
-#elif (defined(_M_X64) || defined(_M_IX86)) && \
-      defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 150030729  // >= VS2008 SP1
+#elif defined(_MSC_VER) && (defined(_M_X64) || defined(_M_IX86))
+
+#if defined(_MSC_FULL_VER) && _MSC_FULL_VER >= 150030729  // >= VS2008 SP1
 #include <intrin.h>
 #define GetCPUInfo(info, type) __cpuidex(info, type, 0)  // set ecx=0
-#elif defined(WEBP_MSC_SSE2)
+#define WEBP_HAVE_MSC_CPUID
+#elif _MSC_VER > 1310
+#include <intrin.h>
 #define GetCPUInfo __cpuid
+#define WEBP_HAVE_MSC_CPUID
+#endif
+
 #endif
 
 // NaCl has no support for xgetbv or the raw opcode.
@@ -94,7 +100,7 @@ static WEBP_INLINE uint64_t xgetbv(void) {
 #define xgetbv() 0U  // no AVX for older x64 or unrecognized toolchains.
 #endif
 
-#if defined(__i386__) || defined(__x86_64__) || defined(WEBP_MSC_SSE2)
+#if defined(__i386__) || defined(__x86_64__) || defined(WEBP_HAVE_MSC_CPUID)
 
 // helper function for run-time detection of slow SSSE3 platforms
 static int CheckSlowModel(int info) {
@@ -179,6 +185,30 @@ static int AndroidCPUInfo(CPUFeature feature) {
   return 0;
 }
 VP8CPUInfo VP8GetCPUInfo = AndroidCPUInfo;
+#elif defined(EMSCRIPTEN) // also needs to be before generic NEON test
+// Use compile flags as an indicator of SIMD support instead of a runtime check.
+static int wasmCPUInfo(CPUFeature feature) {
+  switch (feature) {
+#ifdef WEBP_USE_SSE2
+    case kSSE2:
+      return 1;
+#endif
+#ifdef WEBP_USE_SSE41
+    case kSSE3:
+    case kSlowSSSE3:
+    case kSSE4_1:
+      return 1;
+#endif
+#ifdef WEBP_USE_NEON
+    case kNEON:
+      return 1;
+#endif
+    default:
+      break;
+  }
+  return 0;
+}
+VP8CPUInfo VP8GetCPUInfo = wasmCPUInfo;
 #elif defined(WEBP_USE_NEON)
 // define a dummy function to enable turning off NEON at runtime by setting
 // VP8DecGetCPUInfo = NULL

src/dsp/dec_neon.c

@@ -1283,12 +1283,12 @@ static void DC4_NEON(uint8_t* dst) {    // DC
   const uint8x8_t A = vld1_u8(dst - BPS);  // top row
   const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
   const uint16x4_t p1 = vpadd_u16(p0, p0);
-  const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
-  const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
-  const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
-  const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
-  const uint16x8_t s0 = vaddq_u16(L0, L1);
-  const uint16x8_t s1 = vaddq_u16(L2, L3);
+  const uint8x8_t L0 = vld1_u8(dst + 0 * BPS - 1);
+  const uint8x8_t L1 = vld1_u8(dst + 1 * BPS - 1);
+  const uint8x8_t L2 = vld1_u8(dst + 2 * BPS - 1);
+  const uint8x8_t L3 = vld1_u8(dst + 3 * BPS - 1);
+  const uint16x8_t s0 = vaddl_u8(L0, L1);
+  const uint16x8_t s1 = vaddl_u8(L2, L3);
   const uint16x8_t s01 = vaddq_u16(s0, s1);
   const uint16x8_t sum = vaddq_u16(s01, vcombine_u16(p1, p1));
   const uint8x8_t dc0 = vrshrn_n_u16(sum, 3);  // (sum + 4) >> 3
@@ -1361,7 +1361,8 @@ static void RD4_NEON(uint8_t* dst) {   // Down-right
   const uint32_t J = dst[-1 + 1 * BPS];
   const uint32_t K = dst[-1 + 2 * BPS];
   const uint32_t L = dst[-1 + 3 * BPS];
-  const uint64x1_t LKJI____ = vcreate_u64(L | (K << 8) | (J << 16) | (I << 24));
+  const uint64x1_t LKJI____ =
+      vcreate_u64((uint64_t)L | (K << 8) | (J << 16) | (I << 24));
   const uint64x1_t LKJIXABC = vorr_u64(LKJI____, ____XABC);
   const uint8x8_t KJIXABC_ = vreinterpret_u8_u64(vshr_n_u64(LKJIXABC, 8));
   const uint8x8_t JIXABC__ = vreinterpret_u8_u64(vshr_n_u64(LKJIXABC, 16));
@@ -1427,25 +1428,30 @@ static WEBP_INLINE void DC8_NEON(uint8_t* dst, int do_top, int do_left) {
 
   if (do_top) {
     const uint8x8_t A = vld1_u8(dst - BPS);  // top row
+#if defined(__aarch64__)
+    const uint16_t p2 = vaddlv_u8(A);
+    sum_top = vdupq_n_u16(p2);
+#else
     const uint16x4_t p0 = vpaddl_u8(A);  // cascading summation of the top
     const uint16x4_t p1 = vpadd_u16(p0, p0);
     const uint16x4_t p2 = vpadd_u16(p1, p1);
     sum_top = vcombine_u16(p2, p2);
+#endif
   }
 
   if (do_left) {
-    const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + 0 * BPS - 1));
-    const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + 1 * BPS - 1));
-    const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + 2 * BPS - 1));
-    const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + 3 * BPS - 1));
-    const uint16x8_t L4 = vmovl_u8(vld1_u8(dst + 4 * BPS - 1));
-    const uint16x8_t L5 = vmovl_u8(vld1_u8(dst + 5 * BPS - 1));
-    const uint16x8_t L6 = vmovl_u8(vld1_u8(dst + 6 * BPS - 1));
-    const uint16x8_t L7 = vmovl_u8(vld1_u8(dst + 7 * BPS - 1));
-    const uint16x8_t s0 = vaddq_u16(L0, L1);
-    const uint16x8_t s1 = vaddq_u16(L2, L3);
-    const uint16x8_t s2 = vaddq_u16(L4, L5);
-    const uint16x8_t s3 = vaddq_u16(L6, L7);
+    const uint8x8_t L0 = vld1_u8(dst + 0 * BPS - 1);
+    const uint8x8_t L1 = vld1_u8(dst + 1 * BPS - 1);
+    const uint8x8_t L2 = vld1_u8(dst + 2 * BPS - 1);
+    const uint8x8_t L3 = vld1_u8(dst + 3 * BPS - 1);
+    const uint8x8_t L4 = vld1_u8(dst + 4 * BPS - 1);
+    const uint8x8_t L5 = vld1_u8(dst + 5 * BPS - 1);
+    const uint8x8_t L6 = vld1_u8(dst + 6 * BPS - 1);
+    const uint8x8_t L7 = vld1_u8(dst + 7 * BPS - 1);
+    const uint16x8_t s0 = vaddl_u8(L0, L1);
+    const uint16x8_t s1 = vaddl_u8(L2, L3);
+    const uint16x8_t s2 = vaddl_u8(L4, L5);
+    const uint16x8_t s3 = vaddl_u8(L6, L7);
     const uint16x8_t s01 = vaddq_u16(s0, s1);
     const uint16x8_t s23 = vaddq_u16(s2, s3);
     sum_left = vaddq_u16(s01, s23);
@@ -1505,29 +1511,34 @@ static WEBP_INLINE void DC16_NEON(uint8_t* dst, int do_top, int do_left) {
 
   if (do_top) {
     const uint8x16_t A = vld1q_u8(dst - BPS);  // top row
+#if defined(__aarch64__)
+    const uint16_t p3 = vaddlvq_u8(A);
+    sum_top = vdupq_n_u16(p3);
+#else
     const uint16x8_t p0 = vpaddlq_u8(A);  // cascading summation of the top
     const uint16x4_t p1 = vadd_u16(vget_low_u16(p0), vget_high_u16(p0));
     const uint16x4_t p2 = vpadd_u16(p1, p1);
     const uint16x4_t p3 = vpadd_u16(p2, p2);
     sum_top = vcombine_u16(p3, p3);
+#endif
   }
 
   if (do_left) {
     int i;
     sum_left = vdupq_n_u16(0);
     for (i = 0; i < 16; i += 8) {
-      const uint16x8_t L0 = vmovl_u8(vld1_u8(dst + (i + 0) * BPS - 1));
-      const uint16x8_t L1 = vmovl_u8(vld1_u8(dst + (i + 1) * BPS - 1));
-      const uint16x8_t L2 = vmovl_u8(vld1_u8(dst + (i + 2) * BPS - 1));
-      const uint16x8_t L3 = vmovl_u8(vld1_u8(dst + (i + 3) * BPS - 1));
-      const uint16x8_t L4 = vmovl_u8(vld1_u8(dst + (i + 4) * BPS - 1));
-      const uint16x8_t L5 = vmovl_u8(vld1_u8(dst + (i + 5) * BPS - 1));
-      const uint16x8_t L6 = vmovl_u8(vld1_u8(dst + (i + 6) * BPS - 1));
-      const uint16x8_t L7 = vmovl_u8(vld1_u8(dst + (i + 7) * BPS - 1));
-      const uint16x8_t s0 = vaddq_u16(L0, L1);
-      const uint16x8_t s1 = vaddq_u16(L2, L3);
-      const uint16x8_t s2 = vaddq_u16(L4, L5);
-      const uint16x8_t s3 = vaddq_u16(L6, L7);
+      const uint8x8_t L0 = vld1_u8(dst + (i + 0) * BPS - 1);
+      const uint8x8_t L1 = vld1_u8(dst + (i + 1) * BPS - 1);
+      const uint8x8_t L2 = vld1_u8(dst + (i + 2) * BPS - 1);
+      const uint8x8_t L3 = vld1_u8(dst + (i + 3) * BPS - 1);
+      const uint8x8_t L4 = vld1_u8(dst + (i + 4) * BPS - 1);
+      const uint8x8_t L5 = vld1_u8(dst + (i + 5) * BPS - 1);
+      const uint8x8_t L6 = vld1_u8(dst + (i + 6) * BPS - 1);
+      const uint8x8_t L7 = vld1_u8(dst + (i + 7) * BPS - 1);
+      const uint16x8_t s0 = vaddl_u8(L0, L1);
+      const uint16x8_t s1 = vaddl_u8(L2, L3);
+      const uint16x8_t s2 = vaddl_u8(L4, L5);
+      const uint16x8_t s3 = vaddl_u8(L6, L7);
       const uint16x8_t s01 = vaddq_u16(s0, s1);
       const uint16x8_t s23 = vaddq_u16(s2, s3);
       const uint16x8_t sum = vaddq_u16(s01, s23);

src/dsp/dsp.h

@@ -51,9 +51,7 @@ extern "C" {
 # define __has_builtin(x) 0
 #endif
 
-// for now, none of the optimizations below are available in emscripten
-#if !defined(EMSCRIPTEN)
-
+#if !defined(HAVE_CONFIG_H)
 #if defined(_MSC_VER) && _MSC_VER > 1310 && \
     (defined(_M_X64) || defined(_M_IX86))
 #define WEBP_MSC_SSE2  // Visual C++ SSE2 targets
@@ -63,6 +61,7 @@ extern "C" {
     (defined(_M_X64) || defined(_M_IX86))
 #define WEBP_MSC_SSE41  // Visual C++ SSE4.1 targets
 #endif
+#endif
 
 // WEBP_HAVE_* are used to indicate the presence of the instruction set in dsp
 // files without intrinsics, allowing the corresponding Init() to be called.
@@ -76,6 +75,9 @@ extern "C" {
 #define WEBP_USE_SSE41
 #endif
 
+#undef WEBP_MSC_SSE41
+#undef WEBP_MSC_SSE2
+
 // The intrinsics currently cause compiler errors with arm-nacl-gcc and the
 // inline assembly would need to be modified for use with Native Client.
 #if (defined(__ARM_NEON__) || \
@@ -110,8 +112,6 @@ extern "C" {
 #define WEBP_USE_MSA
 #endif
 
-#endif  /* EMSCRIPTEN */
-
 #ifndef WEBP_DSP_OMIT_C_CODE
 #define WEBP_DSP_OMIT_C_CODE 1
 #endif
@@ -193,6 +193,12 @@ extern "C" {
 #endif
 #endif
 
+// If 'ptr' is NULL, returns NULL. Otherwise returns 'ptr + off'.
+// Prevents undefined behavior sanitizer nullptr-with-nonzero-offset warning.
+#if !defined(WEBP_OFFSET_PTR)
+#define WEBP_OFFSET_PTR(ptr, off) (((ptr) == NULL) ? NULL : ((ptr) + (off)))
+#endif
+
 // Regularize the definition of WEBP_SWAP_16BIT_CSP (backward compatibility)
 #if !defined(WEBP_SWAP_16BIT_CSP)
 #define WEBP_SWAP_16BIT_CSP 0
@@ -246,9 +252,9 @@ extern VP8Fdct VP8FTransform2;   // performs two transforms at a time
 extern VP8WHT VP8FTransformWHT;
 // Predictions
 // *dst is the destination block. *top and *left can be NULL.
-typedef void (*VP8IntraPreds)(uint8_t *dst, const uint8_t* left,
+typedef void (*VP8IntraPreds)(uint8_t* dst, const uint8_t* left,
                               const uint8_t* top);
-typedef void (*VP8Intra4Preds)(uint8_t *dst, const uint8_t* top);
+typedef void (*VP8Intra4Preds)(uint8_t* dst, const uint8_t* top);
 extern VP8Intra4Preds VP8EncPredLuma4;
 extern VP8IntraPreds VP8EncPredLuma16;
 extern VP8IntraPreds VP8EncPredChroma8;
@@ -632,6 +638,8 @@ extern void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
 extern int (*WebPHasAlpha8b)(const uint8_t* src, int length);
 // This function returns true if src[4*i] contains a value different from 0xff.
 extern int (*WebPHasAlpha32b)(const uint8_t* src, int length);
+// replaces transparent values in src[] by 'color'.
+extern void (*WebPAlphaReplace)(uint32_t* src, int length, uint32_t color);
 
 // To be called first before using the above.
 void WebPInitAlphaProcessing(void);

src/dsp/lossless.c

@@ -81,7 +81,7 @@ static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
 
 // gcc <= 4.9 on ARM generates incorrect code in Select() when Sub3() is
 // inlined.
-#if defined(__arm__) && LOCAL_GCC_VERSION <= 0x409
+#if defined(__arm__) && defined(__GNUC__) && LOCAL_GCC_VERSION <= 0x409
 # define LOCAL_INLINE __attribute__ ((noinline))
 #else
 # define LOCAL_INLINE WEBP_INLINE
@@ -107,88 +107,93 @@ static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
 //------------------------------------------------------------------------------
 // Predictors
 
-static uint32_t Predictor0_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor0_C(uint32_t left, const uint32_t* const top) {
   (void)top;
   (void)left;
   return ARGB_BLACK;
 }
-static uint32_t Predictor1_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor1_C(uint32_t left, const uint32_t* const top) {
   (void)top;
   return left;
 }
-static uint32_t Predictor2_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor2_C(uint32_t left, const uint32_t* const top) {
   (void)left;
   return top[0];
 }
-static uint32_t Predictor3_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor3_C(uint32_t left, const uint32_t* const top) {
   (void)left;
   return top[1];
 }
-static uint32_t Predictor4_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor4_C(uint32_t left, const uint32_t* const top) {
   (void)left;
   return top[-1];
 }
-static uint32_t Predictor5_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor5_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Average3(left, top[0], top[1]);
   return pred;
 }
-static uint32_t Predictor6_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor6_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Average2(left, top[-1]);
   return pred;
 }
-static uint32_t Predictor7_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor7_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Average2(left, top[0]);
   return pred;
 }
-static uint32_t Predictor8_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor8_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Average2(top[-1], top[0]);
   (void)left;
   return pred;
 }
-static uint32_t Predictor9_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor9_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Average2(top[0], top[1]);
   (void)left;
   return pred;
 }
-static uint32_t Predictor10_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor10_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Average4(left, top[-1], top[0], top[1]);
   return pred;
 }
-static uint32_t Predictor11_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor11_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = Select(top[0], left, top[-1]);
   return pred;
 }
-static uint32_t Predictor12_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor12_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = ClampedAddSubtractFull(left, top[0], top[-1]);
   return pred;
 }
-static uint32_t Predictor13_C(uint32_t left, const uint32_t* const top) {
+uint32_t VP8LPredictor13_C(uint32_t left, const uint32_t* const top) {
   const uint32_t pred = ClampedAddSubtractHalf(left, top[0], top[-1]);
   return pred;
 }
 
-GENERATE_PREDICTOR_ADD(Predictor0_C, PredictorAdd0_C)
+static void PredictorAdd0_C(const uint32_t* in, const uint32_t* upper,
+                            int num_pixels, uint32_t* out) {
+  int x;
+  (void)upper;
+  for (x = 0; x < num_pixels; ++x) out[x] = VP8LAddPixels(in[x], ARGB_BLACK);
+}
 static void PredictorAdd1_C(const uint32_t* in, const uint32_t* upper,
                             int num_pixels, uint32_t* out) {
   int i;
   uint32_t left = out[-1];
+  (void)upper;
   for (i = 0; i < num_pixels; ++i) {
     out[i] = left = VP8LAddPixels(in[i], left);
   }
-  (void)upper;
 }
-GENERATE_PREDICTOR_ADD(Predictor2_C, PredictorAdd2_C)
-GENERATE_PREDICTOR_ADD(Predictor3_C, PredictorAdd3_C)
-GENERATE_PREDICTOR_ADD(Predictor4_C, PredictorAdd4_C)
-GENERATE_PREDICTOR_ADD(Predictor5_C, PredictorAdd5_C)
-GENERATE_PREDICTOR_ADD(Predictor6_C, PredictorAdd6_C)
-GENERATE_PREDICTOR_ADD(Predictor7_C, PredictorAdd7_C)
-GENERATE_PREDICTOR_ADD(Predictor8_C, PredictorAdd8_C)
-GENERATE_PREDICTOR_ADD(Predictor9_C, PredictorAdd9_C)
-GENERATE_PREDICTOR_ADD(Predictor10_C, PredictorAdd10_C)
-GENERATE_PREDICTOR_ADD(Predictor11_C, PredictorAdd11_C)
-GENERATE_PREDICTOR_ADD(Predictor12_C, PredictorAdd12_C)
-GENERATE_PREDICTOR_ADD(Predictor13_C, PredictorAdd13_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor2_C, PredictorAdd2_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor3_C, PredictorAdd3_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor4_C, PredictorAdd4_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor5_C, PredictorAdd5_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor6_C, PredictorAdd6_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor7_C, PredictorAdd7_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor8_C, PredictorAdd8_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor9_C, PredictorAdd9_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor10_C, PredictorAdd10_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor11_C, PredictorAdd11_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor12_C, PredictorAdd12_C)
+GENERATE_PREDICTOR_ADD(VP8LPredictor13_C, PredictorAdd13_C)
 
 //------------------------------------------------------------------------------
 
@@ -557,7 +562,6 @@ VP8LPredictorFunc VP8LPredictors[16];
 
 // exposed plain-C implementations
 VP8LPredictorAddSubFunc VP8LPredictorsAdd_C[16];
-VP8LPredictorFunc VP8LPredictors_C[16];
 
 VP8LTransformColorInverseFunc VP8LTransformColorInverse;
 
@@ -595,8 +599,7 @@ extern void VP8LDspInitMSA(void);
 } while (0);
 
 WEBP_DSP_INIT_FUNC(VP8LDspInit) {
-  COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors)
-  COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors_C)
+  COPY_PREDICTOR_ARRAY(VP8LPredictor, VP8LPredictors)
   COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd)
   COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd_C)
 

src/dsp/lossless.h

@@ -30,7 +30,22 @@ extern "C" {
 
 typedef uint32_t (*VP8LPredictorFunc)(uint32_t left, const uint32_t* const top);
 extern VP8LPredictorFunc VP8LPredictors[16];
-extern VP8LPredictorFunc VP8LPredictors_C[16];
+
+uint32_t VP8LPredictor0_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor1_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor2_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor3_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor4_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor5_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor6_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor7_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor8_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor9_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor10_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor11_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor12_C(uint32_t left, const uint32_t* const top);
+uint32_t VP8LPredictor13_C(uint32_t left, const uint32_t* const top);
+
 // These Add/Sub function expects upper[-1] and out[-1] to be readable.
 typedef void (*VP8LPredictorAddSubFunc)(const uint32_t* in,
                                         const uint32_t* upper, int num_pixels,

src/dsp/lossless_common.h

@@ -177,24 +177,13 @@ uint32_t VP8LSubPixels(uint32_t a, uint32_t b) {
 static void PREDICTOR_ADD(const uint32_t* in, const uint32_t* upper, \
                           int num_pixels, uint32_t* out) {           \
   int x;                                                             \
+  assert(upper != NULL);                                             \
   for (x = 0; x < num_pixels; ++x) {                                 \
     const uint32_t pred = (PREDICTOR)(out[x - 1], upper + x);        \
     out[x] = VP8LAddPixels(in[x], pred);                             \
   }                                                                  \
 }
 
-// It subtracts the prediction from the input pixel and stores the residual
-// in the output pixel.
-#define GENERATE_PREDICTOR_SUB(PREDICTOR, PREDICTOR_SUB)             \
-static void PREDICTOR_SUB(const uint32_t* in, const uint32_t* upper, \
-                          int num_pixels, uint32_t* out) {           \
-  int x;                                                             \
-  for (x = 0; x < num_pixels; ++x) {                                 \
-    const uint32_t pred = (PREDICTOR)(in[x - 1], upper + x);         \
-    out[x] = VP8LSubPixels(in[x], pred);                             \
-  }                                                                  \
-}
-
 #ifdef __cplusplus
 }    // extern "C"
 #endif

src/dsp/lossless_enc.c

@@ -702,140 +702,6 @@ void VP8LHistogramAdd(const VP8LHistogram* const a,
 //------------------------------------------------------------------------------
 // Image transforms.
 
-static WEBP_INLINE uint32_t Average2(uint32_t a0, uint32_t a1) {
-  return (((a0 ^ a1) & 0xfefefefeu) >> 1) + (a0 & a1);
-}
-
-static WEBP_INLINE uint32_t Average3(uint32_t a0, uint32_t a1, uint32_t a2) {
-  return Average2(Average2(a0, a2), a1);
-}
-
-static WEBP_INLINE uint32_t Average4(uint32_t a0, uint32_t a1,
-                                     uint32_t a2, uint32_t a3) {
-  return Average2(Average2(a0, a1), Average2(a2, a3));
-}
-
-static WEBP_INLINE uint32_t Clip255(uint32_t a) {
-  if (a < 256) {
-    return a;
-  }
-  // return 0, when a is a negative integer.
-  // return 255, when a is positive.
-  return ~a >> 24;
-}
-
-static WEBP_INLINE int AddSubtractComponentFull(int a, int b, int c) {
-  return Clip255(a + b - c);
-}
-
-static WEBP_INLINE uint32_t ClampedAddSubtractFull(uint32_t c0, uint32_t c1,
-                                                   uint32_t c2) {
-  const int a = AddSubtractComponentFull(c0 >> 24, c1 >> 24, c2 >> 24);
-  const int r = AddSubtractComponentFull((c0 >> 16) & 0xff,
-                                         (c1 >> 16) & 0xff,
-                                         (c2 >> 16) & 0xff);
-  const int g = AddSubtractComponentFull((c0 >> 8) & 0xff,
-                                         (c1 >> 8) & 0xff,
-                                         (c2 >> 8) & 0xff);
-  const int b = AddSubtractComponentFull(c0 & 0xff, c1 & 0xff, c2 & 0xff);
-  return ((uint32_t)a << 24) | (r << 16) | (g << 8) | b;
-}
-
-static WEBP_INLINE int AddSubtractComponentHalf(int a, int b) {
-  return Clip255(a + (a - b) / 2);
-}
-
-static WEBP_INLINE uint32_t ClampedAddSubtractHalf(uint32_t c0, uint32_t c1,
-                                                   uint32_t c2) {
-  const uint32_t ave = Average2(c0, c1);
-  const int a = AddSubtractComponentHalf(ave >> 24, c2 >> 24);
-  const int r = AddSubtractComponentHalf((ave >> 16) & 0xff, (c2 >> 16) & 0xff);
-  const int g = AddSubtractComponentHalf((ave >> 8) & 0xff, (c2 >> 8) & 0xff);
-  const int b = AddSubtractComponentHalf((ave >> 0) & 0xff, (c2 >> 0) & 0xff);
-  return ((uint32_t)a << 24) | (r << 16) | (g << 8) | b;
-}
-
-// gcc-4.9 on ARM generates incorrect code in Select() when Sub3() is inlined.
-#if defined(__arm__) && \
-    (LOCAL_GCC_VERSION == 0x409 || LOCAL_GCC_VERSION == 0x408)
-# define LOCAL_INLINE __attribute__ ((noinline))
-#else
-# define LOCAL_INLINE WEBP_INLINE
-#endif
-
-static LOCAL_INLINE int Sub3(int a, int b, int c) {
-  const int pb = b - c;
-  const int pa = a - c;
-  return abs(pb) - abs(pa);
-}
-
-#undef LOCAL_INLINE
-
-static WEBP_INLINE uint32_t Select(uint32_t a, uint32_t b, uint32_t c) {
-  const int pa_minus_pb =
-      Sub3((a >> 24)       , (b >> 24)       , (c >> 24)       ) +
-      Sub3((a >> 16) & 0xff, (b >> 16) & 0xff, (c >> 16) & 0xff) +
-      Sub3((a >>  8) & 0xff, (b >>  8) & 0xff, (c >>  8) & 0xff) +
-      Sub3((a      ) & 0xff, (b      ) & 0xff, (c      ) & 0xff);
-  return (pa_minus_pb <= 0) ? a : b;
-}
-
-//------------------------------------------------------------------------------
-// Predictors
-
-static uint32_t Predictor2(uint32_t left, const uint32_t* const top) {
-  (void)left;
-  return top[0];
-}
-static uint32_t Predictor3(uint32_t left, const uint32_t* const top) {
-  (void)left;
-  return top[1];
-}
-static uint32_t Predictor4(uint32_t left, const uint32_t* const top) {
-  (void)left;
-  return top[-1];
-}
-static uint32_t Predictor5(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average3(left, top[0], top[1]);
-  return pred;
-}
-static uint32_t Predictor6(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average2(left, top[-1]);
-  return pred;
-}
-static uint32_t Predictor7(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average2(left, top[0]);
-  return pred;
-}
-static uint32_t Predictor8(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average2(top[-1], top[0]);
-  (void)left;
-  return pred;
-}
-static uint32_t Predictor9(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average2(top[0], top[1]);
-  (void)left;
-  return pred;
-}
-static uint32_t Predictor10(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Average4(left, top[-1], top[0], top[1]);
-  return pred;
-}
-static uint32_t Predictor11(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = Select(top[0], left, top[-1]);
-  return pred;
-}
-static uint32_t Predictor12(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = ClampedAddSubtractFull(left, top[0], top[-1]);
-  return pred;
-}
-static uint32_t Predictor13(uint32_t left, const uint32_t* const top) {
-  const uint32_t pred = ClampedAddSubtractHalf(left, top[0], top[-1]);
-  return pred;
-}
-
-//------------------------------------------------------------------------------
-
 static void PredictorSub0_C(const uint32_t* in, const uint32_t* upper,
                             int num_pixels, uint32_t* out) {
   int i;
@@ -850,18 +716,33 @@ static void PredictorSub1_C(const uint32_t* in, const uint32_t* upper,
   (void)upper;
 }
 
-GENERATE_PREDICTOR_SUB(Predictor2, PredictorSub2_C)
-GENERATE_PREDICTOR_SUB(Predictor3, PredictorSub3_C)
-GENERATE_PREDICTOR_SUB(Predictor4, PredictorSub4_C)
-GENERATE_PREDICTOR_SUB(Predictor5, PredictorSub5_C)
-GENERATE_PREDICTOR_SUB(Predictor6, PredictorSub6_C)
-GENERATE_PREDICTOR_SUB(Predictor7, PredictorSub7_C)
-GENERATE_PREDICTOR_SUB(Predictor8, PredictorSub8_C)
-GENERATE_PREDICTOR_SUB(Predictor9, PredictorSub9_C)
-GENERATE_PREDICTOR_SUB(Predictor10, PredictorSub10_C)
-GENERATE_PREDICTOR_SUB(Predictor11, PredictorSub11_C)
-GENERATE_PREDICTOR_SUB(Predictor12, PredictorSub12_C)
-GENERATE_PREDICTOR_SUB(Predictor13, PredictorSub13_C)
+// It subtracts the prediction from the input pixel and stores the residual
+// in the output pixel.
+#define GENERATE_PREDICTOR_SUB(PREDICTOR_I)                                \
+static void PredictorSub##PREDICTOR_I##_C(const uint32_t* in,              \
+                                          const uint32_t* upper,           \
+                                          int num_pixels, uint32_t* out) { \
+  int x;                                                                   \
+  assert(upper != NULL);                                                   \
+  for (x = 0; x < num_pixels; ++x) {                                       \
+    const uint32_t pred =                                                  \
+        VP8LPredictor##PREDICTOR_I##_C(in[x - 1], upper + x);              \
+    out[x] = VP8LSubPixels(in[x], pred);                                   \
+  }                                                                        \
+}
+
+GENERATE_PREDICTOR_SUB(2)
+GENERATE_PREDICTOR_SUB(3)
+GENERATE_PREDICTOR_SUB(4)
+GENERATE_PREDICTOR_SUB(5)
+GENERATE_PREDICTOR_SUB(6)
+GENERATE_PREDICTOR_SUB(7)
+GENERATE_PREDICTOR_SUB(8)
+GENERATE_PREDICTOR_SUB(9)
+GENERATE_PREDICTOR_SUB(10)
+GENERATE_PREDICTOR_SUB(11)
+GENERATE_PREDICTOR_SUB(12)
+GENERATE_PREDICTOR_SUB(13)
 
 //------------------------------------------------------------------------------
 

src/dsp/lossless_enc_sse2.c

@@ -249,6 +249,7 @@ static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) {
     }                                  \
   } while (0)
 
+#if !(defined(__i386__) || defined(_M_IX86))
 static float CombinedShannonEntropy_SSE2(const int X[256], const int Y[256]) {
   int i;
   double retval = 0.;
@@ -300,6 +301,8 @@ static float CombinedShannonEntropy_SSE2(const int X[256], const int Y[256]) {
   retval += VP8LFastSLog2(sumX) + VP8LFastSLog2(sumXY);
   return (float)retval;
 }
+#endif  // !(defined(__i386__) || defined(_M_IX86))
+
 #undef ANALYZE_X_OR_Y
 #undef ANALYZE_XY
 
@@ -455,24 +458,27 @@ static void PredictorSub0_SSE2(const uint32_t* in, const uint32_t* upper,
     _mm_storeu_si128((__m128i*)&out[i], res);
   }
   if (i != num_pixels) {
-    VP8LPredictorsSub_C[0](in + i, upper + i, num_pixels - i, out + i);
+    VP8LPredictorsSub_C[0](in + i, NULL, num_pixels - i, out + i);
   }
+  (void)upper;
 }
 
-#define GENERATE_PREDICTOR_1(X, IN)                                           \
-static void PredictorSub##X##_SSE2(const uint32_t* in, const uint32_t* upper, \
-                                   int num_pixels, uint32_t* out) {           \
-  int i;                                                                      \
-  for (i = 0; i + 4 <= num_pixels; i += 4) {                                  \
-    const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);              \
-    const __m128i pred = _mm_loadu_si128((const __m128i*)&(IN));              \
-    const __m128i res = _mm_sub_epi8(src, pred);                              \
-    _mm_storeu_si128((__m128i*)&out[i], res);                                 \
-  }                                                                           \
-  if (i != num_pixels) {                                                      \
-    VP8LPredictorsSub_C[(X)](in + i, upper + i, num_pixels - i, out + i);     \
-  }                                                                           \
-}
+#define GENERATE_PREDICTOR_1(X, IN)                                         \
+  static void PredictorSub##X##_SSE2(const uint32_t* const in,              \
+                                     const uint32_t* const upper,           \
+                                     int num_pixels, uint32_t* const out) { \
+    int i;                                                                  \
+    for (i = 0; i + 4 <= num_pixels; i += 4) {                              \
+      const __m128i src = _mm_loadu_si128((const __m128i*)&in[i]);          \
+      const __m128i pred = _mm_loadu_si128((const __m128i*)&(IN));          \
+      const __m128i res = _mm_sub_epi8(src, pred);                          \
+      _mm_storeu_si128((__m128i*)&out[i], res);                             \
+    }                                                                       \
+    if (i != num_pixels) {                                                  \
+      VP8LPredictorsSub_C[(X)](in + i, WEBP_OFFSET_PTR(upper, i),           \
+                               num_pixels - i, out + i);                    \
+    }                                                                       \
+  }
 
 GENERATE_PREDICTOR_1(1, in[i - 1])       // Predictor1: L
 GENERATE_PREDICTOR_1(2, upper[i])        // Predictor2: T
@@ -656,7 +662,12 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitSSE2(void) {
   VP8LCollectColorRedTransforms = CollectColorRedTransforms_SSE2;
   VP8LAddVector = AddVector_SSE2;
   VP8LAddVectorEq = AddVectorEq_SSE2;
+  // TODO(https://crbug.com/webp/499): this function produces different results
+  // from the C code due to use of double/float resulting in output differences
+  // when compared to -noasm.
+#if !(defined(__i386__) || defined(_M_IX86))
   VP8LCombinedShannonEntropy = CombinedShannonEntropy_SSE2;
+#endif
   VP8LVectorMismatch = VectorMismatch_SSE2;
   VP8LBundleColorMap = BundleColorMap_SSE2;
 

src/dsp/lossless_sse2.c

@@ -191,8 +191,9 @@ static void PredictorAdd0_SSE2(const uint32_t* in, const uint32_t* upper,
     _mm_storeu_si128((__m128i*)&out[i], res);
   }
   if (i != num_pixels) {
-    VP8LPredictorsAdd_C[0](in + i, upper + i, num_pixels - i, out + i);
+    VP8LPredictorsAdd_C[0](in + i, NULL, num_pixels - i, out + i);
   }
+  (void)upper;
 }
 
 // Predictor1: left.

src/dsp/upsampling_msa.c

@@ -576,9 +576,9 @@ static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bot_y,        \
   const uint32_t l_uv = ((cur_u[0]) | ((cur_v[0]) << 16));               \
   const uint32_t uv0 = (3 * tl_uv + l_uv + 0x00020002u) >> 2;            \
   const uint8_t* ptop_y = &top_y[1];                                     \
-  uint8_t *ptop_dst = top_dst + XSTEP;                                   \
+  uint8_t* ptop_dst = top_dst + XSTEP;                                   \
   const uint8_t* pbot_y = &bot_y[1];                                     \
-  uint8_t *pbot_dst = bot_dst + XSTEP;                                   \
+  uint8_t* pbot_dst = bot_dst + XSTEP;                                   \
                                                                          \
   FUNC(top_y[0], uv0 & 0xff, (uv0 >> 16), top_dst);                      \
   if (bot_y != NULL) {                                                   \

src/dsp/upsampling_neon.c

@@ -58,8 +58,8 @@
 } while (0)
 
 // Turn the macro into a function for reducing code-size when non-critical
-static void Upsample16Pixels_NEON(const uint8_t *r1, const uint8_t *r2,
-                                  uint8_t *out) {
+static void Upsample16Pixels_NEON(const uint8_t* r1, const uint8_t* r2,
+                                  uint8_t* out) {
   UPSAMPLE_16PIXELS(r1, r2, out);
 }
 
@@ -190,14 +190,14 @@ static const int16_t kCoeffs1[4] = { 19077, 26149, 6419, 13320 };
 }
 
 #define NEON_UPSAMPLE_FUNC(FUNC_NAME, FMT, XSTEP)                       \
-static void FUNC_NAME(const uint8_t *top_y, const uint8_t *bottom_y,    \
-                      const uint8_t *top_u, const uint8_t *top_v,       \
-                      const uint8_t *cur_u, const uint8_t *cur_v,       \
-                      uint8_t *top_dst, uint8_t *bottom_dst, int len) { \
+static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,    \
+                      const uint8_t* top_u, const uint8_t* top_v,       \
+                      const uint8_t* cur_u, const uint8_t* cur_v,       \
+                      uint8_t* top_dst, uint8_t* bottom_dst, int len) { \
   int block;                                                            \
   /* 16 byte aligned array to cache reconstructed u and v */            \
   uint8_t uv_buf[2 * 32 + 15];                                          \
-  uint8_t *const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);     \
+  uint8_t* const r_uv = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);     \
   const int uv_len = (len + 1) >> 1;                                    \
   /* 9 pixels must be read-able for each block */                       \
   const int num_blocks = (uv_len - 1) >> 3;                             \

src/enc/alpha_enc.c

@@ -13,7 +13,6 @@
 
 #include <assert.h>
 #include <stdlib.h>
-#include <string.h>
 
 #include "src/enc/vp8i_enc.h"
 #include "src/dsp/dsp.h"
@@ -149,7 +148,6 @@ static int EncodeAlphaInternal(const uint8_t* const data, int width, int height,
       }
     } else {
       VP8LBitWriterWipeOut(&tmp_bw);
-      memset(&result->bw, 0, sizeof(result->bw));
       return 0;
     }
   }
@@ -164,7 +162,7 @@ static int EncodeAlphaInternal(const uint8_t* const data, int width, int height,
   header = method | (filter << 2);
   if (reduce_levels) header |= ALPHA_PREPROCESSED_LEVELS << 4;
 
-  if (!VP8BitWriterInit(&result->bw, ALPHA_HEADER_LEN + output_size)) ok = 0;
+  VP8BitWriterInit(&result->bw, ALPHA_HEADER_LEN + output_size);
   ok = ok && VP8BitWriterAppend(&result->bw, &header, ALPHA_HEADER_LEN);
   ok = ok && VP8BitWriterAppend(&result->bw, output, output_size);
 

src/enc/analysis_enc.c

@@ -126,16 +126,6 @@ static void InitHistogram(VP8Histogram* const histo) {
   histo->last_non_zero = 1;
 }
 
-static void MergeHistograms(const VP8Histogram* const in,
-                            VP8Histogram* const out) {
-  if (in->max_value > out->max_value) {
-    out->max_value = in->max_value;
-  }
-  if (in->last_non_zero > out->last_non_zero) {
-    out->last_non_zero = in->last_non_zero;
-  }
-}
-
 //------------------------------------------------------------------------------
 // Simplified k-Means, to assign Nb segments based on alpha-histogram
 
@@ -285,49 +275,6 @@ static int FastMBAnalyze(VP8EncIterator* const it) {
   return 0;
 }
 
-static int MBAnalyzeBestIntra4Mode(VP8EncIterator* const it,
-                                   int best_alpha) {
-  uint8_t modes[16];
-  const int max_mode = MAX_INTRA4_MODE;
-  int i4_alpha;
-  VP8Histogram total_histo;
-  int cur_histo = 0;
-  InitHistogram(&total_histo);
-
-  VP8IteratorStartI4(it);
-  do {
-    int mode;
-    int best_mode_alpha = DEFAULT_ALPHA;
-    VP8Histogram histos[2];
-    const uint8_t* const src = it->yuv_in_ + Y_OFF_ENC + VP8Scan[it->i4_];
-
-    VP8MakeIntra4Preds(it);
-    for (mode = 0; mode < max_mode; ++mode) {
-      int alpha;
-
-      InitHistogram(&histos[cur_histo]);
-      VP8CollectHistogram(src, it->yuv_p_ + VP8I4ModeOffsets[mode],
-                          0, 1, &histos[cur_histo]);
-      alpha = GetAlpha(&histos[cur_histo]);
-      if (IS_BETTER_ALPHA(alpha, best_mode_alpha)) {
-        best_mode_alpha = alpha;
-        modes[it->i4_] = mode;
-        cur_histo ^= 1;   // keep track of best histo so far.
-      }
-    }
-    // accumulate best histogram
-    MergeHistograms(&histos[cur_histo ^ 1], &total_histo);
-    // Note: we reuse the original samples for predictors
-  } while (VP8IteratorRotateI4(it, it->yuv_in_ + Y_OFF_ENC));
-
-  i4_alpha = GetAlpha(&total_histo);
-  if (IS_BETTER_ALPHA(i4_alpha, best_alpha)) {
-    VP8SetIntra4Mode(it, modes);
-    best_alpha = i4_alpha;
-  }
-  return best_alpha;
-}
-
 static int MBAnalyzeBestUVMode(VP8EncIterator* const it) {
   int best_alpha = DEFAULT_ALPHA;
   int smallest_alpha = 0;
@@ -371,13 +318,6 @@ static void MBAnalyze(VP8EncIterator* const it,
     best_alpha = FastMBAnalyze(it);
   } else {
     best_alpha = MBAnalyzeBestIntra16Mode(it);
-    if (enc->method_ >= 5) {
-      // We go and make a fast decision for intra4/intra16.
-      // It's usually not a good and definitive pick, but helps seeding the
-      // stats about level bit-cost.
-      // TODO(skal): improve criterion.
-      best_alpha = MBAnalyzeBestIntra4Mode(it, best_alpha);
-    }
   }
   best_uv_alpha = MBAnalyzeBestUVMode(it);
 

src/enc/backward_references_enc.c

@@ -11,13 +11,14 @@
 //
 
 #include <assert.h>
+#include <float.h>
 #include <math.h>
 
-#include "src/enc/backward_references_enc.h"
-#include "src/enc/histogram_enc.h"
+#include "src/dsp/dsp.h"
 #include "src/dsp/lossless.h"
 #include "src/dsp/lossless_common.h"
-#include "src/dsp/dsp.h"
+#include "src/enc/backward_references_enc.h"
+#include "src/enc/histogram_enc.h"
 #include "src/utils/color_cache_utils.h"
 #include "src/utils/utils.h"
 
@@ -103,6 +104,20 @@ void VP8LBackwardRefsClear(VP8LBackwardRefs* const refs) {
   }
 }
 
+// Swaps the content of two VP8LBackwardRefs.
+static void BackwardRefsSwap(VP8LBackwardRefs* const refs1,
+                             VP8LBackwardRefs* const refs2) {
+  const int point_to_refs1 =
+      (refs1->tail_ != NULL && refs1->tail_ == &refs1->refs_);
+  const int point_to_refs2 =
+      (refs2->tail_ != NULL && refs2->tail_ == &refs2->refs_);
+  const VP8LBackwardRefs tmp = *refs1;
+  *refs1 = *refs2;
+  *refs2 = tmp;
+  if (point_to_refs2) refs1->tail_ = &refs1->refs_;
+  if (point_to_refs1) refs2->tail_ = &refs2->refs_;
+}
+
 void VP8LBackwardRefsInit(VP8LBackwardRefs* const refs, int block_size) {
   assert(refs != NULL);
   memset(refs, 0, sizeof(*refs));
@@ -154,6 +169,22 @@ static PixOrCopyBlock* BackwardRefsNewBlock(VP8LBackwardRefs* const refs) {
   return b;
 }
 
+// Return 1 on success, 0 on error.
+static int BackwardRefsClone(const VP8LBackwardRefs* const from,
+                             VP8LBackwardRefs* const to) {
+  const PixOrCopyBlock* block_from = from->refs_;
+  VP8LClearBackwardRefs(to);
+  while (block_from != NULL) {
+    PixOrCopyBlock* const block_to = BackwardRefsNewBlock(to);
+    if (block_to == NULL) return 0;
+    memcpy(block_to->start_, block_from->start_,
+           block_from->size_ * sizeof(PixOrCopy));
+    block_to->size_ = block_from->size_;
+    block_from = block_from->next_;
+  }
+  return 1;
+}
+
 extern void VP8LBackwardRefsCursorAdd(VP8LBackwardRefs* const refs,
                                       const PixOrCopy v);
 void VP8LBackwardRefsCursorAdd(VP8LBackwardRefs* const refs,
@@ -753,12 +784,18 @@ static int CalculateBestCacheSize(const uint32_t* argb, int quality,
         }
       }
     } else {
+      int code, extra_bits, extra_bits_value;
       // We should compute the contribution of the (distance,length)
       // histograms but those are the same independently from the cache size.
       // As those constant contributions are in the end added to the other
-      // histogram contributions, we can safely ignore them.
+      // histogram contributions, we can ignore them, except for the length
+      // prefix that is part of the literal_ histogram.
       int len = PixOrCopyLength(v);
       uint32_t argb_prev = *argb ^ 0xffffffffu;
+      VP8LPrefixEncode(len, &code, &extra_bits, &extra_bits_value);
+      for (i = 0; i <= cache_bits_max; ++i) {
+        ++histos[i]->literal_[NUM_LITERAL_CODES + code];
+      }
       // Update the color caches.
       do {
         if (*argb != argb_prev) {
@@ -842,16 +879,21 @@ extern int VP8LBackwardReferencesTraceBackwards(
     int xsize, int ysize, const uint32_t* const argb, int cache_bits,
     const VP8LHashChain* const hash_chain,
     const VP8LBackwardRefs* const refs_src, VP8LBackwardRefs* const refs_dst);
-static VP8LBackwardRefs* GetBackwardReferences(
-    int width, int height, const uint32_t* const argb, int quality,
-    int lz77_types_to_try, int* const cache_bits,
-    const VP8LHashChain* const hash_chain, VP8LBackwardRefs* best,
-    VP8LBackwardRefs* worst) {
-  const int cache_bits_initial = *cache_bits;
-  double bit_cost_best = -1;
+static int GetBackwardReferences(int width, int height,
+                                 const uint32_t* const argb, int quality,
+                                 int lz77_types_to_try, int cache_bits_max,
+                                 int do_no_cache,
+                                 const VP8LHashChain* const hash_chain,
+                                 VP8LBackwardRefs* const refs,
+                                 int* const cache_bits_best) {
   VP8LHistogram* histo = NULL;
-  int lz77_type, lz77_type_best = 0;
+  int i, lz77_type;
+  // Index 0 is for a color cache, index 1 for no cache (if needed).
+  int lz77_types_best[2] = {0, 0};
+  double bit_costs_best[2] = {DBL_MAX, DBL_MAX};
   VP8LHashChain hash_chain_box;
+  VP8LBackwardRefs* const refs_tmp = &refs[do_no_cache ? 2 : 1];
+  int status = 0;
   memset(&hash_chain_box, 0, sizeof(hash_chain_box));
 
   histo = VP8LAllocateHistogram(MAX_COLOR_CACHE_BITS);
@@ -860,87 +902,130 @@ static VP8LBackwardRefs* GetBackwardReferences(
   for (lz77_type = 1; lz77_types_to_try;
        lz77_types_to_try &= ~lz77_type, lz77_type <<= 1) {
     int res = 0;
-    double bit_cost;
-    int cache_bits_tmp = cache_bits_initial;
+    double bit_cost = 0.;
     if ((lz77_types_to_try & lz77_type) == 0) continue;
     switch (lz77_type) {
       case kLZ77RLE:
-        res = BackwardReferencesRle(width, height, argb, 0, worst);
+        res = BackwardReferencesRle(width, height, argb, 0, refs_tmp);
         break;
       case kLZ77Standard:
         // Compute LZ77 with no cache (0 bits), as the ideal LZ77 with a color
         // cache is not that different in practice.
-        res = BackwardReferencesLz77(width, height, argb, 0, hash_chain, worst);
+        res = BackwardReferencesLz77(width, height, argb, 0, hash_chain,
+                                     refs_tmp);
         break;
       case kLZ77Box:
         if (!VP8LHashChainInit(&hash_chain_box, width * height)) goto Error;
         res = BackwardReferencesLz77Box(width, height, argb, 0, hash_chain,
-                                        &hash_chain_box, worst);
+                                        &hash_chain_box, refs_tmp);
         break;
       default:
         assert(0);
     }
     if (!res) goto Error;
 
-    // Next, try with a color cache and update the references.
-    if (!CalculateBestCacheSize(argb, quality, worst, &cache_bits_tmp)) {
-      goto Error;
-    }
-    if (cache_bits_tmp > 0) {
-      if (!BackwardRefsWithLocalCache(argb, cache_bits_tmp, worst)) {
-        goto Error;
+    // Start with the no color cache case.
+    for (i = 1; i >= 0; --i) {
+      int cache_bits = (i == 1) ? 0 : cache_bits_max;
+
+      if (i == 1 && !do_no_cache) continue;
+
+      if (i == 0) {
+        // Try with a color cache.
+        if (!CalculateBestCacheSize(argb, quality, refs_tmp, &cache_bits)) {
+          goto Error;
+        }
+        if (cache_bits > 0) {
+          if (!BackwardRefsWithLocalCache(argb, cache_bits, refs_tmp)) {
+            goto Error;
+          }
+        }
+      }
+
+      if (i == 0 && do_no_cache && cache_bits == 0) {
+        // No need to re-compute bit_cost as it was computed at i == 1.
+      } else {
+        VP8LHistogramCreate(histo, refs_tmp, cache_bits);
+        bit_cost = VP8LHistogramEstimateBits(histo);
+      }
+
+      if (bit_cost < bit_costs_best[i]) {
+        if (i == 1) {
+          // Do not swap as the full cache analysis would have the wrong
+          // VP8LBackwardRefs to start with.
+          if (!BackwardRefsClone(refs_tmp, &refs[1])) goto Error;
+        } else {
+          BackwardRefsSwap(refs_tmp, &refs[0]);
+        }
+        bit_costs_best[i] = bit_cost;
+        lz77_types_best[i] = lz77_type;
+        if (i == 0) *cache_bits_best = cache_bits;
       }
     }
-
-    // Keep the best backward references.
-    VP8LHistogramCreate(histo, worst, cache_bits_tmp);
-    bit_cost = VP8LHistogramEstimateBits(histo);
-    if (lz77_type_best == 0 || bit_cost < bit_cost_best) {
-      VP8LBackwardRefs* const tmp = worst;
-      worst = best;
-      best = tmp;
-      bit_cost_best = bit_cost;
-      *cache_bits = cache_bits_tmp;
-      lz77_type_best = lz77_type;
-    }
   }
-  assert(lz77_type_best > 0);
+  assert(lz77_types_best[0] > 0);
+  assert(!do_no_cache || lz77_types_best[1] > 0);
 
   // Improve on simple LZ77 but only for high quality (TraceBackwards is
   // costly).
-  if ((lz77_type_best == kLZ77Standard || lz77_type_best == kLZ77Box) &&
-      quality >= 25) {
-    const VP8LHashChain* const hash_chain_tmp =
-        (lz77_type_best == kLZ77Standard) ? hash_chain : &hash_chain_box;
-    double bit_cost_trace;
-    if (!VP8LBackwardReferencesTraceBackwards(width, height, argb, *cache_bits,
-                                              hash_chain_tmp, best, worst)) {
-      goto Error;
+  for (i = 1; i >= 0; --i) {
+    if (i == 1 && !do_no_cache) continue;
+    if ((lz77_types_best[i] == kLZ77Standard ||
+         lz77_types_best[i] == kLZ77Box) &&
+        quality >= 25) {
+      const VP8LHashChain* const hash_chain_tmp =
+          (lz77_types_best[i] == kLZ77Standard) ? hash_chain : &hash_chain_box;
+      const int cache_bits = (i == 1) ? 0 : *cache_bits_best;
+      double bit_cost_trace;
+      if (!VP8LBackwardReferencesTraceBackwards(width, height, argb, cache_bits,
+                                                hash_chain_tmp, &refs[i],
+                                                refs_tmp)) {
+        goto Error;
+      }
+      VP8LHistogramCreate(histo, refs_tmp, cache_bits);
+      bit_cost_trace = VP8LHistogramEstimateBits(histo);
+      if (bit_cost_trace < bit_costs_best[i]) {
+        BackwardRefsSwap(refs_tmp, &refs[i]);
+      }
     }
-    VP8LHistogramCreate(histo, worst, *cache_bits);
-    bit_cost_trace = VP8LHistogramEstimateBits(histo);
-    if (bit_cost_trace < bit_cost_best) best = worst;
-  }
 
-  BackwardReferences2DLocality(width, best);
+    BackwardReferences2DLocality(width, &refs[i]);
+
+    if (i == 1 && lz77_types_best[0] == lz77_types_best[1] &&
+        *cache_bits_best == 0) {
+      // If the best cache size is 0 and we have the same best LZ77, just copy
+      // the data over and stop here.
+      if (!BackwardRefsClone(&refs[1], &refs[0])) goto Error;
+      break;
+    }
+  }
+  status = 1;
 
 Error:
   VP8LHashChainClear(&hash_chain_box);
   VP8LFreeHistogram(histo);
-  return best;
+  return status;
 }
 
-VP8LBackwardRefs* VP8LGetBackwardReferences(
+WebPEncodingError VP8LGetBackwardReferences(
     int width, int height, const uint32_t* const argb, int quality,
-    int low_effort, int lz77_types_to_try, int* const cache_bits,
-    const VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs_tmp1,
-    VP8LBackwardRefs* const refs_tmp2) {
+    int low_effort, int lz77_types_to_try, int cache_bits_max, int do_no_cache,
+    const VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs,
+    int* const cache_bits_best) {
   if (low_effort) {
-    return GetBackwardReferencesLowEffort(width, height, argb, cache_bits,
-                                          hash_chain, refs_tmp1);
+    VP8LBackwardRefs* refs_best;
+    *cache_bits_best = cache_bits_max;
+    refs_best = GetBackwardReferencesLowEffort(
+        width, height, argb, cache_bits_best, hash_chain, refs);
+    if (refs_best == NULL) return VP8_ENC_ERROR_OUT_OF_MEMORY;
+    // Set it in first position.
+    BackwardRefsSwap(refs_best, &refs[0]);
   } else {
-    return GetBackwardReferences(width, height, argb, quality,
-                                 lz77_types_to_try, cache_bits, hash_chain,
-                                 refs_tmp1, refs_tmp2);
+    if (!GetBackwardReferences(width, height, argb, quality, lz77_types_to_try,
+                               cache_bits_max, do_no_cache, hash_chain, refs,
+                               cache_bits_best)) {
+      return VP8_ENC_ERROR_OUT_OF_MEMORY;
+    }
   }
+  return VP8_ENC_OK;
 }

src/enc/backward_references_enc.h

@@ -16,6 +16,7 @@
 #include <assert.h>
 #include <stdlib.h>
 #include "src/webp/types.h"
+#include "src/webp/encode.h"
 #include "src/webp/format_constants.h"
 
 #ifdef __cplusplus
@@ -218,14 +219,19 @@ enum VP8LLZ77Type {
 // Evaluates best possible backward references for specified quality.
 // The input cache_bits to 'VP8LGetBackwardReferences' sets the maximum cache
 // bits to use (passing 0 implies disabling the local color cache).
-// The optimal cache bits is evaluated and set for the *cache_bits parameter.
-// The return value is the pointer to the best of the two backward refs viz,
-// refs[0] or refs[1].
-VP8LBackwardRefs* VP8LGetBackwardReferences(
+// The optimal cache bits is evaluated and set for the *cache_bits_best
+// parameter with the matching refs_best.
+// If do_no_cache == 0, refs is an array of 2 values and the best
+// VP8LBackwardRefs is put in the first element.
+// If do_no_cache != 0, refs is an array of 3 values and the best
+// VP8LBackwardRefs is put in the first element, the best value with no-cache in
+// the second element.
+// In both cases, the last element is used as temporary internally.
+WebPEncodingError VP8LGetBackwardReferences(
     int width, int height, const uint32_t* const argb, int quality,
-    int low_effort, int lz77_types_to_try, int* const cache_bits,
-    const VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs_tmp1,
-    VP8LBackwardRefs* const refs_tmp2);
+    int low_effort, int lz77_types_to_try, int cache_bits_max, int do_no_cache,
+    const VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs,
+    int* const cache_bits_best);
 
 #ifdef __cplusplus
 }

src/enc/config_enc.c

@@ -39,6 +39,8 @@ int WebPConfigInitInternal(WebPConfig* config,
   config->partitions = 0;
   config->segments = 4;
   config->pass = 1;
+  config->qmin = 0;
+  config->qmax = 100;
   config->show_compressed = 0;
   config->preprocessing = 0;
   config->autofilter = 0;
@@ -106,6 +108,9 @@ int WebPValidateConfig(const WebPConfig* config) {
   if (config->filter_type < 0 || config->filter_type > 1) return 0;
   if (config->autofilter < 0 || config->autofilter > 1) return 0;
   if (config->pass < 1 || config->pass > 10) return 0;
+  if (config->qmin < 0 || config->qmax > 100 || config->qmin > config->qmax) {
+    return 0;
+  }
   if (config->show_compressed < 0 || config->show_compressed > 1) return 0;
   if (config->preprocessing < 0 || config->preprocessing > 7) return 0;
   if (config->partitions < 0 || config->partitions > 3) return 0;

src/enc/frame_enc.c

@@ -31,10 +31,15 @@
 // we allow 2k of extra head-room in PARTITION0 limit.
 #define PARTITION0_SIZE_LIMIT ((VP8_MAX_PARTITION0_SIZE - 2048ULL) << 11)
 
+static float Clamp(float v, float min, float max) {
+  return (v < min) ? min : (v > max) ? max : v;
+}
+
 typedef struct {  // struct for organizing convergence in either size or PSNR
   int is_first;
   float dq;
   float q, last_q;
+  float qmin, qmax;
   double value, last_value;   // PSNR or size
   double target;
   int do_size_search;
@@ -47,7 +52,9 @@ static int InitPassStats(const VP8Encoder* const enc, PassStats* const s) {
 
   s->is_first = 1;
   s->dq = 10.f;
-  s->q = s->last_q = enc->config_->quality;
+  s->qmin = 1.f * enc->config_->qmin;
+  s->qmax = 1.f * enc->config_->qmax;
+  s->q = s->last_q = Clamp(enc->config_->quality, s->qmin, s->qmax);
   s->target = do_size_search ? (double)target_size
             : (target_PSNR > 0.) ? target_PSNR
             : 40.;   // default, just in case
@@ -56,10 +63,6 @@ static int InitPassStats(const VP8Encoder* const enc, PassStats* const s) {
   return do_size_search;
 }
 
-static float Clamp(float v, float min, float max) {
-  return (v < min) ? min : (v > max) ? max : v;
-}
-
 static float ComputeNextQ(PassStats* const s) {
   float dq;
   if (s->is_first) {
@@ -75,7 +78,7 @@ static float ComputeNextQ(PassStats* const s) {
   s->dq = Clamp(dq, -30.f, 30.f);
   s->last_q = s->q;
   s->last_value = s->value;
-  s->q = Clamp(s->q + s->dq, 0.f, 100.f);
+  s->q = Clamp(s->q + s->dq, s->qmin, s->qmax);
   return s->q;
 }
 
@@ -848,9 +851,10 @@ int VP8EncTokenLoop(VP8Encoder* const enc) {
     }
 
 #if (DEBUG_SEARCH > 0)
-    printf("#%2d metric:%.1lf -> %.1lf   last_q=%.2lf q=%.2lf dq=%.2lf\n",
+    printf("#%2d metric:%.1lf -> %.1lf   last_q=%.2lf q=%.2lf dq=%.2lf "
+           " range:[%.1f, %.1f]\n",
            num_pass_left, stats.last_value, stats.value,
-           stats.last_q, stats.q, stats.dq);
+           stats.last_q, stats.q, stats.dq, stats.qmin, stats.qmax);
 #endif
     if (enc->max_i4_header_bits_ > 0 && size_p0 > PARTITION0_SIZE_LIMIT) {
       ++num_pass_left;

src/enc/histogram_enc.c

@@ -208,6 +208,7 @@ void VP8LHistogramAddSinglePixOrCopy(VP8LHistogram* const histo,
   } else if (PixOrCopyIsCacheIdx(v)) {
     const int literal_ix =
         NUM_LITERAL_CODES + NUM_LENGTH_CODES + PixOrCopyCacheIdx(v);
+    assert(histo->palette_code_bits_ != 0);
     ++histo->literal_[literal_ix];
   } else {
     int code, extra_bits;
@@ -641,7 +642,7 @@ static void HistogramAnalyzeEntropyBin(VP8LHistogramSet* const image_histo,
 // Merges some histograms with same bin_id together if it's advantageous.
 // Sets the remaining histograms to NULL.
 static void HistogramCombineEntropyBin(VP8LHistogramSet* const image_histo,
-                                       int *num_used,
+                                       int* num_used,
                                        const uint16_t* const clusters,
                                        uint16_t* const cluster_mappings,
                                        VP8LHistogram* cur_combo,

src/enc/picture_csp_enc.c

@@ -29,11 +29,15 @@
 #define USE_INVERSE_ALPHA_TABLE
 
 #ifdef WORDS_BIGENDIAN
-#define ALPHA_OFFSET 0   // uint32_t 0xff000000 is 0xff,00,00,00 in memory
+// uint32_t 0xff000000 is 0xff,00,00,00 in memory
+#define CHANNEL_OFFSET(i) (i)
 #else
-#define ALPHA_OFFSET 3   // uint32_t 0xff000000 is 0x00,00,00,ff in memory
+// uint32_t 0xff000000 is 0x00,00,00,ff in memory
+#define CHANNEL_OFFSET(i) (3-(i))
 #endif
 
+#define ALPHA_OFFSET CHANNEL_OFFSET(0)
+
 //------------------------------------------------------------------------------
 // Detection of non-trivial transparency
 
@@ -57,16 +61,14 @@ static int CheckNonOpaque(const uint8_t* alpha, int width, int height,
 // Checking for the presence of non-opaque alpha.
 int WebPPictureHasTransparency(const WebPPicture* picture) {
   if (picture == NULL) return 0;
-  if (!picture->use_argb) {
-    return CheckNonOpaque(picture->a, picture->width, picture->height,
-                          1, picture->a_stride);
-  } else {
+  if (picture->use_argb) {
     const int alpha_offset = ALPHA_OFFSET;
     return CheckNonOpaque((const uint8_t*)picture->argb + alpha_offset,
                           picture->width, picture->height,
                           4, picture->argb_stride * sizeof(*picture->argb));
   }
-  return 0;
+  return CheckNonOpaque(picture->a, picture->width, picture->height,
+                        1, picture->a_stride);
 }
 
 //------------------------------------------------------------------------------
@@ -86,8 +88,9 @@ int WebPPictureHasTransparency(const WebPPicture* picture) {
 static int kLinearToGammaTab[kGammaTabSize + 1];
 static uint16_t kGammaToLinearTab[256];
 static volatile int kGammaTablesOk = 0;
+static void InitGammaTables(void);
 
-static WEBP_TSAN_IGNORE_FUNCTION void InitGammaTables(void) {
+WEBP_DSP_INIT_FUNC(InitGammaTables) {
   if (!kGammaTablesOk) {
     int v;
     const double scale = (double)(1 << kGammaTabFix) / kGammaScale;
@@ -177,8 +180,9 @@ static uint32_t kLinearToGammaTabS[kGammaTabSize + 2];
 #define GAMMA_TO_LINEAR_BITS 14
 static uint32_t kGammaToLinearTabS[MAX_Y_T + 1];   // size scales with Y_FIX
 static volatile int kGammaTablesSOk = 0;
+static void InitGammaTablesS(void);
 
-static WEBP_TSAN_IGNORE_FUNCTION void InitGammaTablesS(void) {
+WEBP_DSP_INIT_FUNC(InitGammaTablesS) {
   assert(2 * GAMMA_TO_LINEAR_BITS < 32);  // we use uint32_t intermediate values
   if (!kGammaTablesSOk) {
     int v;
@@ -997,10 +1001,10 @@ static int PictureARGBToYUVA(WebPPicture* picture, WebPEncCSP colorspace,
     return WebPEncodingSetError(picture, VP8_ENC_ERROR_INVALID_CONFIGURATION);
   } else {
     const uint8_t* const argb = (const uint8_t*)picture->argb;
-    const uint8_t* const a = argb + (0 ^ ALPHA_OFFSET);
-    const uint8_t* const r = argb + (1 ^ ALPHA_OFFSET);
-    const uint8_t* const g = argb + (2 ^ ALPHA_OFFSET);
-    const uint8_t* const b = argb + (3 ^ ALPHA_OFFSET);
+    const uint8_t* const a = argb + CHANNEL_OFFSET(0);
+    const uint8_t* const r = argb + CHANNEL_OFFSET(1);
+    const uint8_t* const g = argb + CHANNEL_OFFSET(2);
+    const uint8_t* const b = argb + CHANNEL_OFFSET(3);
 
     picture->colorspace = WEBP_YUV420;
     return ImportYUVAFromRGBA(r, g, b, a, 4, 4 * picture->argb_stride,
@@ -1050,7 +1054,7 @@ int WebPPictureYUVAToARGB(WebPPicture* picture) {
     const int height = picture->height;
     const int argb_stride = 4 * picture->argb_stride;
     uint8_t* dst = (uint8_t*)picture->argb;
-    const uint8_t *cur_u = picture->u, *cur_v = picture->v, *cur_y = picture->y;
+    const uint8_t* cur_u = picture->u, *cur_v = picture->v, *cur_y = picture->y;
     WebPUpsampleLinePairFunc upsample =
         WebPGetLinePairConverter(ALPHA_OFFSET > 0);
 

src/enc/picture_tools_enc.c

@@ -83,6 +83,19 @@ static int SmoothenBlock(const uint8_t* a_ptr, int a_stride, uint8_t* y_ptr,
   return (count == 0);
 }
 
+void WebPReplaceTransparentPixels(WebPPicture* const pic, uint32_t color) {
+  if (pic != NULL && pic->use_argb) {
+    int y = pic->height;
+    uint32_t* argb = pic->argb;
+    color &= 0xffffffu;   // force alpha=0
+    WebPInitAlphaProcessing();
+    while (y-- > 0) {
+      WebPAlphaReplace(argb, pic->width, color);
+      argb += pic->argb_stride;
+    }
+  }
+}
+
 void WebPCleanupTransparentArea(WebPPicture* pic) {
   int x, y, w, h;
   if (pic == NULL) return;
@@ -165,24 +178,6 @@ void WebPCleanupTransparentArea(WebPPicture* pic) {
 #undef SIZE
 #undef SIZE2
 
-void WebPCleanupTransparentAreaLossless(WebPPicture* const pic) {
-  int x, y, w, h;
-  uint32_t* argb;
-  assert(pic != NULL && pic->use_argb);
-  w = pic->width;
-  h = pic->height;
-  argb = pic->argb;
-
-  for (y = 0; y < h; ++y) {
-    for (x = 0; x < w; ++x) {
-      if ((argb[x] & 0xff000000) == 0) {
-        argb[x] = 0x00000000;
-      }
-    }
-    argb += pic->argb_stride;
-  }
-}
-
 //------------------------------------------------------------------------------
 // Blend color and remove transparency info
 

src/enc/vp8i_enc.h

@@ -31,8 +31,8 @@ extern "C" {
 
 // version numbers
 #define ENC_MAJ_VERSION 1
-#define ENC_MIN_VERSION 0
-#define ENC_REV_VERSION 3
+#define ENC_MIN_VERSION 2
+#define ENC_REV_VERSION 0
 
 enum { MAX_LF_LEVELS = 64,       // Maximum loop filter level
        MAX_VARIABLE_LEVEL = 67,  // last (inclusive) level with variable cost
@@ -249,7 +249,7 @@ typedef struct {
   int           percent0_;         // saved initial progress percent
 
   DError        left_derr_;        // left error diffusion (u/v)
-  DError       *top_derr_;         // top diffusion error - NULL if disabled
+  DError*       top_derr_;         // top diffusion error - NULL if disabled
 
   uint8_t* y_left_;    // left luma samples (addressable from index -1 to 15).
   uint8_t* u_left_;    // left u samples (addressable from index -1 to 7)
@@ -505,9 +505,9 @@ int WebPPictureAllocARGB(WebPPicture* const picture, int width, int height);
 // Returns false in case of error (invalid param, out-of-memory).
 int WebPPictureAllocYUVA(WebPPicture* const picture, int width, int height);
 
-// Clean-up the RGB samples under fully transparent area, to help lossless
-// compressibility (no guarantee, though). Assumes that pic->use_argb is true.
-void WebPCleanupTransparentAreaLossless(WebPPicture* const pic);
+// Replace samples that are fully transparent by 'color' to help compressibility
+// (no guarantee, though). Assumes pic->use_argb is true.
+void WebPReplaceTransparentPixels(WebPPicture* const pic, uint32_t color);
 
 //------------------------------------------------------------------------------
 

src/enc/vp8l_enc.c

@@ -144,7 +144,8 @@ typedef enum {
   kSubGreen = 2,
   kSpatialSubGreen = 3,
   kPalette = 4,
-  kNumEntropyIx = 5
+  kPaletteAndSpatial = 5,
+  kNumEntropyIx = 6
 } EntropyIx;
 
 typedef enum {
@@ -354,11 +355,15 @@ static int GetTransformBits(int method, int histo_bits) {
 }
 
 // Set of parameters to be used in each iteration of the cruncher.
-#define CRUNCH_CONFIGS_LZ77_MAX 2
+#define CRUNCH_SUBCONFIGS_MAX 2
+typedef struct {
+  int lz77_;
+  int do_no_cache_;
+} CrunchSubConfig;
 typedef struct {
   int entropy_idx_;
-  int lz77s_types_to_try_[CRUNCH_CONFIGS_LZ77_MAX];
-  int lz77s_types_to_try_size_;
+  CrunchSubConfig sub_configs_[CRUNCH_SUBCONFIGS_MAX];
+  int sub_configs_size_;
 } CrunchConfig;
 
 #define CRUNCH_CONFIGS_MAX kNumEntropyIx
@@ -376,6 +381,9 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
   int i;
   int use_palette;
   int n_lz77s;
+  // If set to 0, analyze the cache with the computed cache value. If 1, also
+  // analyze with no-cache.
+  int do_no_cache = 0;
   assert(pic != NULL && pic->argb != NULL);
 
   use_palette =
@@ -402,10 +410,13 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
       return 0;
     }
     if (method == 6 && config->quality == 100) {
+      do_no_cache = 1;
       // Go brute force on all transforms.
       *crunch_configs_size = 0;
       for (i = 0; i < kNumEntropyIx; ++i) {
-        if (i != kPalette || use_palette) {
+        // We can only apply kPalette or kPaletteAndSpatial if we can indeed use
+        // a palette.
+        if ((i != kPalette && i != kPaletteAndSpatial) || use_palette) {
           assert(*crunch_configs_size < CRUNCH_CONFIGS_MAX);
           crunch_configs[(*crunch_configs_size)++].entropy_idx_ = i;
         }
@@ -414,17 +425,28 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
       // Only choose the guessed best transform.
       *crunch_configs_size = 1;
       crunch_configs[0].entropy_idx_ = min_entropy_ix;
+      if (config->quality >= 75 && method == 5) {
+        // Test with and without color cache.
+        do_no_cache = 1;
+        // If we have a palette, also check in combination with spatial.
+        if (min_entropy_ix == kPalette) {
+          *crunch_configs_size = 2;
+          crunch_configs[1].entropy_idx_ = kPaletteAndSpatial;
+        }
+      }
     }
   }
   // Fill in the different LZ77s.
-  assert(n_lz77s <= CRUNCH_CONFIGS_LZ77_MAX);
+  assert(n_lz77s <= CRUNCH_SUBCONFIGS_MAX);
   for (i = 0; i < *crunch_configs_size; ++i) {
     int j;
     for (j = 0; j < n_lz77s; ++j) {
-      crunch_configs[i].lz77s_types_to_try_[j] =
+      assert(j < CRUNCH_SUBCONFIGS_MAX);
+      crunch_configs[i].sub_configs_[j].lz77_ =
           (j == 0) ? kLZ77Standard | kLZ77RLE : kLZ77Box;
+      crunch_configs[i].sub_configs_[j].do_no_cache_ = do_no_cache;
     }
-    crunch_configs[i].lz77s_types_to_try_size_ = n_lz77s;
+    crunch_configs[i].sub_configs_size_ = n_lz77s;
   }
   return 1;
 }
@@ -440,7 +462,7 @@ static int EncoderInit(VP8LEncoder* const enc) {
   int i;
   if (!VP8LHashChainInit(&enc->hash_chain_, pix_cnt)) return 0;
 
-  for (i = 0; i < 3; ++i) VP8LBackwardRefsInit(&enc->refs_[i], refs_block_size);
+  for (i = 0; i < 4; ++i) VP8LBackwardRefsInit(&enc->refs_[i], refs_block_size);
 
   return 1;
 }
@@ -769,13 +791,10 @@ static WebPEncodingError StoreImageToBitMask(
 }
 
 // Special case of EncodeImageInternal() for cache-bits=0, histo_bits=31
-static WebPEncodingError EncodeImageNoHuffman(VP8LBitWriter* const bw,
-                                              const uint32_t* const argb,
-                                              VP8LHashChain* const hash_chain,
-                                              VP8LBackwardRefs* const refs_tmp1,
-                                              VP8LBackwardRefs* const refs_tmp2,
-                                              int width, int height,
-                                              int quality, int low_effort) {
+static WebPEncodingError EncodeImageNoHuffman(
+    VP8LBitWriter* const bw, const uint32_t* const argb,
+    VP8LHashChain* const hash_chain, VP8LBackwardRefs* const refs_array,
+    int width, int height, int quality, int low_effort) {
   int i;
   int max_tokens = 0;
   WebPEncodingError err = VP8_ENC_OK;
@@ -798,13 +817,11 @@ static WebPEncodingError EncodeImageNoHuffman(VP8LBitWriter* const bw,
     err = VP8_ENC_ERROR_OUT_OF_MEMORY;
     goto Error;
   }
-  refs = VP8LGetBackwardReferences(width, height, argb, quality, 0,
-                                   kLZ77Standard | kLZ77RLE, &cache_bits,
-                                   hash_chain, refs_tmp1, refs_tmp2);
-  if (refs == NULL) {
-    err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-    goto Error;
-  }
+  err = VP8LGetBackwardReferences(
+      width, height, argb, quality, /*low_effort=*/0, kLZ77Standard | kLZ77RLE,
+      cache_bits, /*do_no_cache=*/0, hash_chain, refs_array, &cache_bits);
+  if (err != VP8_ENC_OK) goto Error;
+  refs = &refs_array[0];
   histogram_image = VP8LAllocateHistogramSet(1, cache_bits);
   if (histogram_image == NULL) {
     err = VP8_ENC_ERROR_OUT_OF_MEMORY;
@@ -860,11 +877,11 @@ static WebPEncodingError EncodeImageNoHuffman(VP8LBitWriter* const bw,
 
 static WebPEncodingError EncodeImageInternal(
     VP8LBitWriter* const bw, const uint32_t* const argb,
-    VP8LHashChain* const hash_chain, VP8LBackwardRefs refs_array[3], int width,
+    VP8LHashChain* const hash_chain, VP8LBackwardRefs refs_array[4], int width,
     int height, int quality, int low_effort, int use_cache,
     const CrunchConfig* const config, int* cache_bits, int histogram_bits,
     size_t init_byte_position, int* const hdr_size, int* const data_size) {
-  WebPEncodingError err = VP8_ENC_OK;
+  WebPEncodingError err = VP8_ENC_ERROR_OUT_OF_MEMORY;
   const uint32_t histogram_image_xysize =
       VP8LSubSampleSize(width, histogram_bits) *
       VP8LSubSampleSize(height, histogram_bits);
@@ -876,103 +893,103 @@ static WebPEncodingError EncodeImageInternal(
       3ULL * CODE_LENGTH_CODES, sizeof(*huff_tree));
   HuffmanTreeToken* tokens = NULL;
   HuffmanTreeCode* huffman_codes = NULL;
-  VP8LBackwardRefs* refs_best;
-  VP8LBackwardRefs* refs_tmp;
   uint16_t* const histogram_symbols =
       (uint16_t*)WebPSafeMalloc(histogram_image_xysize,
                                 sizeof(*histogram_symbols));
-  int lz77s_idx;
+  int sub_configs_idx;
+  int cache_bits_init, write_histogram_image;
   VP8LBitWriter bw_init = *bw, bw_best;
   int hdr_size_tmp;
+  VP8LHashChain hash_chain_histogram;  // histogram image hash chain
+  size_t bw_size_best = ~(size_t)0;
   assert(histogram_bits >= MIN_HUFFMAN_BITS);
   assert(histogram_bits <= MAX_HUFFMAN_BITS);
   assert(hdr_size != NULL);
   assert(data_size != NULL);
 
-  if (histogram_symbols == NULL) {
-    err = VP8_ENC_ERROR_OUT_OF_MEMORY;
+  // Make sure we can allocate the different objects.
+  memset(&hash_chain_histogram, 0, sizeof(hash_chain_histogram));
+  if (huff_tree == NULL || histogram_symbols == NULL ||
+      !VP8LHashChainInit(&hash_chain_histogram, histogram_image_xysize) ||
+      !VP8LHashChainFill(hash_chain, quality, argb, width, height,
+                         low_effort)) {
     goto Error;
   }
-
   if (use_cache) {
     // If the value is different from zero, it has been set during the
     // palette analysis.
-    if (*cache_bits == 0) *cache_bits = MAX_COLOR_CACHE_BITS;
+    cache_bits_init = (*cache_bits == 0) ? MAX_COLOR_CACHE_BITS : *cache_bits;
   } else {
-    *cache_bits = 0;
+    cache_bits_init = 0;
   }
-  // 'best_refs' is the reference to the best backward refs and points to one
-  // of refs_array[0] or refs_array[1].
-  // Calculate backward references from ARGB image.
-  if (huff_tree == NULL ||
-      !VP8LHashChainFill(hash_chain, quality, argb, width, height,
-                         low_effort) ||
-      !VP8LBitWriterInit(&bw_best, 0) ||
-      (config->lz77s_types_to_try_size_ > 1 &&
+  // If several iterations will happen, clone into bw_best.
+  if (!VP8LBitWriterInit(&bw_best, 0) ||
+      ((config->sub_configs_size_ > 1 ||
+        config->sub_configs_[0].do_no_cache_) &&
        !VP8LBitWriterClone(bw, &bw_best))) {
-    err = VP8_ENC_ERROR_OUT_OF_MEMORY;
     goto Error;
   }
-  for (lz77s_idx = 0; lz77s_idx < config->lz77s_types_to_try_size_;
-       ++lz77s_idx) {
-    refs_best = VP8LGetBackwardReferences(
-        width, height, argb, quality, low_effort,
-        config->lz77s_types_to_try_[lz77s_idx], cache_bits, hash_chain,
-        &refs_array[0], &refs_array[1]);
-    if (refs_best == NULL) {
-      err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-      goto Error;
-    }
-    // Keep the best references aside and use the other element from the first
-    // two as a temporary for later usage.
-    refs_tmp = &refs_array[refs_best == &refs_array[0] ? 1 : 0];
+  for (sub_configs_idx = 0; sub_configs_idx < config->sub_configs_size_;
+       ++sub_configs_idx) {
+    const CrunchSubConfig* const sub_config =
+        &config->sub_configs_[sub_configs_idx];
+    int cache_bits_best, i_cache;
+    err = VP8LGetBackwardReferences(width, height, argb, quality, low_effort,
+                                    sub_config->lz77_, cache_bits_init,
+                                    sub_config->do_no_cache_, hash_chain,
+                                    &refs_array[0], &cache_bits_best);
+    if (err != VP8_ENC_OK) goto Error;
 
-    histogram_image =
-        VP8LAllocateHistogramSet(histogram_image_xysize, *cache_bits);
-    tmp_histo = VP8LAllocateHistogram(*cache_bits);
-    if (histogram_image == NULL || tmp_histo == NULL) {
-      err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-      goto Error;
-    }
+    for (i_cache = 0; i_cache < (sub_config->do_no_cache_ ? 2 : 1); ++i_cache) {
+      const int cache_bits_tmp = (i_cache == 0) ? cache_bits_best : 0;
+      // Speed-up: no need to study the no-cache case if it was already studied
+      // in i_cache == 0.
+      if (i_cache == 1 && cache_bits_best == 0) break;
 
-    // Build histogram image and symbols from backward references.
-    if (!VP8LGetHistoImageSymbols(width, height, refs_best, quality, low_effort,
-                                  histogram_bits, *cache_bits, histogram_image,
-                                  tmp_histo, histogram_symbols)) {
-      err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-      goto Error;
-    }
-    // Create Huffman bit lengths and codes for each histogram image.
-    histogram_image_size = histogram_image->size;
-    bit_array_size = 5 * histogram_image_size;
-    huffman_codes = (HuffmanTreeCode*)WebPSafeCalloc(bit_array_size,
-                                                     sizeof(*huffman_codes));
-    // Note: some histogram_image entries may point to tmp_histos[], so the
-    // latter need to outlive the following call to GetHuffBitLengthsAndCodes().
-    if (huffman_codes == NULL ||
-        !GetHuffBitLengthsAndCodes(histogram_image, huffman_codes)) {
-      err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-      goto Error;
-    }
-    // Free combined histograms.
-    VP8LFreeHistogramSet(histogram_image);
-    histogram_image = NULL;
+      // Reset the bit writer for this iteration.
+      VP8LBitWriterReset(&bw_init, bw);
 
-    // Free scratch histograms.
-    VP8LFreeHistogram(tmp_histo);
-    tmp_histo = NULL;
+      // Build histogram image and symbols from backward references.
+      histogram_image =
+          VP8LAllocateHistogramSet(histogram_image_xysize, cache_bits_tmp);
+      tmp_histo = VP8LAllocateHistogram(cache_bits_tmp);
+      if (histogram_image == NULL || tmp_histo == NULL ||
+          !VP8LGetHistoImageSymbols(width, height, &refs_array[i_cache],
+                                    quality, low_effort, histogram_bits,
+                                    cache_bits_tmp, histogram_image, tmp_histo,
+                                    histogram_symbols)) {
+        goto Error;
+      }
+      // Create Huffman bit lengths and codes for each histogram image.
+      histogram_image_size = histogram_image->size;
+      bit_array_size = 5 * histogram_image_size;
+      huffman_codes = (HuffmanTreeCode*)WebPSafeCalloc(bit_array_size,
+                                                       sizeof(*huffman_codes));
+      // Note: some histogram_image entries may point to tmp_histos[], so the
+      // latter need to outlive the following call to
+      // GetHuffBitLengthsAndCodes().
+      if (huffman_codes == NULL ||
+          !GetHuffBitLengthsAndCodes(histogram_image, huffman_codes)) {
+        goto Error;
+      }
+      // Free combined histograms.
+      VP8LFreeHistogramSet(histogram_image);
+      histogram_image = NULL;
 
-    // Color Cache parameters.
-    if (*cache_bits > 0) {
-      VP8LPutBits(bw, 1, 1);
-      VP8LPutBits(bw, *cache_bits, 4);
-    } else {
-      VP8LPutBits(bw, 0, 1);
-    }
+      // Free scratch histograms.
+      VP8LFreeHistogram(tmp_histo);
+      tmp_histo = NULL;
 
-    // Huffman image + meta huffman.
-    {
-      const int write_histogram_image = (histogram_image_size > 1);
+      // Color Cache parameters.
+      if (cache_bits_tmp > 0) {
+        VP8LPutBits(bw, 1, 1);
+        VP8LPutBits(bw, cache_bits_tmp, 4);
+      } else {
+        VP8LPutBits(bw, 0, 1);
+      }
+
+      // Huffman image + meta huffman.
+      write_histogram_image = (histogram_image_size > 1);
       VP8LPutBits(bw, write_histogram_image, 1);
       if (write_histogram_image) {
         uint32_t* const histogram_argb =
@@ -980,10 +997,7 @@ static WebPEncodingError EncodeImageInternal(
                                       sizeof(*histogram_argb));
         int max_index = 0;
         uint32_t i;
-        if (histogram_argb == NULL) {
-          err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-          goto Error;
-        }
+        if (histogram_argb == NULL) goto Error;
         for (i = 0; i < histogram_image_xysize; ++i) {
           const int symbol_index = histogram_symbols[i] & 0xffff;
           histogram_argb[i] = (symbol_index << 8);
@@ -995,65 +1009,64 @@ static WebPEncodingError EncodeImageInternal(
 
         VP8LPutBits(bw, histogram_bits - 2, 3);
         err = EncodeImageNoHuffman(
-            bw, histogram_argb, hash_chain, refs_tmp, &refs_array[2],
+            bw, histogram_argb, &hash_chain_histogram, &refs_array[2],
             VP8LSubSampleSize(width, histogram_bits),
             VP8LSubSampleSize(height, histogram_bits), quality, low_effort);
         WebPSafeFree(histogram_argb);
         if (err != VP8_ENC_OK) goto Error;
       }
-    }
 
-    // Store Huffman codes.
-    {
-      int i;
-      int max_tokens = 0;
-      // Find maximum number of symbols for the huffman tree-set.
-      for (i = 0; i < 5 * histogram_image_size; ++i) {
-        HuffmanTreeCode* const codes = &huffman_codes[i];
-        if (max_tokens < codes->num_symbols) {
-          max_tokens = codes->num_symbols;
+      // Store Huffman codes.
+      {
+        int i;
+        int max_tokens = 0;
+        // Find maximum number of symbols for the huffman tree-set.
+        for (i = 0; i < 5 * histogram_image_size; ++i) {
+          HuffmanTreeCode* const codes = &huffman_codes[i];
+          if (max_tokens < codes->num_symbols) {
+            max_tokens = codes->num_symbols;
+          }
+        }
+        tokens = (HuffmanTreeToken*)WebPSafeMalloc(max_tokens, sizeof(*tokens));
+        if (tokens == NULL) goto Error;
+        for (i = 0; i < 5 * histogram_image_size; ++i) {
+          HuffmanTreeCode* const codes = &huffman_codes[i];
+          StoreHuffmanCode(bw, huff_tree, tokens, codes);
+          ClearHuffmanTreeIfOnlyOneSymbol(codes);
         }
       }
-      tokens = (HuffmanTreeToken*)WebPSafeMalloc(max_tokens, sizeof(*tokens));
-      if (tokens == NULL) {
-        err = VP8_ENC_ERROR_OUT_OF_MEMORY;
-        goto Error;
+      // Store actual literals.
+      hdr_size_tmp = (int)(VP8LBitWriterNumBytes(bw) - init_byte_position);
+      err = StoreImageToBitMask(bw, width, histogram_bits, &refs_array[i_cache],
+                                histogram_symbols, huffman_codes);
+      if (err != VP8_ENC_OK) goto Error;
+      // Keep track of the smallest image so far.
+      if (VP8LBitWriterNumBytes(bw) < bw_size_best) {
+        bw_size_best = VP8LBitWriterNumBytes(bw);
+        *cache_bits = cache_bits_tmp;
+        *hdr_size = hdr_size_tmp;
+        *data_size =
+            (int)(VP8LBitWriterNumBytes(bw) - init_byte_position - *hdr_size);
+        VP8LBitWriterSwap(bw, &bw_best);
       }
-      for (i = 0; i < 5 * histogram_image_size; ++i) {
-        HuffmanTreeCode* const codes = &huffman_codes[i];
-        StoreHuffmanCode(bw, huff_tree, tokens, codes);
-        ClearHuffmanTreeIfOnlyOneSymbol(codes);
+      WebPSafeFree(tokens);
+      tokens = NULL;
+      if (huffman_codes != NULL) {
+        WebPSafeFree(huffman_codes->codes);
+        WebPSafeFree(huffman_codes);
+        huffman_codes = NULL;
       }
     }
-    // Store actual literals.
-    hdr_size_tmp = (int)(VP8LBitWriterNumBytes(bw) - init_byte_position);
-    err = StoreImageToBitMask(bw, width, histogram_bits, refs_best,
-                              histogram_symbols, huffman_codes);
-    // Keep track of the smallest image so far.
-    if (lz77s_idx == 0 ||
-        VP8LBitWriterNumBytes(bw) < VP8LBitWriterNumBytes(&bw_best)) {
-      *hdr_size = hdr_size_tmp;
-      *data_size =
-          (int)(VP8LBitWriterNumBytes(bw) - init_byte_position - *hdr_size);
-      VP8LBitWriterSwap(bw, &bw_best);
-    }
-    // Reset the bit writer for the following iteration if any.
-    if (config->lz77s_types_to_try_size_ > 1) VP8LBitWriterReset(&bw_init, bw);
-    WebPSafeFree(tokens);
-    tokens = NULL;
-    if (huffman_codes != NULL) {
-      WebPSafeFree(huffman_codes->codes);
-      WebPSafeFree(huffman_codes);
-      huffman_codes = NULL;
-    }
   }
   VP8LBitWriterSwap(bw, &bw_best);
+  err = VP8_ENC_OK;
 
  Error:
   WebPSafeFree(tokens);
   WebPSafeFree(huff_tree);
   VP8LFreeHistogramSet(histogram_image);
   VP8LFreeHistogram(tmp_histo);
+  VP8LHashChainClear(&hash_chain_histogram);
   if (huffman_codes != NULL) {
     WebPSafeFree(huffman_codes->codes);
     WebPSafeFree(huffman_codes);
@@ -1095,8 +1108,7 @@ static WebPEncodingError ApplyPredictFilter(const VP8LEncoder* const enc,
   VP8LPutBits(bw, pred_bits - 2, 3);
   return EncodeImageNoHuffman(
       bw, enc->transform_data_, (VP8LHashChain*)&enc->hash_chain_,
-      (VP8LBackwardRefs*)&enc->refs_[0],  // cast const away
-      (VP8LBackwardRefs*)&enc->refs_[1], transform_width, transform_height,
+      (VP8LBackwardRefs*)&enc->refs_[0], transform_width, transform_height,
       quality, low_effort);
 }
 
@@ -1116,8 +1128,7 @@ static WebPEncodingError ApplyCrossColorFilter(const VP8LEncoder* const enc,
   VP8LPutBits(bw, ccolor_transform_bits - 2, 3);
   return EncodeImageNoHuffman(
       bw, enc->transform_data_, (VP8LHashChain*)&enc->hash_chain_,
-      (VP8LBackwardRefs*)&enc->refs_[0],  // cast const away
-      (VP8LBackwardRefs*)&enc->refs_[1], transform_width, transform_height,
+      (VP8LBackwardRefs*)&enc->refs_[0], transform_width, transform_height,
       quality, low_effort);
 }
 
@@ -1464,8 +1475,8 @@ static WebPEncodingError EncodePalette(VP8LBitWriter* const bw, int low_effort,
   }
   tmp_palette[0] = palette[0];
   return EncodeImageNoHuffman(bw, tmp_palette, &enc->hash_chain_,
-                              &enc->refs_[0], &enc->refs_[1], palette_size, 1,
-                              20 /* quality */, low_effort);
+                              &enc->refs_[0], palette_size, 1, /*quality=*/20,
+                              low_effort);
 }
 
 // -----------------------------------------------------------------------------
@@ -1491,7 +1502,7 @@ static void VP8LEncoderDelete(VP8LEncoder* enc) {
   if (enc != NULL) {
     int i;
     VP8LHashChainClear(&enc->hash_chain_);
-    for (i = 0; i < 3; ++i) VP8LBackwardRefsClear(&enc->refs_[i]);
+    for (i = 0; i < 4; ++i) VP8LBackwardRefsClear(&enc->refs_[i]);
     ClearTransformBuffer(enc);
     WebPSafeFree(enc);
   }
@@ -1541,7 +1552,7 @@ static int EncodeStreamHook(void* input, void* data2) {
   int data_size = 0;
   int use_delta_palette = 0;
   int idx;
-  size_t best_size = 0;
+  size_t best_size = ~(size_t)0;
   VP8LBitWriter bw_init = *bw, bw_best;
   (void)data2;
 
@@ -1553,11 +1564,13 @@ static int EncodeStreamHook(void* input, void* data2) {
 
   for (idx = 0; idx < num_crunch_configs; ++idx) {
     const int entropy_idx = crunch_configs[idx].entropy_idx_;
-    enc->use_palette_ = (entropy_idx == kPalette);
+    enc->use_palette_ =
+        (entropy_idx == kPalette) || (entropy_idx == kPaletteAndSpatial);
     enc->use_subtract_green_ =
         (entropy_idx == kSubGreen) || (entropy_idx == kSpatialSubGreen);
-    enc->use_predict_ =
-        (entropy_idx == kSpatial) || (entropy_idx == kSpatialSubGreen);
+    enc->use_predict_ = (entropy_idx == kSpatial) ||
+                        (entropy_idx == kSpatialSubGreen) ||
+                        (entropy_idx == kPaletteAndSpatial);
     if (low_effort) {
       enc->use_cross_color_ = 0;
     } else {
@@ -1640,7 +1653,7 @@ static int EncodeStreamHook(void* input, void* data2) {
     if (err != VP8_ENC_OK) goto Error;
 
     // If we are better than what we already have.
-    if (idx == 0 || VP8LBitWriterNumBytes(bw) < best_size) {
+    if (VP8LBitWriterNumBytes(bw) < best_size) {
       best_size = VP8LBitWriterNumBytes(bw);
       // Store the BitWriter.
       VP8LBitWriterSwap(bw, &bw_best);
@@ -1821,7 +1834,7 @@ Error:
 }
 
 #undef CRUNCH_CONFIGS_MAX
-#undef CRUNCH_CONFIGS_LZ77_MAX
+#undef CRUNCH_SUBCONFIGS_MAX
 
 int VP8LEncodeImage(const WebPConfig* const config,
                     const WebPPicture* const picture) {

src/enc/vp8li_enc.h

@@ -71,7 +71,7 @@ typedef struct {
   uint32_t palette_[MAX_PALETTE_SIZE];
 
   // Some 'scratch' (potentially large) objects.
-  struct VP8LBackwardRefs refs_[3];  // Backward Refs array for temporaries.
+  struct VP8LBackwardRefs refs_[4];  // Backward Refs array for temporaries.
   VP8LHashChain hash_chain_;         // HashChain data for constructing
                                      // backward references.
 } VP8LEncoder;

src/enc/webp_enc.c

@@ -400,7 +400,7 @@ int WebPEncode(const WebPConfig* config, WebPPicture* pic) {
     }
 
     if (!config->exact) {
-      WebPCleanupTransparentAreaLossless(pic);
+      WebPReplaceTransparentPixels(pic, 0x000000);
     }
 
     ok = VP8LEncodeImage(config, pic);  // Sets pic->error in case of problem.

src/libwebp.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,0,3
- PRODUCTVERSION 1,0,0,3
+ FILEVERSION 1,0,2,0
+ PRODUCTVERSION 1,0,2,0
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebp DLL"
-            VALUE "FileVersion", "1.0.3"
+            VALUE "FileVersion", "1.2.0"
             VALUE "InternalName", "libwebp.dll"
             VALUE "LegalCopyright", "Copyright (C) 2019"
             VALUE "OriginalFilename", "libwebp.dll"
             VALUE "ProductName", "WebP Image Codec"
-            VALUE "ProductVersion", "1.0.3"
+            VALUE "ProductVersion", "1.2.0"
         END
     END
     BLOCK "VarFileInfo"

src/libwebpdecoder.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,0,3
- PRODUCTVERSION 1,0,0,3
+ FILEVERSION 1,0,2,0
+ PRODUCTVERSION 1,0,2,0
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpdecoder DLL"
-            VALUE "FileVersion", "1.0.3"
+            VALUE "FileVersion", "1.2.0"
             VALUE "InternalName", "libwebpdecoder.dll"
             VALUE "LegalCopyright", "Copyright (C) 2019"
             VALUE "OriginalFilename", "libwebpdecoder.dll"
             VALUE "ProductName", "WebP Image Decoder"
-            VALUE "ProductVersion", "1.0.3"
+            VALUE "ProductVersion", "1.2.0"
         END
     END
     BLOCK "VarFileInfo"

src/mux/Makefile.am

@@ -17,6 +17,6 @@ noinst_HEADERS =
 noinst_HEADERS += ../webp/format_constants.h
 
 libwebpmux_la_LIBADD = ../libwebp.la
-libwebpmux_la_LDFLAGS = -no-undefined -version-info 3:4:0 -lm
+libwebpmux_la_LDFLAGS = -no-undefined -version-info 3:6:0 -lm
 libwebpmuxincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebpmux.pc

src/mux/libwebpmux.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,0,3
- PRODUCTVERSION 1,0,0,3
+ FILEVERSION 1,0,2,0
+ PRODUCTVERSION 1,0,2,0
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpmux DLL"
-            VALUE "FileVersion", "1.0.3"
+            VALUE "FileVersion", "1.2.0"
             VALUE "InternalName", "libwebpmux.dll"
             VALUE "LegalCopyright", "Copyright (C) 2019"
             VALUE "OriginalFilename", "libwebpmux.dll"
             VALUE "ProductName", "WebP Image Muxer"
-            VALUE "ProductVersion", "1.0.3"
+            VALUE "ProductVersion", "1.2.0"
         END
     END
     BLOCK "VarFileInfo"

src/mux/muxi.h

@@ -28,8 +28,8 @@ extern "C" {
 // Defines and constants.
 
 #define MUX_MAJ_VERSION 1
-#define MUX_MIN_VERSION 0
-#define MUX_REV_VERSION 3
+#define MUX_MIN_VERSION 2
+#define MUX_REV_VERSION 0
 
 // Chunk object.
 typedef struct WebPChunk WebPChunk;

src/mux/muxread.c

@@ -100,7 +100,7 @@ static int MuxImageParse(const WebPChunk* const chunk, int copy_data,
                          WebPMuxImage* const wpi) {
   const uint8_t* bytes = chunk->data_.bytes;
   size_t size = chunk->data_.size;
-  const uint8_t* const last = bytes + size;
+  const uint8_t* const last = (bytes == NULL) ? NULL : bytes + size;
   WebPChunk subchunk;
   size_t subchunk_size;
   WebPChunk** unknown_chunk_list = &wpi->unknown_;
@@ -155,7 +155,6 @@ static int MuxImageParse(const WebPChunk* const chunk, int copy_data,
         break;
       default:
         goto Fail;
-        break;
     }
     subchunk_size = ChunkDiskSize(&subchunk);
     bytes += subchunk_size;
@@ -264,7 +263,6 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data,
         if (!MuxImageParse(&chunk, copy_data, wpi)) goto Err;
         ChunkRelease(&chunk);
         goto PushImage;
-        break;
       default:  // A non-image chunk.
         if (wpi->is_partial_) goto Err;  // Encountered a non-image chunk before
                                          // getting all chunks of an image.

src/utils/color_cache_utils.h

@@ -26,7 +26,7 @@ extern "C" {
 
 // Main color cache struct.
 typedef struct {
-  uint32_t *colors_;  // color entries
+  uint32_t* colors_;  // color entries
   int hash_shift_;    // Hash shift: 32 - hash_bits_.
   int hash_bits_;
 } VP8LColorCache;

src/utils/thread_utils.c

@@ -73,7 +73,7 @@ typedef struct {
 #endif
 
 static int pthread_create(pthread_t* const thread, const void* attr,
-                          unsigned int (__stdcall *start)(void*), void* arg) {
+                          unsigned int (__stdcall* start)(void*), void* arg) {
   (void)attr;
 #ifdef USE_CREATE_THREAD
   *thread = CreateThread(NULL,   /* lpThreadAttributes */

src/utils/utils.c

@@ -216,9 +216,14 @@ void WebPSafeFree(void* const ptr) {
   free(ptr);
 }
 
-// Public API function.
+// Public API functions.
+
+void* WebPMalloc(size_t size) {
+  return WebPSafeMalloc(1, size);
+}
+
 void WebPFree(void* ptr) {
-  free(ptr);
+  WebPSafeFree(ptr);
 }
 
 //------------------------------------------------------------------------------
@@ -226,7 +231,7 @@ void WebPFree(void* ptr) {
 void WebPCopyPlane(const uint8_t* src, int src_stride,
                    uint8_t* dst, int dst_stride, int width, int height) {
   assert(src != NULL && dst != NULL);
-  assert(src_stride >= width && dst_stride >= width);
+  assert(abs(src_stride) >= width && abs(dst_stride) >= width);
   while (height-- > 0) {
     memcpy(dst, src, width);
     src += src_stride;

src/webp/decode.h

@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-#define WEBP_DECODER_ABI_VERSION 0x0208    // MAJOR(8b) + MINOR(8b)
+#define WEBP_DECODER_ABI_VERSION 0x0209    // MAJOR(8b) + MINOR(8b)
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
 // the types are left here for reference.
@@ -91,9 +91,6 @@ WEBP_EXTERN uint8_t* WebPDecodeYUV(const uint8_t* data, size_t data_size,
                                    uint8_t** u, uint8_t** v,
                                    int* stride, int* uv_stride);
 
-// Releases memory returned by the WebPDecode*() functions above.
-WEBP_EXTERN void WebPFree(void* ptr);
-
 // These five functions are variants of the above ones, that decode the image
 // directly into a pre-allocated buffer 'output_buffer'. The maximum storage
 // available in this buffer is indicated by 'output_buffer_size'. If this
@@ -456,7 +453,7 @@ struct WebPDecoderOptions {
   int scaled_width, scaled_height;    // final resolution
   int use_threads;                    // if true, use multi-threaded decoding
   int dithering_strength;             // dithering strength (0=Off, 100=full)
-  int flip;                           // flip output vertically
+  int flip;                           // if true, flip output vertically
   int alpha_dithering_strength;       // alpha dithering strength in [0..100]
 
   uint32_t pad[5];                    // padding for later use

src/webp/encode.h

@@ -20,7 +20,7 @@
 extern "C" {
 #endif
 
-#define WEBP_ENCODER_ABI_VERSION 0x020e    // MAJOR(8b) + MINOR(8b)
+#define WEBP_ENCODER_ABI_VERSION 0x020f    // MAJOR(8b) + MINOR(8b)
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
 // the types are left here for reference.
@@ -79,9 +79,6 @@ WEBP_EXTERN size_t WebPEncodeLosslessBGRA(const uint8_t* bgra,
                                           int width, int height, int stride,
                                           uint8_t** output);
 
-// Releases memory returned by the WebPEncode*() functions above.
-WEBP_EXTERN void WebPFree(void* ptr);
-
 //------------------------------------------------------------------------------
 // Coding parameters
 
@@ -151,7 +148,8 @@ struct WebPConfig {
   int use_delta_palette;  // reserved for future lossless feature
   int use_sharp_yuv;      // if needed, use sharp (and slow) RGB->YUV conversion
 
-  uint32_t pad[2];        // padding for later use
+  int qmin;               // minimum permissible quality factor
+  int qmax;               // maximum permissible quality factor
 };
 
 // Enumerate some predefined settings for WebPConfig, depending on the type
@@ -294,6 +292,11 @@ typedef enum WebPEncodingError {
 #define WEBP_MAX_DIMENSION 16383
 
 // Main exchange structure (input samples, output bytes, statistics)
+//
+// Once WebPPictureInit() has been called, it's ok to make all the INPUT fields
+// (use_argb, y/u/v, argb, ...) point to user-owned data, even if
+// WebPPictureAlloc() has been called. Depending on the value use_argb,
+// it's guaranteed that either *argb or *y/*u/*v content will be kept untouched.
 struct WebPPicture {
   //   INPUT
   //////////////
@@ -306,7 +309,7 @@ struct WebPPicture {
   // YUV input (mostly used for input to lossy compression)
   WebPEncCSP colorspace;     // colorspace: should be YUV420 for now (=Y'CbCr).
   int width, height;         // dimensions (less or equal to WEBP_MAX_DIMENSION)
-  uint8_t *y, *u, *v;        // pointers to luma/chroma planes.
+  uint8_t* y, *u, *v;        // pointers to luma/chroma planes.
   int y_stride, uv_stride;   // luma/chroma strides.
   uint8_t* a;                // pointer to the alpha plane
   int a_stride;              // stride of the alpha plane
@@ -350,7 +353,7 @@ struct WebPPicture {
   uint32_t pad3[3];       // padding for later use
 
   // Unused for now
-  uint8_t *pad4, *pad5;
+  uint8_t* pad4, *pad5;
   uint32_t pad6[8];       // padding for later use
 
   // PRIVATE FIELDS

src/webp/mux.h

@@ -57,7 +57,7 @@ extern "C" {
   WebPMuxGetChunk(mux, "ICCP", &icc_profile);
   // ... (Consume icc_data).
   WebPMuxDelete(mux);
-  free(data);
+  WebPFree(data);
 */
 
 // Note: forward declaring enumerations is not allowed in (strict) C and C++,
@@ -245,7 +245,7 @@ WEBP_EXTERN WebPMuxError WebPMuxPushFrame(
     WebPMux* mux, const WebPMuxFrameInfo* frame, int copy_data);
 
 // Gets the nth frame from the mux object.
-// The content of 'frame->bitstream' is allocated using malloc(), and NOT
+// The content of 'frame->bitstream' is allocated using WebPMalloc(), and NOT
 // owned by the 'mux' object. It MUST be deallocated by the caller by calling
 // WebPDataClear().
 // nth=0 has a special meaning - last position.
@@ -376,10 +376,10 @@ WEBP_EXTERN WebPMuxError WebPMuxNumChunks(const WebPMux* mux,
 // Assembles all chunks in WebP RIFF format and returns in 'assembled_data'.
 // This function also validates the mux object.
 // Note: The content of 'assembled_data' will be ignored and overwritten.
-// Also, the content of 'assembled_data' is allocated using malloc(), and NOT
-// owned by the 'mux' object. It MUST be deallocated by the caller by calling
-// WebPDataClear(). It's always safe to call WebPDataClear() upon return,
-// even in case of error.
+// Also, the content of 'assembled_data' is allocated using WebPMalloc(), and
+// NOT owned by the 'mux' object. It MUST be deallocated by the caller by
+// calling WebPDataClear(). It's always safe to call WebPDataClear() upon
+// return, even in case of error.
 // Parameters:
 //   mux - (in/out) object whose chunks are to be assembled
 //   assembled_data - (out) assembled WebP data

src/webp/mux_types.h

@@ -14,7 +14,6 @@
 #ifndef WEBP_WEBP_MUX_TYPES_H_
 #define WEBP_WEBP_MUX_TYPES_H_
 
-#include <stdlib.h>  // free()
 #include <string.h>  // memset()
 #include "./types.h"
 
@@ -56,6 +55,7 @@ typedef enum WebPMuxAnimBlend {
 
 // Data type used to describe 'raw' data, e.g., chunk data
 // (ICC profile, metadata) and WebP compressed image data.
+// 'bytes' memory must be allocated using WebPMalloc() and such.
 struct WebPData {
   const uint8_t* bytes;
   size_t size;
@@ -68,11 +68,11 @@ static WEBP_INLINE void WebPDataInit(WebPData* webp_data) {
   }
 }
 
-// Clears the contents of the 'webp_data' object by calling free(). Does not
-// deallocate the object itself.
+// Clears the contents of the 'webp_data' object by calling WebPFree().
+// Does not deallocate the object itself.
 static WEBP_INLINE void WebPDataClear(WebPData* webp_data) {
   if (webp_data != NULL) {
-    free((void*)webp_data->bytes);
+    WebPFree((void*)webp_data->bytes);
     WebPDataInit(webp_data);
   }
 }
@@ -83,7 +83,7 @@ static WEBP_INLINE int WebPDataCopy(const WebPData* src, WebPData* dst) {
   if (src == NULL || dst == NULL) return 0;
   WebPDataInit(dst);
   if (src->bytes != NULL && src->size != 0) {
-    dst->bytes = (uint8_t*)malloc(src->size);
+    dst->bytes = (uint8_t*)WebPMalloc(src->size);
     if (dst->bytes == NULL) return 0;
     memcpy((void*)dst->bytes, src->bytes, src->size);
     dst->size = src->size;

src/webp/types.h

@@ -7,7 +7,7 @@
 // be found in the AUTHORS file in the root of the source tree.
 // -----------------------------------------------------------------------------
 //
-//  Common types
+//  Common types + memory wrappers
 //
 // Author: Skal (pascal.massimino@gmail.com)
 
@@ -49,4 +49,20 @@ typedef long long int int64_t;
 // Macro to check ABI compatibility (same major revision number)
 #define WEBP_ABI_IS_INCOMPATIBLE(a, b) (((a) >> 8) != ((b) >> 8))
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Allocates 'size' bytes of memory. Returns NULL upon error. Memory
+// must be deallocated by calling WebPFree(). This function is made available
+// by the core 'libwebp' library.
+WEBP_EXTERN void* WebPMalloc(size_t size);
+
+// Releases memory returned by the WebPDecode*() functions (from decode.h).
+WEBP_EXTERN void WebPFree(void* ptr);
+
+#ifdef __cplusplus
+}    // extern "C"
+#endif
+
 #endif  // WEBP_WEBP_TYPES_H_

swig/libwebp.py

@@ -1,14 +1,9 @@
-# Lint as: python2, python3
 # This file was automatically generated by SWIG (http://www.swig.org).
-# Version 3.0.10
+# Version 3.0.12
 #
 # Do not make changes to this file unless you know what you are doing--modify
 # the SWIG interface file instead.
 
-
-
-
-
 from sys import version_info as _swig_python_version_info
 if _swig_python_version_info >= (2, 7, 0):
     def swig_import_helper():
@@ -31,17 +26,18 @@ elif _swig_python_version_info >= (2, 6, 0):
         except ImportError:
             import _libwebp
             return _libwebp
-        if fp is not None:
-            try:
-                _mod = imp.load_module('_libwebp', fp, pathname, description)
-            finally:
+        try:
+            _mod = imp.load_module('_libwebp', fp, pathname, description)
+        finally:
+            if fp is not None:
                 fp.close()
-            return _mod
+        return _mod
     _libwebp = swig_import_helper()
     del swig_import_helper
 else:
     import _libwebp
 del _swig_python_version_info
+
 try:
     _swig_property = property
 except NameError:

swig/libwebp.swig

@@ -32,6 +32,9 @@
 
 #ifdef SWIGPYTHON
 %module(package="com.google.webp") libwebp
+%begin %{
+#define SWIG_PYTHON_STRICT_BYTE_CHAR
+%}
 #else
 %module libwebp
 #endif  /* SWIGPYTHON */