update ChangeLog

Change-Id: I12c2b718e99087201a57f7ab7cc4c16374e205f6

.mailmap

@@ -11,3 +11,4 @@ Vikas Arora <vikasa@google.com>
 Tamar Levy <tamar.levy@intel.com>
 <qrczak@google.com> <qrczak>
 Hui Su <huisu@google.com>
+James Zern <jzern@google.com>

AUTHORS

@@ -4,8 +4,11 @@ Contributors:
 - Charles Munger (clm at google dot com)
 - Cheng Yi (cyi at google dot com)
 - Christian Duvivier (cduvivier at google dot com)
+- Christopher Degawa (ccom at randomderp dot com)
+- Clement Courbet (courbet at google dot com)
 - Djordje Pesut (djordje dot pesut at imgtec dot com)
 - Hui Su (huisu at google dot com)
+- Ilya Kurdyukov (jpegqs at gmail dot com)
 - Ingvar Stepanyan (rreverser at google dot com)
 - James Zern (jzern at google dot com)
 - Jan Engelhardt (jengelh at medozas dot de)
@@ -45,3 +48,4 @@ Contributors:
 - Wan-Teh Chang (wtc at google dot com)
 - Yang Zhang (yang dot zhang at arm dot com)
 - Yannis Guyon (yguyon at google dot com)
+- Zhi An Ng (zhin at chromium dot org)

Android.mk

@@ -74,6 +74,7 @@ dsp_dec_srcs := \
     src/dsp/lossless_msa.c \
     src/dsp/lossless_neon.$(NEON) \
     src/dsp/lossless_sse2.c \
+    src/dsp/lossless_sse41.c \
     src/dsp/rescaler.c \
     src/dsp/rescaler_mips32.c \
     src/dsp/rescaler_mips_dsp_r2.c \

CMakeLists.txt

@@ -1,14 +1,17 @@
-##
-##  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.
+#  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)
 
+if(POLICY CMP0072)
+  cmake_policy(SET CMP0072 NEW)
+endif()
+
 project(WebP C)
 
 # Options for coder / decoder executables.
@@ -26,9 +29,11 @@ option(WEBP_BUILD_GIF2WEBP "Build the gif2webp conversion tool." ON)
 option(WEBP_BUILD_IMG2WEBP "Build the img2webp animation tool." ON)
 option(WEBP_BUILD_VWEBP "Build the vwebp viewer tool." ON)
 option(WEBP_BUILD_WEBPINFO "Build the webpinfo command line tool." ON)
+option(WEBP_BUILD_LIBWEBPMUX "Build the libwebpmux library." ON)
 option(WEBP_BUILD_WEBPMUX "Build the webpmux command line tool." ON)
 option(WEBP_BUILD_EXTRAS "Build extras." ON)
 option(WEBP_BUILD_WEBP_JS "Emscripten build of webp.js." OFF)
+option(WEBP_USE_THREAD "Enable threading support" ON)
 option(WEBP_NEAR_LOSSLESS "Enable near-lossless encoding" ON)
 option(WEBP_ENABLE_SWAP_16BIT_CSP "Enable byte swap for 16 bit colorspaces."
        OFF)
@@ -50,6 +55,7 @@ if(WEBP_BUILD_WEBP_JS)
   set(WEBP_BUILD_WEBPINFO OFF)
   set(WEBP_BUILD_WEBPMUX OFF)
   set(WEBP_BUILD_EXTRAS OFF)
+  set(WEBP_USE_THREAD OFF)
 
   if(WEBP_ENABLE_SIMD)
     message("wasm2js does not support SIMD, disabling webp.js generation.")
@@ -277,10 +283,12 @@ if(XCODE)
   libwebp_add_stub_file(webp)
 endif()
 target_link_libraries(webp ${WEBP_DEP_LIBRARIES})
-target_include_directories(webp
-                           PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
-                                   ${CMAKE_CURRENT_BINARY_DIR}
-                           PUBLIC $<INSTALL_INTERFACE:include>)
+target_include_directories(
+  webp
+  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+          ${CMAKE_CURRENT_BINARY_DIR}
+  PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>
+         $<INSTALL_INTERFACE:include>)
 set_target_properties(
   webp
   PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
@@ -315,10 +323,10 @@ ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
 configure_pkg_config("src/demux/libwebpdemux.pc")
 
 # Set the version numbers.
-function(parse_version FILE NAME VAR)
+macro(set_version FILE TARGET_NAME NAME_IN_MAKEFILE)
   file(READ ${CMAKE_CURRENT_SOURCE_DIR}/src/${FILE} SOURCE_FILE)
   string(REGEX MATCH
-               "${NAME}_la_LDFLAGS[^\n]* -version-info [0-9:]+"
+               "${NAME_IN_MAKEFILE}_la_LDFLAGS[^\n]* -version-info [0-9:]+"
                TMP
                ${SOURCE_FILE})
   string(REGEX MATCH
@@ -327,29 +335,35 @@ function(parse_version FILE NAME VAR)
                ${TMP})
   string(REGEX
          REPLACE ":"
-                 "."
-                 VERSION
+                 " "
+                 LT_VERSION
                  ${TMP})
-  set(${VAR} "${VERSION}" PARENT_SCOPE)
-endfunction()
-parse_version(Makefile.am webp WEBP_WEBP_SOVERSION)
-set_target_properties(webp
-                      PROPERTIES VERSION
-                                 ${PACKAGE_VERSION}
-                                 SOVERSION
-                                 ${WEBP_WEBP_SOVERSION})
-parse_version(Makefile.am webpdecoder WEBP_DECODER_SOVERSION)
-set_target_properties(webpdecoder
-                      PROPERTIES VERSION
-                                 ${PACKAGE_VERSION}
-                                 SOVERSION
-                                 ${WEBP_DECODER_SOVERSION})
-parse_version(demux/Makefile.am webpdemux WEBP_DEMUX_SOVERSION)
-set_target_properties(webpdemux
-                      PROPERTIES VERSION
-                                 ${PACKAGE_VERSION}
-                                 SOVERSION
-                                 ${WEBP_DEMUX_SOVERSION})
+
+  # See the libtool docs for more information:
+  # https://www.gnu.org/software/libtool/manual/libtool.html#Updating-version-info
+  #
+  # c=<current>, a=<age>, r=<revision>
+  #
+  # libtool generates a .so file as .so.[c-a].a.r, while -version-info c:r:a is
+  # passed to libtool.
+  #
+  # We set FULL = [c-a].a.r and MAJOR = [c-a].
+  separate_arguments(LT_VERSION)
+  list(GET LT_VERSION 0 LT_CURRENT)
+  list(GET LT_VERSION 1 LT_REVISION)
+  list(GET LT_VERSION 2 LT_AGE)
+  math(EXPR LT_CURRENT_MINUS_AGE "${LT_CURRENT} - ${LT_AGE}")
+
+  set_target_properties(
+    ${TARGET_NAME}
+    PROPERTIES VERSION
+               ${LT_CURRENT_MINUS_AGE}.${LT_AGE}.${LT_REVISION}
+               SOVERSION
+               ${LT_CURRENT_MINUS_AGE})
+endmacro()
+set_version(Makefile.am webp webp)
+set_version(Makefile.am webpdecoder webpdecoder)
+set_version(demux/Makefile.am webpdemux webpdemux)
 file(READ ${CMAKE_CURRENT_SOURCE_DIR}/configure.ac CONFIGURE_FILE)
 string(REGEX MATCH
              "AC_INIT\\([^\n]*\\[[0-9\\.]+\\]"
@@ -377,6 +391,12 @@ foreach(I_FILE RANGE ${WEBP_SIMD_FILES_TO_INCLUDE_RANGE})
                               ${SIMD_COMPILE_FLAG})
 endforeach()
 
+if(NOT WEBP_BUILD_LIBWEBPMUX)
+  set(WEBP_BUILD_GIF2WEBP OFF)
+  set(WEBP_BUILD_IMG2WEBP OFF)
+  set(WEBP_BUILD_WEBPMUX OFF)
+endif()
+
 if(WEBP_BUILD_GIF2WEBP AND NOT GIF_FOUND)
   set(WEBP_BUILD_GIF2WEBP OFF)
 endif()
@@ -449,19 +469,14 @@ if(WEBP_BUILD_CWEBP)
   install(TARGETS cwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
-if(WEBP_BUILD_GIF2WEBP OR WEBP_BUILD_IMG2WEBP)
+if(WEBP_BUILD_LIBWEBPMUX)
   parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/src/mux "WEBP_MUX_SRCS" "")
   add_library(libwebpmux ${WEBP_MUX_SRCS})
   target_link_libraries(libwebpmux webp)
   target_include_directories(libwebpmux
                              PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
                                      ${CMAKE_CURRENT_SOURCE_DIR})
-  parse_version(mux/Makefile.am webpmux WEBP_MUX_SOVERSION)
-  set_target_properties(libwebpmux
-                        PROPERTIES VERSION
-                                   ${PACKAGE_VERSION}
-                                   SOVERSION
-                                   ${WEBP_MUX_SOVERSION})
+  set_version(mux/Makefile.am libwebpmux webpmux)
   set_target_properties(libwebpmux
                         PROPERTIES PUBLIC_HEADER
                                    "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux.h;\
@@ -567,7 +582,6 @@ if(WEBP_BUILD_EXTRAS)
                              PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
                                      ${CMAKE_CURRENT_SOURCE_DIR}/src
                                      ${CMAKE_CURRENT_BINARY_DIR}/src)
-  install(TARGETS get_disto RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 
   # webp_quality
   add_executable(webp_quality ${WEBP_QUALITY_SRCS} ${WEBP_EXTRAS_SRCS})
@@ -575,7 +589,6 @@ if(WEBP_BUILD_EXTRAS)
   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
   find_package(SDL)
@@ -590,7 +603,6 @@ if(WEBP_BUILD_EXTRAS)
                                        ${SDL_INCLUDE_DIR})
     set(WEBP_HAVE_SDL 1)
     target_compile_definitions(vwebp_sdl PUBLIC WEBP_HAVE_SDL)
-    install(TARGETS vwebp_sdl RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
   endif()
 endif()
 
@@ -606,7 +618,7 @@ if(WEBP_BUILD_WEBP_JS)
       webp_js
       PROPERTIES LINK_FLAGS "-s WASM=0 \
          -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
-         -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
+         -s EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
     set_target_properties(webp_js PROPERTIES OUTPUT_NAME webp)
     target_compile_definitions(webp_js PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
   endif()
@@ -619,7 +631,7 @@ if(WEBP_BUILD_WEBP_JS)
     webp_wasm
     PROPERTIES LINK_FLAGS "-s WASM=1 \
        -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
-       -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
+       -s EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
   target_compile_definitions(webp_wasm PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
 
   target_compile_definitions(webpdspdecode PUBLIC EMSCRIPTEN)
@@ -637,7 +649,6 @@ if(WEBP_BUILD_ANIM_UTILS)
                         imageenc
                         imageioutil
                         webp
-                        libwebpmux
                         webpdemux
                         ${WEBP_DEP_GIF_LIBRARIES})
   target_include_directories(anim_diff PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
@@ -653,7 +664,6 @@ if(WEBP_BUILD_ANIM_UTILS)
                         imageenc
                         imageioutil
                         webp
-                        libwebpmux
                         webpdemux
                         ${WEBP_DEP_GIF_LIBRARIES})
   target_include_directories(anim_dump PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)

CONTRIBUTING.md

@@ -0,0 +1,29 @@
+# How to Contribute
+
+We'd love to accept your patches and contributions to this project. There are
+just a few small guidelines you need to follow.
+
+## Contributor License Agreement
+
+Contributions to this project must be accompanied by a Contributor License
+Agreement. You (or your employer) retain the copyright to your contribution;
+this simply gives us permission to use and redistribute your contributions as
+part of the project. Head over to <https://cla.developers.google.com/> to see
+your current agreements on file or to sign a new one.
+
+You generally only need to submit a CLA once, so if you've already submitted one
+(even if it was for a different project), you probably don't need to do it
+again.
+
+## Code reviews
+
+All submissions, including submissions by project members, require review. We
+use a [Gerrit](https://www.gerritcodereview.com) instance hosted at
+https://chromium-review.googlesource.com for this purpose. See the
+[WebM Project page](https://www.webmproject.org/code/contribute/submitting-patches/)
+for additional details.
+
+## Community Guidelines
+
+This project follows
+[Google's Open Source Community Guidelines](https://opensource.google.com/conduct/).

ChangeLog

@@ -1,10 +1,109 @@
+d9191588 fuzzer/*: normalize src/ includes
+53b6f762 fix indent
+731246ba update ChangeLog (tag: v1.2.1-rc2)
+d250f01d dsp/*: use WEBP_HAVE_* to determine Init availability
+3a4d3ecd update NEWS
+b2bc8093 bump version to 1.2.1
+e542fc7a update AUTHORS
+e0241154 Merge "libwebp/CMake: Add <BUILD_INTERFACE> to webp incl" into main
+edea6444 libwebp/CMake: Add <BUILD_INTERFACE> to webp incl
+ece18e55 dsp.h: respect --disable-sse2/sse4.1/neon
+a89a3230 wicdec: support alpha from WebP WIC decoder
+26f4aa01 Merge "alpha_processing: fix visual studio warnings" into main
+8f594663 alpha_processing: fix visual studio warnings
+46d844e6 Merge "cpu.cmake: fix compiler flag detection w/3.17.0+" into main
+298d26ea Merge changes I593adf92,If20675e7,Ifac68eac into main
+a1e5dae0 alpha_processing*: use WEBP_RESTRICT qualifier
+327ef24f cpu.cmake: fix compiler flag detection w/3.17.0+
+f70819de configure: enable libwebpmux by default
+dc7e2b42 configure: add informational notices when disabling binaries
+9df23ddd configure: move lib flag checks before binaries
+a2e18f10 Merge "WebPConfig.config.in: correct WEBP_INCLUDE_DIRS" into main
+e1a8d4f3 Merge "bit_reader_inl_utils: uniformly apply WEBP_RESTRICT" into main
+4de35f43 rescaler.c: fix alignment
+0f13eec7 bit_reader_inl_utils: uniformly apply WEBP_RESTRICT
+277d3074 Fix size_t overflow in  WebPRescalerInit
+97adbba5 WebPConfig.config.in: correct WEBP_INCLUDE_DIRS
+b60d4603 advanced_api_fuzzer: add extreme config value coverage
+72fe52f6 anim_encode.c,cosmetics: normalize indent
+116d235c anim_encode: Fix encoded_frames_[] overflow
+6f445b3e CMake: set CMP0072 to NEW
+b1cf887f define WEBP_RESTRICT for MSVC
+3e265136 Add WEBP_RESTRICT & use it in VP8BitReader
+f6d29247 vp8l_dec::ProcessRows: fix int overflow in multiply
+de3b4ba8 CMake: add WEBP_BUILD_LIBWEBPMUX
+7f09d3d1 CMakeLists.txt: rm libwebpmux dep from anim_{diff,dump}
+4edea4a6 Init{RGB,YUV}Rescaler: fix a few more int overflows
+c9e26bdb rescaler_utils: set max valid scaled w/h to INT_MAX/2
+28d488e6 utils.h: add SizeOverflow()
+695bdaa2 Export/EmitRescaledRowsRGBA: fix pointer offset int overflow
+685d073e Init{RGB,YUV}Rescaler: fix int overflows in multiplication
+d38bd0dd WebPFlipBuffer: fix integer overflow
+109ff0f1 utils: allow MALLOC_LIMIT to indicate a max
+a2fce867 WebPRescalerImportRowExpand_C: promote some vals before multiply
+776983d4 AllocateBuffer: fix int multiplication overflow check
+315abbd6 Merge "Revert "Do not use a palette for one color images.""
+eae815d0 Merge changes Ica3bbf75,I82f82954
+afbca5a1 Require Emscripten 2.0.18
+3320416b CMakeLists,emscripten: use EXPORTED_RUNTIME_METHODS
+29145ed6 Update README instructions for using Emscripten
+1f579139 cosmetics: remove use of 'sanity' / 'master'
+29b6129c WebPAnimEncoderNewInternal: remove some unnecessary inits
+b60869a1 Revert "Do not use a palette for one color images."
+6fb4cddc demux: move padded size calc post unpadded validation
+05b72d42 vp8l_enc.c: normalize index types
+b6513fba Do not use a palette for one color images.
+98bbe35b Fix multi-threading with palettes.
+b1674240 Add modified Zeng's method to palette sorting.
+88c90c45 add CONTRIBUTING.md
+6a9916d7 WebPRescalerInit: add missing int64_t promotion
+b6cf52d5 WebPIoInitFromOptions: treat use_scaling as a bool
+3b12b7f4 WebPIoInitFromOptions: treat use_cropping as a bool
+595fa13f add WebPCheckCropDimensions()
+8fdaecb0 Disable cross-color when palette is used.
+8933bac2 WebPIoInitFromOptions: respect incoming bypass_filtering val
+7d416ff0 webpdec,cosmetics: match error text to function call
+ec6cfeb5 Fix typo on WebPPictureAlloc() in README
+7e58a1a2 *.cmake: add license header
+5651a6b2 cmake: fix .so versioning
+25ae67b3 xcframeworkbuild.sh: add arm64 simulator target
+5d4ee4c3 cosmetics: remove use of the term 'dummy'
+01b38ee1 faster CollectColorXXXTransforms_SSE41
+652aa344 Merge "Use BitCtz for FastSLog2Slow_C"
+0320e1e3 add the missing default BitsCtz() code
+8886f620 Use BitCtz for FastSLog2Slow_C
+fae41617 faster CombinedShannonEntropy_SSE2
+5bd2704e Introduce the BitCtz() function.
+fee64287 Merge "wicdec,icc: treat unsupported op as non-fatal"
+33ddb894 lossless_sse{2,41}: remove some unneeded includes
+b27ea852 wicdec,icc: treat unsupported op as non-fatal
+b78494a9 Merge "Fix undefined signed shift."
+e79974cd Fix undefined signed shift.
+a8853394 SSE4.1 versions of BGRA to RGB/BGR color-space conversions
+a09a6472 SSE4.1 version of TransformColorInverse
+401da22b Merge "pngdec: check version before using png_get_chunk_malloc_max"
+26907822 pngdec: check version before using png_get_chunk_malloc_max
+06c1e72e Code cleanup
+8f0d41aa Merge changes Id135bbf4,I99e59797
+373eb170 gif2webp: don't store loop-count if there's only 1 frame
+759b9d5a cmake: add WEBP_USE_THREAD option
+926ce921 cmake: don't install binaries from extras/
+9c367bc6 WebPAnimDecoderNewInternal: validate bitstream before alloc
+47f64f6e filters_sse2: import Chromium change
+cc3577e9 fuzzer/*: use src/ based include paths
+004d77ff Merge tag 'v1.2.0'
+fedac6cc update ChangeLog (tag: v1.2.0-rc3, tag: v1.2.0)
 170a8712 Fix check_c_source_compiles with pthread.
+ceddb5fc Fix check_c_source_compiles with pthread.
 85995719 disable CombinedShannonEntropy_SSE2 on x86
+289757fe TiffDec: enforce stricter mem/dimension limit on tiles
 8af7436f Merge "{ios,xcframework}build.sh: make min version(s) more visible" into 1.2.0
 e56c3c5b pngdec: raise memory limit if needed
+8696147d 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
+8df77fb1 animdecoder_fuzzer: fix memory leak
 52ce6333 update NEWS
 28c49820 bump version to 1.2.0
 7363dff2 webp/encode.h: restore WEBP_ENCODER_ABI_VERSION to v1.1.0

Makefile.vc

@@ -215,6 +215,7 @@ DSP_DEC_OBJS = \
     $(DIROBJ)\dsp\lossless_msa.obj \
     $(DIROBJ)\dsp\lossless_neon.obj \
     $(DIROBJ)\dsp\lossless_sse2.obj \
+    $(DIROBJ)\dsp\lossless_sse41.obj \
     $(DIROBJ)\dsp\rescaler.obj \
     $(DIROBJ)\dsp\rescaler_mips32.obj \
     $(DIROBJ)\dsp\rescaler_mips_dsp_r2.obj \

NEWS

@@ -1,3 +1,15 @@
+- 7/20/2021: version 1.2.1
+  This is a binary compatible release.
+  * minor lossless encoder improvements and x86 color conversion speed up
+  * add ARM64 simulator support to xcframeworkbuild.sh (#510)
+  * further security related hardening in libwebp & examples
+    (issues: #497, #508, #518)
+    (chromium: #1196480, #1196773, #1196775, #1196777, #1196778, #1196850)
+    (oss-fuzz: #28658, #28978)
+  * toolchain updates and bug fixes (#498, #501, #502, #504, #505, #506, #509,
+                                     #533)
+  * use more inclusive language within the source (#507)
+
 - 12/23/2020: version 1.2.0
   * API changes:
     - libwebp:

README

@@ -4,7 +4,7 @@
           \__\__/\____/\_____/__/ ____  ___
                 / _/ /    \    \ /  _ \/ _/
                /  \_/   / /   \ \   __/  \__
-               \____/____/\_____/_____/____/v1.2.0
+               \____/____/\_____/_____/____/v1.2.1
 
 Description:
 ============
@@ -619,7 +619,7 @@ The encoding flow looks like:
   pic.width = width;
   pic.height = height;
   // allocated picture of dimension width x height
-  if (!WebPPictureAllocate(&pic)) {
+  if (!WebPPictureAlloc(&pic)) {
     return 0;   // memory error
   }
   // at this point, 'pic' has been initialized as a container,

README.mux

@@ -1,7 +1,7 @@
           __   __  ____  ____  ____  __ __  _     __ __
          /  \\/  \/  _ \/  _ \/  _ \/  \  \/ \___/_ / _\
          \       /   __/  _  \   __/      /  /  (_/  /__
-          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v1.2.0
+          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v1.2.1
 
 
 Description:

README.webp_js

@@ -10,26 +10,18 @@ This file describes the compilation of libwebp into a JavaScript decoder
 using Emscripten and CMake.
 
  - install the Emscripten SDK following the procedure described at:
-   https://kripken.github.io/emscripten-site/docs/getting_started/downloads.html
+   https://emscripten.org/docs/getting_started/downloads.html#installation-instructions-using-the-emsdk-recommended
    After installation, you should have some global variable positioned to the
-   location of the SDK. In particular, $EMSCRIPTEN should point to the
+   location of the SDK. In particular, $EMSDK should point to the
    top-level directory containing Emscripten tools.
 
- - make sure the file $EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake is
-   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/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 \
-       -DCMAKE_TOOLCHAIN_FILE=$EMSCRIPTEN/cmake/Modules/Platform/Emscripten.cmake \
+ emcmake cmake -DWEBP_BUILD_WEBP_JS=ON \
        ../
 
- - compile webp.js using 'make'.
+ - compile webp.js using 'emmake make'.
 
  - that's it! Upon completion, you should have the webp.js and
    webp.wasm files generated.
@@ -58,7 +50,7 @@ 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,
+  You will need a fairly recent version of Emscripten (at least 2.0.18,
   latest-upstream is recommended) and of your WASM-enabled browser to run this
   version.
 

build.gradle

@@ -141,6 +141,7 @@ model {
             include "lossless_msa.c"
             include "lossless_neon.$NEON"
             include "lossless_sse2.c"
+            include "lossless_sse41.c"
             include "rescaler.c"
             include "rescaler_mips32.c"
             include "rescaler_mips_dsp_r2.c"

cmake/WebPConfig.cmake.in

@@ -5,7 +5,7 @@ set(WEBP_VERSION ${WebP_VERSION})
 
 include ("${CMAKE_CURRENT_LIST_DIR}/@PROJECT_NAME@Targets.cmake")
 
-set(WebP_INCLUDE_DIRS "webp")
+set(WebP_INCLUDE_DIRS "@CMAKE_INSTALL_FULL_INCLUDEDIR@")
 set(WEBP_INCLUDE_DIRS ${WebP_INCLUDE_DIRS})
 set(WebP_LIBRARIES "@INSTALLED_LIBRARIES@")
 set(WEBP_LIBRARIES "${WebP_LIBRARIES}")

cmake/cpu.cmake

@@ -1,3 +1,11 @@
+#  Copyright (c) 2021 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.
+
 # Check for SIMD extensions.
 include(CMakePushCheckState)
 
@@ -117,6 +125,12 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
       list(GET SIMD_DISABLE_FLAGS ${I_SIMD} SIMD_COMPILE_FLAG)
       include(CheckCCompilerFlag)
       if(SIMD_COMPILE_FLAG)
+        # Between 3.17.0 and 3.18.2 check_cxx_compiler_flag() sets a normal
+        # variable at parent scope while check_cxx_source_compiles() continues
+        # to set an internal cache variable, so we unset both to avoid the
+        # failure / success state persisting between checks. See
+        # https://gitlab.kitware.com/cmake/cmake/-/issues/21207.
+        unset(HAS_COMPILE_FLAG)
         unset(HAS_COMPILE_FLAG CACHE)
         check_c_compiler_flag(${SIMD_COMPILE_FLAG} HAS_COMPILE_FLAG)
         if(HAS_COMPILE_FLAG)
@@ -134,6 +148,7 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
                                   "warning: argument unused during compilation:"
                                   ${COMMON_PATTERNS})
           if(NOT FLAG_${SIMD_COMPILE_FLAG})
+            unset(HAS_COMPILE_FLAG)
             unset(HAS_COMPILE_FLAG CACHE)
           endif()
         endif()

cmake/deps.cmake

@@ -1,3 +1,11 @@
+#  Copyright (c) 2021 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.
+
 # Generate the config.h to compile with specific intrinsics / libs.
 
 # Check for compiler options.
@@ -22,26 +30,25 @@ check_c_source_compiles("
   " HAVE_BUILTIN_BSWAP64)
 
 # Check for libraries.
-if(NOT WEBP_BUILD_WEBP_JS)
-  # Disable pThreads for WASM.
+if(WEBP_USE_THREAD)
   find_package(Threads)
-endif()
-if(Threads_FOUND)
-  # 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 "${CMAKE_C_FLAGS} -pthread")
+  if(Threads_FOUND)
+    # work around cmake bug on QNX (https://cmake.org/Bug/view.php?id=11333)
+    if(CMAKE_USE_PTHREADS_INIT AND NOT CMAKE_SYSTEM_NAME STREQUAL "QNX")
+      set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -pthread")
+    endif()
+    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()
-  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})
+  set(WEBP_USE_THREAD ${Threads_FOUND})
 endif()
-set(WEBP_USE_THREAD ${Threads_FOUND})
 
 # TODO: this seems unused, check with autotools.
 set(LT_OBJDIR ".libs/")

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([libwebp], [1.2.0],
+AC_INIT([libwebp], [1.2.1],
         [https://bugs.chromium.org/p/webp],,
         [http://developers.google.com/speed/webp])
 AC_CANONICAL_HOST
@@ -32,6 +32,40 @@ AC_ARG_ENABLE([everything],
                SET_IF_UNSET([enable_libwebpextras], [$enableval])
                SET_IF_UNSET([enable_libwebpmux], [$enableval])])
 
+dnl === Check whether libwebpmux should be built
+AC_MSG_CHECKING(whether libwebpmux is to be built)
+AC_ARG_ENABLE([libwebpmux],
+              AS_HELP_STRING([--disable-libwebpmux],
+                             [Disable libwebpmux @<:@default=no@:>@]),
+              [], [enable_libwebpmux=yes])
+AC_MSG_RESULT(${enable_libwebpmux-no})
+AM_CONDITIONAL([BUILD_MUX], [test "$enable_libwebpmux" = "yes"])
+
+dnl === Check whether libwebpdemux should be built
+AC_MSG_CHECKING(whether libwebpdemux is to be built)
+AC_ARG_ENABLE([libwebpdemux],
+              AS_HELP_STRING([--disable-libwebpdemux],
+                             [Disable libwebpdemux @<:@default=no@:>@]),
+              [], [enable_libwebpdemux=yes])
+AC_MSG_RESULT(${enable_libwebpdemux-no})
+AM_CONDITIONAL([BUILD_DEMUX], [test "$enable_libwebpdemux" = "yes"])
+
+dnl === Check whether decoder library should be built.
+AC_MSG_CHECKING(whether decoder library is to be built)
+AC_ARG_ENABLE([libwebpdecoder],
+              AS_HELP_STRING([--enable-libwebpdecoder],
+                             [Build libwebpdecoder @<:@default=no@:>@]))
+AC_MSG_RESULT(${enable_libwebpdecoder-no})
+AM_CONDITIONAL([BUILD_LIBWEBPDECODER], [test "$enable_libwebpdecoder" = "yes"])
+
+dnl === Check whether libwebpextras should be built
+AC_MSG_CHECKING(whether libwebpextras is to be built)
+AC_ARG_ENABLE([libwebpextras],
+              AS_HELP_STRING([--enable-libwebpextras],
+                             [Build libwebpextras @<:@default=no@:>@]))
+AC_MSG_RESULT(${enable_libwebpextras-no})
+AM_CONDITIONAL([BUILD_EXTRAS], [test "$enable_libwebpextras" = "yes"])
+
 dnl === If --enable-asserts is not defined, define NDEBUG
 
 AC_MSG_CHECKING(whether asserts are enabled)
@@ -225,6 +259,9 @@ AS_IF([test "x$enable_neon" != "xno"], [
         *android*) AC_CHECK_HEADERS([cpu-features.h]) ;;
       esac
       ;;
+    aarch64*|arm64*)
+      AC_DEFINE(WEBP_HAVE_NEON, [1], [Set to 1 if NEON is supported])
+      ;;
   esac
   AC_SUBST([NEON_FLAGS])])
 
@@ -409,6 +446,10 @@ AS_IF([test "x$enable_gl" != "xno"], [
 
   if test "$glut_support" = "yes" -a "$enable_libwebpdemux" = "yes"; then
     build_vwebp=yes
+  else
+    AC_MSG_NOTICE(
+      m4_normalize([Not building vwebp.
+                    OpenGL libraries and --enable-libwebpdemux are required.]))
   fi
 ])
 AM_CONDITIONAL([BUILD_VWEBP], [test "$build_vwebp" = "yes"])
@@ -468,7 +509,7 @@ AS_IF([test "x$enable_sdl" != "xno"], [
   if test x"$sdl_support" = "xyes"; then
     build_vwebp_sdl=yes
   else
-    AC_MSG_WARN(Optional SDL library not found)
+    AC_MSG_NOTICE([Not building vwebp-sdl. SDL library is required.])
   fi
 ])
 
@@ -592,11 +633,17 @@ AS_IF([test "x$enable_gif" != "xno"], [
   if test "$gif_support" = "yes" -a \
           "$enable_libwebpdemux" = "yes"; then
     build_anim_diff=yes
+  else
+    AC_MSG_NOTICE(
+      [Not building anim_diff. libgif and --enable-libwebpdemux are required.])
   fi
 
   if test "$gif_support" = "yes" -a \
           "$enable_libwebpmux" = "yes"; then
     build_gif2webp=yes
+  else
+    AC_MSG_NOTICE(
+      [Not building gif2webp. libgif and --enable-libwebpmux are required.])
   fi
 ])
 AM_CONDITIONAL([BUILD_ANIMDIFF], [test "${build_anim_diff}" = "yes"])
@@ -604,11 +651,17 @@ AM_CONDITIONAL([BUILD_GIF2WEBP], [test "${build_gif2webp}" = "yes"])
 
 if test "$enable_libwebpdemux" = "yes" -a "$enable_libwebpmux" = "yes"; then
   build_img2webp=yes
+else
+  AC_MSG_NOTICE(
+    m4_normalize([Not building img2webp.
+                  --enable-libwebpdemux & --enable-libwebpmux are required.]))
 fi
 AM_CONDITIONAL([BUILD_IMG2WEBP], [test "${build_img2webp}" = "yes"])
 
 if test "$enable_libwebpmux" = "yes"; then
   build_webpinfo=yes
+else
+  AC_MSG_NOTICE([Not building webpinfo. --enable-libwebpdemux is required.])
 fi
 AM_CONDITIONAL([BUILD_WEBPINFO], [test "${build_webpinfo}" = "yes"])
 
@@ -692,39 +745,6 @@ else
   AC_MSG_RESULT([no])
 fi
 
-dnl === Check whether libwebpmux should be built
-AC_MSG_CHECKING(whether libwebpmux is to be built)
-AC_ARG_ENABLE([libwebpmux],
-              AS_HELP_STRING([--enable-libwebpmux],
-                             [Build libwebpmux @<:@default=no@:>@]))
-AC_MSG_RESULT(${enable_libwebpmux-no})
-AM_CONDITIONAL([BUILD_MUX], [test "$enable_libwebpmux" = "yes"])
-
-dnl === Check whether libwebpdemux should be built
-AC_MSG_CHECKING(whether libwebpdemux is to be built)
-AC_ARG_ENABLE([libwebpdemux],
-              AS_HELP_STRING([--disable-libwebpdemux],
-                             [Disable libwebpdemux @<:@default=no@:>@]),
-              [], [enable_libwebpdemux=yes])
-AC_MSG_RESULT(${enable_libwebpdemux-no})
-AM_CONDITIONAL([BUILD_DEMUX], [test "$enable_libwebpdemux" = "yes"])
-
-dnl === Check whether decoder library should be built.
-AC_MSG_CHECKING(whether decoder library is to be built)
-AC_ARG_ENABLE([libwebpdecoder],
-              AS_HELP_STRING([--enable-libwebpdecoder],
-                             [Build libwebpdecoder @<:@default=no@:>@]))
-AC_MSG_RESULT(${enable_libwebpdecoder-no})
-AM_CONDITIONAL([BUILD_LIBWEBPDECODER], [test "$enable_libwebpdecoder" = "yes"])
-
-dnl === Check whether libwebpextras should be built
-AC_MSG_CHECKING(whether libwebpextras is to be built)
-AC_ARG_ENABLE([libwebpextras],
-              AS_HELP_STRING([--enable-libwebpextras],
-                             [Build libwebpextras @<:@default=no@:>@]))
-AC_MSG_RESULT(${enable_libwebpextras-no})
-AM_CONDITIONAL([BUILD_EXTRAS], [test "$enable_libwebpextras" = "yes"])
-
 dnl =========================
 
 AC_CONFIG_MACRO_DIR([m4])

doc/webp-container-spec.txt

@@ -816,7 +816,7 @@ RIFF/WEBP
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 
 [vp8spec]:  http://tools.ietf.org/html/rfc6386
-[webpllspec]: https://chromium.googlesource.com/webm/libwebp/+/master/doc/webp-lossless-bitstream-spec.txt
+[webpllspec]: https://chromium.googlesource.com/webm/libwebp/+/HEAD/doc/webp-lossless-bitstream-spec.txt
 [iccspec]: http://www.color.org/icc_specs2.xalter
 [metadata]: http://www.metadataworkinggroup.org/pdf/mwg_guidance.pdf
 [rfc 1166]: http://tools.ietf.org/html/rfc1166

examples/gif2webp.c

@@ -465,8 +465,10 @@ int main(int argc, const char* argv[]) {
     fprintf(stderr, "%s\n", WebPAnimEncoderGetError(enc));
     goto End;
   }
-
-  if (!loop_compatibility) {
+  // If there's only one frame, we don't need to handle loop count.
+  if (frame_number == 1) {
+    loop_count = 0;
+  } else if (!loop_compatibility) {
     if (!stored_loop_count) {
       // if no loop-count element is seen, the default is '1' (loop-once)
       // and we need to signal it explicitly in WebP. Note however that

examples/webpmux.c

@@ -472,11 +472,11 @@ static WebPMux* DuplicateMuxHeader(const WebPMux* const mux) {
 }
 
 static int ParseFrameArgs(const char* args, WebPMuxFrameInfo* const info) {
-  int dispose_method, dummy;
+  int dispose_method, unused;
   char plus_minus, blend_method;
   const int num_args = sscanf(args, "+%d+%d+%d+%d%c%c+%d", &info->duration,
                               &info->x_offset, &info->y_offset, &dispose_method,
-                              &plus_minus, &blend_method, &dummy);
+                              &plus_minus, &blend_method, &unused);
   switch (num_args) {
     case 1:
       info->x_offset = info->y_offset = 0;  // fall through
@@ -495,7 +495,7 @@ static int ParseFrameArgs(const char* args, WebPMuxFrameInfo* const info) {
 
   WarnAboutOddOffset(info);
 
-  // Note: The sanity of the following conversion is checked by
+  // Note: The validity of the following conversion is checked by
   // WebPMuxPushFrame().
   info->dispose_method = (WebPMuxAnimDispose)dispose_method;
 

extras/extras.c

@@ -19,7 +19,7 @@
 
 #define XTRA_MAJ_VERSION 1
 #define XTRA_MIN_VERSION 2
-#define XTRA_REV_VERSION 0
+#define XTRA_REV_VERSION 1
 
 //------------------------------------------------------------------------------
 

imageio/image_enc.c

@@ -155,8 +155,8 @@ int WebPWritePNG(const char* out_file_name, int use_stdout,
 }
 
 #elif defined(WEBP_HAVE_PNG)    // !HAVE_WINCODEC_H
-static void PNGAPI PNGErrorFunction(png_structp png, png_const_charp dummy) {
-  (void)dummy;  // remove variable-unused warning
+static void PNGAPI PNGErrorFunction(png_structp png, png_const_charp unused) {
+  (void)unused;  // remove variable-unused warning
   longjmp(png_jmpbuf(png), 1);
 }
 
@@ -320,7 +320,7 @@ int WebPWriteBMP(FILE* fout, const WebPDecBuffer* const buffer) {
   PutLE16(bmp_header + 26, 1);                    // number of planes
   PutLE16(bmp_header + 28, bytes_per_px * 8);     // bits per pixel
   PutLE32(bmp_header + 30, 0);                    // no compression (BI_RGB)
-  PutLE32(bmp_header + 34, 0);                    // image size (dummy)
+  PutLE32(bmp_header + 34, 0);                    // image size (placeholder)
   PutLE32(bmp_header + 38, 2400);                 // x pixels/meter
   PutLE32(bmp_header + 42, 2400);                 // y pixels/meter
   PutLE32(bmp_header + 46, 0);                    // number of palette colors

imageio/imageio_util.c

@@ -148,9 +148,11 @@ void ImgIoUtilCopyPlane(const uint8_t* src, int src_stride,
 
 // -----------------------------------------------------------------------------
 
-int ImgIoUtilCheckSizeArgumentsOverflow(uint64_t nmemb, size_t size) {
-  const uint64_t total_size = nmemb * size;
+int ImgIoUtilCheckSizeArgumentsOverflow(uint64_t stride, size_t height) {
+  const uint64_t total_size = stride * height;
   int ok = (total_size == (size_t)total_size);
+  // check that 'stride' is representable as int:
+  ok = ok && ((uint64_t)(int)stride == stride);
 #if defined(WEBP_MAX_IMAGE_SIZE)
   ok = ok && (total_size <= (uint64_t)WEBP_MAX_IMAGE_SIZE);
 #endif

imageio/imageio_util.h

@@ -54,8 +54,8 @@ void ImgIoUtilCopyPlane(const uint8_t* src, int src_stride,
 
 //------------------------------------------------------------------------------
 
-// Returns 0 in case of overflow of nmemb * size.
-int ImgIoUtilCheckSizeArgumentsOverflow(uint64_t nmemb, size_t size);
+// Returns 0 in case of overflow, memory over-allocation or excessive dimension.
+int ImgIoUtilCheckSizeArgumentsOverflow(uint64_t stride, size_t height);
 
 #ifdef __cplusplus
 }    // extern "C"

imageio/jpegdec.c

@@ -274,7 +274,7 @@ int ReadJPEG(const uint8_t* const data, size_t data_size,
   ctx.data = data;
   ctx.data_size = data_size;
 
-  memset((j_decompress_ptr)&dinfo, 0, sizeof(dinfo));   // for setjmp sanity
+  memset((j_decompress_ptr)&dinfo, 0, sizeof(dinfo));   // for setjmp safety
   dinfo.err = jpeg_std_error(&jerr.pub);
   jerr.pub.error_exit = my_error_exit;
 

imageio/pngdec.c

@@ -259,11 +259,14 @@ 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;

imageio/tiffdec.c

@@ -157,6 +157,7 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
   MyData my_data = { data, (toff_t)data_size, 0 };
   TIFF* tif;
   uint32_t image_width, image_height, tile_width, tile_height;
+  uint64_t stride;
   uint16_t samples_per_px = 0;
   uint16_t extra_samples = 0;
   uint16_t* extra_samples_ptr = NULL;
@@ -194,20 +195,31 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
     fprintf(stderr, "Error! Cannot retrieve TIFF image dimensions.\n");
     goto End;
   }
-  if (!ImgIoUtilCheckSizeArgumentsOverflow((uint64_t)image_width * image_height,
-                                           sizeof(*raster))) {
+  stride = (uint64_t)image_width * sizeof(*raster);
+  if (!ImgIoUtilCheckSizeArgumentsOverflow(stride, image_height)) {
+    fprintf(stderr, "Error! TIFF image dimension (%d x %d) is too large.\n",
+            image_width, image_height);
     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;
+  // 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. We also check that they respect WebP's dimension and memory limit.
+  // Note that a tile can be 6byte/px in some cases. Here we assume
+  // 4byte/px with sizeof(*raster), to be conservative.
+  if (TIFFGetField(tif, TIFFTAG_TILEWIDTH, &tile_width) &&
+      TIFFGetField(tif, TIFFTAG_TILELENGTH, &tile_height)) {
+    if ((tile_width > 32 && tile_width / 2 > image_width) ||
+        (tile_height > 32 && tile_height / 2 > image_height) ||
+        ImgIoUtilCheckSizeArgumentsOverflow(
+            (uint64_t)tile_width * sizeof(*raster), tile_height)) {
+      fprintf(stderr, "Error! TIFF tile dimension (%d x %d) is too large.\n",
+              tile_width, tile_height);
+      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");
@@ -215,15 +227,13 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
   }
 
   // _Tiffmalloc uses a signed type for size.
-  alloc_size =
-      (int64_t)((uint64_t)image_width * image_height * sizeof(*raster));
+  alloc_size = (int64_t)(stride * image_height);
   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, image_width, image_height, raster,
                                   ORIENTATION_TOPLEFT, 1)) {
-      const int stride = image_width * sizeof(*raster);
       pic->width = image_width;
       pic->height = image_height;
       // TIFF data is ABGR
@@ -241,8 +251,8 @@ int ReadTIFF(const uint8_t* const data, size_t data_size,
         }
       }
       ok = keep_alpha
-         ? WebPPictureImportRGBA(pic, (const uint8_t*)raster, stride)
-         : WebPPictureImportRGBX(pic, (const uint8_t*)raster, stride);
+         ? WebPPictureImportRGBA(pic, (const uint8_t*)raster, (int)stride)
+         : WebPPictureImportRGBX(pic, (const uint8_t*)raster, (int)stride);
     }
     _TIFFfree(raster);
   } else {

imageio/webpdec.c

@@ -95,7 +95,7 @@ VP8StatusCode DecodeWebPIncremental(
   {
     WebPIDecoder* const idec = WebPIDecode(data, data_size, config);
     if (idec == NULL) {
-      fprintf(stderr, "Failed during WebPINewDecoder().\n");
+      fprintf(stderr, "Failed during WebPIDecode().\n");
       return VP8_STATUS_OUT_OF_MEMORY;
     } else {
       status = WebPIUpdate(idec, data, data_size);

imageio/wicdec.c

@@ -134,7 +134,10 @@ static HRESULT ExtractICCP(IWICImagingFactory* const factory,
   IWICColorContext** color_contexts;
 
   IFS(IWICBitmapFrameDecode_GetColorContexts(frame, 0, NULL, &count));
-  if (FAILED(hr) || count == 0) return hr;
+  if (FAILED(hr) || count == 0) {
+    // Treat unsupported operation as a non-fatal error. See crbug.com/webp/506.
+    return (hr == WINCODEC_ERR_UNSUPPORTEDOPERATION) ? S_OK : hr;
+  }
 
   color_contexts = (IWICColorContext**)calloc(count, sizeof(*color_contexts));
   if (color_contexts == NULL) return E_OUTOFMEMORY;
@@ -270,10 +273,15 @@ int ReadPictureWithWIC(const char* const filename,
   WICPixelFormatGUID src_pixel_format = GUID_WICPixelFormatUndefined;
   const WICFormatImporter* importer = NULL;
   GUID src_container_format = GUID_NULL_;
+  // From Windows Kits\10\Include\10.0.19041.0\um\wincodec.h
+  WEBP_DEFINE_GUID(GUID_ContainerFormatWebp_,
+                   0xe094b0e2, 0x67f2, 0x45b3,
+                   0xb0, 0xea, 0x11, 0x53, 0x37, 0xca, 0x7c, 0xf3);
   static const GUID* kAlphaContainers[] = {
     &GUID_ContainerFormatBmp,
     &GUID_ContainerFormatPng,
     &GUID_ContainerFormatTiff,
+    &GUID_ContainerFormatWebp_,
     NULL
   };
   int has_alpha = 0;

makefile.unix

@@ -166,6 +166,7 @@ DSP_DEC_OBJS = \
     src/dsp/lossless_msa.o \
     src/dsp/lossless_neon.o \
     src/dsp/lossless_sse2.o \
+    src/dsp/lossless_sse41.o \
     src/dsp/rescaler.o \
     src/dsp/rescaler_mips32.o \
     src/dsp/rescaler_mips_dsp_r2.o \

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 8:1:1
+libwebp_la_LDFLAGS = -no-undefined -version-info 8:2:1
 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 4:1:1
+  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 4:2:1
   pkgconfig_DATA += libwebpdecoder.pc
 endif
 

src/dec/alpha_dec.c

@@ -183,7 +183,7 @@ const uint8_t* VP8DecompressAlphaRows(VP8Decoder* const dec,
   assert(dec != NULL && io != NULL);
 
   if (row < 0 || num_rows <= 0 || row + num_rows > height) {
-    return NULL;    // sanity check.
+    return NULL;
   }
 
   if (!dec->is_alpha_decoded_) {

src/dec/buffer_dec.c

@@ -102,7 +102,7 @@ static VP8StatusCode AllocateBuffer(WebPDecBuffer* const buffer) {
     int stride;
     uint64_t size;
 
-    if ((uint64_t)w * kModeBpp[mode] >= (1ull << 32)) {
+    if ((uint64_t)w * kModeBpp[mode] >= (1ull << 31)) {
       return VP8_STATUS_INVALID_PARAM;
     }
     stride = w * kModeBpp[mode];
@@ -117,7 +117,6 @@ static VP8StatusCode AllocateBuffer(WebPDecBuffer* const buffer) {
     }
     total_size = size + 2 * uv_size + a_size;
 
-    // Security/sanity checks
     output = (uint8_t*)WebPSafeMalloc(total_size, sizeof(*output));
     if (output == NULL) {
       return VP8_STATUS_OUT_OF_MEMORY;
@@ -156,11 +155,11 @@ VP8StatusCode WebPFlipBuffer(WebPDecBuffer* const buffer) {
   }
   if (WebPIsRGBMode(buffer->colorspace)) {
     WebPRGBABuffer* const buf = &buffer->u.RGBA;
-    buf->rgba += (buffer->height - 1) * buf->stride;
+    buf->rgba += (int64_t)(buffer->height - 1) * buf->stride;
     buf->stride = -buf->stride;
   } else {
     WebPYUVABuffer* const buf = &buffer->u.YUVA;
-    const int H = buffer->height;
+    const int64_t H = buffer->height;
     buf->y += (H - 1) * buf->y_stride;
     buf->y_stride = -buf->y_stride;
     buf->u += ((H - 1) >> 1) * buf->u_stride;
@@ -188,8 +187,7 @@ VP8StatusCode WebPAllocateDecBuffer(int width, int height,
       const int ch = options->crop_height;
       const int x = options->crop_left & ~1;
       const int y = options->crop_top & ~1;
-      if (x < 0 || y < 0 || cw <= 0 || ch <= 0 ||
-          x + cw > width || y + ch > height) {
+      if (!WebPCheckCropDimensions(width, height, x, y, cw, ch)) {
         return VP8_STATUS_INVALID_PARAM;   // out of frame boundary.
       }
       width = cw;

src/dec/frame_dec.c

@@ -705,7 +705,7 @@ static int AllocateMemory(VP8Decoder* const dec) {
                         + cache_size + alpha_size + WEBP_ALIGN_CST;
   uint8_t* mem;
 
-  if (needed != (size_t)needed) return 0;  // check for overflow
+  if (!CheckSizeOverflow(needed)) return 0;  // check for overflow
   if (needed > dec->mem_size_) {
     WebPSafeFree(dec->mem_);
     dec->mem_size_ = 0;

src/dec/io_dec.c

@@ -298,46 +298,57 @@ static int InitYUVRescaler(const VP8Io* const io, WebPDecParams* const p) {
   const int uv_out_height = (out_height + 1) >> 1;
   const int uv_in_width  = (io->mb_w + 1) >> 1;
   const int uv_in_height = (io->mb_h + 1) >> 1;
-  const size_t work_size = 2 * out_width;   // scratch memory for luma rescaler
+  // scratch memory for luma rescaler
+  const size_t work_size = 2 * (size_t)out_width;
   const size_t uv_work_size = 2 * uv_out_width;  // and for each u/v ones
-  size_t tmp_size, rescaler_size;
+  uint64_t total_size;
+  size_t rescaler_size;
   rescaler_t* work;
   WebPRescaler* scalers;
   const int num_rescalers = has_alpha ? 4 : 3;
 
-  tmp_size = (work_size + 2 * uv_work_size) * sizeof(*work);
+  total_size = ((uint64_t)work_size + 2 * uv_work_size) * sizeof(*work);
   if (has_alpha) {
-    tmp_size += work_size * sizeof(*work);
+    total_size += (uint64_t)work_size * sizeof(*work);
   }
   rescaler_size = num_rescalers * sizeof(*p->scaler_y) + WEBP_ALIGN_CST;
+  total_size += rescaler_size;
+  if (!CheckSizeOverflow(total_size)) {
+    return 0;
+  }
 
-  p->memory = WebPSafeMalloc(1ULL, tmp_size + rescaler_size);
+  p->memory = WebPSafeMalloc(1ULL, (size_t)total_size);
   if (p->memory == NULL) {
     return 0;   // memory error
   }
   work = (rescaler_t*)p->memory;
 
-  scalers = (WebPRescaler*)WEBP_ALIGN((const uint8_t*)work + tmp_size);
+  scalers = (WebPRescaler*)WEBP_ALIGN(
+      (const uint8_t*)work + total_size - rescaler_size);
   p->scaler_y = &scalers[0];
   p->scaler_u = &scalers[1];
   p->scaler_v = &scalers[2];
   p->scaler_a = has_alpha ? &scalers[3] : NULL;
 
-  WebPRescalerInit(p->scaler_y, io->mb_w, io->mb_h,
-                   buf->y, out_width, out_height, buf->y_stride, 1,
-                   work);
-  WebPRescalerInit(p->scaler_u, uv_in_width, uv_in_height,
-                   buf->u, uv_out_width, uv_out_height, buf->u_stride, 1,
-                   work + work_size);
-  WebPRescalerInit(p->scaler_v, uv_in_width, uv_in_height,
-                   buf->v, uv_out_width, uv_out_height, buf->v_stride, 1,
-                   work + work_size + uv_work_size);
+  if (!WebPRescalerInit(p->scaler_y, io->mb_w, io->mb_h,
+                        buf->y, out_width, out_height, buf->y_stride, 1,
+                        work) ||
+      !WebPRescalerInit(p->scaler_u, uv_in_width, uv_in_height,
+                        buf->u, uv_out_width, uv_out_height, buf->u_stride, 1,
+                        work + work_size) ||
+      !WebPRescalerInit(p->scaler_v, uv_in_width, uv_in_height,
+                        buf->v, uv_out_width, uv_out_height, buf->v_stride, 1,
+                        work + work_size + uv_work_size)) {
+    return 0;
+  }
   p->emit = EmitRescaledYUV;
 
   if (has_alpha) {
-    WebPRescalerInit(p->scaler_a, io->mb_w, io->mb_h,
-                     buf->a, out_width, out_height, buf->a_stride, 1,
-                     work + work_size + 2 * uv_work_size);
+    if (!WebPRescalerInit(p->scaler_a, io->mb_w, io->mb_h,
+                          buf->a, out_width, out_height, buf->a_stride, 1,
+                          work + work_size + 2 * uv_work_size)) {
+      return 0;
+    }
     p->emit_alpha = EmitRescaledAlphaYUV;
     WebPInitAlphaProcessing();
   }
@@ -480,51 +491,58 @@ static int InitRGBRescaler(const VP8Io* const io, WebPDecParams* const p) {
   const int out_height = io->scaled_height;
   const int uv_in_width  = (io->mb_w + 1) >> 1;
   const int uv_in_height = (io->mb_h + 1) >> 1;
-  const size_t work_size = 2 * out_width;   // scratch memory for one rescaler
+  // scratch memory for one rescaler
+  const size_t work_size = 2 * (size_t)out_width;
   rescaler_t* work;  // rescalers work area
   uint8_t* tmp;   // tmp storage for scaled YUV444 samples before RGB conversion
-  size_t tmp_size1, tmp_size2, total_size, rescaler_size;
+  uint64_t tmp_size1, tmp_size2, total_size;
+  size_t rescaler_size;
   WebPRescaler* scalers;
   const int num_rescalers = has_alpha ? 4 : 3;
 
-  tmp_size1 = 3 * work_size;
-  tmp_size2 = 3 * out_width;
-  if (has_alpha) {
-    tmp_size1 += work_size;
-    tmp_size2 += out_width;
-  }
+  tmp_size1 = (uint64_t)num_rescalers * work_size;
+  tmp_size2 = (uint64_t)num_rescalers * out_width;
   total_size = tmp_size1 * sizeof(*work) + tmp_size2 * sizeof(*tmp);
   rescaler_size = num_rescalers * sizeof(*p->scaler_y) + WEBP_ALIGN_CST;
+  total_size += rescaler_size;
+  if (!CheckSizeOverflow(total_size)) {
+    return 0;
+  }
 
-  p->memory = WebPSafeMalloc(1ULL, total_size + rescaler_size);
+  p->memory = WebPSafeMalloc(1ULL, (size_t)total_size);
   if (p->memory == NULL) {
     return 0;   // memory error
   }
   work = (rescaler_t*)p->memory;
   tmp = (uint8_t*)(work + tmp_size1);
 
-  scalers = (WebPRescaler*)WEBP_ALIGN((const uint8_t*)work + total_size);
+  scalers = (WebPRescaler*)WEBP_ALIGN(
+      (const uint8_t*)work + total_size - rescaler_size);
   p->scaler_y = &scalers[0];
   p->scaler_u = &scalers[1];
   p->scaler_v = &scalers[2];
   p->scaler_a = has_alpha ? &scalers[3] : NULL;
 
-  WebPRescalerInit(p->scaler_y, io->mb_w, io->mb_h,
-                   tmp + 0 * out_width, out_width, out_height, 0, 1,
-                   work + 0 * work_size);
-  WebPRescalerInit(p->scaler_u, uv_in_width, uv_in_height,
-                   tmp + 1 * out_width, out_width, out_height, 0, 1,
-                   work + 1 * work_size);
-  WebPRescalerInit(p->scaler_v, uv_in_width, uv_in_height,
-                   tmp + 2 * out_width, out_width, out_height, 0, 1,
-                   work + 2 * work_size);
+  if (!WebPRescalerInit(p->scaler_y, io->mb_w, io->mb_h,
+                        tmp + 0 * out_width, out_width, out_height, 0, 1,
+                        work + 0 * work_size) ||
+      !WebPRescalerInit(p->scaler_u, uv_in_width, uv_in_height,
+                        tmp + 1 * out_width, out_width, out_height, 0, 1,
+                        work + 1 * work_size) ||
+      !WebPRescalerInit(p->scaler_v, uv_in_width, uv_in_height,
+                        tmp + 2 * out_width, out_width, out_height, 0, 1,
+                        work + 2 * work_size)) {
+    return 0;
+  }
   p->emit = EmitRescaledRGB;
   WebPInitYUV444Converters();
 
   if (has_alpha) {
-    WebPRescalerInit(p->scaler_a, io->mb_w, io->mb_h,
-                     tmp + 3 * out_width, out_width, out_height, 0, 1,
-                     work + 3 * work_size);
+    if (!WebPRescalerInit(p->scaler_a, io->mb_w, io->mb_h,
+                          tmp + 3 * out_width, out_width, out_height, 0, 1,
+                          work + 3 * work_size)) {
+      return 0;
+    }
     p->emit_alpha = EmitRescaledAlphaRGB;
     if (p->output->colorspace == MODE_RGBA_4444 ||
         p->output->colorspace == MODE_rgbA_4444) {

src/dec/vp8_dec.c

@@ -335,7 +335,7 @@ int VP8GetHeaders(VP8Decoder* const dec, VP8Io* const io) {
     io->scaled_width = io->width;
     io->scaled_height = io->height;
 
-    io->mb_w = io->width;   // sanity check
+    io->mb_w = io->width;   // for soundness
     io->mb_h = io->height;  // ditto
 
     VP8ResetProba(&dec->proba_);

src/dec/vp8i_dec.h

@@ -32,7 +32,7 @@ extern "C" {
 // version numbers
 #define DEC_MAJ_VERSION 1
 #define DEC_MIN_VERSION 2
-#define DEC_REV_VERSION 0
+#define DEC_REV_VERSION 1
 
 // 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

@@ -559,8 +559,11 @@ static int AllocateAndInitRescaler(VP8LDecoder* const dec, VP8Io* const io) {
   memory += work_size * sizeof(*work);
   scaled_data = (uint32_t*)memory;
 
-  WebPRescalerInit(dec->rescaler, in_width, in_height, (uint8_t*)scaled_data,
-                   out_width, out_height, 0, num_channels, work);
+  if (!WebPRescalerInit(dec->rescaler, in_width, in_height,
+                        (uint8_t*)scaled_data, out_width, out_height,
+                        0, num_channels, work)) {
+    return 0;
+  }
   return 1;
 }
 #endif   // WEBP_REDUCE_SIZE
@@ -574,13 +577,14 @@ static int AllocateAndInitRescaler(VP8LDecoder* const dec, VP8Io* const io) {
 static int Export(WebPRescaler* const rescaler, WEBP_CSP_MODE colorspace,
                   int rgba_stride, uint8_t* const rgba) {
   uint32_t* const src = (uint32_t*)rescaler->dst;
+  uint8_t* dst = rgba;
   const int dst_width = rescaler->dst_width;
   int num_lines_out = 0;
   while (WebPRescalerHasPendingOutput(rescaler)) {
-    uint8_t* const dst = rgba + num_lines_out * rgba_stride;
     WebPRescalerExportRow(rescaler);
     WebPMultARGBRow(src, dst_width, 1);
     VP8LConvertFromBGRA(src, dst_width, colorspace, dst);
+    dst += rgba_stride;
     ++num_lines_out;
   }
   return num_lines_out;
@@ -594,8 +598,8 @@ static int EmitRescaledRowsRGBA(const VP8LDecoder* const dec,
   int num_lines_in = 0;
   int num_lines_out = 0;
   while (num_lines_in < mb_h) {
-    uint8_t* const row_in = in + num_lines_in * in_stride;
-    uint8_t* const row_out = out + num_lines_out * out_stride;
+    uint8_t* const row_in = in + (uint64_t)num_lines_in * in_stride;
+    uint8_t* const row_out = out + (uint64_t)num_lines_out * out_stride;
     const int lines_left = mb_h - num_lines_in;
     const int needed_lines = WebPRescaleNeededLines(dec->rescaler, lines_left);
     int lines_imported;
@@ -796,7 +800,8 @@ static void ProcessRows(VP8LDecoder* const dec, int row) {
       const WebPDecBuffer* const output = dec->output_;
       if (WebPIsRGBMode(output->colorspace)) {  // convert to RGBA
         const WebPRGBABuffer* const buf = &output->u.RGBA;
-        uint8_t* const rgba = buf->rgba + dec->last_out_row_ * buf->stride;
+        uint8_t* const rgba =
+            buf->rgba + (int64_t)dec->last_out_row_ * buf->stride;
         const int num_rows_out =
 #if !defined(WEBP_REDUCE_SIZE)
          io->use_scaling ?
@@ -1514,7 +1519,7 @@ static int AllocateInternalBuffers32b(VP8LDecoder* const dec, int final_width) {
   assert(dec->width_ <= final_width);
   dec->pixels_ = (uint32_t*)WebPSafeMalloc(total_num_pixels, sizeof(uint32_t));
   if (dec->pixels_ == NULL) {
-    dec->argb_cache_ = NULL;    // for sanity check
+    dec->argb_cache_ = NULL;    // for soundness
     dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
     return 0;
   }
@@ -1524,7 +1529,7 @@ static int AllocateInternalBuffers32b(VP8LDecoder* const dec, int final_width) {
 
 static int AllocateInternalBuffers8b(VP8LDecoder* const dec) {
   const uint64_t total_num_pixels = (uint64_t)dec->width_ * dec->height_;
-  dec->argb_cache_ = NULL;    // for sanity check
+  dec->argb_cache_ = NULL;    // for soundness
   dec->pixels_ = (uint32_t*)WebPSafeMalloc(total_num_pixels, sizeof(uint8_t));
   if (dec->pixels_ == NULL) {
     dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
@@ -1666,7 +1671,6 @@ int VP8LDecodeImage(VP8LDecoder* const dec) {
   VP8Io* io = NULL;
   WebPDecParams* params = NULL;
 
-  // Sanity checks.
   if (dec == NULL) return 0;
 
   assert(dec->hdr_.huffman_tables_ != NULL);

src/dec/webp_dec.c

@@ -785,6 +785,13 @@ VP8StatusCode WebPDecode(const uint8_t* data, size_t data_size,
 //------------------------------------------------------------------------------
 // Cropping and rescaling.
 
+int WebPCheckCropDimensions(int image_width, int image_height,
+                            int x, int y, int w, int h) {
+  return !(x < 0 || y < 0 || w <= 0 || h <= 0 ||
+           x >= image_width || w > image_width || w > image_width - x ||
+           y >= image_height || h > image_height || h > image_height - y);
+}
+
 int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
                           VP8Io* const io, WEBP_CSP_MODE src_colorspace) {
   const int W = io->width;
@@ -792,7 +799,7 @@ int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
   int x = 0, y = 0, w = W, h = H;
 
   // Cropping
-  io->use_cropping = (options != NULL) && (options->use_cropping > 0);
+  io->use_cropping = (options != NULL) && options->use_cropping;
   if (io->use_cropping) {
     w = options->crop_width;
     h = options->crop_height;
@@ -802,7 +809,7 @@ int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
       x &= ~1;
       y &= ~1;
     }
-    if (x < 0 || y < 0 || w <= 0 || h <= 0 || x + w > W || y + h > H) {
+    if (!WebPCheckCropDimensions(W, H, x, y, w, h)) {
       return 0;  // out of frame boundary error
     }
   }
@@ -814,7 +821,7 @@ int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
   io->mb_h = h;
 
   // Scaling
-  io->use_scaling = (options != NULL) && (options->use_scaling > 0);
+  io->use_scaling = (options != NULL) && options->use_scaling;
   if (io->use_scaling) {
     int scaled_width = options->scaled_width;
     int scaled_height = options->scaled_height;
@@ -835,8 +842,8 @@ int WebPIoInitFromOptions(const WebPDecoderOptions* const options,
 
   if (io->use_scaling) {
     // disable filter (only for large downscaling ratio).
-    io->bypass_filtering = (io->scaled_width < W * 3 / 4) &&
-                           (io->scaled_height < H * 3 / 4);
+    io->bypass_filtering |= (io->scaled_width < W * 3 / 4) &&
+                            (io->scaled_height < H * 3 / 4);
     io->fancy_upsampling = 0;
   }
   return 1;

src/dec/webpi_dec.h

@@ -77,6 +77,10 @@ VP8StatusCode WebPParseHeaders(WebPHeaderStructure* const headers);
 //------------------------------------------------------------------------------
 // Misc utils
 
+// Returns true if crop dimensions are within image bounds.
+int WebPCheckCropDimensions(int image_width, int image_height,
+                            int x, int y, int w, int h);
+
 // Initializes VP8Io with custom setup, io and teardown functions. The default
 // hooks will use the supplied 'params' as io->opaque handle.
 void WebPInitCustomIo(WebPDecParams* const params, VP8Io* const io);

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 2:7:0
+libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:8:0
 libwebpdemuxincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebpdemux.pc

src/demux/anim_decode.c

@@ -87,11 +87,19 @@ WebPAnimDecoder* WebPAnimDecoderNewInternal(
     int abi_version) {
   WebPAnimDecoderOptions options;
   WebPAnimDecoder* dec = NULL;
+  WebPBitstreamFeatures features;
   if (webp_data == NULL ||
       WEBP_ABI_IS_INCOMPATIBLE(abi_version, WEBP_DEMUX_ABI_VERSION)) {
     return NULL;
   }
 
+  // Validate the bitstream before doing expensive allocations. The demuxer may
+  // be more tolerant than the decoder.
+  if (WebPGetFeatures(webp_data->bytes, webp_data->size, &features) !=
+      VP8_STATUS_OK) {
+    return NULL;
+  }
+
   // Note: calloc() so that the pointer members are initialized to NULL.
   dec = (WebPAnimDecoder*)WebPSafeCalloc(1ULL, sizeof(*dec));
   if (dec == NULL) goto Error;
@@ -145,7 +153,7 @@ static int ZeroFillCanvas(uint8_t* buf, uint32_t canvas_width,
                           uint32_t canvas_height) {
   const uint64_t size =
       (uint64_t)canvas_width * canvas_height * NUM_CHANNELS * sizeof(*buf);
-  if (size != (size_t)size) return 0;
+  if (!CheckSizeOverflow(size)) return 0;
   memset(buf, 0, (size_t)size);
   return 1;
 }
@@ -166,7 +174,7 @@ static void ZeroFillFrameRect(uint8_t* buf, int buf_stride, int x_offset,
 static int CopyCanvas(const uint8_t* src, uint8_t* dst,
                       uint32_t width, uint32_t height) {
   const uint64_t size = (uint64_t)width * height * NUM_CHANNELS;
-  if (size != (size_t)size) return 0;
+  if (!CheckSizeOverflow(size)) return 0;
   assert(src != NULL && dst != NULL);
   memcpy(dst, src, (size_t)size);
   return 1;

src/demux/demux.c

@@ -25,7 +25,7 @@
 
 #define DMUX_MAJ_VERSION 1
 #define DMUX_MIN_VERSION 2
-#define DMUX_REV_VERSION 0
+#define DMUX_REV_VERSION 1
 
 typedef struct {
   size_t start_;        // start location of the data
@@ -221,12 +221,16 @@ static ParseStatus StoreFrame(int frame_num, uint32_t min_size,
     const size_t chunk_start_offset = mem->start_;
     const uint32_t fourcc = ReadLE32(mem);
     const uint32_t payload_size = ReadLE32(mem);
-    const uint32_t payload_size_padded = payload_size + (payload_size & 1);
-    const size_t payload_available = (payload_size_padded > MemDataSize(mem))
-                                   ? MemDataSize(mem) : payload_size_padded;
-    const size_t chunk_size = CHUNK_HEADER_SIZE + payload_available;
+    uint32_t payload_size_padded;
+    size_t payload_available;
+    size_t chunk_size;
 
     if (payload_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
+
+    payload_size_padded = payload_size + (payload_size & 1);
+    payload_available = (payload_size_padded > MemDataSize(mem))
+                      ? MemDataSize(mem) : payload_size_padded;
+    chunk_size = CHUNK_HEADER_SIZE + payload_available;
     if (SizeIsInvalid(mem, payload_size_padded)) return PARSE_ERROR;
     if (payload_size_padded > MemDataSize(mem)) status = PARSE_NEED_MORE_DATA;
 
@@ -451,9 +455,11 @@ static ParseStatus ParseVP8XChunks(WebPDemuxer* const dmux) {
     const size_t chunk_start_offset = mem->start_;
     const uint32_t fourcc = ReadLE32(mem);
     const uint32_t chunk_size = ReadLE32(mem);
-    const uint32_t chunk_size_padded = chunk_size + (chunk_size & 1);
+    uint32_t chunk_size_padded;
 
     if (chunk_size > MAX_CHUNK_PAYLOAD) return PARSE_ERROR;
+
+    chunk_size_padded = chunk_size + (chunk_size & 1);
     if (SizeIsInvalid(mem, chunk_size_padded)) return PARSE_ERROR;
 
     switch (fourcc) {

src/demux/libwebpdemux.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,2,0
- PRODUCTVERSION 1,0,2,0
+ FILEVERSION 1,0,2,1
+ PRODUCTVERSION 1,0,2,1
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpdemux DLL"
-            VALUE "FileVersion", "1.2.0"
+            VALUE "FileVersion", "1.2.1"
             VALUE "InternalName", "libwebpdemux.dll"
-            VALUE "LegalCopyright", "Copyright (C) 2019"
+            VALUE "LegalCopyright", "Copyright (C) 2021"
             VALUE "OriginalFilename", "libwebpdemux.dll"
             VALUE "ProductName", "WebP Image Demuxer"
-            VALUE "ProductVersion", "1.2.0"
+            VALUE "ProductVersion", "1.2.1"
         END
     END
     BLOCK "VarFileInfo"

src/dsp/Makefile.am

@@ -46,6 +46,7 @@ ENC_SOURCES += ssim.c
 libwebpdspdecode_sse41_la_SOURCES =
 libwebpdspdecode_sse41_la_SOURCES += alpha_processing_sse41.c
 libwebpdspdecode_sse41_la_SOURCES += dec_sse41.c
+libwebpdspdecode_sse41_la_SOURCES += lossless_sse41.c
 libwebpdspdecode_sse41_la_SOURCES += upsampling_sse41.c
 libwebpdspdecode_sse41_la_SOURCES += yuv_sse41.c
 libwebpdspdecode_sse41_la_CPPFLAGS = $(libwebpdsp_la_CPPFLAGS)

src/dsp/alpha_processing.c

@@ -157,7 +157,8 @@ void WebPMultARGBRow_C(uint32_t* const ptr, int width, int inverse) {
   }
 }
 
-void WebPMultRow_C(uint8_t* const ptr, const uint8_t* const alpha,
+void WebPMultRow_C(uint8_t* WEBP_RESTRICT const ptr,
+                   const uint8_t* WEBP_RESTRICT const alpha,
                    int width, int inverse) {
   int x;
   for (x = 0; x < width; ++x) {
@@ -178,7 +179,8 @@ void WebPMultRow_C(uint8_t* const ptr, const uint8_t* const alpha,
 #undef MFIX
 
 void (*WebPMultARGBRow)(uint32_t* const ptr, int width, int inverse);
-void (*WebPMultRow)(uint8_t* const ptr, const uint8_t* const alpha,
+void (*WebPMultRow)(uint8_t* WEBP_RESTRICT const ptr,
+                    const uint8_t* WEBP_RESTRICT const alpha,
                     int width, int inverse);
 
 //------------------------------------------------------------------------------
@@ -193,8 +195,8 @@ void WebPMultARGBRows(uint8_t* ptr, int stride, int width, int num_rows,
   }
 }
 
-void WebPMultRows(uint8_t* ptr, int stride,
-                  const uint8_t* alpha, int alpha_stride,
+void WebPMultRows(uint8_t* WEBP_RESTRICT ptr, int stride,
+                  const uint8_t* WEBP_RESTRICT alpha, int alpha_stride,
                   int width, int num_rows, int inverse) {
   int n;
   for (n = 0; n < num_rows; ++n) {
@@ -290,9 +292,9 @@ static void ApplyAlphaMultiply_16b_C(uint8_t* rgba4444,
 }
 
 #if !WEBP_NEON_OMIT_C_CODE
-static int DispatchAlpha_C(const uint8_t* alpha, int alpha_stride,
+static int DispatchAlpha_C(const uint8_t* WEBP_RESTRICT alpha, int alpha_stride,
                            int width, int height,
-                           uint8_t* dst, int dst_stride) {
+                           uint8_t* WEBP_RESTRICT dst, int dst_stride) {
   uint32_t alpha_mask = 0xff;
   int i, j;
 
@@ -309,9 +311,10 @@ static int DispatchAlpha_C(const uint8_t* alpha, int alpha_stride,
   return (alpha_mask != 0xff);
 }
 
-static void DispatchAlphaToGreen_C(const uint8_t* alpha, int alpha_stride,
-                                   int width, int height,
-                                   uint32_t* dst, int dst_stride) {
+static void DispatchAlphaToGreen_C(const uint8_t* WEBP_RESTRICT alpha,
+                                   int alpha_stride, int width, int height,
+                                   uint32_t* WEBP_RESTRICT dst,
+                                   int dst_stride) {
   int i, j;
   for (j = 0; j < height; ++j) {
     for (i = 0; i < width; ++i) {
@@ -322,9 +325,9 @@ static void DispatchAlphaToGreen_C(const uint8_t* alpha, int alpha_stride,
   }
 }
 
-static int ExtractAlpha_C(const uint8_t* argb, int argb_stride,
+static int ExtractAlpha_C(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
                           int width, int height,
-                          uint8_t* alpha, int alpha_stride) {
+                          uint8_t* WEBP_RESTRICT alpha, int alpha_stride) {
   uint8_t alpha_mask = 0xff;
   int i, j;
 
@@ -340,7 +343,8 @@ static int ExtractAlpha_C(const uint8_t* argb, int argb_stride,
   return (alpha_mask == 0xff);
 }
 
-static void ExtractGreen_C(const uint32_t* argb, uint8_t* alpha, int size) {
+static void ExtractGreen_C(const uint32_t* WEBP_RESTRICT argb,
+                           uint8_t* WEBP_RESTRICT alpha, int size) {
   int i;
   for (i = 0; i < size; ++i) alpha[i] = argb[i] >> 8;
 }
@@ -372,8 +376,11 @@ static WEBP_INLINE uint32_t MakeARGB32(int a, int r, int g, int b) {
 }
 
 #ifdef WORDS_BIGENDIAN
-static void PackARGB_C(const uint8_t* a, const uint8_t* r, const uint8_t* g,
-                       const uint8_t* b, int len, uint32_t* out) {
+static void PackARGB_C(const uint8_t* WEBP_RESTRICT a,
+                       const uint8_t* WEBP_RESTRICT r,
+                       const uint8_t* WEBP_RESTRICT g,
+                       const uint8_t* WEBP_RESTRICT b,
+                       int len, uint32_t* WEBP_RESTRICT out) {
   int i;
   for (i = 0; i < len; ++i) {
     out[i] = MakeARGB32(a[4 * i], r[4 * i], g[4 * i], b[4 * i]);
@@ -381,8 +388,10 @@ static void PackARGB_C(const uint8_t* a, const uint8_t* r, const uint8_t* g,
 }
 #endif
 
-static void PackRGB_C(const uint8_t* r, const uint8_t* g, const uint8_t* b,
-                      int len, int step, uint32_t* out) {
+static void PackRGB_C(const uint8_t* WEBP_RESTRICT r,
+                      const uint8_t* WEBP_RESTRICT g,
+                      const uint8_t* WEBP_RESTRICT b,
+                      int len, int step, uint32_t* WEBP_RESTRICT out) {
   int i, offset = 0;
   for (i = 0; i < len; ++i) {
     out[i] = MakeARGB32(0xff, r[offset], g[offset], b[offset]);
@@ -392,16 +401,22 @@ static void PackRGB_C(const uint8_t* r, const uint8_t* g, const uint8_t* b,
 
 void (*WebPApplyAlphaMultiply)(uint8_t*, int, int, int, int);
 void (*WebPApplyAlphaMultiply4444)(uint8_t*, int, int, int);
-int (*WebPDispatchAlpha)(const uint8_t*, int, int, int, uint8_t*, int);
-void (*WebPDispatchAlphaToGreen)(const uint8_t*, int, int, int, uint32_t*, int);
-int (*WebPExtractAlpha)(const uint8_t*, int, int, int, uint8_t*, int);
-void (*WebPExtractGreen)(const uint32_t* argb, uint8_t* alpha, int size);
+int (*WebPDispatchAlpha)(const uint8_t* WEBP_RESTRICT, int, int, int,
+                         uint8_t* WEBP_RESTRICT, int);
+void (*WebPDispatchAlphaToGreen)(const uint8_t* WEBP_RESTRICT, int, int, int,
+                                 uint32_t* WEBP_RESTRICT, int);
+int (*WebPExtractAlpha)(const uint8_t* WEBP_RESTRICT, int, int, int,
+                        uint8_t* WEBP_RESTRICT, int);
+void (*WebPExtractGreen)(const uint32_t* WEBP_RESTRICT argb,
+                         uint8_t* WEBP_RESTRICT alpha, int size);
 #ifdef WORDS_BIGENDIAN
 void (*WebPPackARGB)(const uint8_t* a, const uint8_t* r, const uint8_t* g,
                      const uint8_t* b, int, uint32_t*);
 #endif
-void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
-                    int len, int step, uint32_t* out);
+void (*WebPPackRGB)(const uint8_t* WEBP_RESTRICT r,
+                    const uint8_t* WEBP_RESTRICT g,
+                    const uint8_t* WEBP_RESTRICT b,
+                    int len, int step, uint32_t* WEBP_RESTRICT out);
 
 int (*WebPHasAlpha8b)(const uint8_t* src, int length);
 int (*WebPHasAlpha32b)(const uint8_t* src, int length);
@@ -438,10 +453,10 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitAlphaProcessingSSE2();
-#if defined(WEBP_USE_SSE41)
+#if defined(WEBP_HAVE_SSE41)
       if (VP8GetCPUInfo(kSSE4_1)) {
         WebPInitAlphaProcessingSSE41();
       }
@@ -455,7 +470,7 @@ WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     WebPInitAlphaProcessingNEON();

src/dsp/alpha_processing_neon.c

@@ -80,9 +80,9 @@ static void ApplyAlphaMultiply_NEON(uint8_t* rgba, int alpha_first,
 
 //------------------------------------------------------------------------------
 
-static int DispatchAlpha_NEON(const uint8_t* alpha, int alpha_stride,
-                              int width, int height,
-                              uint8_t* dst, int dst_stride) {
+static int DispatchAlpha_NEON(const uint8_t* WEBP_RESTRICT alpha,
+                              int alpha_stride, int width, int height,
+                              uint8_t* WEBP_RESTRICT dst, int dst_stride) {
   uint32_t alpha_mask = 0xffffffffu;
   uint8x8_t mask8 = vdup_n_u8(0xff);
   uint32_t tmp[2];
@@ -112,9 +112,10 @@ static int DispatchAlpha_NEON(const uint8_t* alpha, int alpha_stride,
   return (alpha_mask != 0xffffffffu);
 }
 
-static void DispatchAlphaToGreen_NEON(const uint8_t* alpha, int alpha_stride,
-                                      int width, int height,
-                                      uint32_t* dst, int dst_stride) {
+static void DispatchAlphaToGreen_NEON(const uint8_t* WEBP_RESTRICT alpha,
+                                      int alpha_stride, int width, int height,
+                                      uint32_t* WEBP_RESTRICT dst,
+                                      int dst_stride) {
   int i, j;
   uint8x8x4_t greens;   // leave A/R/B channels zero'd.
   greens.val[0] = vdup_n_u8(0);
@@ -131,9 +132,9 @@ static void DispatchAlphaToGreen_NEON(const uint8_t* alpha, int alpha_stride,
   }
 }
 
-static int ExtractAlpha_NEON(const uint8_t* argb, int argb_stride,
+static int ExtractAlpha_NEON(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
                              int width, int height,
-                             uint8_t* alpha, int alpha_stride) {
+                             uint8_t* WEBP_RESTRICT alpha, int alpha_stride) {
   uint32_t alpha_mask = 0xffffffffu;
   uint8x8_t mask8 = vdup_n_u8(0xff);
   uint32_t tmp[2];
@@ -161,8 +162,8 @@ static int ExtractAlpha_NEON(const uint8_t* argb, int argb_stride,
   return (alpha_mask == 0xffffffffu);
 }
 
-static void ExtractGreen_NEON(const uint32_t* argb,
-                              uint8_t* alpha, int size) {
+static void ExtractGreen_NEON(const uint32_t* WEBP_RESTRICT argb,
+                              uint8_t* WEBP_RESTRICT alpha, int size) {
   int i;
   for (i = 0; i + 16 <= size; i += 16) {
     const uint8x16x4_t rgbX = vld4q_u8((const uint8_t*)(argb + i));

src/dsp/alpha_processing_sse2.c

@@ -18,9 +18,9 @@
 
 //------------------------------------------------------------------------------
 
-static int DispatchAlpha_SSE2(const uint8_t* alpha, int alpha_stride,
-                              int width, int height,
-                              uint8_t* dst, int dst_stride) {
+static int DispatchAlpha_SSE2(const uint8_t* WEBP_RESTRICT alpha,
+                              int alpha_stride, int width, int height,
+                              uint8_t* WEBP_RESTRICT dst, int dst_stride) {
   // alpha_and stores an 'and' operation of all the alpha[] values. The final
   // value is not 0xff if any of the alpha[] is not equal to 0xff.
   uint32_t alpha_and = 0xff;
@@ -72,9 +72,10 @@ static int DispatchAlpha_SSE2(const uint8_t* alpha, int alpha_stride,
   return (alpha_and != 0xff);
 }
 
-static void DispatchAlphaToGreen_SSE2(const uint8_t* alpha, int alpha_stride,
-                                      int width, int height,
-                                      uint32_t* dst, int dst_stride) {
+static void DispatchAlphaToGreen_SSE2(const uint8_t* WEBP_RESTRICT alpha,
+                                      int alpha_stride, int width, int height,
+                                      uint32_t* WEBP_RESTRICT dst,
+                                      int dst_stride) {
   int i, j;
   const __m128i zero = _mm_setzero_si128();
   const int limit = width & ~15;
@@ -98,9 +99,9 @@ static void DispatchAlphaToGreen_SSE2(const uint8_t* alpha, int alpha_stride,
   }
 }
 
-static int ExtractAlpha_SSE2(const uint8_t* argb, int argb_stride,
+static int ExtractAlpha_SSE2(const uint8_t* WEBP_RESTRICT argb, int argb_stride,
                              int width, int height,
-                             uint8_t* alpha, int alpha_stride) {
+                             uint8_t* WEBP_RESTRICT alpha, int alpha_stride) {
   // alpha_and stores an 'and' operation of all the alpha[] values. The final
   // value is not 0xff if any of the alpha[] is not equal to 0xff.
   uint32_t alpha_and = 0xff;
@@ -317,7 +318,8 @@ static void MultARGBRow_SSE2(uint32_t* const ptr, int width, int inverse) {
   if (width > 0) WebPMultARGBRow_C(ptr + x, width, inverse);
 }
 
-static void MultRow_SSE2(uint8_t* const ptr, const uint8_t* const alpha,
+static void MultRow_SSE2(uint8_t* WEBP_RESTRICT const ptr,
+                         const uint8_t* WEBP_RESTRICT const alpha,
                          int width, int inverse) {
   int x = 0;
   if (!inverse) {

src/dsp/alpha_processing_sse41.c

@@ -19,9 +19,9 @@
 
 //------------------------------------------------------------------------------
 
-static int ExtractAlpha_SSE41(const uint8_t* argb, int argb_stride,
-                              int width, int height,
-                              uint8_t* alpha, int alpha_stride) {
+static int ExtractAlpha_SSE41(const uint8_t* WEBP_RESTRICT argb,
+                              int argb_stride, int width, int height,
+                              uint8_t* WEBP_RESTRICT alpha, int alpha_stride) {
   // alpha_and stores an 'and' operation of all the alpha[] values. The final
   // value is not 0xff if any of the alpha[] is not equal to 0xff.
   uint32_t alpha_and = 0xff;

src/dsp/cost.c

@@ -395,12 +395,12 @@ WEBP_DSP_INIT_FUNC(VP8EncDspCostInit) {
       VP8EncDspCostInitMIPSdspR2();
     }
 #endif
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8EncDspCostInitSSE2();
     }
 #endif
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
     if (VP8GetCPUInfo(kNEON)) {
       VP8EncDspCostInitNEON();
     }

src/dsp/cpu.c

@@ -189,17 +189,17 @@ VP8CPUInfo VP8GetCPUInfo = AndroidCPUInfo;
 // 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
+#ifdef WEBP_HAVE_SSE2
     case kSSE2:
       return 1;
 #endif
-#ifdef WEBP_USE_SSE41
+#ifdef WEBP_HAVE_SSE41
     case kSSE3:
     case kSlowSSSE3:
     case kSSE4_1:
       return 1;
 #endif
-#ifdef WEBP_USE_NEON
+#ifdef WEBP_HAVE_NEON
     case kNEON:
       return 1;
 #endif
@@ -209,9 +209,10 @@ static int wasmCPUInfo(CPUFeature feature) {
   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
+#elif defined(WEBP_HAVE_NEON)
+// In most cases this function doesn't check for NEON support (it's assumed by
+// the configuration), but enables turning off NEON at runtime, for testing
+// purposes, by setting VP8DecGetCPUInfo = NULL.
 static int armCPUInfo(CPUFeature feature) {
   if (feature != kNEON) return 0;
 #if defined(__linux__) && defined(WEBP_HAVE_NEON_RTCD)

src/dsp/dec.c

@@ -807,10 +807,10 @@ WEBP_DSP_INIT_FUNC(VP8DspInit) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8DspInitSSE2();
-#if defined(WEBP_USE_SSE41)
+#if defined(WEBP_HAVE_SSE41)
       if (VP8GetCPUInfo(kSSE4_1)) {
         VP8DspInitSSE41();
       }
@@ -834,7 +834,7 @@ WEBP_DSP_INIT_FUNC(VP8DspInit) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     VP8DspInitNEON();

src/dsp/dsp.h

@@ -26,6 +26,23 @@ extern "C" {
 
 #define BPS 32   // this is the common stride for enc/dec
 
+//------------------------------------------------------------------------------
+// WEBP_RESTRICT
+
+// Declares a pointer with the restrict type qualifier if available.
+// This allows code to hint to the compiler that only this pointer references a
+// particular object or memory region within the scope of the block in which it
+// is declared. This may allow for improved optimizations due to the lack of
+// pointer aliasing. See also:
+// https://en.cppreference.com/w/c/language/restrict
+#if defined(__GNUC__)
+#define WEBP_RESTRICT __restrict__
+#elif defined(_MSC_VER)
+#define WEBP_RESTRICT __restrict
+#else
+#define WEBP_RESTRICT
+#endif
+
 //------------------------------------------------------------------------------
 // CPU detection
 
@@ -67,21 +84,31 @@ extern "C" {
 // files without intrinsics, allowing the corresponding Init() to be called.
 // Files containing intrinsics will need to be built targeting the instruction
 // set so should succeed on one of the earlier tests.
-#if defined(__SSE2__) || defined(WEBP_MSC_SSE2) || defined(WEBP_HAVE_SSE2)
+#if (defined(__SSE2__) || defined(WEBP_MSC_SSE2)) && \
+    (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_SSE2))
 #define WEBP_USE_SSE2
 #endif
 
-#if defined(__SSE4_1__) || defined(WEBP_MSC_SSE41) || defined(WEBP_HAVE_SSE41)
+#if defined(WEBP_USE_SSE2) && !defined(WEBP_HAVE_SSE2)
+#define WEBP_HAVE_SSE2
+#endif
+
+#if (defined(__SSE4_1__) || defined(WEBP_MSC_SSE41)) && \
+    (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_SSE41))
 #define WEBP_USE_SSE41
 #endif
 
+#if defined(WEBP_USE_SSE41) && !defined(WEBP_HAVE_SSE41)
+#define WEBP_HAVE_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__) || \
-     defined(__aarch64__) || defined(WEBP_HAVE_NEON)) && \
+#if ((defined(__ARM_NEON__) || defined(__aarch64__)) && \
+     (!defined(HAVE_CONFIG_H) || defined(WEBP_HAVE_NEON))) && \
     !defined(__native_client__)
 #define WEBP_USE_NEON
 #endif
@@ -97,6 +124,10 @@ extern "C" {
 #define WEBP_USE_INTRINSICS
 #endif
 
+#if defined(WEBP_USE_NEON) && !defined(WEBP_HAVE_NEON)
+#define WEBP_HAVE_NEON
+#endif
+
 #if defined(__mips__) && !defined(__mips64) && \
     defined(__mips_isa_rev) && (__mips_isa_rev >= 1) && (__mips_isa_rev < 6)
 #define WEBP_USE_MIPS32
@@ -116,7 +147,7 @@ extern "C" {
 #define WEBP_DSP_OMIT_C_CODE 1
 #endif
 
-#if (defined(__aarch64__) || defined(__ARM_NEON__)) && WEBP_DSP_OMIT_C_CODE
+#if defined(WEBP_USE_NEON) && WEBP_DSP_OMIT_C_CODE
 #define WEBP_NEON_OMIT_C_CODE 1
 #else
 #define WEBP_NEON_OMIT_C_CODE 0
@@ -578,26 +609,29 @@ extern void (*WebPApplyAlphaMultiply4444)(
 
 // Dispatch the values from alpha[] plane to the ARGB destination 'dst'.
 // Returns true if alpha[] plane has non-trivial values different from 0xff.
-extern int (*WebPDispatchAlpha)(const uint8_t* alpha, int alpha_stride,
-                                int width, int height,
-                                uint8_t* dst, int dst_stride);
+extern int (*WebPDispatchAlpha)(const uint8_t* WEBP_RESTRICT alpha,
+                                int alpha_stride, int width, int height,
+                                uint8_t* WEBP_RESTRICT dst, int dst_stride);
 
 // Transfer packed 8b alpha[] values to green channel in dst[], zero'ing the
 // A/R/B values. 'dst_stride' is the stride for dst[] in uint32_t units.
-extern void (*WebPDispatchAlphaToGreen)(const uint8_t* alpha, int alpha_stride,
-                                        int width, int height,
-                                        uint32_t* dst, int dst_stride);
+extern void (*WebPDispatchAlphaToGreen)(const uint8_t* WEBP_RESTRICT alpha,
+                                        int alpha_stride, int width, int height,
+                                        uint32_t* WEBP_RESTRICT dst,
+                                        int dst_stride);
 
 // Extract the alpha values from 32b values in argb[] and pack them into alpha[]
 // (this is the opposite of WebPDispatchAlpha).
 // Returns true if there's only trivial 0xff alpha values.
-extern int (*WebPExtractAlpha)(const uint8_t* argb, int argb_stride,
-                               int width, int height,
-                               uint8_t* alpha, int alpha_stride);
+extern int (*WebPExtractAlpha)(const uint8_t* WEBP_RESTRICT argb,
+                               int argb_stride, int width, int height,
+                               uint8_t* WEBP_RESTRICT alpha,
+                               int alpha_stride);
 
 // Extract the green values from 32b values in argb[] and pack them into alpha[]
 // (this is the opposite of WebPDispatchAlphaToGreen).
-extern void (*WebPExtractGreen)(const uint32_t* argb, uint8_t* alpha, int size);
+extern void (*WebPExtractGreen)(const uint32_t* WEBP_RESTRICT argb,
+                                uint8_t* WEBP_RESTRICT alpha, int size);
 
 // Pre-Multiply operation transforms x into x * A / 255  (where x=Y,R,G or B).
 // Un-Multiply operation transforms x into x * 255 / A.
@@ -610,29 +644,35 @@ void WebPMultARGBRows(uint8_t* ptr, int stride, int width, int num_rows,
                       int inverse);
 
 // Same for a row of single values, with side alpha values.
-extern void (*WebPMultRow)(uint8_t* const ptr, const uint8_t* const alpha,
+extern void (*WebPMultRow)(uint8_t* WEBP_RESTRICT const ptr,
+                           const uint8_t* WEBP_RESTRICT const alpha,
                            int width, int inverse);
 
 // Same a WebPMultRow(), but for several 'num_rows' rows.
-void WebPMultRows(uint8_t* ptr, int stride,
-                  const uint8_t* alpha, int alpha_stride,
+void WebPMultRows(uint8_t* WEBP_RESTRICT ptr, int stride,
+                  const uint8_t* WEBP_RESTRICT alpha, int alpha_stride,
                   int width, int num_rows, int inverse);
 
 // Plain-C versions, used as fallback by some implementations.
-void WebPMultRow_C(uint8_t* const ptr, const uint8_t* const alpha,
+void WebPMultRow_C(uint8_t* WEBP_RESTRICT const ptr,
+                   const uint8_t* WEBP_RESTRICT const alpha,
                    int width, int inverse);
 void WebPMultARGBRow_C(uint32_t* const ptr, int width, int inverse);
 
 #ifdef WORDS_BIGENDIAN
 // ARGB packing function: a/r/g/b input is rgba or bgra order.
-extern void (*WebPPackARGB)(const uint8_t* a, const uint8_t* r,
-                            const uint8_t* g, const uint8_t* b, int len,
-                            uint32_t* out);
+extern void (*WebPPackARGB)(const uint8_t* WEBP_RESTRICT a,
+                            const uint8_t* WEBP_RESTRICT r,
+                            const uint8_t* WEBP_RESTRICT g,
+                            const uint8_t* WEBP_RESTRICT b,
+                            int len, uint32_t* WEBP_RESTRICT out);
 #endif
 
 // RGB packing function. 'step' can be 3 or 4. r/g/b input is rgb or bgr order.
-extern void (*WebPPackRGB)(const uint8_t* r, const uint8_t* g, const uint8_t* b,
-                           int len, int step, uint32_t* out);
+extern void (*WebPPackRGB)(const uint8_t* WEBP_RESTRICT r,
+                           const uint8_t* WEBP_RESTRICT g,
+                           const uint8_t* WEBP_RESTRICT b,
+                           int len, int step, uint32_t* WEBP_RESTRICT out);
 
 // This function returns true if src[i] contains a value different from 0xff.
 extern int (*WebPHasAlpha8b)(const uint8_t* src, int length);

src/dsp/enc.c

@@ -773,10 +773,10 @@ WEBP_DSP_INIT_FUNC(VP8EncDspInit) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8EncDspInitSSE2();
-#if defined(WEBP_USE_SSE41)
+#if defined(WEBP_HAVE_SSE41)
       if (VP8GetCPUInfo(kSSE4_1)) {
         VP8EncDspInitSSE41();
       }
@@ -800,7 +800,7 @@ WEBP_DSP_INIT_FUNC(VP8EncDspInit) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     VP8EncDspInitNEON();

src/dsp/filters.c

@@ -254,7 +254,7 @@ WEBP_DSP_INIT_FUNC(VP8FiltersInit) {
 #endif
 
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8FiltersInitSSE2();
     }
@@ -271,7 +271,7 @@ WEBP_DSP_INIT_FUNC(VP8FiltersInit) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     VP8FiltersInitNEON();

src/dsp/filters_sse2.c

@@ -320,7 +320,12 @@ extern void VP8FiltersInitSSE2(void);
 
 WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInitSSE2(void) {
   WebPUnfilters[WEBP_FILTER_HORIZONTAL] = HorizontalUnfilter_SSE2;
+#if defined(CHROMIUM)
+  // TODO(crbug.com/654974)
+  (void)VerticalUnfilter_SSE2;
+#else
   WebPUnfilters[WEBP_FILTER_VERTICAL] = VerticalUnfilter_SSE2;
+#endif
   WebPUnfilters[WEBP_FILTER_GRADIENT] = GradientUnfilter_SSE2;
 
   WebPFilters[WEBP_FILTER_HORIZONTAL] = HorizontalFilter_SSE2;

src/dsp/lossless.c

@@ -575,6 +575,7 @@ VP8LMapARGBFunc VP8LMapColor32b;
 VP8LMapAlphaFunc VP8LMapColor8b;
 
 extern void VP8LDspInitSSE2(void);
+extern void VP8LDspInitSSE41(void);
 extern void VP8LDspInitNEON(void);
 extern void VP8LDspInitMIPSdspR2(void);
 extern void VP8LDspInitMSA(void);
@@ -621,9 +622,14 @@ WEBP_DSP_INIT_FUNC(VP8LDspInit) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8LDspInitSSE2();
+#if defined(WEBP_HAVE_SSE41)
+      if (VP8GetCPUInfo(kSSE4_1)) {
+        VP8LDspInitSSE41();
+      }
+#endif
     }
 #endif
 #if defined(WEBP_USE_MIPS_DSP_R2)
@@ -638,7 +644,7 @@ WEBP_DSP_INIT_FUNC(VP8LDspInit) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     VP8LDspInitNEON();

src/dsp/lossless_enc.c

@@ -329,6 +329,15 @@ const uint8_t kPrefixEncodeExtraBitsValue[PREFIX_LOOKUP_IDX_MAX] = {
 static float FastSLog2Slow_C(uint32_t v) {
   assert(v >= LOG_LOOKUP_IDX_MAX);
   if (v < APPROX_LOG_WITH_CORRECTION_MAX) {
+#if !defined(WEBP_HAVE_SLOW_CLZ_CTZ)
+    // use clz if available
+    const int log_cnt = BitsLog2Floor(v) - 7;
+    const uint32_t y = 1 << log_cnt;
+    int correction = 0;
+    const float v_f = (float)v;
+    const uint32_t orig_v = v;
+    v >>= log_cnt;
+#else
     int log_cnt = 0;
     uint32_t y = 1;
     int correction = 0;
@@ -339,6 +348,7 @@ static float FastSLog2Slow_C(uint32_t v) {
       v = v >> 1;
       y = y << 1;
     } while (v >= LOG_LOOKUP_IDX_MAX);
+#endif
     // vf = (2^log_cnt) * Xf; where y = 2^log_cnt and Xf < 256
     // Xf = floor(Xf) * (1 + (v % y) / v)
     // log2(Xf) = log2(floor(Xf)) + log2(1 + (v % y) / v)
@@ -355,6 +365,14 @@ static float FastSLog2Slow_C(uint32_t v) {
 static float FastLog2Slow_C(uint32_t v) {
   assert(v >= LOG_LOOKUP_IDX_MAX);
   if (v < APPROX_LOG_WITH_CORRECTION_MAX) {
+#if !defined(WEBP_HAVE_SLOW_CLZ_CTZ)
+    // use clz if available
+    const int log_cnt = BitsLog2Floor(v) - 7;
+    const uint32_t y = 1 << log_cnt;
+    const uint32_t orig_v = v;
+    double log_2;
+    v >>= log_cnt;
+#else
     int log_cnt = 0;
     uint32_t y = 1;
     const uint32_t orig_v = v;
@@ -364,6 +382,7 @@ static float FastLog2Slow_C(uint32_t v) {
       v = v >> 1;
       y = y << 1;
     } while (v >= LOG_LOOKUP_IDX_MAX);
+#endif
     log_2 = kLog2Table[v] + log_cnt;
     if (orig_v >= APPROX_LOG_MAX) {
       // Since the division is still expensive, add this correction factor only
@@ -843,10 +862,10 @@ WEBP_DSP_INIT_FUNC(VP8LEncDspInit) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8LEncDspInitSSE2();
-#if defined(WEBP_USE_SSE41)
+#if defined(WEBP_HAVE_SSE41)
       if (VP8GetCPUInfo(kSSE4_1)) {
         VP8LEncDspInitSSE41();
       }
@@ -870,7 +889,7 @@ WEBP_DSP_INIT_FUNC(VP8LEncDspInit) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     VP8LEncDspInitNEON();

src/dsp/lossless_enc_sse2.c

@@ -232,79 +232,55 @@ static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) {
 //------------------------------------------------------------------------------
 // Entropy
 
-// Checks whether the X or Y contribution is worth computing and adding.
-// Used in loop unrolling.
-#define ANALYZE_X_OR_Y(x_or_y, j)                                           \
-  do {                                                                      \
-    if ((x_or_y)[i + (j)] != 0) retval -= VP8LFastSLog2((x_or_y)[i + (j)]); \
-  } while (0)
+// 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(WEBP_HAVE_SLOW_CLZ_CTZ) || defined(__i386__) || defined(_M_IX86))
 
-// Checks whether the X + Y contribution is worth computing and adding.
-// Used in loop unrolling.
-#define ANALYZE_XY(j)                  \
-  do {                                 \
-    if (tmp[j] != 0) {                 \
-      retval -= VP8LFastSLog2(tmp[j]); \
-      ANALYZE_X_OR_Y(X, j);            \
-    }                                  \
-  } 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.;
-  int sumX, sumXY;
-  int32_t tmp[4];
-  __m128i zero = _mm_setzero_si128();
-  // Sums up X + Y, 4 ints at a time (and will merge it at the end for sumXY).
-  __m128i sumXY_128 = zero;
-  __m128i sumX_128 = zero;
+  int sumX = 0, sumXY = 0;
+  const __m128i zero = _mm_setzero_si128();
 
-  for (i = 0; i < 256; i += 4) {
-    const __m128i x = _mm_loadu_si128((const __m128i*)(X + i));
-    const __m128i y = _mm_loadu_si128((const __m128i*)(Y + i));
-
-    // Check if any X is non-zero: this actually provides a speedup as X is
-    // usually sparse.
-    if (_mm_movemask_epi8(_mm_cmpeq_epi32(x, zero)) != 0xFFFF) {
-      const __m128i xy_128 = _mm_add_epi32(x, y);
-      sumXY_128 = _mm_add_epi32(sumXY_128, xy_128);
-
-      sumX_128 = _mm_add_epi32(sumX_128, x);
-
-      // Analyze the different X + Y.
-      _mm_storeu_si128((__m128i*)tmp, xy_128);
-
-      ANALYZE_XY(0);
-      ANALYZE_XY(1);
-      ANALYZE_XY(2);
-      ANALYZE_XY(3);
-    } else {
-      // X is fully 0, so only deal with Y.
-      sumXY_128 = _mm_add_epi32(sumXY_128, y);
-
-      ANALYZE_X_OR_Y(Y, 0);
-      ANALYZE_X_OR_Y(Y, 1);
-      ANALYZE_X_OR_Y(Y, 2);
-      ANALYZE_X_OR_Y(Y, 3);
+  for (i = 0; i < 256; i += 16) {
+    const __m128i x0 = _mm_loadu_si128((const __m128i*)(X + i +  0));
+    const __m128i y0 = _mm_loadu_si128((const __m128i*)(Y + i +  0));
+    const __m128i x1 = _mm_loadu_si128((const __m128i*)(X + i +  4));
+    const __m128i y1 = _mm_loadu_si128((const __m128i*)(Y + i +  4));
+    const __m128i x2 = _mm_loadu_si128((const __m128i*)(X + i +  8));
+    const __m128i y2 = _mm_loadu_si128((const __m128i*)(Y + i +  8));
+    const __m128i x3 = _mm_loadu_si128((const __m128i*)(X + i + 12));
+    const __m128i y3 = _mm_loadu_si128((const __m128i*)(Y + i + 12));
+    const __m128i x4 = _mm_packs_epi16(_mm_packs_epi32(x0, x1),
+                                       _mm_packs_epi32(x2, x3));
+    const __m128i y4 = _mm_packs_epi16(_mm_packs_epi32(y0, y1),
+                                       _mm_packs_epi32(y2, y3));
+    const int32_t mx = _mm_movemask_epi8(_mm_cmpgt_epi8(x4, zero));
+    int32_t my = _mm_movemask_epi8(_mm_cmpgt_epi8(y4, zero)) | mx;
+    while (my) {
+      const int32_t j = BitsCtz(my);
+      int xy;
+      if ((mx >> j) & 1) {
+        const int x = X[i + j];
+        sumXY += x;
+        retval -= VP8LFastSLog2(x);
+      }
+      xy = X[i + j] + Y[i + j];
+      sumX += xy;
+      retval -= VP8LFastSLog2(xy);
+      my &= my - 1;
     }
   }
-
-  // Sum up sumX_128 to get sumX.
-  _mm_storeu_si128((__m128i*)tmp, sumX_128);
-  sumX = tmp[3] + tmp[2] + tmp[1] + tmp[0];
-
-  // Sum up sumXY_128 to get sumXY.
-  _mm_storeu_si128((__m128i*)tmp, sumXY_128);
-  sumXY = tmp[3] + tmp[2] + tmp[1] + tmp[0];
-
   retval += VP8LFastSLog2(sumX) + VP8LFastSLog2(sumXY);
   return (float)retval;
 }
-#endif  // !(defined(__i386__) || defined(_M_IX86))
 
-#undef ANALYZE_X_OR_Y
-#undef ANALYZE_XY
+#else
+
+#define DONT_USE_COMBINED_SHANNON_ENTROPY_SSE2_FUNC   // won't be faster
+
+#endif
 
 //------------------------------------------------------------------------------
 
@@ -662,10 +638,7 @@ 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))
+#if !defined(DONT_USE_COMBINED_SHANNON_ENTROPY_SSE2_FUNC)
   VP8LCombinedShannonEntropy = CombinedShannonEntropy_SSE2;
 #endif
   VP8LVectorMismatch = VectorMismatch_SSE2;

src/dsp/lossless_enc_sse41.c

@@ -44,46 +44,47 @@ static void SubtractGreenFromBlueAndRed_SSE41(uint32_t* argb_data,
 //------------------------------------------------------------------------------
 // Color Transform
 
-#define SPAN 8
+#define MK_CST_16(HI, LO) \
+  _mm_set1_epi32((int)(((uint32_t)(HI) << 16) | ((LO) & 0xffff)))
+
 static void CollectColorBlueTransforms_SSE41(const uint32_t* argb, int stride,
                                              int tile_width, int tile_height,
                                              int green_to_blue, int red_to_blue,
                                              int histo[]) {
-  const __m128i mults_r = _mm_set1_epi16(CST_5b(red_to_blue));
-  const __m128i mults_g = _mm_set1_epi16(CST_5b(green_to_blue));
-  const __m128i mask_g = _mm_set1_epi16((short)0xff00);   // green mask
-  const __m128i mask_gb = _mm_set1_epi32(0xffff);         // green/blue mask
-  const __m128i mask_b = _mm_set1_epi16(0x00ff);          // blue mask
-  const __m128i shuffler_lo = _mm_setr_epi8(-1, 2, -1, 6, -1, 10, -1, 14, -1,
-                                            -1, -1, -1, -1, -1, -1, -1);
-  const __m128i shuffler_hi = _mm_setr_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1,
-                                            2, -1, 6, -1, 10, -1, 14);
-  int y;
-  for (y = 0; y < tile_height; ++y) {
-    const uint32_t* const src = argb + y * stride;
-    int i, x;
-    for (x = 0; x + SPAN <= tile_width; x += SPAN) {
-      uint16_t values[SPAN];
-      const __m128i in0 = _mm_loadu_si128((__m128i*)&src[x + 0]);
-      const __m128i in1 = _mm_loadu_si128((__m128i*)&src[x + SPAN / 2]);
-      const __m128i r0 = _mm_shuffle_epi8(in0, shuffler_lo);
-      const __m128i r1 = _mm_shuffle_epi8(in1, shuffler_hi);
-      const __m128i r = _mm_or_si128(r0, r1);         // r 0
-      const __m128i gb0 = _mm_and_si128(in0, mask_gb);
-      const __m128i gb1 = _mm_and_si128(in1, mask_gb);
-      const __m128i gb = _mm_packus_epi32(gb0, gb1);  // g b
-      const __m128i g = _mm_and_si128(gb, mask_g);    // g 0
-      const __m128i A = _mm_mulhi_epi16(r, mults_r);  // x dbr
-      const __m128i B = _mm_mulhi_epi16(g, mults_g);  // x dbg
-      const __m128i C = _mm_sub_epi8(gb, B);          // x b'
-      const __m128i D = _mm_sub_epi8(C, A);           // x b''
-      const __m128i E = _mm_and_si128(D, mask_b);     // 0 b''
-      _mm_storeu_si128((__m128i*)values, E);
-      for (i = 0; i < SPAN; ++i) ++histo[values[i]];
+  const __m128i mult =
+      MK_CST_16(CST_5b(red_to_blue) + 256,CST_5b(green_to_blue));
+  const __m128i perm =
+      _mm_setr_epi8(-1, 1, -1, 2, -1, 5, -1, 6, -1, 9, -1, 10, -1, 13, -1, 14);
+  if (tile_width >= 4) {
+    int y;
+    for (y = 0; y < tile_height; ++y) {
+      const uint32_t* const src = argb + y * stride;
+      const __m128i A1 = _mm_loadu_si128((const __m128i*)src);
+      const __m128i B1 = _mm_shuffle_epi8(A1, perm);
+      const __m128i C1 = _mm_mulhi_epi16(B1, mult);
+      const __m128i D1 = _mm_sub_epi16(A1, C1);
+      __m128i E = _mm_add_epi16(_mm_srli_epi32(D1, 16), D1);
+      int x;
+      for (x = 4; x + 4 <= tile_width; x += 4) {
+        const __m128i A2 = _mm_loadu_si128((const __m128i*)(src + x));
+        __m128i B2, C2, D2;
+        ++histo[_mm_extract_epi8(E,  0)];
+        B2 = _mm_shuffle_epi8(A2, perm);
+        ++histo[_mm_extract_epi8(E,  4)];
+        C2 = _mm_mulhi_epi16(B2, mult);
+        ++histo[_mm_extract_epi8(E,  8)];
+        D2 = _mm_sub_epi16(A2, C2);
+        ++histo[_mm_extract_epi8(E, 12)];
+        E = _mm_add_epi16(_mm_srli_epi32(D2, 16), D2);
+      }
+      ++histo[_mm_extract_epi8(E,  0)];
+      ++histo[_mm_extract_epi8(E,  4)];
+      ++histo[_mm_extract_epi8(E,  8)];
+      ++histo[_mm_extract_epi8(E, 12)];
     }
   }
   {
-    const int left_over = tile_width & (SPAN - 1);
+    const int left_over = tile_width & 3;
     if (left_over > 0) {
       VP8LCollectColorBlueTransforms_C(argb + tile_width - left_over, stride,
                                        left_over, tile_height,
@@ -95,33 +96,37 @@ static void CollectColorBlueTransforms_SSE41(const uint32_t* argb, int stride,
 static void CollectColorRedTransforms_SSE41(const uint32_t* argb, int stride,
                                             int tile_width, int tile_height,
                                             int green_to_red, int histo[]) {
-  const __m128i mults_g = _mm_set1_epi16(CST_5b(green_to_red));
-  const __m128i mask_g = _mm_set1_epi32(0x00ff00);  // green mask
-  const __m128i mask = _mm_set1_epi16(0xff);
 
-  int y;
-  for (y = 0; y < tile_height; ++y) {
-    const uint32_t* const src = argb + y * stride;
-    int i, x;
-    for (x = 0; x + SPAN <= tile_width; x += SPAN) {
-      uint16_t values[SPAN];
-      const __m128i in0 = _mm_loadu_si128((__m128i*)&src[x + 0]);
-      const __m128i in1 = _mm_loadu_si128((__m128i*)&src[x + SPAN / 2]);
-      const __m128i g0 = _mm_and_si128(in0, mask_g);  // 0 0  | g 0
-      const __m128i g1 = _mm_and_si128(in1, mask_g);
-      const __m128i g = _mm_packus_epi32(g0, g1);     // g 0
-      const __m128i A0 = _mm_srli_epi32(in0, 16);     // 0 0  | x r
-      const __m128i A1 = _mm_srli_epi32(in1, 16);
-      const __m128i A = _mm_packus_epi32(A0, A1);     // x r
-      const __m128i B = _mm_mulhi_epi16(g, mults_g);  // x dr
-      const __m128i C = _mm_sub_epi8(A, B);           // x r'
-      const __m128i D = _mm_and_si128(C, mask);       // 0 r'
-      _mm_storeu_si128((__m128i*)values, D);
-      for (i = 0; i < SPAN; ++i) ++histo[values[i]];
+  const __m128i mult = MK_CST_16(0, CST_5b(green_to_red));
+  const __m128i mask_g = _mm_set1_epi32(0x0000ff00);
+  if (tile_width >= 4) {
+    int y;
+    for (y = 0; y < tile_height; ++y) {
+      const uint32_t* const src = argb + y * stride;
+      const __m128i A1 = _mm_loadu_si128((const __m128i*)src);
+      const __m128i B1 = _mm_and_si128(A1, mask_g);
+      const __m128i C1 = _mm_madd_epi16(B1, mult);
+      __m128i D = _mm_sub_epi16(A1, C1);
+      int x;
+      for (x = 4; x + 4 <= tile_width; x += 4) {
+        const __m128i A2 = _mm_loadu_si128((const __m128i*)(src + x));
+        __m128i B2, C2;
+        ++histo[_mm_extract_epi8(D,  2)];
+        B2 = _mm_and_si128(A2, mask_g);
+        ++histo[_mm_extract_epi8(D,  6)];
+        C2 = _mm_madd_epi16(B2, mult);
+        ++histo[_mm_extract_epi8(D, 10)];
+        ++histo[_mm_extract_epi8(D, 14)];
+        D = _mm_sub_epi16(A2, C2);
+      }
+      ++histo[_mm_extract_epi8(D,  2)];
+      ++histo[_mm_extract_epi8(D,  6)];
+      ++histo[_mm_extract_epi8(D, 10)];
+      ++histo[_mm_extract_epi8(D, 14)];
     }
   }
   {
-    const int left_over = tile_width & (SPAN - 1);
+    const int left_over = tile_width & 3;
     if (left_over > 0) {
       VP8LCollectColorRedTransforms_C(argb + tile_width - left_over, stride,
                                       left_over, tile_height, green_to_red,
@@ -130,6 +135,8 @@ static void CollectColorRedTransforms_SSE41(const uint32_t* argb, int stride,
   }
 }
 
+#undef MK_CST_16
+
 //------------------------------------------------------------------------------
 // Entry point
 

src/dsp/lossless_sse2.c

@@ -18,7 +18,6 @@
 #include "src/dsp/common_sse2.h"
 #include "src/dsp/lossless.h"
 #include "src/dsp/lossless_common.h"
-#include <assert.h>
 #include <emmintrin.h>
 
 //------------------------------------------------------------------------------

src/dsp/lossless_sse41.c

@@ -0,0 +1,132 @@
+// Copyright 2021 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING 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.
+// -----------------------------------------------------------------------------
+//
+// SSE41 variant of methods for lossless decoder
+
+#include "src/dsp/dsp.h"
+
+#if defined(WEBP_USE_SSE41)
+
+#include "src/dsp/common_sse41.h"
+#include "src/dsp/lossless.h"
+#include "src/dsp/lossless_common.h"
+
+//------------------------------------------------------------------------------
+// Color-space conversion functions
+
+static void TransformColorInverse_SSE41(const VP8LMultipliers* const m,
+                                        const uint32_t* const src,
+                                        int num_pixels, uint32_t* dst) {
+// sign-extended multiplying constants, pre-shifted by 5.
+#define CST(X)  (((int16_t)(m->X << 8)) >> 5)   // sign-extend
+  const __m128i mults_rb = _mm_set1_epi32((uint32_t)CST(green_to_red_) << 16 |
+                                          (CST(green_to_blue_) & 0xffff));
+  const __m128i mults_b2 = _mm_set1_epi32(CST(red_to_blue_));
+#undef CST
+  const __m128i mask_ag = _mm_set1_epi32(0xff00ff00);
+  const __m128i perm1 = _mm_setr_epi8(-1, 1, -1, 1, -1, 5, -1, 5,
+                                      -1, 9, -1, 9, -1, 13, -1, 13);
+  const __m128i perm2 = _mm_setr_epi8(-1, 2, -1, -1, -1, 6, -1, -1,
+                                      -1, 10, -1, -1, -1, 14, -1, -1);
+  int i;
+  for (i = 0; i + 4 <= num_pixels; i += 4) {
+    const __m128i A = _mm_loadu_si128((const __m128i*)(src + i));
+    const __m128i B = _mm_shuffle_epi8(A, perm1); // argb -> g0g0
+    const __m128i C = _mm_mulhi_epi16(B, mults_rb);
+    const __m128i D = _mm_add_epi8(A, C);
+    const __m128i E = _mm_shuffle_epi8(D, perm2);
+    const __m128i F = _mm_mulhi_epi16(E, mults_b2);
+    const __m128i G = _mm_add_epi8(D, F);
+    const __m128i out = _mm_blendv_epi8(G, A, mask_ag);
+    _mm_storeu_si128((__m128i*)&dst[i], out);
+  }
+  // Fall-back to C-version for left-overs.
+  if (i != num_pixels) {
+    VP8LTransformColorInverse_C(m, src + i, num_pixels - i, dst + i);
+  }
+}
+
+//------------------------------------------------------------------------------
+
+#define ARGB_TO_RGB_SSE41 do {                        \
+  while (num_pixels >= 16) {                          \
+    const __m128i in0 = _mm_loadu_si128(in + 0);      \
+    const __m128i in1 = _mm_loadu_si128(in + 1);      \
+    const __m128i in2 = _mm_loadu_si128(in + 2);      \
+    const __m128i in3 = _mm_loadu_si128(in + 3);      \
+    const __m128i a0 = _mm_shuffle_epi8(in0, perm0);  \
+    const __m128i a1 = _mm_shuffle_epi8(in1, perm1);  \
+    const __m128i a2 = _mm_shuffle_epi8(in2, perm2);  \
+    const __m128i a3 = _mm_shuffle_epi8(in3, perm3);  \
+    const __m128i b0 = _mm_blend_epi16(a0, a1, 0xc0); \
+    const __m128i b1 = _mm_blend_epi16(a1, a2, 0xf0); \
+    const __m128i b2 = _mm_blend_epi16(a2, a3, 0xfc); \
+    _mm_storeu_si128(out + 0, b0);                    \
+    _mm_storeu_si128(out + 1, b1);                    \
+    _mm_storeu_si128(out + 2, b2);                    \
+    in += 4;                                          \
+    out += 3;                                         \
+    num_pixels -= 16;                                 \
+  }                                                   \
+} while (0)
+
+static void ConvertBGRAToRGB_SSE41(const uint32_t* src, int num_pixels,
+                                   uint8_t* dst) {
+  const __m128i* in = (const __m128i*)src;
+  __m128i* out = (__m128i*)dst;
+  const __m128i perm0 = _mm_setr_epi8(2, 1, 0, 6, 5, 4, 10, 9,
+                                      8, 14, 13, 12, -1, -1, -1, -1);
+  const __m128i perm1 = _mm_shuffle_epi32(perm0, 0x39);
+  const __m128i perm2 = _mm_shuffle_epi32(perm0, 0x4e);
+  const __m128i perm3 = _mm_shuffle_epi32(perm0, 0x93);
+
+  ARGB_TO_RGB_SSE41;
+
+  // left-overs
+  if (num_pixels > 0) {
+    VP8LConvertBGRAToRGB_C((const uint32_t*)in, num_pixels, (uint8_t*)out);
+  }
+}
+
+static void ConvertBGRAToBGR_SSE41(const uint32_t* src,
+                                   int num_pixels, uint8_t* dst) {
+  const __m128i* in = (const __m128i*)src;
+  __m128i* out = (__m128i*)dst;
+  const __m128i perm0 = _mm_setr_epi8(0, 1, 2, 4, 5, 6, 8, 9, 10,
+                                      12, 13, 14, -1, -1, -1, -1);
+  const __m128i perm1 = _mm_shuffle_epi32(perm0, 0x39);
+  const __m128i perm2 = _mm_shuffle_epi32(perm0, 0x4e);
+  const __m128i perm3 = _mm_shuffle_epi32(perm0, 0x93);
+
+  ARGB_TO_RGB_SSE41;
+
+  // left-overs
+  if (num_pixels > 0) {
+    VP8LConvertBGRAToBGR_C((const uint32_t*)in, num_pixels, (uint8_t*)out);
+  }
+}
+
+#undef ARGB_TO_RGB_SSE41
+
+//------------------------------------------------------------------------------
+// Entry point
+
+extern void VP8LDspInitSSE41(void);
+
+WEBP_TSAN_IGNORE_FUNCTION void VP8LDspInitSSE41(void) {
+  VP8LTransformColorInverse = TransformColorInverse_SSE41;
+  VP8LConvertBGRAToRGB = ConvertBGRAToRGB_SSE41;
+  VP8LConvertBGRAToBGR = ConvertBGRAToBGR_SSE41;
+}
+
+#else  // !WEBP_USE_SSE41
+
+WEBP_DSP_INIT_STUB(VP8LDspInitSSE41)
+
+#endif  // WEBP_USE_SSE41

src/dsp/rescaler.c

@@ -38,8 +38,9 @@ void WebPRescalerImportRowExpand_C(WebPRescaler* const wrk,
     int x_out = channel;
     // simple bilinear interpolation
     int accum = wrk->x_add;
-    int left = src[x_in];
-    int right = (wrk->src_width > 1) ? src[x_in + x_stride] : left;
+    rescaler_t left = (rescaler_t)src[x_in];
+    rescaler_t right =
+        (wrk->src_width > 1) ? (rescaler_t)src[x_in + x_stride] : left;
     x_in += x_stride;
     while (1) {
       wrk->frow[x_out] = right * wrk->x_add + (left - right) * accum;
@@ -50,7 +51,7 @@ void WebPRescalerImportRowExpand_C(WebPRescaler* const wrk,
         left = right;
         x_in += x_stride;
         assert(x_in < wrk->src_width * x_stride);
-        right = src[x_in];
+        right = (rescaler_t)src[x_in];
         accum += wrk->x_add;
       }
     }
@@ -213,7 +214,7 @@ WEBP_DSP_INIT_FUNC(WebPRescalerDspInit) {
   WebPRescalerImportRowShrink = WebPRescalerImportRowShrink_C;
 
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       WebPRescalerDspInitSSE2();
     }
@@ -235,7 +236,7 @@ WEBP_DSP_INIT_FUNC(WebPRescalerDspInit) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     WebPRescalerDspInitNEON();

src/dsp/ssim.c

@@ -150,7 +150,7 @@ WEBP_DSP_INIT_FUNC(VP8SSIMDspInit) {
 #endif
 
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       VP8SSIMDspInitSSE2();
     }

src/dsp/upsampling.c

@@ -233,12 +233,12 @@ WEBP_DSP_INIT_FUNC(WebPInitYUV444Converters) {
   WebPYUV444Converters[MODE_rgbA_4444] = WebPYuv444ToRgba4444_C;
 
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitYUV444ConvertersSSE2();
     }
 #endif
-#if defined(WEBP_USE_SSE41)
+#if defined(WEBP_HAVE_SSE41)
     if (VP8GetCPUInfo(kSSE4_1)) {
       WebPInitYUV444ConvertersSSE41();
     }
@@ -278,12 +278,12 @@ WEBP_DSP_INIT_FUNC(WebPInitUpsamplers) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitUpsamplersSSE2();
     }
 #endif
-#if defined(WEBP_USE_SSE41)
+#if defined(WEBP_HAVE_SSE41)
     if (VP8GetCPUInfo(kSSE4_1)) {
       WebPInitUpsamplersSSE41();
     }
@@ -300,7 +300,7 @@ WEBP_DSP_INIT_FUNC(WebPInitUpsamplers) {
 #endif
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     WebPInitUpsamplersNEON();

src/dsp/yuv.c

@@ -90,16 +90,16 @@ WEBP_DSP_INIT_FUNC(WebPInitSamplers) {
 
   // If defined, use CPUInfo() to overwrite some pointers with faster versions.
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitSamplersSSE2();
     }
-#endif  // WEBP_USE_SSE2
-#if defined(WEBP_USE_SSE41)
+#endif  // WEBP_HAVE_SSE2
+#if defined(WEBP_HAVE_SSE41)
     if (VP8GetCPUInfo(kSSE4_1)) {
       WebPInitSamplersSSE41();
     }
-#endif  // WEBP_USE_SSE41
+#endif  // WEBP_HAVE_SSE41
 #if defined(WEBP_USE_MIPS32)
     if (VP8GetCPUInfo(kMIPS32)) {
       WebPInitSamplersMIPS32();
@@ -276,26 +276,26 @@ WEBP_DSP_INIT_FUNC(WebPInitConvertARGBToYUV) {
 #endif
 
   if (VP8GetCPUInfo != NULL) {
-#if defined(WEBP_USE_SSE2)
+#if defined(WEBP_HAVE_SSE2)
     if (VP8GetCPUInfo(kSSE2)) {
       WebPInitConvertARGBToYUVSSE2();
       WebPInitSharpYUVSSE2();
     }
-#endif  // WEBP_USE_SSE2
-#if defined(WEBP_USE_SSE41)
+#endif  // WEBP_HAVE_SSE2
+#if defined(WEBP_HAVE_SSE41)
     if (VP8GetCPUInfo(kSSE4_1)) {
       WebPInitConvertARGBToYUVSSE41();
     }
-#endif  // WEBP_USE_SSE41
+#endif  // WEBP_HAVE_SSE41
   }
 
-#if defined(WEBP_USE_NEON)
+#if defined(WEBP_HAVE_NEON)
   if (WEBP_NEON_OMIT_C_CODE ||
       (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
     WebPInitConvertARGBToYUVNEON();
     WebPInitSharpYUVNEON();
   }
-#endif  // WEBP_USE_NEON
+#endif  // WEBP_HAVE_NEON
 
   assert(WebPConvertARGBToY != NULL);
   assert(WebPConvertARGBToUV != NULL);

src/enc/alpha_enc.c

@@ -303,7 +303,7 @@ static int EncodeAlpha(VP8Encoder* const enc,
   int ok = 1;
   const int reduce_levels = (quality < 100);
 
-  // quick sanity checks
+  // quick correctness checks
   assert((uint64_t)data_size == (uint64_t)width * height);  // as per spec
   assert(enc != NULL && pic != NULL && pic->a != NULL);
   assert(output != NULL && output_size != NULL);
@@ -361,7 +361,7 @@ static int EncodeAlpha(VP8Encoder* const enc,
 //------------------------------------------------------------------------------
 // Main calls
 
-static int CompressAlphaJob(void* arg1, void* dummy) {
+static int CompressAlphaJob(void* arg1, void* unused) {
   VP8Encoder* const enc = (VP8Encoder*)arg1;
   const WebPConfig* config = enc->config_;
   uint8_t* alpha_data = NULL;
@@ -375,13 +375,13 @@ static int CompressAlphaJob(void* arg1, void* dummy) {
                    filter, effort_level, &alpha_data, &alpha_size)) {
     return 0;
   }
-  if (alpha_size != (uint32_t)alpha_size) {  // Sanity check.
+  if (alpha_size != (uint32_t)alpha_size) {  // Soundness check.
     WebPSafeFree(alpha_data);
     return 0;
   }
   enc->alpha_data_size_ = (uint32_t)alpha_size;
   enc->alpha_data_ = alpha_data;
-  (void)dummy;
+  (void)unused;
   return 1;
 }
 

src/enc/histogram_enc.c

@@ -1171,13 +1171,15 @@ static void RemoveEmptyHistograms(VP8LHistogramSet* const image_histo) {
 int VP8LGetHistoImageSymbols(int xsize, int ysize,
                              const VP8LBackwardRefs* const refs,
                              int quality, int low_effort,
-                             int histo_bits, int cache_bits,
+                             int histogram_bits, int cache_bits,
                              VP8LHistogramSet* const image_histo,
                              VP8LHistogram* const tmp_histo,
                              uint16_t* const histogram_symbols) {
   int ok = 0;
-  const int histo_xsize = histo_bits ? VP8LSubSampleSize(xsize, histo_bits) : 1;
-  const int histo_ysize = histo_bits ? VP8LSubSampleSize(ysize, histo_bits) : 1;
+  const int histo_xsize =
+      histogram_bits ? VP8LSubSampleSize(xsize, histogram_bits) : 1;
+  const int histo_ysize =
+      histogram_bits ? VP8LSubSampleSize(ysize, histogram_bits) : 1;
   const int image_histo_raw_size = histo_xsize * histo_ysize;
   VP8LHistogramSet* const orig_histo =
       VP8LAllocateHistogramSet(image_histo_raw_size, cache_bits);
@@ -1193,7 +1195,7 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize,
   if (orig_histo == NULL || map_tmp == NULL) goto Error;
 
   // Construct the histograms from backward references.
-  HistogramBuild(xsize, histo_bits, refs, orig_histo);
+  HistogramBuild(xsize, histogram_bits, refs, orig_histo);
   // Copies the histograms and computes its bit_cost.
   // histogram_symbols is optimized
   HistogramCopyAndAnalyze(orig_histo, image_histo, &num_used,

src/enc/histogram_enc.h

@@ -64,8 +64,8 @@ void VP8LHistogramCreate(VP8LHistogram* const p,
                          const VP8LBackwardRefs* const refs,
                          int palette_code_bits);
 
-// Return the size of the histogram for a given palette_code_bits.
-int VP8LGetHistogramSize(int palette_code_bits);
+// Return the size of the histogram for a given cache_bits.
+int VP8LGetHistogramSize(int cache_bits);
 
 // Set the palette_code_bits and reset the stats.
 // If init_arrays is true, the arrays are also filled with 0's.
@@ -110,7 +110,7 @@ int VP8LGetHistoImageSymbols(int xsize, int ysize,
                              const VP8LBackwardRefs* const refs,
                              int quality, int low_effort,
                              int histogram_bits, int cache_bits,
-                             VP8LHistogramSet* const image_in,
+                             VP8LHistogramSet* const image_histo,
                              VP8LHistogram* const tmp_histo,
                              uint16_t* const histogram_symbols);
 

src/enc/picture_rescale_enc.c

@@ -164,22 +164,25 @@ int WebPPictureCrop(WebPPicture* pic,
 //------------------------------------------------------------------------------
 // Simple picture rescaler
 
-static void RescalePlane(const uint8_t* src,
-                         int src_width, int src_height, int src_stride,
-                         uint8_t* dst,
-                         int dst_width, int dst_height, int dst_stride,
-                         rescaler_t* const work,
-                         int num_channels) {
+static int RescalePlane(const uint8_t* src,
+                        int src_width, int src_height, int src_stride,
+                        uint8_t* dst,
+                        int dst_width, int dst_height, int dst_stride,
+                        rescaler_t* const work,
+                        int num_channels) {
   WebPRescaler rescaler;
   int y = 0;
-  WebPRescalerInit(&rescaler, src_width, src_height,
-                   dst, dst_width, dst_height, dst_stride,
-                   num_channels, work);
+  if (!WebPRescalerInit(&rescaler, src_width, src_height,
+                        dst, dst_width, dst_height, dst_stride,
+                        num_channels, work)) {
+    return 0;
+  }
   while (y < src_height) {
     y += WebPRescalerImport(&rescaler, src_height - y,
                             src + y * src_stride, src_stride);
     WebPRescalerExport(&rescaler);
   }
+  return 1;
 }
 
 static void AlphaMultiplyARGB(WebPPicture* const pic, int inverse) {
@@ -222,25 +225,28 @@ int WebPPictureRescale(WebPPicture* pic, int width, int height) {
     // If present, we need to rescale alpha first (for AlphaMultiplyY).
     if (pic->a != NULL) {
       WebPInitAlphaProcessing();
-      RescalePlane(pic->a, prev_width, prev_height, pic->a_stride,
-                   tmp.a, width, height, tmp.a_stride, work, 1);
+      if (!RescalePlane(pic->a, prev_width, prev_height, pic->a_stride,
+                        tmp.a, width, height, tmp.a_stride, work, 1)) {
+        return 0;
+      }
     }
 
     // We take transparency into account on the luma plane only. That's not
     // totally exact blending, but still is a good approximation.
     AlphaMultiplyY(pic, 0);
-    RescalePlane(pic->y, prev_width, prev_height, pic->y_stride,
-                 tmp.y, width, height, tmp.y_stride, work, 1);
+    if (!RescalePlane(pic->y, prev_width, prev_height, pic->y_stride,
+                      tmp.y, width, height, tmp.y_stride, work, 1) ||
+        !RescalePlane(pic->u,
+                      HALVE(prev_width), HALVE(prev_height), pic->uv_stride,
+                      tmp.u,
+                      HALVE(width), HALVE(height), tmp.uv_stride, work, 1) ||
+        !RescalePlane(pic->v,
+                      HALVE(prev_width), HALVE(prev_height), pic->uv_stride,
+                      tmp.v,
+                      HALVE(width), HALVE(height), tmp.uv_stride, work, 1)) {
+      return 0;
+    }
     AlphaMultiplyY(&tmp, 1);
-
-    RescalePlane(pic->u,
-                 HALVE(prev_width), HALVE(prev_height), pic->uv_stride,
-                 tmp.u,
-                 HALVE(width), HALVE(height), tmp.uv_stride, work, 1);
-    RescalePlane(pic->v,
-                 HALVE(prev_width), HALVE(prev_height), pic->uv_stride,
-                 tmp.v,
-                 HALVE(width), HALVE(height), tmp.uv_stride, work, 1);
   } else {
     work = (rescaler_t*)WebPSafeMalloc(2ULL * width * 4, sizeof(*work));
     if (work == NULL) {
@@ -252,11 +258,12 @@ int WebPPictureRescale(WebPPicture* pic, int width, int height) {
     // the premultiplication afterward (while preserving the alpha channel).
     WebPInitAlphaProcessing();
     AlphaMultiplyARGB(pic, 0);
-    RescalePlane((const uint8_t*)pic->argb, prev_width, prev_height,
-                 pic->argb_stride * 4,
-                 (uint8_t*)tmp.argb, width, height,
-                 tmp.argb_stride * 4,
-                 work, 4);
+    if (!RescalePlane((const uint8_t*)pic->argb, prev_width, prev_height,
+                      pic->argb_stride * 4,
+                      (uint8_t*)tmp.argb, width, height,
+                      tmp.argb_stride * 4, work, 4)) {
+      return 0;
+    }
     AlphaMultiplyARGB(&tmp, 1);
   }
   WebPPictureFree(pic);

src/enc/syntax_enc.c

@@ -349,7 +349,7 @@ int VP8EncWrite(VP8Encoder* const enc) {
                                        (enc->alpha_data_size_ & 1);
     riff_size += CHUNK_HEADER_SIZE + padded_alpha_size;
   }
-  // Sanity check.
+  // RIFF size should fit in 32-bits.
   if (riff_size > 0xfffffffeU) {
     return WebPEncodingSetError(pic, VP8_ENC_ERROR_FILE_TOO_BIG);
   }

src/enc/vp8i_enc.h

@@ -32,7 +32,7 @@ extern "C" {
 // version numbers
 #define ENC_MAJ_VERSION 1
 #define ENC_MIN_VERSION 2
-#define ENC_REV_VERSION 0
+#define ENC_REV_VERSION 1
 
 enum { MAX_LF_LEVELS = 64,       // Maximum loop filter level
        MAX_VARIABLE_LEVEL = 67,  // last (inclusive) level with variable cost
@@ -286,8 +286,7 @@ int VP8IteratorNext(VP8EncIterator* const it);
 // save the yuv_out_ boundary values to top_/left_ arrays for next iterations.
 void VP8IteratorSaveBoundary(VP8EncIterator* const it);
 // Report progression based on macroblock rows. Return 0 for user-abort request.
-int VP8IteratorProgress(const VP8EncIterator* const it,
-                        int final_delta_percent);
+int VP8IteratorProgress(const VP8EncIterator* const it, int delta);
 // Intra4x4 iterations
 void VP8IteratorStartI4(VP8EncIterator* const it);
 // returns true if not done.

src/enc/vp8l_enc.c

@@ -65,25 +65,22 @@ static WEBP_INLINE void SwapColor(uint32_t* const col1, uint32_t* const col2) {
   *col2 = tmp;
 }
 
-static void GreedyMinimizeDeltas(uint32_t palette[], int num_colors) {
-  // Find greedily always the closest color of the predicted color to minimize
-  // deltas in the palette. This reduces storage needs since the
-  // palette is stored with delta encoding.
-  uint32_t predict = 0x00000000;
-  int i, k;
-  for (i = 0; i < num_colors; ++i) {
-    int best_ix = i;
-    uint32_t best_score = ~0U;
-    for (k = i; k < num_colors; ++k) {
-      const uint32_t cur_score = PaletteColorDistance(palette[k], predict);
-      if (best_score > cur_score) {
-        best_score = cur_score;
-        best_ix = k;
-      }
+static WEBP_INLINE int SearchColorNoIdx(const uint32_t sorted[], uint32_t color,
+                                        int num_colors) {
+  int low = 0, hi = num_colors;
+  if (sorted[low] == color) return low;  // loop invariant: sorted[low] != color
+  while (1) {
+    const int mid = (low + hi) >> 1;
+    if (sorted[mid] == color) {
+      return mid;
+    } else if (sorted[mid] < color) {
+      low = mid;
+    } else {
+      hi = mid;
     }
-    SwapColor(&palette[best_ix], &palette[i]);
-    predict = palette[i];
   }
+  assert(0);
+  return 0;
 }
 
 // The palette has been sorted by alpha. This function checks if the other
@@ -92,7 +89,8 @@ static void GreedyMinimizeDeltas(uint32_t palette[], int num_colors) {
 // no benefit to re-organize them greedily. A monotonic development
 // would be spotted in green-only situations (like lossy alpha) or gray-scale
 // images.
-static int PaletteHasNonMonotonousDeltas(uint32_t palette[], int num_colors) {
+static int PaletteHasNonMonotonousDeltas(const uint32_t* const palette,
+                                         int num_colors) {
   uint32_t predict = 0x000000;
   int i;
   uint8_t sign_found = 0x00;
@@ -115,28 +113,215 @@ static int PaletteHasNonMonotonousDeltas(uint32_t palette[], int num_colors) {
   return (sign_found & (sign_found << 1)) != 0;  // two consequent signs.
 }
 
+static void PaletteSortMinimizeDeltas(const uint32_t* const palette_sorted,
+                                      int num_colors, uint32_t* const palette) {
+  uint32_t predict = 0x00000000;
+  int i, k;
+  memcpy(palette, palette_sorted, num_colors * sizeof(*palette));
+  if (!PaletteHasNonMonotonousDeltas(palette_sorted, num_colors)) return;
+  // Find greedily always the closest color of the predicted color to minimize
+  // deltas in the palette. This reduces storage needs since the
+  // palette is stored with delta encoding.
+  for (i = 0; i < num_colors; ++i) {
+    int best_ix = i;
+    uint32_t best_score = ~0U;
+    for (k = i; k < num_colors; ++k) {
+      const uint32_t cur_score = PaletteColorDistance(palette[k], predict);
+      if (best_score > cur_score) {
+        best_score = cur_score;
+        best_ix = k;
+      }
+    }
+    SwapColor(&palette[best_ix], &palette[i]);
+    predict = palette[i];
+  }
+}
+
+// Sort palette in increasing order and prepare an inverse mapping array.
+static void PrepareMapToPalette(const uint32_t palette[], uint32_t num_colors,
+                                uint32_t sorted[], uint32_t idx_map[]) {
+  uint32_t i;
+  memcpy(sorted, palette, num_colors * sizeof(*sorted));
+  qsort(sorted, num_colors, sizeof(*sorted), PaletteCompareColorsForQsort);
+  for (i = 0; i < num_colors; ++i) {
+    idx_map[SearchColorNoIdx(sorted, palette[i], num_colors)] = i;
+  }
+}
+
+// -----------------------------------------------------------------------------
+// Modified Zeng method from "A Survey on Palette Reordering
+// Methods for Improving the Compression of Color-Indexed Images" by Armando J.
+// Pinho and Antonio J. R. Neves.
+
+// Finds the biggest cooccurrence in the matrix.
+static void CoOccurrenceFindMax(const uint32_t* const cooccurrence,
+                                uint32_t num_colors, uint8_t* const c1,
+                                uint8_t* const c2) {
+  // Find the index that is most frequently located adjacent to other
+  // (different) indexes.
+  uint32_t best_sum = 0u;
+  uint32_t i, j, best_cooccurrence;
+  *c1 = 0u;
+  for (i = 0; i < num_colors; ++i) {
+    uint32_t sum = 0;
+    for (j = 0; j < num_colors; ++j) sum += cooccurrence[i * num_colors + j];
+    if (sum > best_sum) {
+      best_sum = sum;
+      *c1 = i;
+    }
+  }
+  // Find the index that is most frequently found adjacent to *c1.
+  *c2 = 0u;
+  best_cooccurrence = 0u;
+  for (i = 0; i < num_colors; ++i) {
+    if (cooccurrence[*c1 * num_colors + i] > best_cooccurrence) {
+      best_cooccurrence = cooccurrence[*c1 * num_colors + i];
+      *c2 = i;
+    }
+  }
+  assert(*c1 != *c2);
+}
+
+// Builds the cooccurrence matrix
+static WebPEncodingError CoOccurrenceBuild(const WebPPicture* const pic,
+                                           const uint32_t* const palette,
+                                           uint32_t num_colors,
+                                           uint32_t* cooccurrence) {
+  uint32_t *lines, *line_top, *line_current, *line_tmp;
+  int x, y;
+  const uint32_t* src = pic->argb;
+  uint32_t prev_pix = ~src[0];
+  uint32_t prev_idx = 0u;
+  uint32_t idx_map[MAX_PALETTE_SIZE] = {0};
+  uint32_t palette_sorted[MAX_PALETTE_SIZE];
+  lines = (uint32_t*)WebPSafeMalloc(2 * pic->width, sizeof(*lines));
+  if (lines == NULL) return VP8_ENC_ERROR_OUT_OF_MEMORY;
+  line_top = &lines[0];
+  line_current = &lines[pic->width];
+  PrepareMapToPalette(palette, num_colors, palette_sorted, idx_map);
+  for (y = 0; y < pic->height; ++y) {
+    for (x = 0; x < pic->width; ++x) {
+      const uint32_t pix = src[x];
+      if (pix != prev_pix) {
+        prev_idx = idx_map[SearchColorNoIdx(palette_sorted, pix, num_colors)];
+        prev_pix = pix;
+      }
+      line_current[x] = prev_idx;
+      // 4-connectivity is what works best as mentioned in "On the relation
+      // between Memon's and the modified Zeng's palette reordering methods".
+      if (x > 0 && prev_idx != line_current[x - 1]) {
+        const uint32_t left_idx = line_current[x - 1];
+        ++cooccurrence[prev_idx * num_colors + left_idx];
+        ++cooccurrence[left_idx * num_colors + prev_idx];
+      }
+      if (y > 0 && prev_idx != line_top[x]) {
+        const uint32_t top_idx = line_top[x];
+        ++cooccurrence[prev_idx * num_colors + top_idx];
+        ++cooccurrence[top_idx * num_colors + prev_idx];
+      }
+    }
+    line_tmp = line_top;
+    line_top = line_current;
+    line_current = line_tmp;
+    src += pic->argb_stride;
+  }
+  WebPSafeFree(lines);
+  return VP8_ENC_OK;
+}
+
+struct Sum {
+  uint8_t index;
+  uint32_t sum;
+};
+
+// Implements the modified Zeng method from "A Survey on Palette Reordering
+// Methods for Improving the Compression of Color-Indexed Images" by Armando J.
+// Pinho and Antonio J. R. Neves.
+static WebPEncodingError PaletteSortModifiedZeng(
+    const WebPPicture* const pic, const uint32_t* const palette_sorted,
+    uint32_t num_colors, uint32_t* const palette) {
+  uint32_t i, j, ind;
+  uint8_t remapping[MAX_PALETTE_SIZE];
+  uint32_t* cooccurrence;
+  struct Sum sums[MAX_PALETTE_SIZE];
+  uint32_t first, last;
+  uint32_t num_sums;
+  // TODO(vrabaud) check whether one color images should use palette or not.
+  if (num_colors <= 1) return VP8_ENC_OK;
+  // Build the co-occurrence matrix.
+  cooccurrence =
+      (uint32_t*)WebPSafeCalloc(num_colors * num_colors, sizeof(*cooccurrence));
+  if (cooccurrence == NULL) return VP8_ENC_ERROR_OUT_OF_MEMORY;
+  if (CoOccurrenceBuild(pic, palette_sorted, num_colors, cooccurrence) !=
+      VP8_ENC_OK) {
+    WebPSafeFree(cooccurrence);
+    return VP8_ENC_ERROR_OUT_OF_MEMORY;
+  }
+
+  // Initialize the mapping list with the two best indices.
+  CoOccurrenceFindMax(cooccurrence, num_colors, &remapping[0], &remapping[1]);
+
+  // We need to append and prepend to the list of remapping. To this end, we
+  // actually define the next start/end of the list as indices in a vector (with
+  // a wrap around when the end is reached).
+  first = 0;
+  last = 1;
+  num_sums = num_colors - 2;  // -2 because we know the first two values
+  if (num_sums > 0) {
+    // Initialize the sums with the first two remappings and find the best one
+    struct Sum* best_sum = &sums[0];
+    best_sum->index = 0u;
+    best_sum->sum = 0u;
+    for (i = 0, j = 0; i < num_colors; ++i) {
+      if (i == remapping[0] || i == remapping[1]) continue;
+      sums[j].index = i;
+      sums[j].sum = cooccurrence[i * num_colors + remapping[0]] +
+                    cooccurrence[i * num_colors + remapping[1]];
+      if (sums[j].sum > best_sum->sum) best_sum = &sums[j];
+      ++j;
+    }
+
+    while (num_sums > 0) {
+      const uint8_t best_index = best_sum->index;
+      // Compute delta to know if we need to prepend or append the best index.
+      int32_t delta = 0;
+      const int32_t n = num_colors - num_sums;
+      for (ind = first, j = 0; (ind + j) % num_colors != last + 1; ++j) {
+        const uint16_t l_j = remapping[(ind + j) % num_colors];
+        delta += (n - 1 - 2 * (int32_t)j) *
+                 (int32_t)cooccurrence[best_index * num_colors + l_j];
+      }
+      if (delta > 0) {
+        first = (first == 0) ? num_colors - 1 : first - 1;
+        remapping[first] = best_index;
+      } else {
+        ++last;
+        remapping[last] = best_index;
+      }
+      // Remove best_sum from sums.
+      *best_sum = sums[num_sums - 1];
+      --num_sums;
+      // Update all the sums and find the best one.
+      best_sum = &sums[0];
+      for (i = 0; i < num_sums; ++i) {
+        sums[i].sum += cooccurrence[best_index * num_colors + sums[i].index];
+        if (sums[i].sum > best_sum->sum) best_sum = &sums[i];
+      }
+    }
+  }
+  assert((last + 1) % num_colors == first);
+  WebPSafeFree(cooccurrence);
+
+  // Re-map the palette.
+  for (i = 0; i < num_colors; ++i) {
+    palette[i] = palette_sorted[remapping[(first + i) % num_colors]];
+  }
+  return VP8_ENC_OK;
+}
+
 // -----------------------------------------------------------------------------
 // Palette
 
-// If number of colors in the image is less than or equal to MAX_PALETTE_SIZE,
-// creates a palette and returns true, else returns false.
-static int AnalyzeAndCreatePalette(const WebPPicture* const pic,
-                                   int low_effort,
-                                   uint32_t palette[MAX_PALETTE_SIZE],
-                                   int* const palette_size) {
-  const int num_colors = WebPGetColorPalette(pic, palette);
-  if (num_colors > MAX_PALETTE_SIZE) {
-    *palette_size = 0;
-    return 0;
-  }
-  *palette_size = num_colors;
-  qsort(palette, num_colors, sizeof(*palette), PaletteCompareColorsForQsort);
-  if (!low_effort && PaletteHasNonMonotonousDeltas(palette, num_colors)) {
-    GreedyMinimizeDeltas(palette, num_colors);
-  }
-  return 1;
-}
-
 // These five modes are evaluated and their respective entropy is computed.
 typedef enum {
   kDirect = 0,
@@ -148,6 +333,13 @@ typedef enum {
   kNumEntropyIx = 6
 } EntropyIx;
 
+typedef enum {
+  kSortedDefault = 0,
+  kMinimizeDelta = 1,
+  kModifiedZeng = 2,
+  kUnusedPalette = 3,
+} PaletteSorting;
+
 typedef enum {
   kHistoAlpha = 0,
   kHistoAlphaPred,
@@ -362,11 +554,14 @@ typedef struct {
 } CrunchSubConfig;
 typedef struct {
   int entropy_idx_;
+  PaletteSorting palette_sorting_type_;
   CrunchSubConfig sub_configs_[CRUNCH_SUBCONFIGS_MAX];
   int sub_configs_size_;
 } CrunchConfig;
 
-#define CRUNCH_CONFIGS_MAX kNumEntropyIx
+// +2 because we add a palette sorting configuration for kPalette and
+// kPaletteAndSpatial.
+#define CRUNCH_CONFIGS_MAX (kNumEntropyIx + 2)
 
 static int EncoderAnalyze(VP8LEncoder* const enc,
                           CrunchConfig crunch_configs[CRUNCH_CONFIGS_MAX],
@@ -386,9 +581,15 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
   int do_no_cache = 0;
   assert(pic != NULL && pic->argb != NULL);
 
-  use_palette =
-      AnalyzeAndCreatePalette(pic, low_effort,
-                              enc->palette_, &enc->palette_size_);
+  // Check whether a palette is possible.
+  enc->palette_size_ = WebPGetColorPalette(pic, enc->palette_sorted_);
+  use_palette = (enc->palette_size_ <= MAX_PALETTE_SIZE);
+  if (!use_palette) {
+    enc->palette_size_ = 0;
+  } else {
+    qsort(enc->palette_sorted_, enc->palette_size_,
+          sizeof(*enc->palette_sorted_), PaletteCompareColorsForQsort);
+  }
 
   // Empirical bit sizes.
   enc->histo_bits_ = GetHistoBits(method, use_palette,
@@ -398,6 +599,8 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
   if (low_effort) {
     // AnalyzeEntropy is somewhat slow.
     crunch_configs[0].entropy_idx_ = use_palette ? kPalette : kSpatialSubGreen;
+    crunch_configs[0].palette_sorting_type_ =
+        use_palette ? kSortedDefault : kUnusedPalette;
     n_lz77s = 1;
     *crunch_configs_size = 1;
   } else {
@@ -418,13 +621,28 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
         // a palette.
         if ((i != kPalette && i != kPaletteAndSpatial) || use_palette) {
           assert(*crunch_configs_size < CRUNCH_CONFIGS_MAX);
-          crunch_configs[(*crunch_configs_size)++].entropy_idx_ = i;
+          crunch_configs[(*crunch_configs_size)].entropy_idx_ = i;
+          if (use_palette && (i == kPalette || i == kPaletteAndSpatial)) {
+            crunch_configs[(*crunch_configs_size)].palette_sorting_type_ =
+                kMinimizeDelta;
+            ++*crunch_configs_size;
+            // Also add modified Zeng's method.
+            crunch_configs[(*crunch_configs_size)].entropy_idx_ = i;
+            crunch_configs[(*crunch_configs_size)].palette_sorting_type_ =
+                kModifiedZeng;
+          } else {
+            crunch_configs[(*crunch_configs_size)].palette_sorting_type_ =
+                kUnusedPalette;
+          }
+          ++*crunch_configs_size;
         }
       }
     } else {
       // Only choose the guessed best transform.
       *crunch_configs_size = 1;
       crunch_configs[0].entropy_idx_ = min_entropy_ix;
+      crunch_configs[0].palette_sorting_type_ =
+          use_palette ? kMinimizeDelta : kUnusedPalette;
       if (config->quality >= 75 && method == 5) {
         // Test with and without color cache.
         do_no_cache = 1;
@@ -432,6 +650,7 @@ static int EncoderAnalyze(VP8LEncoder* const enc,
         if (min_entropy_ix == kPalette) {
           *crunch_configs_size = 2;
           crunch_configs[1].entropy_idx_ = kPaletteAndSpatial;
+          crunch_configs[1].palette_sorting_type_ = kMinimizeDelta;
         }
       }
     }
@@ -1283,22 +1502,6 @@ static WebPEncodingError MakeInputImageCopy(VP8LEncoder* const enc) {
 
 // -----------------------------------------------------------------------------
 
-static WEBP_INLINE int SearchColorNoIdx(const uint32_t sorted[], uint32_t color,
-                                        int hi) {
-  int low = 0;
-  if (sorted[low] == color) return low;  // loop invariant: sorted[low] != color
-  while (1) {
-    const int mid = (low + hi) >> 1;
-    if (sorted[mid] == color) {
-      return mid;
-    } else if (sorted[mid] < color) {
-      low = mid;
-    } else {
-      hi = mid;
-    }
-  }
-}
-
 #define APPLY_PALETTE_GREEDY_MAX 4
 
 static WEBP_INLINE uint32_t SearchColorGreedy(const uint32_t palette[],
@@ -1333,17 +1536,6 @@ static WEBP_INLINE uint32_t ApplyPaletteHash2(uint32_t color) {
          (32 - PALETTE_INV_SIZE_BITS);
 }
 
-// Sort palette in increasing order and prepare an inverse mapping array.
-static void PrepareMapToPalette(const uint32_t palette[], int num_colors,
-                                uint32_t sorted[], uint32_t idx_map[]) {
-  int i;
-  memcpy(sorted, palette, num_colors * sizeof(*sorted));
-  qsort(sorted, num_colors, sizeof(*sorted), PaletteCompareColorsForQsort);
-  for (i = 0; i < num_colors; ++i) {
-    idx_map[SearchColorNoIdx(sorted, palette[i], num_colors)] = i;
-  }
-}
-
 // Use 1 pixel cache for ARGB pixels.
 #define APPLY_PALETTE_FOR(COLOR_INDEX) do {         \
   uint32_t prev_pix = palette[0];                   \
@@ -1571,7 +1763,8 @@ static int EncodeStreamHook(void* input, void* data2) {
     enc->use_predict_ = (entropy_idx == kSpatial) ||
                         (entropy_idx == kSpatialSubGreen) ||
                         (entropy_idx == kPaletteAndSpatial);
-    if (low_effort) {
+    // When using a palette, R/B==0, hence no need to test for cross-color.
+    if (low_effort || enc->use_palette_) {
       enc->use_cross_color_ = 0;
     } else {
       enc->use_cross_color_ = red_and_blue_always_zero ? 0 : enc->use_predict_;
@@ -1603,6 +1796,19 @@ static int EncodeStreamHook(void* input, void* data2) {
 
     // Encode palette
     if (enc->use_palette_) {
+      if (crunch_configs[idx].palette_sorting_type_ == kSortedDefault) {
+        // Nothing to do, we have already sorted the palette.
+        memcpy(enc->palette_, enc->palette_sorted_,
+               enc->palette_size_ * sizeof(*enc->palette_));
+      } else if (crunch_configs[idx].palette_sorting_type_ == kMinimizeDelta) {
+        PaletteSortMinimizeDeltas(enc->palette_sorted_, enc->palette_size_,
+                                  enc->palette_);
+      } else {
+        assert(crunch_configs[idx].palette_sorting_type_ == kModifiedZeng);
+        err = PaletteSortModifiedZeng(enc->pic_, enc->palette_sorted_,
+                                      enc->palette_size_, enc->palette_);
+        if (err != VP8_ENC_OK) goto Error;
+      }
       err = EncodePalette(bw, low_effort, enc);
       if (err != VP8_ENC_OK) goto Error;
       err = MapImageFromPalette(enc, use_delta_palette);
@@ -1767,6 +1973,8 @@ WebPEncodingError VP8LEncodeStream(const WebPConfig* const config,
         enc_side->palette_size_ = enc_main->palette_size_;
         memcpy(enc_side->palette_, enc_main->palette_,
                sizeof(enc_main->palette_));
+        memcpy(enc_side->palette_sorted_, enc_main->palette_sorted_,
+               sizeof(enc_main->palette_sorted_));
         param->enc_ = enc_side;
       }
       // Create the workers.

src/enc/vp8li_enc.h

@@ -69,6 +69,8 @@ typedef struct {
   int use_palette_;
   int palette_size_;
   uint32_t palette_[MAX_PALETTE_SIZE];
+  // Sorted version of palette_ for cache purposes.
+  uint32_t palette_sorted_[MAX_PALETTE_SIZE];
 
   // Some 'scratch' (potentially large) objects.
   struct VP8LBackwardRefs refs_[4];  // Backward Refs array for temporaries.

src/libwebp.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,2,0
- PRODUCTVERSION 1,0,2,0
+ FILEVERSION 1,0,2,1
+ PRODUCTVERSION 1,0,2,1
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebp DLL"
-            VALUE "FileVersion", "1.2.0"
+            VALUE "FileVersion", "1.2.1"
             VALUE "InternalName", "libwebp.dll"
-            VALUE "LegalCopyright", "Copyright (C) 2019"
+            VALUE "LegalCopyright", "Copyright (C) 2021"
             VALUE "OriginalFilename", "libwebp.dll"
             VALUE "ProductName", "WebP Image Codec"
-            VALUE "ProductVersion", "1.2.0"
+            VALUE "ProductVersion", "1.2.1"
         END
     END
     BLOCK "VarFileInfo"

src/libwebpdecoder.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,2,0
- PRODUCTVERSION 1,0,2,0
+ FILEVERSION 1,0,2,1
+ PRODUCTVERSION 1,0,2,1
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpdecoder DLL"
-            VALUE "FileVersion", "1.2.0"
+            VALUE "FileVersion", "1.2.1"
             VALUE "InternalName", "libwebpdecoder.dll"
-            VALUE "LegalCopyright", "Copyright (C) 2019"
+            VALUE "LegalCopyright", "Copyright (C) 2021"
             VALUE "OriginalFilename", "libwebpdecoder.dll"
             VALUE "ProductName", "WebP Image Decoder"
-            VALUE "ProductVersion", "1.2.0"
+            VALUE "ProductVersion", "1.2.1"
         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 3:6:0 -lm
+libwebpmux_la_LDFLAGS = -no-undefined -version-info 3:7:0 -lm
 libwebpmuxincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebpmux.pc

src/mux/anim_encode.c

@@ -248,9 +248,6 @@ WebPAnimEncoder* WebPAnimEncoderNewInternal(
 
   enc = (WebPAnimEncoder*)WebPSafeCalloc(1, sizeof(*enc));
   if (enc == NULL) return NULL;
-  // sanity inits, so we can call WebPAnimEncoderDelete():
-  enc->encoded_frames_ = NULL;
-  enc->mux_ = NULL;
   MarkNoError(enc);
 
   // Dimensions and options.
@@ -421,7 +418,7 @@ static void MinimizeChangeRectangle(const WebPPicture* const src,
   const int max_allowed_diff_lossy = QualityToMaxDiff(quality);
   const int max_allowed_diff = is_lossless ? 0 : max_allowed_diff_lossy;
 
-  // Sanity checks.
+  // Assumption/correctness checks.
   assert(src->width == dst->width && src->height == dst->height);
   assert(rect->x_offset_ + rect->width_ <= dst->width);
   assert(rect->y_offset_ + rect->height_ <= dst->height);
@@ -949,7 +946,8 @@ static int IncreasePreviousDuration(WebPAnimEncoder* const enc, int duration) {
   int new_duration;
 
   assert(enc->count_ >= 1);
-  assert(prev_enc_frame->sub_frame_.duration ==
+  assert(!prev_enc_frame->is_key_frame_ ||
+         prev_enc_frame->sub_frame_.duration ==
          prev_enc_frame->key_frame_.duration);
   assert(prev_enc_frame->sub_frame_.duration ==
          (prev_enc_frame->sub_frame_.duration & (MAX_DURATION - 1)));
@@ -966,7 +964,7 @@ static int IncreasePreviousDuration(WebPAnimEncoder* const enc, int duration) {
       0x10, 0x88, 0x88, 0x08
     };
     const WebPData lossless_1x1 = {
-        lossless_1x1_bytes, sizeof(lossless_1x1_bytes)
+      lossless_1x1_bytes, sizeof(lossless_1x1_bytes)
     };
     const uint8_t lossy_1x1_bytes[] = {
       0x52, 0x49, 0x46, 0x46, 0x40, 0x00, 0x00, 0x00, 0x57, 0x45, 0x42, 0x50,
@@ -1358,6 +1356,12 @@ int WebPAnimEncoderAdd(WebPAnimEncoder* enc, WebPPicture* frame, int timestamp,
     if (!IncreasePreviousDuration(enc, (int)prev_frame_duration)) {
       return 0;
     }
+    // IncreasePreviousDuration() may add a frame to avoid exceeding
+    // MAX_DURATION which could cause CacheFrame() to over read encoded_frames_
+    // before the next flush.
+    if (enc->count_ == enc->size_ && !FlushFrames(enc)) {
+      return 0;
+    }
   } else {
     enc->first_timestamp_ = timestamp;
   }

src/mux/libwebpmux.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 1,0,2,0
- PRODUCTVERSION 1,0,2,0
+ FILEVERSION 1,0,2,1
+ PRODUCTVERSION 1,0,2,1
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpmux DLL"
-            VALUE "FileVersion", "1.2.0"
+            VALUE "FileVersion", "1.2.1"
             VALUE "InternalName", "libwebpmux.dll"
-            VALUE "LegalCopyright", "Copyright (C) 2019"
+            VALUE "LegalCopyright", "Copyright (C) 2021"
             VALUE "OriginalFilename", "libwebpmux.dll"
             VALUE "ProductName", "WebP Image Muxer"
-            VALUE "ProductVersion", "1.2.0"
+            VALUE "ProductVersion", "1.2.1"
         END
     END
     BLOCK "VarFileInfo"

src/mux/muxedit.c

@@ -235,7 +235,6 @@ WebPMuxError WebPMuxSetImage(WebPMux* mux, const WebPData* bitstream,
   WebPMuxImage wpi;
   WebPMuxError err;
 
-  // Sanity checks.
   if (mux == NULL || bitstream == NULL || bitstream->bytes == NULL ||
       bitstream->size > MAX_CHUNK_PAYLOAD) {
     return WEBP_MUX_INVALID_ARGUMENT;
@@ -267,7 +266,6 @@ WebPMuxError WebPMuxPushFrame(WebPMux* mux, const WebPMuxFrameInfo* info,
   WebPMuxImage wpi;
   WebPMuxError err;
 
-  // Sanity checks.
   if (mux == NULL || info == NULL) return WEBP_MUX_INVALID_ARGUMENT;
 
   if (info->id != WEBP_CHUNK_ANMF) return WEBP_MUX_INVALID_ARGUMENT;

src/mux/muxi.h

@@ -29,7 +29,7 @@ extern "C" {
 
 #define MUX_MAJ_VERSION 1
 #define MUX_MIN_VERSION 2
-#define MUX_REV_VERSION 0
+#define MUX_REV_VERSION 1
 
 // Chunk object.
 typedef struct WebPChunk WebPChunk;

src/mux/muxread.c

@@ -56,7 +56,7 @@ static WebPMuxError ChunkVerifyAndAssign(WebPChunk* chunk,
   uint32_t chunk_size;
   WebPData chunk_data;
 
-  // Sanity checks.
+  // Correctness checks.
   if (data_size < CHUNK_HEADER_SIZE) return WEBP_MUX_NOT_ENOUGH_DATA;
   chunk_size = GetLE32(data + TAG_SIZE);
   if (chunk_size > MAX_CHUNK_PAYLOAD) return WEBP_MUX_BAD_DATA;
@@ -186,7 +186,6 @@ WebPMux* WebPMuxCreateInternal(const WebPData* bitstream, int copy_data,
   WebPChunk** chunk_list_ends[WEBP_CHUNK_NIL + 1] = { NULL };
   ChunkInit(&chunk);
 
-  // Sanity checks.
   if (WEBP_ABI_IS_INCOMPATIBLE(version, WEBP_MUX_ABI_VERSION)) {
     return NULL;  // version mismatch
   }
@@ -481,7 +480,6 @@ WebPMuxError WebPMuxGetFrame(
   WebPMuxError err;
   WebPMuxImage* wpi;
 
-  // Sanity checks.
   if (mux == NULL || frame == NULL) {
     return WEBP_MUX_INVALID_ARGUMENT;
   }

src/utils/bit_reader_inl_utils.h

@@ -55,7 +55,7 @@ void VP8LoadFinalBytes(VP8BitReader* const br);
 
 // makes sure br->value_ has at least BITS bits worth of data
 static WEBP_UBSAN_IGNORE_UNDEF WEBP_INLINE
-void VP8LoadNewBytes(VP8BitReader* const br) {
+void VP8LoadNewBytes(VP8BitReader* WEBP_RESTRICT const br) {
   assert(br != NULL && br->buf_ != NULL);
   // Read 'BITS' bits at a time if possible.
   if (br->buf_ < br->buf_max_) {
@@ -104,7 +104,7 @@ void VP8LoadNewBytes(VP8BitReader* const br) {
 }
 
 // Read a bit with proba 'prob'. Speed-critical function!
-static WEBP_INLINE int VP8GetBit(VP8BitReader* const br,
+static WEBP_INLINE int VP8GetBit(VP8BitReader* WEBP_RESTRICT const br,
                                  int prob, const char label[]) {
   // Don't move this declaration! It makes a big speed difference to store
   // 'range' *before* calling VP8LoadNewBytes(), even if this function doesn't
@@ -137,7 +137,8 @@ static WEBP_INLINE int VP8GetBit(VP8BitReader* const br,
 
 // simplified version of VP8GetBit() for prob=0x80 (note shift is always 1 here)
 static WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW WEBP_INLINE
-int VP8GetSigned(VP8BitReader* const br, int v, const char label[]) {
+int VP8GetSigned(VP8BitReader* WEBP_RESTRICT const br, int v,
+                 const char label[]) {
   if (br->bits_ < 0) {
     VP8LoadNewBytes(br);
   }
@@ -155,7 +156,7 @@ int VP8GetSigned(VP8BitReader* const br, int v, const char label[]) {
   }
 }
 
-static WEBP_INLINE int VP8GetBitAlt(VP8BitReader* const br,
+static WEBP_INLINE int VP8GetBitAlt(VP8BitReader* WEBP_RESTRICT const br,
                                     int prob, const char label[]) {
   // Don't move this declaration! It makes a big speed difference to store
   // 'range' *before* calling VP8LoadNewBytes(), even if this function doesn't

src/utils/bit_writer_utils.c

@@ -278,7 +278,7 @@ void VP8LPutBitsFlushBits(VP8LBitWriter* const bw) {
   // If needed, make some room by flushing some bits out.
   if (bw->cur_ + VP8L_WRITER_BYTES > bw->end_) {
     const uint64_t extra_size = (bw->end_ - bw->buf_) + MIN_EXTRA_SIZE;
-    if (extra_size != (size_t)extra_size ||
+    if (!CheckSizeOverflow(extra_size) ||
         !VP8LBitWriterResize(bw, (size_t)extra_size)) {
       bw->cur_ = bw->buf_;
       bw->error_ = 1;
@@ -314,7 +314,7 @@ void VP8LPutBitsInternal(VP8LBitWriter* const bw, uint32_t bits, int n_bits) {
     while (used >= VP8L_WRITER_BITS) {
       if (bw->cur_ + VP8L_WRITER_BYTES > bw->end_) {
         const uint64_t extra_size = (bw->end_ - bw->buf_) + MIN_EXTRA_SIZE;
-        if (extra_size != (size_t)extra_size ||
+        if (!CheckSizeOverflow(extra_size) ||
             !VP8LBitWriterResize(bw, (size_t)extra_size)) {
           bw->cur_ = bw->buf_;
           bw->error_ = 1;

src/utils/color_cache_utils.c

@@ -20,22 +20,22 @@
 //------------------------------------------------------------------------------
 // VP8LColorCache.
 
-int VP8LColorCacheInit(VP8LColorCache* const cc, int hash_bits) {
+int VP8LColorCacheInit(VP8LColorCache* const color_cache, int hash_bits) {
   const int hash_size = 1 << hash_bits;
-  assert(cc != NULL);
+  assert(color_cache != NULL);
   assert(hash_bits > 0);
-  cc->colors_ = (uint32_t*)WebPSafeCalloc((uint64_t)hash_size,
-                                          sizeof(*cc->colors_));
-  if (cc->colors_ == NULL) return 0;
-  cc->hash_shift_ = 32 - hash_bits;
-  cc->hash_bits_ = hash_bits;
+  color_cache->colors_ = (uint32_t*)WebPSafeCalloc(
+      (uint64_t)hash_size, sizeof(*color_cache->colors_));
+  if (color_cache->colors_ == NULL) return 0;
+  color_cache->hash_shift_ = 32 - hash_bits;
+  color_cache->hash_bits_ = hash_bits;
   return 1;
 }
 
-void VP8LColorCacheClear(VP8LColorCache* const cc) {
-  if (cc != NULL) {
-    WebPSafeFree(cc->colors_);
-    cc->colors_ = NULL;
+void VP8LColorCacheClear(VP8LColorCache* const color_cache) {
+  if (color_cache != NULL) {
+    WebPSafeFree(color_cache->colors_);
+    color_cache->colors_ = NULL;
   }
 }
 

src/utils/huffman_encode_utils.c

@@ -404,8 +404,7 @@ static void ConvertBitDepthsToSymbols(HuffmanTreeCode* const tree) {
 // Main entry point
 
 void VP8LCreateHuffmanTree(uint32_t* const histogram, int tree_depth_limit,
-                           uint8_t* const buf_rle,
-                           HuffmanTree* const huff_tree,
+                           uint8_t* const buf_rle, HuffmanTree* const huff_tree,
                            HuffmanTreeCode* const huff_code) {
   const int num_symbols = huff_code->num_symbols;
   memset(buf_rle, 0, num_symbols * sizeof(*buf_rle));

src/utils/huffman_encode_utils.h

@@ -51,7 +51,7 @@ int VP8LCreateCompressedHuffmanTree(const HuffmanTreeCode* const tree,
 // huffman code tree.
 void VP8LCreateHuffmanTree(uint32_t* const histogram, int tree_depth_limit,
                            uint8_t* const buf_rle, HuffmanTree* const huff_tree,
-                           HuffmanTreeCode* const tree);
+                           HuffmanTreeCode* const huff_code);
 
 #ifdef __cplusplus
 }

src/utils/rescaler_utils.c

@@ -12,66 +12,74 @@
 // Author: Skal (pascal.massimino@gmail.com)
 
 #include <assert.h>
+#include <limits.h>
 #include <stdlib.h>
 #include <string.h>
 #include "src/dsp/dsp.h"
 #include "src/utils/rescaler_utils.h"
+#include "src/utils/utils.h"
 
 //------------------------------------------------------------------------------
 
-void WebPRescalerInit(WebPRescaler* const wrk, int src_width, int src_height,
-                      uint8_t* const dst,
-                      int dst_width, int dst_height, int dst_stride,
-                      int num_channels, rescaler_t* const work) {
+int WebPRescalerInit(WebPRescaler* const rescaler,
+                     int src_width, int src_height,
+                     uint8_t* const dst,
+                     int dst_width, int dst_height, int dst_stride,
+                     int num_channels, rescaler_t* const work) {
   const int x_add = src_width, x_sub = dst_width;
   const int y_add = src_height, y_sub = dst_height;
-  wrk->x_expand = (src_width < dst_width);
-  wrk->y_expand = (src_height < dst_height);
-  wrk->src_width = src_width;
-  wrk->src_height = src_height;
-  wrk->dst_width = dst_width;
-  wrk->dst_height = dst_height;
-  wrk->src_y = 0;
-  wrk->dst_y = 0;
-  wrk->dst = dst;
-  wrk->dst_stride = dst_stride;
-  wrk->num_channels = num_channels;
+  const uint64_t total_size = 2ull * dst_width * num_channels * sizeof(*work);
+  if (!CheckSizeOverflow(total_size)) return 0;
+
+  rescaler->x_expand = (src_width < dst_width);
+  rescaler->y_expand = (src_height < dst_height);
+  rescaler->src_width = src_width;
+  rescaler->src_height = src_height;
+  rescaler->dst_width = dst_width;
+  rescaler->dst_height = dst_height;
+  rescaler->src_y = 0;
+  rescaler->dst_y = 0;
+  rescaler->dst = dst;
+  rescaler->dst_stride = dst_stride;
+  rescaler->num_channels = num_channels;
 
   // for 'x_expand', we use bilinear interpolation
-  wrk->x_add = wrk->x_expand ? (x_sub - 1) : x_add;
-  wrk->x_sub = wrk->x_expand ? (x_add - 1) : x_sub;
-  if (!wrk->x_expand) {  // fx_scale is not used otherwise
-    wrk->fx_scale = WEBP_RESCALER_FRAC(1, wrk->x_sub);
+  rescaler->x_add = rescaler->x_expand ? (x_sub - 1) : x_add;
+  rescaler->x_sub = rescaler->x_expand ? (x_add - 1) : x_sub;
+  if (!rescaler->x_expand) {  // fx_scale is not used otherwise
+    rescaler->fx_scale = WEBP_RESCALER_FRAC(1, rescaler->x_sub);
   }
   // vertical scaling parameters
-  wrk->y_add = wrk->y_expand ? y_add - 1 : y_add;
-  wrk->y_sub = wrk->y_expand ? y_sub - 1 : y_sub;
-  wrk->y_accum = wrk->y_expand ? wrk->y_sub : wrk->y_add;
-  if (!wrk->y_expand) {
+  rescaler->y_add = rescaler->y_expand ? y_add - 1 : y_add;
+  rescaler->y_sub = rescaler->y_expand ? y_sub - 1 : y_sub;
+  rescaler->y_accum = rescaler->y_expand ? rescaler->y_sub : rescaler->y_add;
+  if (!rescaler->y_expand) {
     // This is WEBP_RESCALER_FRAC(dst_height, x_add * y_add) without the cast.
-    // Its value is <= WEBP_RESCALER_ONE, because dst_height <= wrk->y_add, and
-    // wrk->x_add >= 1;
-    const uint64_t ratio =
-        (uint64_t)dst_height * WEBP_RESCALER_ONE / (wrk->x_add * wrk->y_add);
+    // Its value is <= WEBP_RESCALER_ONE, because dst_height <= rescaler->y_add
+    // and rescaler->x_add >= 1;
+    const uint64_t num = (uint64_t)dst_height * WEBP_RESCALER_ONE;
+    const uint64_t den = (uint64_t)rescaler->x_add * rescaler->y_add;
+    const uint64_t ratio = num / den;
     if (ratio != (uint32_t)ratio) {
       // When ratio == WEBP_RESCALER_ONE, we can't represent the ratio with the
       // current fixed-point precision. This happens when src_height ==
-      // wrk->y_add (which == src_height), and wrk->x_add == 1.
+      // rescaler->y_add (which == src_height), and rescaler->x_add == 1.
       // => We special-case fxy_scale = 0, in WebPRescalerExportRow().
-      wrk->fxy_scale = 0;
+      rescaler->fxy_scale = 0;
     } else {
-      wrk->fxy_scale = (uint32_t)ratio;
+      rescaler->fxy_scale = (uint32_t)ratio;
     }
-    wrk->fy_scale = WEBP_RESCALER_FRAC(1, wrk->y_sub);
+    rescaler->fy_scale = WEBP_RESCALER_FRAC(1, rescaler->y_sub);
   } else {
-    wrk->fy_scale = WEBP_RESCALER_FRAC(1, wrk->x_add);
-    // wrk->fxy_scale is unused here.
+    rescaler->fy_scale = WEBP_RESCALER_FRAC(1, rescaler->x_add);
+    // rescaler->fxy_scale is unused here.
   }
-  wrk->irow = work;
-  wrk->frow = work + num_channels * dst_width;
-  memset(work, 0, 2 * dst_width * num_channels * sizeof(*work));
+  rescaler->irow = work;
+  rescaler->frow = work + num_channels * dst_width;
+  memset(work, 0, (size_t)total_size);
 
   WebPRescalerDspInit();
+  return 1;
 }
 
 int WebPRescalerGetScaledDimensions(int src_width, int src_height,
@@ -82,6 +90,7 @@ int WebPRescalerGetScaledDimensions(int src_width, int src_height,
   {
     int width = *scaled_width;
     int height = *scaled_height;
+    const int max_size = INT_MAX / 2;
 
     // if width is unspecified, scale original proportionally to height ratio.
     if (width == 0 && src_height > 0) {
@@ -94,7 +103,7 @@ int WebPRescalerGetScaledDimensions(int src_width, int src_height,
           (int)(((uint64_t)src_height * width + src_width - 1) / src_width);
     }
     // Check if the overall dimensions still make sense.
-    if (width <= 0 || height <= 0) {
+    if (width <= 0 || height <= 0 || width > max_size || height > max_size) {
       return 0;
     }
 
@@ -107,31 +116,34 @@ int WebPRescalerGetScaledDimensions(int src_width, int src_height,
 //------------------------------------------------------------------------------
 // all-in-one calls
 
-int WebPRescaleNeededLines(const WebPRescaler* const wrk, int max_num_lines) {
-  const int num_lines = (wrk->y_accum + wrk->y_sub - 1) / wrk->y_sub;
+int WebPRescaleNeededLines(const WebPRescaler* const rescaler,
+                           int max_num_lines) {
+  const int num_lines =
+      (rescaler->y_accum + rescaler->y_sub - 1) / rescaler->y_sub;
   return (num_lines > max_num_lines) ? max_num_lines : num_lines;
 }
 
-int WebPRescalerImport(WebPRescaler* const wrk, int num_lines,
+int WebPRescalerImport(WebPRescaler* const rescaler, int num_lines,
                        const uint8_t* src, int src_stride) {
   int total_imported = 0;
-  while (total_imported < num_lines && !WebPRescalerHasPendingOutput(wrk)) {
-    if (wrk->y_expand) {
-      rescaler_t* const tmp = wrk->irow;
-      wrk->irow = wrk->frow;
-      wrk->frow = tmp;
+  while (total_imported < num_lines &&
+         !WebPRescalerHasPendingOutput(rescaler)) {
+    if (rescaler->y_expand) {
+      rescaler_t* const tmp = rescaler->irow;
+      rescaler->irow = rescaler->frow;
+      rescaler->frow = tmp;
     }
-    WebPRescalerImportRow(wrk, src);
-    if (!wrk->y_expand) {     // Accumulate the contribution of the new row.
+    WebPRescalerImportRow(rescaler, src);
+    if (!rescaler->y_expand) {    // Accumulate the contribution of the new row.
       int x;
-      for (x = 0; x < wrk->num_channels * wrk->dst_width; ++x) {
-        wrk->irow[x] += wrk->frow[x];
+      for (x = 0; x < rescaler->num_channels * rescaler->dst_width; ++x) {
+        rescaler->irow[x] += rescaler->frow[x];
       }
     }
-    ++wrk->src_y;
+    ++rescaler->src_y;
     src += src_stride;
     ++total_imported;
-    wrk->y_accum -= wrk->y_sub;
+    rescaler->y_accum -= rescaler->y_sub;
   }
   return total_imported;
 }

src/utils/rescaler_utils.h

@@ -47,12 +47,13 @@ struct WebPRescaler {
 };
 
 // Initialize a rescaler given scratch area 'work' and dimensions of src & dst.
-void WebPRescalerInit(WebPRescaler* const rescaler,
-                      int src_width, int src_height,
-                      uint8_t* const dst,
-                      int dst_width, int dst_height, int dst_stride,
-                      int num_channels,
-                      rescaler_t* const work);
+// Returns false in case of error.
+int WebPRescalerInit(WebPRescaler* const rescaler,
+                     int src_width, int src_height,
+                     uint8_t* const dst,
+                     int dst_width, int dst_height, int dst_stride,
+                     int num_channels,
+                     rescaler_t* const work);
 
 // If either 'scaled_width' or 'scaled_height' (but not both) is 0 the value
 // will be calculated preserving the aspect ratio, otherwise the values are

src/utils/utils.c

@@ -101,6 +101,9 @@ static void Increment(int* const v) {
 #if defined(MALLOC_LIMIT)
     {
       const char* const malloc_limit_str = getenv("MALLOC_LIMIT");
+#if MALLOC_LIMIT > 1
+      mem_limit = (size_t)MALLOC_LIMIT;
+#endif
       if (malloc_limit_str != NULL) {
         mem_limit = atoi(malloc_limit_str);
       }
@@ -169,16 +172,16 @@ static int CheckSizeArgumentsOverflow(uint64_t nmemb, size_t size) {
   const uint64_t total_size = nmemb * size;
   if (nmemb == 0) return 1;
   if ((uint64_t)size > WEBP_MAX_ALLOCABLE_MEMORY / nmemb) return 0;
-  if (total_size != (size_t)total_size) return 0;
+  if (!CheckSizeOverflow(total_size)) return 0;
 #if defined(PRINT_MEM_INFO) && defined(MALLOC_FAIL_AT)
   if (countdown_to_fail > 0 && --countdown_to_fail == 0) {
     return 0;    // fake fail!
   }
 #endif
-#if defined(MALLOC_LIMIT)
+#if defined(PRINT_MEM_INFO) && defined(MALLOC_LIMIT)
   if (mem_limit > 0) {
     const uint64_t new_total_mem = (uint64_t)total_mem + total_size;
-    if (new_total_mem != (size_t)new_total_mem ||
+    if (!CheckSizeOverflow(new_total_mem) ||
         new_total_mem > mem_limit) {
       return 0;   // fake fail!
     }

src/utils/utils.h

@@ -42,6 +42,10 @@ extern "C" {
 #endif
 #endif  // WEBP_MAX_ALLOCABLE_MEMORY
 
+static WEBP_INLINE int CheckSizeOverflow(uint64_t size) {
+  return size == (size_t)size;
+}
+
 // size-checking safe malloc/calloc: verify that the requested size is not too
 // large, or return NULL. You don't need to call these for constructs like
 // malloc(sizeof(foo)), but only if there's picture-dependent size involved
@@ -107,24 +111,33 @@ static WEBP_INLINE void PutLE32(uint8_t* const data, uint32_t val) {
   PutLE16(data + 2, (int)(val >> 16));
 }
 
-// Returns (int)floor(log2(n)). n must be > 0.
 // use GNU builtins where available.
 #if defined(__GNUC__) && \
     ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4)
+// Returns (int)floor(log2(n)). n must be > 0.
 static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
   return 31 ^ __builtin_clz(n);
 }
+// counts the number of trailing zero
+static WEBP_INLINE int BitsCtz(uint32_t n) { return __builtin_ctz(n); }
 #elif defined(_MSC_VER) && _MSC_VER > 1310 && \
       (defined(_M_X64) || defined(_M_IX86))
 #include <intrin.h>
 #pragma intrinsic(_BitScanReverse)
+#pragma intrinsic(_BitScanForward)
 
 static WEBP_INLINE int BitsLog2Floor(uint32_t n) {
-  unsigned long first_set_bit;
+  unsigned long first_set_bit;  // NOLINT (runtime/int)
   _BitScanReverse(&first_set_bit, n);
   return first_set_bit;
 }
-#else   // default: use the C-version.
+static WEBP_INLINE int BitsCtz(uint32_t n) {
+  unsigned long first_set_bit;  // NOLINT (runtime/int)
+  _BitScanForward(&first_set_bit, n);
+  return first_set_bit;
+}
+#else   // default: use the (slow) C-version.
+#define WEBP_HAVE_SLOW_CLZ_CTZ   // signal that the Clz/Ctz function are slow
 // Returns 31 ^ clz(n) = log2(n). This is the default C-implementation, either
 // based on table or not. Can be used as fallback if clz() is not available.
 #define WEBP_NEED_LOG_TABLE_8BIT
@@ -139,6 +152,15 @@ static WEBP_INLINE int WebPLog2FloorC(uint32_t n) {
 }
 
 static WEBP_INLINE int BitsLog2Floor(uint32_t n) { return WebPLog2FloorC(n); }
+
+static WEBP_INLINE int BitsCtz(uint32_t n) {
+  int i;
+  for (i = 0; i < 32; ++i, n >>= 1) {
+    if (n & 1) return i;
+  }
+  return 32;
+}
+
 #endif
 
 //------------------------------------------------------------------------------

swig/libwebp_python_wrap.c

@@ -2764,7 +2764,7 @@ SWIG_Python_DestroyModule(void *vptr)
 SWIGRUNTIME void
 SWIG_Python_SetModule(swig_module_info *swig_module) {
 #if PY_VERSION_HEX >= 0x03000000
- /* Add a dummy module object into sys.modules */
+ /* Add a placeholder module object into sys.modules */
   PyObject *module = PyImport_AddModule((char*)"swig_runtime_data" SWIG_RUNTIME_VERSION);
 #else
   static PyMethodDef swig_empty_runtime_method_table[] = { {NULL, NULL, 0, NULL} }; /* Sentinel */

tests/fuzzer/advanced_api_fuzzer.c

@@ -14,10 +14,13 @@
 //
 ////////////////////////////////////////////////////////////////////////////////
 
+#include <string.h>
+
 #include "./fuzz_utils.h"
-#include "webp/decode.h"
+#include "src/webp/decode.h"
 
 int LLVMFuzzerTestOneInput(const uint8_t* const data, size_t size) {
+  int i;
   WebPDecoderConfig config;
   if (!WebPInitDecoderConfig(&config)) return 0;
   if (WebPGetFeatures(data, size, &config.input) != VP8_STATUS_OK) return 0;
@@ -59,39 +62,53 @@ int LLVMFuzzerTestOneInput(const uint8_t* const data, size_t size) {
   config.output.colorspace = (WEBP_CSP_MODE)(value % MODE_LAST);
 #endif  // WEBP_REDUCE_CSP
 
-  if (size % 3) {
-    // Decodes incrementally in chunks of increasing size.
-    WebPIDecoder* idec = WebPIDecode(NULL, 0, &config);
-    if (!idec) return 0;
-    VP8StatusCode status;
-    if (size & 8) {
-      size_t available_size = value + 1;
-      while (1) {
-        if (available_size > size) available_size = size;
-        status = WebPIUpdate(idec, data, available_size);
-        if (status != VP8_STATUS_SUSPENDED || available_size == size) break;
-        available_size *= 2;
-      }
-    } else {
-      // WebPIAppend expects new data and its size with each call.
-      // Implemented here by simply advancing the pointer into data.
-      const uint8_t* new_data = data;
-      size_t new_size = value + 1;
-      while (1) {
-        if (new_data + new_size > data + size) {
-          new_size = data + size - new_data;
-        }
-        status = WebPIAppend(idec, new_data, new_size);
-        if (status != VP8_STATUS_SUSPENDED || new_size == 0) break;
-        new_data += new_size;
-        new_size *= 2;
+  for (i = 0; i < 2; ++i) {
+    if (i == 1) {
+      // Use the bitstream data to generate extreme ranges for the options. An
+      // alternative approach would be to use a custom corpus containing webp
+      // files prepended with sizeof(config.options) zeroes to allow the fuzzer
+      // to modify these independently.
+      const int data_offset = 50;
+      if (size > data_offset + sizeof(config.options)) {
+        memcpy(&config.options, data + data_offset, sizeof(config.options));
+      } else {
+        break;
       }
     }
-    WebPIDelete(idec);
-  } else {
-    WebPDecode(data, size, &config);
-  }
+    if (size % 3) {
+      // Decodes incrementally in chunks of increasing size.
+      WebPIDecoder* idec = WebPIDecode(NULL, 0, &config);
+      if (!idec) return 0;
+      VP8StatusCode status;
+      if (size & 8) {
+        size_t available_size = value + 1;
+        while (1) {
+          if (available_size > size) available_size = size;
+          status = WebPIUpdate(idec, data, available_size);
+          if (status != VP8_STATUS_SUSPENDED || available_size == size) break;
+          available_size *= 2;
+        }
+      } else {
+        // WebPIAppend expects new data and its size with each call.
+        // Implemented here by simply advancing the pointer into data.
+        const uint8_t* new_data = data;
+        size_t new_size = value + 1;
+        while (1) {
+          if (new_data + new_size > data + size) {
+            new_size = data + size - new_data;
+          }
+          status = WebPIAppend(idec, new_data, new_size);
+          if (status != VP8_STATUS_SUSPENDED || new_size == 0) break;
+          new_data += new_size;
+          new_size *= 2;
+        }
+      }
+      WebPIDelete(idec);
+    } else {
+      WebPDecode(data, size, &config);
+    }
 
-  WebPFreeDecBuffer(&config.output);
+    WebPFreeDecBuffer(&config.output);
+  }
   return 0;
 }

tests/fuzzer/animation_api_fuzzer.c

@@ -15,9 +15,9 @@
 ////////////////////////////////////////////////////////////////////////////////
 
 #include "./fuzz_utils.h"
-#include "webp/decode.h"
-#include "webp/demux.h"
-#include "webp/mux_types.h"
+#include "src/webp/decode.h"
+#include "src/webp/demux.h"
+#include "src/webp/mux_types.h"
 
 int LLVMFuzzerTestOneInput(const uint8_t* const data, size_t size) {
   WebPData webp_data;

tests/fuzzer/animdecoder_fuzzer.cc

@@ -16,7 +16,7 @@
 
 #include "examples/anim_util.h"
 #include "imageio/imageio_util.h"
-#include "webp/demux.h"
+#include "src/webp/demux.h"
 
 extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   // WebPAnimDecoderGetInfo() is too late to check the canvas size as

tests/fuzzer/animencoder_fuzzer.cc

@@ -18,8 +18,8 @@
 #include <stdlib.h>
 
 #include "./fuzz_utils.h"
-#include "webp/encode.h"
-#include "webp/mux.h"
+#include "src/webp/encode.h"
+#include "src/webp/mux.h"
 
 namespace {
 

tests/fuzzer/enc_dec_fuzzer.cc

@@ -18,8 +18,8 @@
 #include <stdlib.h>
 
 #include "./fuzz_utils.h"
-#include "webp/decode.h"
-#include "webp/encode.h"
+#include "src/webp/decode.h"
+#include "src/webp/encode.h"
 
 namespace {
 

tests/fuzzer/fuzz_utils.h

@@ -23,8 +23,8 @@
 #include "./img_alpha.h"
 #include "./img_grid.h"
 #include "./img_peak.h"
-#include "dsp/dsp.h"
-#include "webp/encode.h"
+#include "src/dsp/dsp.h"
+#include "src/webp/encode.h"
 
 //------------------------------------------------------------------------------
 // Arbitrary limits to prevent OOM, timeout, or slow execution.

tests/fuzzer/mux_demux_api_fuzzer.c

@@ -15,8 +15,8 @@
 ////////////////////////////////////////////////////////////////////////////////
 
 #include "./fuzz_utils.h"
-#include "webp/demux.h"
-#include "webp/mux.h"
+#include "src/webp/demux.h"
+#include "src/webp/mux.h"
 
 int LLVMFuzzerTestOneInput(const uint8_t* const data, size_t size) {
   WebPData webp_data;

tests/fuzzer/simple_api_fuzzer.c

@@ -15,7 +15,7 @@
 ////////////////////////////////////////////////////////////////////////////////
 
 #include "./fuzz_utils.h"
-#include "webp/decode.h"
+#include "src/webp/decode.h"
 
 int LLVMFuzzerTestOneInput(const uint8_t* const data, size_t size) {
   int w, h;

xcframeworkbuild.sh

@@ -48,6 +48,7 @@ PLATFORMS[$MACOS_CATALYST]="MacOSX-Catalyst-x86_64"
 if [[ "${XCODE%%.*}" -ge 12 ]]; then
   PLATFORMS[$MACOS]+=" MacOSX-arm64"
   PLATFORMS[$MACOS_CATALYST]+=" MacOSX-Catalyst-arm64"
+  PLATFORMS[$IOS_SIMULATOR]+=" iPhoneSimulator-arm64"
 elif [[ "${XCODE%%.*}" -eq 11 ]]; then
   cat << EOF
 WARNING: Xcode 12.0 or higher is required to build targets for
@@ -140,7 +141,9 @@ for (( i = 0; i < $NUM_PLATFORMS; ++i )); do
     CFLAGS="-pipe -isysroot ${SDKROOT} -O3 -DNDEBUG"
     case "${PLATFORM}" in
       iPhone*)
-        CFLAGS+=" -miphoneos-version-min=${IOS_MIN_VERSION} -fembed-bitcode"
+        CFLAGS+=" -fembed-bitcode"
+        CFLAGS+=" -target ${ARCH}-apple-ios${IOS_MIN_VERSION}"
+        [[ "${PLATFORM}" == *Simulator* ]] && CFLAGS+="-simulator"
         ;;
       MacOSX-Catalyst*)
         CFLAGS+=" -target"