ffmpeg.gyp - third_party/ffmpeg - Git at Google

 # Copyright (c) 2011 The Chromium Authors. All rights reserved.
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.

 # There's a couple key GYP variables that control how FFmpeg is built:
 #   ffmpeg_branding
 #     Controls whether we build the Chromium or Google Chrome version of
 #     FFmpeg.  The Google Chrome version contains additional codecs.
 #     Typical values are Chromium, Chrome, ChromiumOS, and ChromeOS.
 #   ffmpeg_component
 #     Set true to build ffmpeg as a shared library. NOTE: this means we should
 #     always consult the value of 'ffmpeg_component' instead of 'component' for
 #     this file. This helps linux chromium packagers that swap out our
 #     ffmpeg.so with their own. See discussion here
 #     https://groups.google.com/a/chromium.org/forum/#!msg/chromium-packagers/R5rcZXWxBEQ/B6k0zzmJbvcJ
 #

 {
   'target_defaults': {
     'variables': {
       # Since we are not often debugging FFmpeg, and performance is
       # unacceptable without optimization, freeze the optimizations to -O2.
       # If someone really wants -O1 , they can change these in their checkout.
       # If you want -O0, see the Gotchas in README.Chromium for why that
       # won't work.
       'release_optimize': '2',
       'debug_optimize': '2',
       'mac_debug_optimization': '2',
       # In addition to the above reasons, /Od optimization won't remove symbols
       # that are under "if (0)" style sections.  Which lead to link time errors
       # when for example it tries to link an ARM symbol on X86.
       'win_debug_Optimization': '2',
       # Run time checks are incompatible with any level of optimizations.
       'win_debug_RuntimeChecks': '0',
       'conditions': [
         ['OS == "win"', {
           # Setting the optimizations to 'speed' or to 'max' results in a lot of
           # unresolved symbols. The only supported mode is 'size_no_ltcg' (see
           # crbug.com/264459).
           'optimize' :'size_no_ltcg',
         }],
       ],
     },
   },
   'variables': {
     # Make sure asm_sources is always defined even if an arch doesn't have any
     # asm sources (e.g. mips or x86 with forcefully disabled asm).
     'asm_sources': [
     ],


     # Allow overriding the selection of which FFmpeg binaries to copy via an
     # environment variable.  Affects the ffmpeg_binaries target.
     'conditions': [
       # Android ia32 can't handle textrels and ffmpeg can't compile without them.
       # http://crbug.com/559379
       ['(OS == "android" and target_arch == "ia32") or msan==1', {
         'disable_ffmpeg_asm%': 1,
       }, {
         'disable_ffmpeg_asm%': 0,
       }],
       ['target_arch == "arm" and arm_version >= 7 and arm_neon == 1', {
         # Need a separate config for arm+neon vs arm
         'ffmpeg_config%': 'arm-neon',
       }, {
         'ffmpeg_config%': '<(target_arch)',
       }],
       ['OS == "mac" or OS == "win" or OS == "openbsd" or OS == "android"', {
         'os_config%': '<(OS)',
       }, {  # all other Unix OS's use the linux config
         'conditions': [
           ['msan==1', {
             # MemorySanitizer doesn't like assembly code.
             'os_config%': 'linux-noasm',
           }, {
             'os_config%': 'linux',
           }]
         ],
       }],
       ['chromeos == 1', {
         'ffmpeg_branding%': '<(branding)OS',
       }, {  # otherwise, assume Chrome/Chromium.
         'ffmpeg_branding%': '<(branding)',
       }],
     ],

     'ffmpeg_component%': '<(component)',

     # Locations for generated artifacts.
     'shared_generated_dir': '<(SHARED_INTERMEDIATE_DIR)/third_party/ffmpeg',

     # Stub generator script and signatures of all functions used by Chrome.
     'generate_stubs_script': '../../tools/generate_stubs/generate_stubs.py',
     'sig_files': ['chromium/ffmpeg.sigs'],
   },
   'conditions': [
     ['(target_arch == "ia32" or target_arch == "x64") and disable_ffmpeg_asm == 0', {
       'targets': [
         {
           'target_name': 'ffmpeg_yasm',
           'type': 'static_library',
           'includes': [
             'ffmpeg_generated.gypi',
             '../yasm/yasm_compile.gypi',
           ],
           'sources': [
             '<@(asm_sources)',
             # XCode doesn't want to link a pure assembly target and will fail
             # to link when it creates an empty file list.  So add a dummy file
             # keep the linker happy.  See http://crbug.com/157073
             'xcode_hack.c',
           ],
           'variables': {
             # Path to platform configuration files.
             'platform_config_root': 'chromium/config/<(ffmpeg_branding)/<(os_config)/<(ffmpeg_config)',

             'conditions': [
               ['target_arch == "ia32"', {
                 'more_yasm_flags': [
                   '-DARCH_X86_32',
                  ],
               }, {
                 'more_yasm_flags': [
                   '-DARCH_X86_64',
                 ],
               }],
               ['OS == "mac"', {
                 'more_yasm_flags': [
                   # Necessary to ensure symbols end up with a _ prefix; added by
                   # yasm_compile.gypi for Windows, but not Mac.
                   '-DPREFIX',
                 ]
               }],
             ],
             'yasm_flags': [
               '-DPIC',
               '>@(more_yasm_flags)',
               '-I', '<(platform_config_root)',
               '-I', 'libavcodec/x86/',
               '-I', 'libavutil/x86/',
               '-I', '.',
               # Disable warnings, prevents log spam for things we won't fix.
               '-w',
               '-P', 'config.asm',
             ],
             'yasm_includes': [
               # Sets visibility hidden for cglobal functions. Explicitly included
               # to avoid overlooking changes to this file in incremental builds.
               'libavutil/x86/x86inc.asm',
             ],
             'yasm_output_path': '<(shared_generated_dir)/yasm'
           },
         },
       ] # targets
     }], # (target_arch == "ia32" or target_arch == "x64")
   ],

   'targets': [{
     'target_name': 'ffmpeg',
     'type': '<(ffmpeg_component)',
     'variables': {
       # Path to platform configuration files.
       'platform_config_root': 'chromium/config/<(ffmpeg_branding)/<(os_config)/<(ffmpeg_config)',
     },
     'conditions': [
       # This looks crazy, but it seems to be the only way to make this work. Without this
       # conditions block, platform_config_root will be undefined when expanded below.
       ['1 == 1',
         {
           'direct_dependent_settings': {
             'include_dirs': [
               '../..',  # The chromium 'src' directory.
               '<(platform_config_root)',
               '.',
             ],
           },
           'includes': [
             'ffmpeg_generated.gypi',
           ],
           'sources': [
             '<@(c_sources)',
             '<(platform_config_root)/config.h',
             '<(platform_config_root)/libavutil/avconfig.h',
           ],
           'include_dirs': [
             '<(platform_config_root)',
             '.',
           ],
           'defines': [
             'HAVE_AV_CONFIG_H',
             '_POSIX_C_SOURCE=200112',
             '_XOPEN_SOURCE=600',
             'PIC',
             # Disable deprecated features that generate spammy warnings.
             # BUILD.gn & media/ffmpeg/ffmpeg_common.h must be kept in sync.
             'FF_API_CONVERGENCE_DURATION=0',
             # Upstream libavcodec/utils.c still uses the deprecated
             # av_dup_packet(), causing deprecation warnings.
             # The normal fix for such things is to disable the feature as below,
             # but the upstream code does not yet compile with it disabled.
             # (In this case, the fix is replacing the call with a new function.)
             # In the meantime, we directly disable those warnings in the C file.
             # 'FF_API_AVPACKET_OLD_API=0',
           ],
           'variables': {
             'clang_warning_flags': [
               '-Wno-absolute-value',
               # ffmpeg uses its own deprecated functions.
               '-Wno-deprecated-declarations',
               # ffmpeg doesn't care about pointer constness.
               '-Wno-incompatible-pointer-types',
               # ffmpeg doesn't follow usual parentheses conventions.
               '-Wno-parentheses',
               # ffmpeg doesn't care about pointer signedness.
               '-Wno-pointer-sign',
               # ffmpeg doesn't believe in exhaustive switch statements.
               '-Wno-switch',
               # matroskadec.c has a "failed:" label that's only used if some
               # CONFIG_ flags we don't set are set.
               '-Wno-unused-label',
               # ffmpeg has a lot of unused variables.
               '-Wno-unused-variable',
               # This fires on `av_assert0(!"valid element size")` in utils.c
               '-Wno-string-conversion',
               # This fires on `pos_min` and `pos_max` in
               # autorename_libavformat_utils.c
               '-Wno-sometimes-uninitialized',
               # ffmpeg contains static functions in header files, which lead
               # to unused function warnings. There are a few legit unused
               # functions too.
               '-Wno-unused-function',
               # vp3data.h's vp31_inter_dequant stores '128' in an int8_t array.
               '-Wno-constant-conversion',
             ],
           },
           'cflags': [
             '-fPIC',
             '-fomit-frame-pointer',
             # ffmpeg uses its own deprecated functions.
             '-Wno-deprecated-declarations',
           ],
           # Silence a warning in libc++ builds (C code doesn't need this flag).
           'ldflags!': [ '-stdlib=libc++', ],
           'conditions': [
             ['(target_arch == "ia32" or target_arch == "x64") and disable_ffmpeg_asm == 0', {
               'dependencies': [
                 'ffmpeg_yasm',
               ],
             }],
             ['clang != 1', {
               'cflags': [
                 # gcc doesn't have flags for specific warnings, so disable them
                 # all.
                 '-w',
               ],
             }],
             ['target_arch == "ia32"', {
               # Turn off valgrind build option that breaks ffmpeg builds.
               'cflags!': [
                 '-fno-omit-frame-pointer',
               ],
               'debug_extra_cflags!': [
                 '-fno-omit-frame-pointer',
               ],
               'release_extra_cflags!': [
                 '-fno-omit-frame-pointer',
               ],
             }],  # target_arch == "ia32"
             ['target_arch == "arm" or target_arch == "arm64"', {
               # On arm we use gcc to compile the assembly.
               'sources': [
                 '<@(asm_sources)',
               ],
               'conditions': [
                 ['arm_float_abi == "hard"', {
                   'cflags': [
                     '-DHAVE_VFP_ARGS=1'
                   ],
                 }, {
                   'cflags': [
                     '-DHAVE_VFP_ARGS=0'
                   ],
                 }],
                 ['clang==1', {
                   # TODO(hans) Enable integrated-as (crbug.com/124610).
                   'cflags': [ '-fno-integrated-as' ],
                   'conditions': [
                     ['OS == "android"', {
                       # Else /usr/bin/as gets picked up.
                       'cflags': [ '-B<(android_toolchain)' ],
                     }],
                   ],
                 }],
               ],
             }],
             ['os_posix == 1 and OS != "mac"', {
               'defines': [
                 '_ISOC99_SOURCE',
                 '_LARGEFILE_SOURCE',
                 # BUG(ihf): ffmpeg compiles with this define. But according to
                 # ajwong: I wouldn't change _FILE_OFFSET_BITS.  That's a scary change
                 # because it affects the default length of off_t, and fpos_t,
                 # which can cause strange problems if the loading code doesn't
                 # have it set and you start passing FILE*s or file descriptors
                 # between symbol contexts.
                 # '_FILE_OFFSET_BITS=64',
               ],
               'cflags': [
                 '-std=c99',
                 '-pthread',
                 '-fno-math-errno',
                 '-fno-signed-zeros',
                 '-fno-tree-vectorize',
               ],
               'link_settings': {
                 'libraries': [
                 ],
               },
               'conditions': [
                 ['OS != "android"', {
                   'link_settings': {
                     # OS=android requires that both -lz and -lm occur
                     # after -lc++_shared on the link command
                     # line. Android link rules already include -lm, and
                     # we get -lz as a transitive dependency of
                     # libandroid.so, so simply moving both to the
                     # non-Android section solves the problem.
                     #
                     # The root cause of this problem is certain system
                     # libraries (libm starting with MNC and libz before
                     # MNC, among others) re-export the libgcc unwinder,
                     # and libc++ exports the libc++abi unwinder. As we
                     # build against libc++ headers, libc++ must be the
                     # first in the runtime symbol lookup order (among
                     # all unwinder-providing libraries).
                     'libraries': [
                       '-lm',
                       '-lz',
                       '-lrt',
                     ],
                   },
                 }],
                 ['ffmpeg_component == "shared_library"', {
                   # Export all symbols when building as component.
                   'cflags!': [
                     '-fvisibility=hidden',
                   ],
                   # Fixes warnings PIC relocation when building as component.
                   # *WARNING* -- DO NOT put this inside of a link_settings
                   # section or these flags will be propagated outside of the
                   # ffmpeg target and cause debug allocator crashes.
                   'ldflags': [
                     '-Wl,-Bsymbolic',
                     '-L<(shared_generated_dir)',
                   ],
                 }],
               ],
             }],  # os_posix == 1 and OS != "mac"
             ['OS == "openbsd"', {
               # OpenBSD's gcc (4.2.1) does not support this flag
               'cflags!': [
                 '-fno-signed-zeros',
               ],
             }],
             ['OS == "mac"', {
               'defines': [
                 '_DARWIN_C_SOURCE',
               ],
               'conditions': [
                 ['ffmpeg_component == "shared_library"', {
                   'xcode_settings': {
                     # GCC version of no -fvisiliity=hidden. Ensures that all
                     # symbols are exported for component builds.
                     'GCC_SYMBOLS_PRIVATE_EXTERN': 'NO',
                   },
                 }],
               ],
             }],  # OS == "mac"
             ['OS == "win"', {
               # Disable /analyze because of internal compiler error on log.c
               # with VS 2015 Update 1 when building with /analyze:
               # https://connect.microsoft.com/VisualStudio/feedback/details/2299303
               'msvs_settings': {
                 'VCCLCompilerTool': {
                   'AdditionalOptions!': [ '/analyze:WX-' ]
                 },
               },
               # TODO(dalecurtis): We should fix these.  http://crbug.com/154421
               'msvs_disabled_warnings': [
                 4996, 4018, 4090, 4305, 4133, 4146, 4554, 4028, 4334, 4101, 4102,
                 4116, 4307, 4273, 4005, 4056, 4756,
               ],
               'conditions': [
                 ['clang == 1', {
                   'msvs_settings': {
                     'VCCLCompilerTool': {
                       # This corresponds to msvs_disabled_warnings 4273 above.
                       'AdditionalOptions': [ '-Wno-inconsistent-dllimport' ],
                     },
                   },
                 }],
                 ['MSVS_VERSION == "2013" or MSVS_VERSION == "2013e"', {
                   'defines': [
                     'inline=__inline',
                     'strtoll=_strtoi64',
                     '_ISOC99_SOURCE',
                     '_LARGEFILE_SOURCE',
                     'HAVE_AV_CONFIG_H',
                     'strtod=avpriv_strtod',
                     'snprintf=avpriv_snprintf',
                     '_snprintf=avpriv_snprintf',
                     'vsnprintf=avpriv_vsnprintf',
                   ],
                 }],
                 ['target_arch == "x64"', {
                   # TODO(wolenetz): We should fix this.  http://crbug.com/171009
                   'msvs_disabled_warnings' : [
                     4267
                   ],
                 }],
                 ['ffmpeg_component == "shared_library"', {
                   # Fix warnings about a local symbol being inefficiently imported.
                   'msvs_settings': {
                     'VCCLCompilerTool': {
                       'AdditionalOptions': [
                         '/FIcompat/msvcrt/snprintf.h',
                         '/FIcompat/msvcrt/strtod.h',
                       ],
                     },
                   },
                   'sources': [
                     '<(shared_generated_dir)/ffmpeg.def',
                   ],
                   'conditions': [
                     ['OS=="win" and win_use_allocator_shim==1', {
                       'dependencies': [
                         '../../base/allocator/allocator.gyp:allocator',
                       ],
                       'sources': [
                         'chromium/dllmain.cc',
                       ],
                     }],
                   ],
                   'actions': [
                     {
                       'action_name': 'generate_def',
                       'inputs': [
                         '<(generate_stubs_script)',
                         '<@(sig_files)',
                       ],
                       'outputs': [
                         '<(shared_generated_dir)/ffmpeg.def',
                       ],
                       'action': ['python',
                                  '<(generate_stubs_script)',
                                  '-i', '<(INTERMEDIATE_DIR)',
                                  '-o', '<(shared_generated_dir)',
                                  '-t', 'windows_def',
                                  '-m', 'ffmpeg.dll',
                                  '<@(_inputs)',
                       ],
                       'message': 'Generating FFmpeg export definitions',
                     },
                   ],
                 }],
               ],
             }],
           ],
         },
       ],
     ],
   }],
 }
	# Copyright (c) 2011 The Chromium Authors. All rights reserved.
	# Use of this source code is governed by a BSD-style license that can be
	# found in the LICENSE file.

	# There's a couple key GYP variables that control how FFmpeg is built:
	# ffmpeg_branding
	# Controls whether we build the Chromium or Google Chrome version of
	# FFmpeg. The Google Chrome version contains additional codecs.
	# Typical values are Chromium, Chrome, ChromiumOS, and ChromeOS.
	# ffmpeg_component
	# Set true to build ffmpeg as a shared library. NOTE: this means we should
	# always consult the value of 'ffmpeg_component' instead of 'component' for
	# this file. This helps linux chromium packagers that swap out our
	# ffmpeg.so with their own. See discussion here
	# https://groups.google.com/a/chromium.org/forum/#!msg/chromium-packagers/R5rcZXWxBEQ/B6k0zzmJbvcJ
	#

	{
	'target_defaults': {
	'variables': {
	# Since we are not often debugging FFmpeg, and performance is
	# unacceptable without optimization, freeze the optimizations to -O2.
	# If someone really wants -O1 , they can change these in their checkout.
	# If you want -O0, see the Gotchas in README.Chromium for why that
	# won't work.
	'release_optimize': '2',
	'debug_optimize': '2',
	'mac_debug_optimization': '2',
	# In addition to the above reasons, /Od optimization won't remove symbols
	# that are under "if (0)" style sections. Which lead to link time errors
	# when for example it tries to link an ARM symbol on X86.
	'win_debug_Optimization': '2',
	# Run time checks are incompatible with any level of optimizations.
	'win_debug_RuntimeChecks': '0',
	'conditions': [
	['OS == "win"', {
	# Setting the optimizations to 'speed' or to 'max' results in a lot of
	# unresolved symbols. The only supported mode is 'size_no_ltcg' (see
	# crbug.com/264459).
	'optimize' :'size_no_ltcg',
	}],
	],
	},
	},
	'variables': {
	# Make sure asm_sources is always defined even if an arch doesn't have any
	# asm sources (e.g. mips or x86 with forcefully disabled asm).
	'asm_sources': [
	],


	# Allow overriding the selection of which FFmpeg binaries to copy via an
	# environment variable. Affects the ffmpeg_binaries target.
	'conditions': [
	# Android ia32 can't handle textrels and ffmpeg can't compile without them.
	# http://crbug.com/559379
	['(OS == "android" and target_arch == "ia32") or msan==1', {
	'disable_ffmpeg_asm%': 1,
	}, {
	'disable_ffmpeg_asm%': 0,
	}],
	['target_arch == "arm" and arm_version >= 7 and arm_neon == 1', {
	# Need a separate config for arm+neon vs arm
	'ffmpeg_config%': 'arm-neon',
	}, {
	'ffmpeg_config%': '<(target_arch)',
	}],
	['OS == "mac" or OS == "win" or OS == "openbsd" or OS == "android"', {
	'os_config%': '<(OS)',
	}, { # all other Unix OS's use the linux config
	'conditions': [
	['msan==1', {
	# MemorySanitizer doesn't like assembly code.
	'os_config%': 'linux-noasm',
	}, {
	'os_config%': 'linux',
	}]
	],
	}],
	['chromeos == 1', {
	'ffmpeg_branding%': '<(branding)OS',
	}, { # otherwise, assume Chrome/Chromium.
	'ffmpeg_branding%': '<(branding)',
	}],
	],

	'ffmpeg_component%': '<(component)',

	# Locations for generated artifacts.
	'shared_generated_dir': '<(SHARED_INTERMEDIATE_DIR)/third_party/ffmpeg',

	# Stub generator script and signatures of all functions used by Chrome.
	'generate_stubs_script': '../../tools/generate_stubs/generate_stubs.py',
	'sig_files': ['chromium/ffmpeg.sigs'],
	},
	'conditions': [
	['(target_arch == "ia32" or target_arch == "x64") and disable_ffmpeg_asm == 0', {
	'targets': [
	{
	'target_name': 'ffmpeg_yasm',
	'type': 'static_library',
	'includes': [
	'ffmpeg_generated.gypi',
	'../yasm/yasm_compile.gypi',
	],
	'sources': [
	'<@(asm_sources)',
	# XCode doesn't want to link a pure assembly target and will fail
	# to link when it creates an empty file list. So add a dummy file
	# keep the linker happy. See http://crbug.com/157073
	'xcode_hack.c',
	],
	'variables': {
	# Path to platform configuration files.
	'platform_config_root': 'chromium/config/<(ffmpeg_branding)/<(os_config)/<(ffmpeg_config)',

	'conditions': [
	['target_arch == "ia32"', {
	'more_yasm_flags': [
	'-DARCH_X86_32',
	],
	}, {
	'more_yasm_flags': [
	'-DARCH_X86_64',
	],
	}],
	['OS == "mac"', {
	'more_yasm_flags': [
	# Necessary to ensure symbols end up with a _ prefix; added by
	# yasm_compile.gypi for Windows, but not Mac.
	'-DPREFIX',
	]
	}],
	],
	'yasm_flags': [
	'-DPIC',
	'>@(more_yasm_flags)',
	'-I', '<(platform_config_root)',
	'-I', 'libavcodec/x86/',
	'-I', 'libavutil/x86/',
	'-I', '.',
	# Disable warnings, prevents log spam for things we won't fix.
	'-w',
	'-P', 'config.asm',
	],
	'yasm_includes': [
	# Sets visibility hidden for cglobal functions. Explicitly included
	# to avoid overlooking changes to this file in incremental builds.
	'libavutil/x86/x86inc.asm',
	],
	'yasm_output_path': '<(shared_generated_dir)/yasm'
	},
	},
	] # targets
	}], # (target_arch == "ia32" or target_arch == "x64")
	],

	'targets': [{
	'target_name': 'ffmpeg',
	'type': '<(ffmpeg_component)',
	'variables': {
	# Path to platform configuration files.
	'platform_config_root': 'chromium/config/<(ffmpeg_branding)/<(os_config)/<(ffmpeg_config)',
	},
	'conditions': [
	# This looks crazy, but it seems to be the only way to make this work. Without this
	# conditions block, platform_config_root will be undefined when expanded below.
	['1 == 1',
	{
	'direct_dependent_settings': {
	'include_dirs': [
	'../..', # The chromium 'src' directory.
	'<(platform_config_root)',
	'.',
	],
	},
	'includes': [
	'ffmpeg_generated.gypi',
	],
	'sources': [
	'<@(c_sources)',
	'<(platform_config_root)/config.h',
	'<(platform_config_root)/libavutil/avconfig.h',
	],
	'include_dirs': [
	'<(platform_config_root)',
	'.',
	],
	'defines': [
	'HAVE_AV_CONFIG_H',
	'_POSIX_C_SOURCE=200112',
	'_XOPEN_SOURCE=600',
	'PIC',
	# Disable deprecated features that generate spammy warnings.
	# BUILD.gn & media/ffmpeg/ffmpeg_common.h must be kept in sync.
	'FF_API_CONVERGENCE_DURATION=0',
	# Upstream libavcodec/utils.c still uses the deprecated
	# av_dup_packet(), causing deprecation warnings.
	# The normal fix for such things is to disable the feature as below,
	# but the upstream code does not yet compile with it disabled.
	# (In this case, the fix is replacing the call with a new function.)
	# In the meantime, we directly disable those warnings in the C file.
	# 'FF_API_AVPACKET_OLD_API=0',
	],
	'variables': {
	'clang_warning_flags': [
	'-Wno-absolute-value',
	# ffmpeg uses its own deprecated functions.
	'-Wno-deprecated-declarations',
	# ffmpeg doesn't care about pointer constness.
	'-Wno-incompatible-pointer-types',
	# ffmpeg doesn't follow usual parentheses conventions.
	'-Wno-parentheses',
	# ffmpeg doesn't care about pointer signedness.
	'-Wno-pointer-sign',
	# ffmpeg doesn't believe in exhaustive switch statements.
	'-Wno-switch',
	# matroskadec.c has a "failed:" label that's only used if some
	# CONFIG_ flags we don't set are set.
	'-Wno-unused-label',
	# ffmpeg has a lot of unused variables.
	'-Wno-unused-variable',
	# This fires on `av_assert0(!"valid element size")` in utils.c
	'-Wno-string-conversion',
	# This fires on `pos_min` and `pos_max` in
	# autorename_libavformat_utils.c
	'-Wno-sometimes-uninitialized',
	# ffmpeg contains static functions in header files, which lead
	# to unused function warnings. There are a few legit unused
	# functions too.
	'-Wno-unused-function',
	# vp3data.h's vp31_inter_dequant stores '128' in an int8_t array.
	'-Wno-constant-conversion',
	],
	},
	'cflags': [
	'-fPIC',
	'-fomit-frame-pointer',
	# ffmpeg uses its own deprecated functions.
	'-Wno-deprecated-declarations',
	],
	# Silence a warning in libc++ builds (C code doesn't need this flag).
	'ldflags!': [ '-stdlib=libc++', ],
	'conditions': [
	['(target_arch == "ia32" or target_arch == "x64") and disable_ffmpeg_asm == 0', {
	'dependencies': [
	'ffmpeg_yasm',
	],
	}],
	['clang != 1', {
	'cflags': [
	# gcc doesn't have flags for specific warnings, so disable them
	# all.
	'-w',
	],
	}],
	['target_arch == "ia32"', {
	# Turn off valgrind build option that breaks ffmpeg builds.
	'cflags!': [
	'-fno-omit-frame-pointer',
	],
	'debug_extra_cflags!': [
	'-fno-omit-frame-pointer',
	],
	'release_extra_cflags!': [
	'-fno-omit-frame-pointer',
	],
	}], # target_arch == "ia32"
	['target_arch == "arm" or target_arch == "arm64"', {
	# On arm we use gcc to compile the assembly.
	'sources': [
	'<@(asm_sources)',
	],
	'conditions': [
	['arm_float_abi == "hard"', {
	'cflags': [
	'-DHAVE_VFP_ARGS=1'
	],
	}, {
	'cflags': [
	'-DHAVE_VFP_ARGS=0'
	],
	}],
	['clang==1', {
	# TODO(hans) Enable integrated-as (crbug.com/124610).
	'cflags': [ '-fno-integrated-as' ],
	'conditions': [
	['OS == "android"', {
	# Else /usr/bin/as gets picked up.
	'cflags': [ '-B<(android_toolchain)' ],
	}],
	],
	}],
	],
	}],
	['os_posix == 1 and OS != "mac"', {
	'defines': [
	'_ISOC99_SOURCE',
	'_LARGEFILE_SOURCE',
	# BUG(ihf): ffmpeg compiles with this define. But according to
	# ajwong: I wouldn't change _FILE_OFFSET_BITS. That's a scary change
	# because it affects the default length of off_t, and fpos_t,
	# which can cause strange problems if the loading code doesn't
	# have it set and you start passing FILE*s or file descriptors
	# between symbol contexts.
	# '_FILE_OFFSET_BITS=64',
	],
	'cflags': [
	'-std=c99',
	'-pthread',
	'-fno-math-errno',
	'-fno-signed-zeros',
	'-fno-tree-vectorize',
	],
	'link_settings': {
	'libraries': [
	],
	},
	'conditions': [
	['OS != "android"', {
	'link_settings': {
	# OS=android requires that both -lz and -lm occur
	# after -lc++_shared on the link command
	# line. Android link rules already include -lm, and
	# we get -lz as a transitive dependency of
	# libandroid.so, so simply moving both to the
	# non-Android section solves the problem.
	#
	# The root cause of this problem is certain system
	# libraries (libm starting with MNC and libz before
	# MNC, among others) re-export the libgcc unwinder,
	# and libc++ exports the libc++abi unwinder. As we
	# build against libc++ headers, libc++ must be the
	# first in the runtime symbol lookup order (among
	# all unwinder-providing libraries).
	'libraries': [
	'-lm',
	'-lz',
	'-lrt',
	],
	},
	}],
	['ffmpeg_component == "shared_library"', {
	# Export all symbols when building as component.
	'cflags!': [
	'-fvisibility=hidden',
	],
	# Fixes warnings PIC relocation when building as component.
	# WARNING -- DO NOT put this inside of a link_settings
	# section or these flags will be propagated outside of the
	# ffmpeg target and cause debug allocator crashes.
	'ldflags': [
	'-Wl,-Bsymbolic',
	'-L<(shared_generated_dir)',
	],
	}],
	],
	}], # os_posix == 1 and OS != "mac"
	['OS == "openbsd"', {
	# OpenBSD's gcc (4.2.1) does not support this flag
	'cflags!': [
	'-fno-signed-zeros',
	],
	}],
	['OS == "mac"', {
	'defines': [
	'_DARWIN_C_SOURCE',
	],
	'conditions': [
	['ffmpeg_component == "shared_library"', {
	'xcode_settings': {
	# GCC version of no -fvisiliity=hidden. Ensures that all
	# symbols are exported for component builds.
	'GCC_SYMBOLS_PRIVATE_EXTERN': 'NO',
	},
	}],
	],
	}], # OS == "mac"
	['OS == "win"', {
	# Disable /analyze because of internal compiler error on log.c
	# with VS 2015 Update 1 when building with /analyze:
	# https://connect.microsoft.com/VisualStudio/feedback/details/2299303
	'msvs_settings': {
	'VCCLCompilerTool': {
	'AdditionalOptions!': [ '/analyze:WX-' ]
	},
	},
	# TODO(dalecurtis): We should fix these. http://crbug.com/154421
	'msvs_disabled_warnings': [
	4996, 4018, 4090, 4305, 4133, 4146, 4554, 4028, 4334, 4101, 4102,
	4116, 4307, 4273, 4005, 4056, 4756,
	],
	'conditions': [
	['clang == 1', {
	'msvs_settings': {
	'VCCLCompilerTool': {
	# This corresponds to msvs_disabled_warnings 4273 above.
	'AdditionalOptions': [ '-Wno-inconsistent-dllimport' ],
	},
	},
	}],
	['MSVS_VERSION == "2013" or MSVS_VERSION == "2013e"', {
	'defines': [
	'inline=__inline',
	'strtoll=_strtoi64',
	'_ISOC99_SOURCE',
	'_LARGEFILE_SOURCE',
	'HAVE_AV_CONFIG_H',
	'strtod=avpriv_strtod',
	'snprintf=avpriv_snprintf',
	'_snprintf=avpriv_snprintf',
	'vsnprintf=avpriv_vsnprintf',
	],
	}],
	['target_arch == "x64"', {
	# TODO(wolenetz): We should fix this. http://crbug.com/171009
	'msvs_disabled_warnings' : [
	4267
	],
	}],
	['ffmpeg_component == "shared_library"', {
	# Fix warnings about a local symbol being inefficiently imported.
	'msvs_settings': {
	'VCCLCompilerTool': {
	'AdditionalOptions': [
	'/FIcompat/msvcrt/snprintf.h',
	'/FIcompat/msvcrt/strtod.h',
	],
	},
	},
	'sources': [
	'<(shared_generated_dir)/ffmpeg.def',
	],
	'conditions': [
	['OS=="win" and win_use_allocator_shim==1', {
	'dependencies': [
	'../../base/allocator/allocator.gyp:allocator',
	],
	'sources': [
	'chromium/dllmain.cc',
	],
	}],
	],
	'actions': [
	{
	'action_name': 'generate_def',
	'inputs': [
	'<(generate_stubs_script)',
	'<@(sig_files)',
	],
	'outputs': [
	'<(shared_generated_dir)/ffmpeg.def',
	],
	'action': ['python',
	'<(generate_stubs_script)',
	'-i', '<(INTERMEDIATE_DIR)',
	'-o', '<(shared_generated_dir)',
	'-t', 'windows_def',
	'-m', 'ffmpeg.dll',
	'<@(_inputs)',
	],
	'message': 'Generating FFmpeg export definitions',
	},
	],
	}],
	],
	}],
	],
	},
	],
	],
	}],
	}