在“WebRTC 构建系统介绍之gn和ninja”中，大概介绍了 gn 和 ninja 的简单用法，这次来看看 gn 用到的项目文件 .gn 、 .gni 和 DEPS ，它们指导了如何生成 ninja 构建文件。
借用 C++ 的概念，如果把 gn 看成一个编译系统， .gn 就是源文件， .gni 就是头文件。我们姑且这么理解就好了（其实 gni 里做的事情， gn 都可以做）。DEPS 主要用来设定包含路径。
gn 和 gni 文件都在源码树中，比如 src 目录。当执行 gn gen 时，gn 工具根据 gn 和 gni 生成 ninja 文件并将这些 ninja 文件放到指定的构建目录中。
.gn 文件是 GN build 的 “源文件”，在这里可以做各种条件判断和配置，gn 会根据这些配置生成特定的 ninja 文件。
.gn 文件中可以使用预定义的参数，比如 is_debug ， target_os ， rtc_use_h264 等。
.gn 中可以 import .gni 文件。
看一下 src/BUILD.gn ：
import("webrtc/webrtc.gni") group("default") { testonly = true deps = [ "//webrtc", "//webrtc/examples", "//webrtc/tools", ] if (rtc_include_tests) { deps += [ "//webrtc:webrtc_tests" ] } }.gn 和 .gni 文件中用到各种指令，都在这里有说明：GN Reference。
这个 gn 文件中，导入了 webrtc/webrtc.gni 文件。
这个 gn 文件，用 group 指令声明了一个 default 目标，这个目标依赖 webrtc 、 webrtc/examples 和 webrtc/tools ，你可以在 webrtc 、 webrtc/examples 、 webrtc/tools 目录下找到对应的 BUILD.gn 。你可以把 group 当做 VS 的 solution ，或者 QtCreator 的 dir 项目。
gn 文件中也可以通过 defines 来定义宏，通过 cflags 来指定传递给编译器的标记，通过 ldflags 指定传递给链接器的标记，还可以使用 sources 指定源文件。下面是 webrtc/BUILD.gn 文件的部分内容：
if (is_win) { defines += [ "WEBRTC_WIN", "_CRT_SECURE_NO_WARNINGS", # Suppress warnings about _vsnprinf ] } if (is_android) { defines += [ "WEBRTC_LINUX", "WEBRTC_ANDROID", ] } if (is_chromeos) { defines += [ "CHROMEOS" ] } if (rtc_sanitize_coverage != "") { assert(is_clang, "sanitizer coverage requires clang") cflags += [ "-fsanitize-coverage=${rtc_sanitize_coverage}" ] ldflags += [ "-fsanitize-coverage=${rtc_sanitize_coverage}" ] }gni 文件是 GN build 使用的头文件，它里面可以做各种事情，比如定义变量、宏、定义配置、定义模板等。
看下 webrtc/webrtc.gni 文件：
# Copyright (c) 2014 The WebRTC project authors. All Rights Reserved. # # 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. import("//build/config/arm.gni") import("//build/config/features.gni") import("//build/config/mips.gni") import("//build/config/sanitizers/sanitizers.gni") import("//build_overrides/build.gni") import("//testing/test.gni") declare_args() { # Disable this to avoid building the Opus audio codec. rtc_include_opus = true # Enable this if the Opus version upon which WebRTC is built supports direct # encoding of 120 ms packets. rtc_opus_support_120ms_ptime = false # Enable this to let the Opus audio codec change complexity on the fly. rtc_opus_variable_complexity = false # Disable to use absolute header paths for some libraries. rtc_relative_path = true # Used to specify an external Jsoncpp include path when not compiling the # library that comes with WebRTC (i.e. rtc_build_json == 0). rtc_jsoncpp_root = "//third_party/jsoncpp/source/include" # Used to specify an external OpenSSL include path when not compiling the # library that comes with WebRTC (i.e. rtc_build_ssl == 0). rtc_ssl_root = "" # Selects fixed-point code where possible. rtc_prefer_fixed_point = false # Enables the use of protocol buffers for debug recordings. rtc_enable_protobuf = true # Disable the code for the intelligibility enhancer by default. rtc_enable_intelligibility_enhancer = false # Enable when an external authentication mechanism is used for performing # packet authentication for RTP packets instead of libsrtp. rtc_enable_external_auth = build_with_chromium # Selects whether debug dumps for the audio processing module # should be generated. apm_debug_dump = false # Set this to true to enable BWE test logging. rtc_enable_bwe_test_logging = false # Set this to disable building with support for SCTP data channels. rtc_enable_sctp = true # Disable these to not build components which can be externally provided. rtc_build_expat = true rtc_build_json = true rtc_build_libjpeg = true rtc_build_libsrtp = true rtc_build_libvpx = true rtc_libvpx_build_vp9 = true rtc_build_libyuv = true rtc_build_openmax_dl = true rtc_build_opus = true rtc_build_ssl = true rtc_build_usrsctp = true # Enable to use the Mozilla internal settings. build_with_mozilla = false rtc_enable_android_opensl = false # Link-Time Optimizations. # Executes code generation at link-time instead of compile-time. # https://gcc.gnu.org/wiki/LinkTimeOptimization rtc_use_lto = false # Set to "func", "block", "edge" for coverage generation. # At unit test runtime set UBSAN_OPTIONS="coverage=1". # It is recommend to set include_examples=0. # Use llvm's sancov -html-report for human readable reports. # See http://clang.llvm.org/docs/SanitizerCoverage.html . rtc_sanitize_coverage = "" # Enable libevent task queues on platforms that support it. if (is_win || is_mac || is_ios || is_nacl) { rtc_enable_libevent = false rtc_build_libevent = false } else { rtc_enable_libevent = true rtc_build_libevent = true } if (current_cpu == "arm" || current_cpu == "arm64") { rtc_prefer_fixed_point = true } if (!is_ios && (current_cpu != "arm" || arm_version >= 7) && current_cpu != "mips64el") { rtc_use_openmax_dl = true } else { rtc_use_openmax_dl = false } # Determines whether NEON code will be built. rtc_build_with_neon = (current_cpu == "arm" && arm_use_neon) || current_cpu == "arm64" # Enable this to build OpenH264 encoder/FFmpeg decoder. This is supported on # all platforms except Android and iOS. Because FFmpeg can be built # with/without H.264 support, |ffmpeg_branding| has to separately be set to a # value that includes H.264, for example "Chrome". If FFmpeg is built without # H.264, compilation succeeds but |H264DecoderImpl| fails to initialize. See # also: |rtc_initialize_ffmpeg|. # CHECK THE OPENH264, FFMPEG AND H.264 LICENSES/PATENTS BEFORE BUILDING. # http://www.openh264.org, https://www.ffmpeg.org/ rtc_use_h264 = proprietary_codecs && !is_android && !is_ios # Determines whether QUIC code will be built. rtc_use_quic = false # By default, use normal platform audio support or dummy audio, but don't # use file-based audio playout and record. rtc_use_dummy_audio_file_devices = false # When set to true, test targets will declare the files needed to run memcheck # as data dependencies. This is to enable memcheck execution on swarming bots. rtc_use_memcheck = false # FFmpeg must be initialized for |H264DecoderImpl| to work. This can be done # by WebRTC during |H264DecoderImpl::InitDecode| or externally. FFmpeg must # only be initialized once. Projects that initialize FFmpeg externally, such # as Chromium, must turn this flag off so that WebRTC does not also # initialize. rtc_initialize_ffmpeg = !build_with_chromium # Build sources requiring GTK. NOTICE: This is not present in Chrome OS # build environments, even if available for Chromium builds. rtc_use_gtk = !build_with_chromium } # A second declare_args block, so that declarations within it can # depend on the possibly overridden variables in the first # declare_args block. declare_args() { # Include the iLBC audio codec? rtc_include_ilbc = !(build_with_chromium || build_with_mozilla) rtc_restrict_logging = build_with_chromium # Excluded in Chromium since its prerequisites don't require Pulse Audio. rtc_include_pulse_audio = !build_with_chromium # Chromium uses its own IO handling, so the internal ADM is only built for # standalone WebRTC. rtc_include_internal_audio_device = !build_with_chromium # Include tests in standalone checkout. rtc_include_tests = !build_with_chromium } # Make it possible to provide custom locations for some libraries (move these # up into declare_args should we need to actually use them for the GN build). rtc_libvpx_dir = "//third_party/libvpx" rtc_libyuv_dir = "//third_party/libyuv" rtc_opus_dir = "//third_party/opus" # Desktop capturer is supported only on Windows, OSX and Linux. rtc_desktop_capture_supported = is_win || is_mac || is_linux ############################################################################### # Templates # # Points to //webrtc/ in webrtc stand-alone or to //third_party/webrtc/ in # chromium. # We need absolute paths for all configs in templates as they are shared in # different subdirectories. webrtc_root = get_path_info(".", "abspath") # Global configuration that should be applied to all WebRTC targets. # You normally shouldn't need to include this in your target as it's # automatically included when using the rtc_* templates. # It sets defines, include paths and compilation warnings accordingly, # both for WebRTC stand-alone builds and for the scenario when WebRTC # native code is built as part of Chromium. rtc_common_configs = [ webrtc_root + ":common_config" ] # Global public configuration that should be applied to all WebRTC targets. You # normally shouldn't need to include this in your target as it's automatically # included when using the rtc_* templates. It set the defines, include paths and # compilation warnings that should be propagated to dependents of the targets # depending on the target having this config. rtc_common_inherited_config = webrtc_root + ":common_inherited_config" # Common configs to remove or add in all rtc targets. rtc_remove_configs = [] rtc_add_configs = rtc_common_configs set_defaults("rtc_test") { configs = rtc_add_configs suppressed_configs = [] } set_defaults("rtc_source_set") { configs = rtc_add_configs suppressed_configs = [] } set_defaults("rtc_executable") { configs = rtc_add_configs suppressed_configs = [] } set_defaults("rtc_static_library") { configs = rtc_add_configs suppressed_configs = [] } set_defaults("rtc_shared_library") { configs = rtc_add_configs suppressed_configs = [] } template("rtc_test") { test(target_name) { forward_variables_from(invoker, "*", [ "configs", "public_configs", "suppressed_configs", ]) configs += invoker.configs configs -= rtc_remove_configs configs -= invoker.suppressed_configs public_configs = [ rtc_common_inherited_config ] if (defined(invoker.public_configs)) { public_configs += invoker.public_configs } } } template("rtc_source_set") { source_set(target_name) { forward_variables_from(invoker, "*", [ "configs", "public_configs", "suppressed_configs", ]) configs += invoker.configs configs -= rtc_remove_configs configs -= invoker.suppressed_configs public_configs = [ rtc_common_inherited_config ] if (defined(invoker.public_configs)) { public_configs += invoker.public_configs } } } template("rtc_executable") { executable(target_name) { forward_variables_from(invoker, "*", [ "deps", "configs", "public_configs", "suppressed_configs", ]) configs += invoker.configs configs -= rtc_remove_configs configs -= invoker.suppressed_configs deps = [ "//build/config/sanitizers:deps", ] deps += invoker.deps public_configs = [ rtc_common_inherited_config ] if (defined(invoker.public_configs)) { public_configs += invoker.public_configs } } } template("rtc_static_library") { static_library(target_name) { forward_variables_from(invoker, "*", [ "configs", "public_configs", "suppressed_configs", ]) configs += invoker.configs configs -= rtc_remove_configs configs -= invoker.suppressed_configs public_configs = [ rtc_common_inherited_config ] if (defined(invoker.public_configs)) { public_configs += invoker.public_configs } } } template("rtc_shared_library") { shared_library(target_name) { forward_variables_from(invoker, "*", [ "configs", "public_configs", "suppressed_configs", ]) configs += invoker.configs configs -= rtc_remove_configs configs -= invoker.suppressed_configs public_configs = [ rtc_common_inherited_config ] if (defined(invoker.public_configs)) { public_configs += invoker.public_configs } } }webrtc.gni 是一个比较特殊的 gni 文件，你可以把它看做全局配置文件。
webrtc.gni 定义了 WebRTC 项目用到的一些标记，比如 rtc_build_libvpx、rtc_build_ssl、rtc_use_h264 等。
还使用 template 语句定义了几个模板，比如 rtc_executable 、 rtc_static_library 、 rtc_shared_library ，这几个模板定义了生成可执行文件、静态库、动态库的规则。在 webrtc/examples/BUILD.gn 中就有用到这些模板，用它们来指导如何生成可执行文件、静态库等。
你也可以直接使用 gn 内置的 shared_library 和 static_library 来声明目标，比如 third_party/ffmpeg/BUILD.gn 就使用 shared_library 来生成动态库。
给个例子看看吧，webrtc/examples/DEPS ：
include_rules = [ "+WebRTC", "+webrtc/api", "+webrtc/base", "+webrtc/media", "+webrtc/modules/audio_device", "+webrtc/modules/video_capture", "+webrtc/p2p", "+webrtc/pc", ]include_rules 定义了包含路径。
了解 .gn 和 .gni 文件的目的是修改它们。比如你想打开 WebRTC 对 H264 的支持，就可以修改 webrtc/webrtc.gni ，直接把 rtc_use_h264 设置为 true 。
比如你想为某个模块加一些文件，就可以修改 .gn 文件，修改 sources 变量，直接把你的源文件加进去。
好啦，到这儿吧。
相关阅读：
WebRTC学习资料大全Ubuntu 14.04下编译WebRTCWebRTC源码中turnserver的使用方法打开 WebRTC 的日志（native api）让WebRTC支持H264编解码WebRTC编译系统之gn和ninja