webrtc.org stable branch 3.50 is being used in Chrome 34, to be released soon. There is no newer stable branch. Process is being documented here: https://wiki.mozilla.org/Media/WebRTC/Updating_Process

Green try on desktop (Android/B2G sections not merged yet) https://tbpl.mozilla.org/?tree=Try&rev=ad14572d8f6d

Green try on desktop/Android/B2G https://tbpl.mozilla.org/?tree=Try&rev=64f6f555d209 https://tbpl.mozilla.org/?tree=Try&rev=d88d99080f28

Most of this is forward porting existing code.

Same, forward porting existing code.

Do we still need this? Probably not as we bumped to platform-17 very recently?

Comment on attachment 8428685 [details] [diff] [review] Patch 1. Various build fixes Review of attachment 8428685 [details] [diff] [review]: ----------------------------------------------------------------- r+ with nits ::: media/webrtc/signaling/signaling.gyp @@ +167,5 @@ > '_NO_LOG4CXX', > 'USE_SSLEAY', > '_CPR_USE_EXTERNAL_LOGGER', > 'WEBRTC_RELATIVE_PATH', > + 'HAVE_WEBRTC_VIDEO', No change, remove (unless maybe it's a tab-removal) @@ +220,5 @@ > '-I$(ANDROID_SOURCE)/frameworks/native/include', > '-I$(ANDROID_SOURCE)/frameworks/native/opengl/include', > ], > 'defines' : [ > + 'MOZ_WEBRTC_OMX', no change (tabs?) ::: media/webrtc/trunk/build/all.gyp @@ -103,5 @@ > ], > }], > ], > }], > - ['toolkit_uses_gtk==1', { Instead of removing it, add a test against build_with_mozilla==0 (and maybe comment why) ::: media/webrtc/trunk/build/filename_rules.gypi @@ -65,5 @@ > ['exclude', '_(x|x11)(_unittest)?\\.(h|cc)$'], > ['exclude', '(^|/)x11_[^/]*\\.(h|cc)$'], > ], > }], > - ['<(toolkit_uses_gtk)!=1 or >(nacl_untrusted_build)==1', { Instead, build_with_mozilla==0 (and maybe comment why)

Comment on attachment 8428686 [details] [diff] [review] Patch 2. Android Audio related fixes Review of attachment 8428686 [details] [diff] [review]: ----------------------------------------------------------------- ::: media/webrtc/trunk/webrtc/modules/audio_device/android/audio_record_jni.h @@ +111,5 @@ > int32_t SetRecordingSampleRate(const uint32_t samplesPerSec); > > + static const uint32_t N_REC_SAMPLES_PER_SEC = 16000; // Default is 16 kHz > + static const uint32_t N_REC_CHANNELS = 1; // default is mono recording > + static const uint32_t REC_BUF_SIZE_IN_SAMPLES = 480; // Handle max 10 ms @ 48 kHz Ok, but we need to be careful if we pass these as arguments to functions (to avoid needing to do tricks like "+N_REC_CHANNELS" to force it to be referenceable). If it builds clean on all platforms, we're good.

Comment on attachment 8428687 [details] [diff] [review] Patch 3. OpenSLES dummy backend, fixes Review of attachment 8428687 [details] [diff] [review]: ----------------------------------------------------------------- r+ with use_openssl moved ::: media/webrtc/trunk/build/common.gypi @@ +1049,5 @@ > # Uses Android's crash report system > 'linux_breakpad%': 0, > > # Always uses openssl. > + 'use_openssl%': 0, Move to build/gyp.mozbuild: use_openssl: 0, ::: media/webrtc/trunk/supplement/supplement.gypi @@ +9,5 @@ > 'include_internal_audio_device%': 1, > 'include_internal_video_capture%': 1, > 'include_internal_video_render%': 1, > 'include_pulse_audio%': 1, > + 'use_openssl%': 0, gyp.mozbuild will override this ::: media/webrtc/trunk/webrtc/modules/audio_device/android/opensles_output.h @@ +267,5 @@ > +// Dummy OpenSlesOutput > +class OpenSlesOutput : public PlayoutDelayProvider { > + public: > + explicit OpenSlesOutput(const int32_t id) : > + initialized_(false), speaker_initialized_(false), trailing spaces ::: media/webrtc/trunk/webrtc/modules/audio_device/audio_device.gypi @@ -154,5 @@ > 'conditions': [ > ['enable_android_opensl==1', { > 'sources': [ > - 'opensl/audio_device_opensles.cc', > - 'opensl/audio_device_opensles.h', Are any of these files removed? I see 3 (empty) files removed. Did these just go elsewhere?

Comment on attachment 8428688 [details] [diff] [review] Patch 4. Include and build NDK cpu-features Review of attachment 8428688 [details] [diff] [review]: ----------------------------------------------------------------- ::: media/webrtc/trunk/webrtc/system_wrappers/source/droid-cpu-features.c @@ +275,5 @@ > + archNumber = strtol(cpuArch, &end, 10); > + > + /* Here we assume that ARMv8 will be upwards compatible with v7 > + * in the future. Unfortunately, there is no 'Features' field to > + * indicate that Thumb-2 is supported. ARM1156 (which is ARMv6) introduced Thumb-2.

Comment on attachment 8428688 [details] [diff] [review] Patch 4. Include and build NDK cpu-features Review of attachment 8428688 [details] [diff] [review]: ----------------------------------------------------------------- ::: media/webrtc/trunk/webrtc/system_wrappers/source/droid-cpu-features.c @@ +275,5 @@ > + archNumber = strtol(cpuArch, &end, 10); > + > + /* Here we assume that ARMv8 will be upwards compatible with v7 > + * in the future. Unfortunately, there is no 'Features' field to > + * indicate that Thumb-2 is supported. Upon further review, the ARMv6 chips that support Thumb-2 don't support the entire instruction set. Better to restrict to ARMv7 and higher. http://en.wikipedia.org/wiki/ARM_Cortex-M0#Instruction_sets

Green Try: (not waiting on Mac 10.8 or 2.2/2.3 runs which haven't even started yet) https://tbpl.mozilla.org/?tree=Try&rev=aaa2a429dc5c

https://hg.mozilla.org/integration/mozilla-inbound/rev/d33accc6a20e https://hg.mozilla.org/integration/mozilla-inbound/rev/ad8a86bfd860 https://hg.mozilla.org/integration/mozilla-inbound/rev/7a835877bc62 https://hg.mozilla.org/integration/mozilla-inbound/rev/0dd0c20bea92 https://hg.mozilla.org/integration/mozilla-inbound/rev/69e5c91e3017 https://hg.mozilla.org/integration/mozilla-inbound/rev/8eed969ead48 https://hg.mozilla.org/integration/mozilla-inbound/rev/88a8f94e28e3 https://hg.mozilla.org/integration/mozilla-inbound/rev/72fa06ef1bc9 https://hg.mozilla.org/integration/mozilla-inbound/rev/7e68e14f4d3b https://hg.mozilla.org/integration/mozilla-inbound/rev/7dbea1b7abe8 https://hg.mozilla.org/integration/mozilla-inbound/rev/b43898f38d5f https://hg.mozilla.org/integration/mozilla-inbound/rev/424a23313d10

Non-unified build bustage fix: https://hg.mozilla.org/integration/mozilla-inbound/rev/f939b0491228

https://hg.mozilla.org/mozilla-central/rev/d33accc6a20e https://hg.mozilla.org/mozilla-central/rev/ad8a86bfd860 https://hg.mozilla.org/mozilla-central/rev/7a835877bc62 https://hg.mozilla.org/mozilla-central/rev/0dd0c20bea92 https://hg.mozilla.org/mozilla-central/rev/69e5c91e3017 https://hg.mozilla.org/mozilla-central/rev/8eed969ead48 https://hg.mozilla.org/mozilla-central/rev/88a8f94e28e3 https://hg.mozilla.org/mozilla-central/rev/72fa06ef1bc9 https://hg.mozilla.org/mozilla-central/rev/7e68e14f4d3b https://hg.mozilla.org/mozilla-central/rev/7dbea1b7abe8 https://hg.mozilla.org/mozilla-central/rev/b43898f38d5f https://hg.mozilla.org/mozilla-central/rev/424a23313d10 https://hg.mozilla.org/mozilla-central/rev/f939b0491228

I've added a note to our Mercurial documentation on how to properly set up qcrecord so it doesn't lose patch authorship information :-/

This bug added a non-existent symlink (media/webrtc/trunk/webrtc/tools/e2e_quality/audio/perf) which interfered with grep. I removed it, hope that's ok. https://hg.mozilla.org/integration/mozilla-inbound/rev/0ca8b92f0266