1654112 - [meta] Update libwebrtc to new stable branch 2H2020

Status: RESOLVED FIXED

(Core :: WebRTC, enhancement, P2)

Description

[Nico Grunbaum [:ng, @chew:mozilla.org]]

This bug tracks the work necessary for the next libwebrtc update (2H 2020).

Information needed to perform this update is documented at https://wiki.mozilla.org/Media/WebRTC/libwebrtc_Update_2H2020 .

The process is documented at https://wiki.mozilla.org/Media/WebRTC/libwebrtc_Update_Process .

Comment 2

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove dead rtp_rtcp code that was previously just rolling along. r?ng

This code comes from the rollup of previous local patches to upstream, see
https://hg.mozilla.org/mozilla-central/rev/c8e80efab6fb

Comment 3

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Plumb RemoteRTCPSenderInfo through also for audio and change unit to ms. r?ng

This is needed to not regress stats. The milliseconds unit is more prevalent in
upstream stats code for timestamps.

Comment 4

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Include RtcpPacketTypeCounter in audio send stats, to not regress nackCount. r?ng

This is similar to how it's already included for video send.

Comment 5

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Plumb InsertAudioLevelForContributingSource into upstream, for mochitests. r?ng

Comment 6

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Change order of kCsrcAudioLevelUri to better align with upstream. r?ng

Comment 7

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Move AudioConduit to use upstream for audioLevel and remove unused code. r?ng

Comment 8

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove CODEC_PLNAME_SIZE as CodecInst no longer exists upstream. r?ng

Comment 9

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove obsolete ChannelProxy and VoiceEngine API integrations. r?ng

This switches those integrations to the new Call API, with AudioSendStream and
AudioReceiveStream as main API surface.

Comment 10

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update threading invariant of AudioConduit::GetLocalSSRCs to match mRecvStreamConfig comment. r?ng

Comment 11

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix AudioConduit::SendAudioFrame to use the new AudioSendStream API. r?ng

Comment 12

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix AudioCondut::GetAudioFrame to use the new AudioReceiveStream API. r?ng

Comment 13

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Track base_seq in MediaPipeline for stats purposes. r?ng!,r?bwc!

This will be used in PeerConnectionImpl for packetsReceived on the send side (from
rtcp).

Comment 14

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update Audio- and VideoConduit stats API and integration. r?ng!,r?bwc!

For AudioConduit this patch:

• Uses correct threads per upstream, pending changes due to the threading model
  update that's coming
• Uses new upstream APIs for getting data for the stats

For VideoConduit this patch:

• Removes the stats polling 1-second timer on main thread
• Removes the internal stats classes for bridging stats between main and sts
  threads
• Uses correct threads for direct access per upstream's thread checkers, pending
  changes due to the threading model

For PeerConnectionImpl and RTCRtpReceiver this patch:

• Manages thread access into conduits for thread safety
• Translates from the upstream Stats data models into our DOM data model

Comment 15

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix AudioConduit comments, member naming and other minor things. r?ng

These fixes are made to reflect the current state of upstream.

Comment 16

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Create Call instances using the new Call config. r?ng!,r?bwc!

This sets up our code for creating Call instances so that it works. A number of
things are left to do. Most notable fixing the threading model.

Until that is fixed we use main thread as a "worker" but this is not ideal.
It both leads to excessive work loads on main, and us having to set main as
"current" (using upstream's API for this) whenever we call into main from what
is meant to be the worker thread.

Comment 17

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Set main thread as current TaskQueue when calling into libwebrtc so that audio calling works. r?bwc!

Comment 18

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove RtcpEventObserver. r?ng

Comment 19

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Don't register/unregister conduit with the call wrapper in ctor/dtor. r?ng!,r?bwc!

Doing this outside of the dtor, especially, removes the requirement of the
conduit having to be destroyed on the worker thread (main).

Not having threading requirements on the conduits' dtor means we can constify
conduit members everywhere, so this patch tries to do that, for added compiler
guarantees.

Comment 20

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Clean up and sort Audio/Video/MediaConduit includes. r?ng

Comment 21

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove overzealous assert.

This assert was intended for when an input stream is started with processing on.
However, when an input stream is started with processing off, then later when
processing is turned on it triggers this assert because mSegment contains data
from InsertInGraph().

Depends on D102289

Comment 22

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Always start feeding input to audio processing, and always end on output.

The new AudioProcessingImpl, and in particular AEC3 it seems, is more picky on
the ordering here. After an ApplyConfig() it needs ProcessStream() to properly
set up its internals. ProcessReverseStream sets off various asserts, for
instance because the processing sample rate is not divisible by 16k if
ProcessReverseStream() was fed 44.1k.

Comment 23

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Don't feed the reverse stream in passthrough.

The input stream is not processed in passthrough, so we should skip feeding the
reverse stream too, to stay in sync.

Comment 24

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Reset AudioProcessing when turning it off to clear old state.

Comment 25

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Minor tidying around lambdas and ControlMessages in MediaEngineWebRTCAudio.

Comment 26

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update microphone DeviceChanged handler to new AudioProcessing API.

This should also work with libwebrtc 64 currently in-tree.

Comment 27

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Adapt to new AudioProcessing config.

This removes some prefs for things that are now unconfigurable, and adds some
others. This also re-works the plumbing for updating the config on the fly
through applyConstraints(). It is now simpler, with a single infallible call
to apply the new config.

Comment 28

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Always iterate the graph on a data callback.

Before this patch, if the data callback contained less data than was left over
from the last iteration, we'd skip iterating the graph again. The graph is
however is capable of 0-iterations.

This patch removes that optimization so that every data callback results in an
iteration.

This makes it easier to defer processing input data to the track processing
step, as is needed by the next patch in this stack, to keep input and output as
handed to AudioProcessing in sync as config changes are applied with
ControlMessages (in between input and output data notifications). With input
data processing deferred, ControlMessages are run first, then processing of
input data and last processing of output data.

Comment 29

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Tighten AudioTrackGraph gtest discontinuity checks, and fix related comment/log.

These previously allowed a discontinuity at the beginning of the output stream.
This will no longer occur.

Comment 30

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Report video frame-/bitrate telemetry once per call instead of every second. r?ng

This is similar to the behavior prior to bug 1532898, which introduced telemetry
reporting every second. There's a bit of a difference in that we only report
data if we had collected some though the 1 second interval telemetry timer in
PeerConnectionCtx.

Comment 31

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Don't access GetStats() in AudioConduit::GetRemoteSSRC(). r?ng

Comment 32

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Implement AudioConduit::DeleteStreams. r?ng

Comment 33

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix up AudioConduit audio thread access model, and the channel count for silence.

This patch uses mMutex to guarantee that the audio thread does not get things
pulled away from under its feet. When on the AudioProxyThread we allow blocking
the thread during negotiation. On the real audio thread we avoid blocking on
the mutex and return an error instead. This leads to us inserting silence --
which is fine since we're going through negotiation (call setup, teardown or
re-negotiation).

This also fixes the channel count when GetAudioFrame fails and we insert
silence, to use the same channel count as when we last appended real data. Since
changing the channel count can in the worst case cause a GraphDriver switch.

Comment 34

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Tidy up around PeerConnectionCtx and use accessors for mPeerConnections. r?ng!,r?bwc!

Comment 35

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Destroy Call instances as their associated PeerConnections close. r?ng!,r?bwc!

Prior to this patch, WebRtcCallWrapper would never clear the call instance,
meaning that it went away as the classes having references to WebRtcCallWrapper
also went away. These are mainly TransceiverImpl and PeerConnectionMedia. There
are other classes too, but these are the ones that have a strong-ref to
WebRtcCallWrapper and are cycle collected.

With the worker threads in libwebrtc now using our own TaskQueues on top of
our SharedThreadPool instances, the ownership model of the Call instances
becomes problematic.

On shutdown, xpcom-shutdown-threads happens before the cycle collector is shut
down.

xpcom-shutdown-threads will however, block shutdown until all SharedThreadPools
have been shut down.

If a TransceiverImpl sits around until shutdown it will only be destroyed
during cycle collector shutdown. And, alas, we reach a deadlock as we're stuck
in xpcom-shutdown-threads until that TransceiverImpl gets destroyed.

This patch decouples the Call instance lifetime from the cycle collector, and
avoids the shutdown hang.

Comment 36

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Don't run delayed runnables from libwebrtc after the TaskQueue was deleted. r?ng!

This is especially prompted by a recurring task from libwebrtc that when run
will schedule itself again, which on a shut down TaskQueue will fail an assert.
The behavior with this patch better mirrors blink's behavior, and is what
libwebrtc expects.

Comment 37

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Move static functions in VideoConduit.cpp to anonymous namespace. r?ng

Comment 38

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Rework GMP PluginID plumbing in VideoCondit. r?bwc

This prepares VideoConduit for the decoder/encoder factory and its handling of
PluginID.

Comment 39

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Adapt VideoConduit to the new libwebrtc version. r?ng!,r?bwc!

This hooks up VideoReceiveStream::Config with a decoder factory and
VideoSendStream::Config with both a bitrate allocator factory and an encoder
factory.

The factories are new and wraps all the types of decoders and encoders that we
could create prior to this patch. This patch lets libwebrtc create our codecs
on demand, allowing for reconfiguring the streams and changing codec on the fly.

Some extra care had to be taken to ensure that GMP PluginIDs can be properly
plumbed. This has been implemented with MediaEvent and MozPromise.

Some other changes are made too, as warranted by runtime or threading changes.

Comment 40

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Set main thread as current TaskQueue when calling into libwebrtc so that video calling works. r?bwc

Comment 41

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove WebrtcGmpPCHandleSetter and plumb pc handles explicitly. r?bwc!

With the libwebrtc update codecs will be created off-main.
WebrtcGmpPCHandleSetter not only adds a lot of indirection, it is also
main-thread-only. With this patch we can create gmp codecs on any thread.

Comment 42

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Add a MediaEventForwarder for aggregating multiple MediaEventSources into one. r?alwu

Comment 43

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Re-wire GMP PluginID between codecs and VideoConduit. r?bwc!

Comment 44

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Adapt the webrtc codecs to the updated libwebrtc version. r?bwc!

Comment 45

[Byron Campen [:bwc]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=48ee2324cf10f39f6d55d4bed668403e0752e9c3

Comment 46

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Adapt GMP-OpenH264 encoder to new libwebrtc RtpPacketizerH264 expectations. r?dminor!,r?bwc!

The RTPFragmentationHeader is gone. Instead the packetizer scans for NALU start
sequences. OpenH264 encodes them natively, but converts them to NALU sizes
as expected by GMP.

This patch puts the start sequences back in, in-place, and mitigates bug 1533001
by allowing the start sequences that leak through from OpenH264 straight through
to libwebrtc, as an audit of OpenH264 code shows this can happen when internally
OpenH264 reports a NAL count of 0.

Comment 47

[Byron Campen [:bwc]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=a97cc1147107fb94d211d502aeebe386b5ec055b

Comment 48

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix MediaPipelineFilter stream id parsing. r?ng

Comment 49

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Reformat some moz.build files to satisfy linters on try.

Comment 50

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Use SprintfLiteral instead of snprintf to satisfy linter.

Comment 51

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Reformat some moz.build files to satisfy linters on try.

Depends on D102953

Comment 52

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Use SprintfLiteral instead of snprintf to satisfy linter.

Depends on D105009

Comment 53

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Unset execute permissions on this script to keep linters happy.

Depends on D105010

Comment 54

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Exempt code in webrtc namespace from implicit conversion operator checks.

Depends on D105011

Comment 55

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Disable internal_audio_device build.

Depends on D105012

Comment 56

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Generated files from previous commit (Disable internal_audio_device build).

Depends on D105013

Comment 57

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Get OS X build working.

• Pull in sdk/objc/base and sdk/objc/helpers
• Ensure that -fobjc-arc is set (have to use cflags, because our gn -> json scripting ignores things like cflags_objc)
• Add gclient_args.gni to keep build happy.
• Add a missing include path for libyuv
• Support .m files in build.

Depends on D105014

Comment 58

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Generated files (linux x64, OS X x64) for previous commit (Get OS X build working)

Depends on D105015

Comment 59

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Specify template param to fix build.

Depends on D105016

Comment 60

[Byron Campen [:bwc]]

Attachment: Bug 1654112: OS X include fixes, and add a necessary source file for OS X.

Depends on D105017

Comment 61

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Get vp9 build working again.

Depends on D105018

Comment 62

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Linux and OS X x64 generated files for previous commit (Get vp9 build working again).

Depends on D105019

Comment 63

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Un-nest VideoConduit SetRemoteSSRC runnables. r?bwc

Comment 64

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Use regular threadsafe refcounting for WebRtcCallWrapper. r?bwc

Comment 65

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove mozilla:: prefixes from mediapipeline_unittest.cpp. r?bwc!

Comment 66

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Modify media::ShutdownTicket to make it more convenient to fire and forget. r?jib

Comment 67

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Use a ShutdownBlockingTicket in WebRtcCallWrapper. r?jib!,r?bwc!

Peer connections are cleaned up on shutdown through an xpcom-shutdown observer.
It doesn't block shutdown throughout async operations.

webrtc::Call instances are currently used on the main thread and so destroying
them could be a sync operation. However with upcoming threading model changes
they are moving to a worker thread, making destruction of them inevitably async.

Currently they are even destroyed async, even though still on the main thread,
to better simulate the behavior we'd see with the threading model updated.

Destroying Call instances async can lead to a race between destroying all
TaskQueues tied to the Call instance, and reaching xpcom-shutdown-threads.
xpcom-shutdown-threads will shut down the thread pools backing the TaskQueues,
making future dispatches fail. We have asserts checking that dispatches always
succeed.

This patch adds a ShutdownBlocker to the Call wrapper, so progressing to
xpcom-shutdown-threads is blocked until Call instances are cleaned up and no
more dispatches are expected.

Comment 68

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix AudioConduit unittests. r?ng!

Comment 69

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - libwebrtc: Fix re-starting VideoReceiveStreams. r?ng!

Comment 70

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - libwebrtc: Add a way to set keyframe request method on VideoReceiveStream. r?ng!

Comment 71

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - libwebrtc: Add a REMB on/off switch to VideoReceiveStream. r?ng!

Comment 72

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Set keyframe request method in VideoConduit, based on negotiation. r?ng!

Comment 73

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Set REMB switch in VideoConduit, based on negotiation. r?ng!

Comment 74

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix VideoConduit unittests. r?ng!

Comment 75

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Big fat hack to get asan working.

Comment 76

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Check whether this is an audio conduit before treating it as such.

Comment 77

[Byron Campen [:bwc]]

Attachment: Bug 1654112: (libwebrtc modification) Fix comparison so capture timestamps aren't allowed to be 0.

Comment 78

[Byron Campen [:bwc]]

Attachment: Bug 1654112: These can fail if the task queue is shutting down.

Comment 79

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Get RTCP BYE and RTP timeout handling working again (from Bug 1595479) r?mjf

Comment 80

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Cherry-pick abseil-cpp commit 310c82cd97b3f1f0d1ee93a0ee2b0aee828b2a93. r?bwc!

This helps with ASAN link errors. Cherry-picked commit's commit message:

310c82cd97b3f1f0d1ee93a0ee2b0aee828b2a93 by Abseil Team <[e-mail]>:

Simplify unaligned memory access functions.

The #ifdef to produce calls to __sanitizer_unaligned_load16 etc were needed in past versions of this code, when we were lying to the compiler about the alignment of the loads/stores, by using a reinterpret_cast.

However, a year ago, absl switched to simply use memcpy. Sanitizers support this correctly by default, nothing extra is required.

PiperOrigin-RevId: 343159883

Comment 81

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Add sanitizer/common_interface_defs.h as system-header. r?glandium!

This patch fixes a linker error where __sanitizer_annotate_contiguous_container
is an undefined hidden symbol referenced by
third_party/libwebrtc/third_party/abseil-cpp/absl/container/fixed_array.h.

Comment 82

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Add compile-time errors if WEBRTC_USE_X11 or MOZ_X11 are not defined when compiling X11 desktop capture. r?ng!

Comment 83

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove obsolete desktop capture comments. r?ng!

Comment 84

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update webrtc.mozbuild with more defines. r?ng!

Comment 85

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Get contributing/synchronization sources working again. r?ng

Comment 86

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Clean up shared webrtc state in PeerConnectionCtx::Cleanup. r?mjf

Comment 87

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Plug leak of WebrtcVideoConduit, and release encoder/decoder factories on main.

Comment 88

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Make sure the audio device module is refcounted.

Comment 89

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Clean up delayed tasks when the task queue is shut down. Also clean up the task queue itself.

Comment 90

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Get tsan build working, and get wpt running under tsan (still needs some race fixes).

Comment 91

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Add a mutex lock that was somehow lost here. r?ng

Comment 92

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Make MediaSessionConduit::DeliverPacket async. r?bwc!

Comment 93

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update USE_X11 to WEBRTC_USE_X11 like rest of libwebrtc, in desktop_capturer.cc. r?ng!

Comment 94

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - allow pulling MSVS paths and version from environment for WebRTC gn generation;r?dminor

Comment 95

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Use newer thread run callback, and adapt PlatformUIThread;r?pehrsons!

Depends on D107878

Comment 96

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Removing old mozilla specific carry overs from video_capture GN file;r?pehrsons!,dminor!

Depends on D107879

Comment 97

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Update include paths for desktop_device_info_win;r?pehrsons

Depends on D107880

Comment 98

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Add pid_t to desktop_capture_types.h;r?pehrsons

Upstreaming bug 1697385

Depends on D107881

Comment 99

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Fix typo in window_capture_utils;r?dminor

Depends on D107897

Comment 100

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - fix device_info_ds pid and Windows constants includes;r?pehrsons

Upstreaming bug 1697385

Depends on D107898

Comment 101

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Add check for mozilla build that was in the previous import, restore changes;r?dminor

Depends on D107899

Comment 102

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Do not set WEBRTC_ENABLE_AVX2 if build_with_mozilla is set;r?dminor

WIP Note: originally I commented this block out instead of checking build_with_mozilla. This needs to be checked.

Depends on D107900

Comment 103

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Add build_with_mozilla to BUILDCONFIG.gn;r?dminor

Depends on D107901

Comment 104

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Disable AVX2 in common_audio and audio_processing;r?pehrsons!

Depends on D107902

Comment 105

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Fix static-analysis errors. r?bwc!

Comment 106

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Implement dummy PluginEvent getters in non-gmp encoder. r?bwc!

Comment 107

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update libwebrtc third party paths. r?ng!

This removes media/signaling/src from ThirdPartyPaths.txt as that directory no
longer exists.

It also adds media/signaling/gtest/MockCall.h as the new SetAndGetRecordingState
from upstream libwebrtc causes a static-analysis error because RecordingState is
a non-param type.

Comment 108

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Don't create two WebRtcCallWrappers in VideoConduitTest. r?ng!

Comment 109

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Add the last committime file from WebRTC for GN;r?dminor!

Depends on D107878

Comment 110

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - include dir is not previously defined here;r?pehrsons!

Depends on D108087

Comment 111

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Abandoned - Bug 1654112 - Add protobuff for AEC dumping;r?pehrsons

Comment 112

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Reimplement ssrc/csrc stats cache. r?ng

The cache can be a lot simpler than the old one, since libwebrtc handles
pruning old entries for us, and since we're now telling libwebrtc when we
render frames (so we don't have to make our own stats with guesses of when
frames will be rendered). We also now have a single implementation of the
cache in the base class.

This patch also removes the old CSRC stat insertion stuff we hacked into
libwebrtc for testing, and handles that stuff in our cache directly.

Comment 113

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Get simulcast test code working again. r?mjf

Involves teaching MediaPipelineFilter to learn ssrc->rid mappings more
aggressively since libwebrtc does not emit RTP RID more than once.

Also involves modifying a couple of mochitests to look for RTP RID in
the first packet, instead of whichever one comes in next once we care.

Comment 114

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Bring our packetsReceived estimate up-to-spec. r?ng

• Make sure this is 0 if we have not received a RTCP RR yet.
• Subtract the packet lost count from the result.

Comment 115

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Fix this conversion. r?ng

The divisor here was already in ticks/sec, which results in a conversion to seconds.

Comment 116

[Byron Campen [:bwc]]

Attachment: Bug 1654112: libwebrtc modification: Init some stats that were being passed to us uninitialized. r?ng

Comment 117

[Byron Campen [:bwc]]

Attachment: Bug 1654112: libwebrtc modification: Surface video RTCP SR stats again. r?ng

libwebrtc has stopped surfacing these, and Chromium does not support
these stats at all.

Comment 118

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Remove tests for packetsDiscarded/discardedPackets. r?ng

We've never supported the spec spelling (packetsDiscarded), and libwebrtc
no longer supports discarded packet count anyway.

Comment 119

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Disallow insane values in stats. r?ng

Alsol, make sure errors note what kind of stats (audio/video) are being checked.

Comment 120

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Make sure this waits for RTCP SR. r?jib

Comment 121

[Byron Campen [:bwc]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=449123173f2edfe89657407ae5234ca6868d4511

Comment 122

[Byron Campen [:bwc]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=2015dfba62fe8d9a38339cedfe986a5c2944100b

Comment 123

[Byron Campen [:bwc]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=fd3a9e65b9dddfe81bc4dbe4cb91eaf3702c070e

Comment 124

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - add script to generate gn json and moz.build files. r?ng!

Comment 125

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Unbust base_capturer_pipewire. r?stransky

This makes the latest changes to base_capturer_pipewire in our tree compile
with the libwebrtc 86 update.

Comment 126

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - update json and moz.build files for Apple Silicon builds. r?ng!

Comment 127

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - add script to remove dev-machine specific info from generated json gn files. r?ng!

Comment 128

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - cleaned up json and moz.build files. r?ng!

Comment 129

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Adapt AVX2 build in libwebrtc;r?dminor!

Comment 130

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Add missing -mavx;r?dminor

Depends on D107902

Comment 131

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Workaround type deduction in windows build;r?pehrsons

Depends on D110897

Comment 132

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Include AEC dump interface on Windows;r?pehrsons

Depends on D110898

Comment 133

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: WIP: Temp Bug 1654112 - windows generate script

Depends on D110895

Comment 134

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - don't look for windows tools whose values will be ignored anyway;r?dminor

Depends on D111294

Comment 135

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Clarifying prev. rev that moved pid_t into the global namespace;r?pehrsons

This updates the previous revision: https://phabricator.services.mozilla.com/D107897 .

Comment 136

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Reorder include to fulfill interdepencies;r?pehrsons

It turns out these were in a specific order. This restores the order from a previous patch.

Comment 137

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Update generate-gn-build-files.sh for Windows dev;r?mjf

Additionally:

• Trys to detect non-generated build file changes to the tree and exits to prevent a clean checkout and clobber
• Adds the DEBUG_GEN variable that prints what is happening in the script
• tee(s) the output of mach into the logs so one doesn't have to check them to see what went wrong
• Removes the symlinked directories if they were left over by a failed run
• Exits as soon as anything exits unexpectedly
• Adds an optional variable so that the git checkout and the gclient sync checkout can live in different locations

Comment 138

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Adding windows mozconfigs;r?mjf

Depends on D112131

Comment 139

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - Switching to a precomputed list of values for win setup_tools.py;r?dminor

Depends on D112132

Comment 140

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Allow using IsOnCurrentThread on AbstractThread. r?#xpcom-reviewers

Comment 141

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Move WebRtcCallWrapper to dedicated files. r?ng!

This patch also renames it WebrtcCallWrapper, to better match neighboring
classes starting on "Webrtc".

Comment 142

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Create CallWorkerThread which implements AbstractThread for a TaskQueueWrapper. r?ng!

Ideally TaskQueueWrapper would implement both webrtc::TaskQueueBase and
AbstractThread, but webrtc::TaskQueueBase is not refcounted so implementing both
in the same class is not possible.

This patch implements AbstractThread in a refcounted class CallWorkerThread on
top of TaskQueueWrapper. Because of SharedModuleThread not having a thread-safe
refcount, the same CallWorkerThread must be used for all Call instances.
PeerConnectionCtx's SharedWebrtcState becomes the canonical owner of the
CallWorkerThread, but WebrtcCallWrappers will all hold a strong reference to it.

The new AbstractThread allows us to use webrtc TaskQueues with Mozilla
specific higher order threading functions, like MozPromise, MediaEventSource,
and whatever else could come in handy, without having to use an explicit
CurrentTaskQueueSetter (which is not public) everywhere.

Comment 143

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Update conduit unittests to work with the Call worker thread. r?ng!

Comment 144

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Avoid `using namespace` in headers. r?ng!

Comment 145

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Constify RTCStatsTimestampMaker to make it obviously threadsafe. r?bwc!

Comment 146

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Constify VideoConduit members that are based on prefs. r?ng!

Comment 147

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Reinstate usage of mLockScaling in VideoConduit. r?ng!

Comment 148

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Make the GmpPluginNotifier owner thread injectable. r?bwc!

Comment 149

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Move VideoConduit to call worker thread for mCall and related access. r?ng!,r?bwc!

Comment 150

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Move AudioConduit to call worker thread for mCall and related access. r?ng!,r?bwc!

Comment 151

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Make mediapipeline_unittest pass and remove TaskQueueWrapper::MainAsCurrent. r?ng!,r?bwc!

This patch makes mediapipeline_unittest pass an event target to the Call wrapper
but keeps using main thread for that event target, so that it can keep the test
cases simpler by allowing sybnc calls into the conduits.

TaskQueueWrapper::MainAsCurrent is no longer needed as mediapipeline_unittest
was the last user of it. A simpler version remains dedicated in
mediapipeline_unittest instead.

Comment 152

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Make MediaPipeline natively async with WatchManager. r?bwc!

On its own this patch should mainly be seen as a refactor of the way we handle
races in MediaPipelineTransmit::SetTrack. Namely that Stop() immediately
followed by SetTrack is still exposed to racing.

This patch makes MediaPipelines use a WatchManager to manage their tracks.
For receive pipelines, mActive triggers whether we enable the receive listener.
For transmit pipelines, mActive and mDomTrack triggers whether mDomTrack's
underlying track is hooked up to feed data to the conduit. A hole is also
punched through this allowing for unittests to set an overriden send track
to avoid mocking a MediaStreamTrack (DOM object) and MediaTrackGraph (several
non-virtual methods we'd need to override).

This patch also moves the responsibility of starting and stopping conduits to
the transceiver/receiver objects, where most of the interaction with the
conduits is already happening.

Comment 153

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Pass codec configs by value/reference instead of pointers. r?ng!

This patch mainly sets the stage for using codec configs with StateMirroring.
It also makes the code a bit simpler, and removes the need to handle the nullptr
case.

Comment 154

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Rename MediaSessionConduit::DeleteStreams to Shutdown, as VideoConduit does more than just delete streams. r?ng!

Comment 155

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Move NoTrialsConfig into SharedWebrtcState and make WebrtcCallWrapper keep a ref to it. r?ng!

SharedWebrtcState when managed by PeerConnectionCtx is destroyed on main thread.
With interactions with call instances being async, there is no guarantee that
the objects owned by SharedWebrtcState (and referenced through raw pointers by
Call instances) are alive until all Call instances have been destroyed.

Making the WebrtcCallWrapper keep a strong-ref to the SharedWebrtcState will
ensure it can manage the lifetime of the SharedWebrtcState for as long as
necessary.

Comment 156

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Deliberately choose DTMF codec. r?ng!

The logic used for choosing codec in this patch mimics the logic of Chromium.

Comment 157

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Make RTCDTMFSender signal DTMF tones through a MediaEvent. r?ng!

Comment 158

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Rework AudioConduit DTMF API to be closer to the upstream API. r?ng!

Comment 159

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Make SharedWebrtcState able to take responsibility of destroying the Call worker thread. r?ng!

The case added by this patch will be used to destroy the worker thread on
shutdown, and have it release its resources. In particular the reference to its
underlying thread pool, which could be blocking shutdown.

Unittests might still use the case where responsibility of destroying the worker
thread is left external, to allow using SharedWebrtcState with main thread as
the worker thread.

Comment 160

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Allow TaskQueueWrapper creators to decide whether the underlying TaskQueue should support tail dispatch. r?ng!

This patch makes the Call worker thread support tail dispatch, since it is
required for state mirroring. Since tail dispatch is only used when both the
source and target threads support it, this will only be used for dispatch to or
from other Mozilla threads.

TaskQueueWrappers created by libwebrtc are made to not support tail dispatch,
since upstream code assumes that dispatches are direct. In some places this is
seen by one webrtc::TaskQueue dispatching a task to another webrtc::TaskQueue
and blocks the source thread until the task has run on the target.

Comment 161

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Alias Ssrc types. r?ng!

Comment 162

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Implement operator== for RtpRtcpConfig. r?ng!

Comment 163

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Remove unused WebrtcCallWrapper ctor. r?ng!

Comment 164

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Don't assert which thread we destroy AudioConduit on. r?bwc!

This patch makes AudioConduit mimic VideoConduit. We no longer do any
thread-specific actions in the dtor, so this assert is unnecessary.

Comment 165

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - In VideoConduit::SendVideoFrame protect against racy setters. r?bwc!

With VideoConduit setters being called on a worker thread rather than main, we
no longer have a guarantee that the first SendVideoFrame comes in after ssrcs
or the send codec config are set.

Setters have so far been called on main, and the listener that triggers
SendVideoFrame is also managed by main. Using a call worker thread would change
the thread the setters run on, but not the SendVideoFrame triggering listener.
This breaks the guarantee.

Comment 166

[Andreas Pehrson [:pehrsons]]

Attachment: Bug 1654112 - Implement MediaConduitControl for controlling conduits async through state mirroring. r?ng!, r?bwc!

Comment 167

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - fix try builds for macos. r?ng!

• Our build didn't like the non-explictness of these conversion
  operators.
• Changing to explicit required casts in various places.

Comment 168

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Adding androidx for building webrtc for android. r?ng!

Comment 169

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Remove unnecessary files. r?ng!

Depends on D113433

Comment 170

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Fix paths in mobile/android/geckoview/build.gradle. r?ng!

Depends on D113434

Comment 171

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - BUILD.gn and webrtc.gni changes. r?ng!

Depends on D113435

Comment 172

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Bulk change 'android.support.annotation' to 'androidx.annotation'. r?ng!

Depends on D113436

Comment 173

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - General build fixes for paths and naming changes. r?ng!

Depends on D113437

Comment 174

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Hack WebrtcMediaCodecVP8VideoCodec.* to build, but not work yet. r?ng!

Depends on D113438

Comment 175

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Fix a couple issues in generate-gn-build-files.sh. r?ng!

• pipefail doesn't work in sh on linux, so we're running under bash now.
• removing a link for third_party/libwebrtc/.git didn't work due to a typo.
• needed export on MOZCONFIG line.
• only grab the specific .json file based on the build config.

Depends on D113439

Comment 176

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Updated json files with changes from fixup_json.py included. r?ng!

Depends on D113440

Comment 177

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - moz.build files generated after json fixup. r?ng!

Depends on D113441

Comment 178

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Add android to linux configs in generate-gn-build-files.sh. r?ng!

Depends on D113442

Comment 179

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Generated json and mozconfig for android build. r?ng!

Depends on D113443

Comment 180

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - moz.build file changes for android. r?ng!

Depends on D113444

Comment 181

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - changing windows version define to the literal behind NTDDI_WIN10;r?mjf

Depends on D112133

Comment 182

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - adding windows generated gn json files;r?mjf!

Depends on D113604

Comment 183

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - Reformat fixup_json.py with black. r?ng!

Comment 184

[Nico Grunbaum [:ng, @chew:mozilla.org]]

Attachment: Bug 1654112 - moz.build changes for win-x64;r?mjf!

Depends on D113605

Comment 185

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Work around the old (<1.5) libxrandr headers on our build machines.

Comment 186

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Remove a bogus assertion.

Comment 187

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Don't stomp the config object created earlier.

Comment 188

[Byron Campen [:bwc]]

Attachment: Bug 1654112: libwebrtc modification: Init this to avoid crashes.

Only works on Chrome because they zero out alloced memory?

Comment 189

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Make sure the RunLoop that feeds screen capture frames gets cycles.

Involves calling CFRunLoopRunInMode on the capture thread (so the RunLoop gets
cycles), and also calling Start() on the capture thread (so the callbacks are
connected to the same RunLoop we're giving cycles to).

Comment 190

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Remove a step that waits for media flow in a test where an m-section is marked inactive.

Comment 191

[Byron Campen [:bwc]]

Attachment: Bug 1654112: Add some pipewire stubs.

Comment 192

[Michael Froman [:mjf]]

Attachment: Bug 1654112 - fix type mismatch in desktop_device_info.cc. r?ng!

Fixing error: format specifies type 'long' but the argument has type 'webrtc::ScreenId' (aka 'int')