Fixed the racing in the resampler of webrtc when two streams are calling OnData()

Fixed the racing in the resampler of webrtc when two streams are calling OnData().

There can be multiple streams calling OnData() when the users setup multiple getUserMedia with WebAudio and microphone. We have no way to avoid this situation since it is a valid user cases from the JS perspective.
But unfortunately webrtc does not support it since it has one global transmit mixer, and it will crash when more than 1 stream getting into its resampler.

This CL introduces a lock to sequence the calls to avoid the racing, which should fix the crash.


BUG=332126
TEST=https://devpool-13.talkgadget.google.com/hangouts, go to setting page and play the rington, verify no crash and sound good.
Or go to https://x20web.corp.google.com/~xians/multiple_sources.html, start the call, AddWebAUdio, then Playout a clip, verify the audio sounds good and no crash.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=243902

[no longer working on chromium, 2014-01-09 09:55:21]

Henrik, would you please review this CL?

SX

[henrika (OOO until Aug 14), 2014-01-09 10:07:14]

What happens in the demo that crashed before using this patch? Does it now work
as intended or do you simply avoid a crash but something is still missing?

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
File content/renderer/media/webrtc_audio_device_impl.cc (right):

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
content/renderer/media/webrtc_audio_device_impl.cc:86: // there are more than
one input streams calling OnData(), which can happen
nit, ...input stream calling...

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
content/renderer/media/webrtc_audio_device_impl.cc:90: // TODO(xians): Remove
this workaround after we move the
Add link to crbug issue here.

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
File content/renderer/media/webrtc_audio_device_impl.h (right):

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
content/renderer/media/webrtc_audio_device_impl.h:389: // than one input streams
calling OnData().
Do we know on which threads these calls are made? If so, should we document it?

[no longer working on chromium, 2014-01-09 10:18:32]

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
File content/renderer/media/webrtc_audio_device_impl.cc (right):

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
content/renderer/media/webrtc_audio_device_impl.cc:86: // there are more than
one input streams calling OnData(), which can happen
On 2014/01/09 10:07:14, henrika wrote:
> nit, ...input stream calling...

Done.

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
content/renderer/media/webrtc_audio_device_impl.cc:90: // TODO(xians): Remove
this workaround after we move the
On 2014/01/09 10:07:14, henrika wrote:
> Add link to crbug issue here.

Done.

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
File content/renderer/media/webrtc_audio_device_impl.h (right):

https://codereview.chromium.org/131213003/diff/1/content/renderer/media/webrt...
content/renderer/media/webrtc_audio_device_impl.h:389: // than one input streams
calling OnData().
On 2014/01/09 10:07:14, henrika wrote:
> Do we know on which threads these calls are made? If so, should we document
it?

Threads can be the from WebRtcAudioCapturer and from WebAudio, or from different
WebAudio threads.

I am not sure if can define which threads they are since it is up to the JS.

[henrika (OOO until Aug 14), 2014-01-09 10:35:55]

LGTM. Perhaps you should make some ML tests and additional WebRTC + WebAudio
tests as well just in case.

[no longer working on chromium, 2014-01-09 10:58:33]

On 2014/01/09 10:35:55, henrika wrote:
> LGTM. Perhaps you should make some ML tests and additional WebRTC + WebAudio
> tests as well just in case.

Done with adding additional WebRtc + WebAudio test case:
go to https://x20web.corp.google.com/~xians/multiple_sources.html, start the
call, AddWebAUdio, then Playout a clip, verify the audio sounds good and no
crash.

[commit-bot: I haz the power, 2014-01-09 11:00:48]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/xians@chromium.org/131213003/60001

[commit-bot: I haz the power, 2014-01-09 14:13:35]

Retried try job too often on win_rel for step(s) browser_tests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...

[commit-bot: I haz the power, 2014-01-09 14:15:30]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/xians@chromium.org/131213003/60001

[commit-bot: I haz the power, 2014-01-09 15:17:37]

Change committed as 243902

[tommi (sloooow) - chröme, 2014-01-10 15:48:00]

shouldn't this be fixed in WebRTC?

[no longer working on chromium, 2014-01-10 15:51:43]

On 2014/01/10 15:48:00, tommi wrote:
> shouldn't this be fixed in WebRTC?

It is right that the problem can be fixed either in WebRtc or in Chrome. But
this is Chrome specific problem, so it is better to fix in Chrome.
Also, the issue exists from M31, if possible, I would like to try merging it to
the previous milestones.