Fix the way locks are acquired in WebRtcAudioDeviceImpl::SetAudioRenderer.

WebRtcAudioDeviceImpl and WebRtcAudioRenderer reference each other and are invoked in different threads.
This CL ensures that locks are taken in the same order in both threads in order to avoid deadlocks.

BUG=433993
TEST=Existing WebRtc content_browsertests with TSAN enabled.

Committed: https://crrev.com/5330f31cf74e3756dc2103f8c3d5840b5f6a2884
Cr-Commit-Position: refs/heads/master@{#322995}

[Guido Urdaneta, 2015-03-26 17:08:15]

guidou@chromium.org changed reviewers:
+ hbos@chromium.org, tommi@chromium.org

[Guido Urdaneta, 2015-03-26 17:44:06]

guidou@chromium.org changed reviewers:
+ henrika@chromium.org

[Guido Urdaneta, 2015-03-27 08:44:27]

The CQ bit was checked by guidou@chromium.org

[Guido Urdaneta, 2015-03-27 08:44:28]

The CQ bit was unchecked by guidou@chromium.org

[Guido Urdaneta, 2015-03-30 10:03:25]

Hi,

Please take a look at my first CL.

[henrika (OOO until Aug 14), 2015-03-30 10:37:50]

LGTM.

Given that you now have analyzed this code. Would it also be possible to improve
the comments in the header file regarding the assumed threading model?

[henrika (OOO until Aug 14), 2015-03-30 10:39:46]

Not sure why, but I have two comments that are in draft state but I am unable to
publish them. They are just nits. When you mention locks_ in comments, please
write |locks_| instead.

[hbos_chromium, 2015-03-30 11:11:30]

lgtm

[Guido Urdaneta, 2015-03-30 11:27:20]

On 2015/03/30 10:39:46, henrika wrote:
> Not sure why, but I have two comments that are in draft state but I am unable
to
> publish them. They are just nits. When you mention locks_ in comments, please
> write |locks_| instead.

Patch set 4 uses the || convention to refer to all variables.

[Guido Urdaneta, 2015-03-30 11:29:18]

On 2015/03/30 10:37:50, henrika wrote:
> LGTM.
> 
> Given that you now have analyzed this code. Would it also be possible to
improve
> the comments in the header file regarding the assumed threading model?

I did a minor update to a header file comment about threading. Let me know if
it's a valid improvement or if it's better to leave it as it wa.

[tommi (sloooow) - chröme, 2015-03-31 08:20:50]

The CQ bit was checked by tommi@chromium.org

[tommi (sloooow) - chröme, 2015-03-31 08:20:51]

The patchset sent to the CQ was uploaded after l-g-t-m from hbos@chromium.org,
henrika@chromium.org
Link to the patchset: https://codereview.chromium.org/1036993003/#ps60001
(title: "Name variables in comments according to convention, extra header
comment.")

[commit-bot: I haz the power, 2015-03-31 08:21:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1036993003/60001

[tommi (sloooow) - chröme, 2015-03-31 08:29:13]

lgtm w/comment addressed.

https://codereview.chromium.org/1036993003/diff/60001/content/renderer/media/...
File content/renderer/media/webrtc_audio_device_impl.cc (right):

https://codereview.chromium.org/1036993003/diff/60001/content/renderer/media/...
content/renderer/media/webrtc_audio_device_impl.cc:423:
DCHECK(!renderer_.get());
Would also be good to have a note here about how it's guaranteed that only one
thread will be calling SetAudioRenderer() at a time. (i.e. how we guarantee to
never hit this check).  Actually, can you change this to be a regular CHECK(),
not just DCHECK()?

[commit-bot: I haz the power, 2015-03-31 08:31:48]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-03-31 08:32:39]

Patchset 4 (id:??) landed as
https://crrev.com/5330f31cf74e3756dc2103f8c3d5840b5f6a2884
Cr-Commit-Position: refs/heads/master@{#322995}