Reland 15721002: Hook up the device selection to the WebAudio live audio

Try relanding this CL, the original CL passed the try bots but failed the mac 10.7 bot because the oroginal CL queried the device info while there is no device on the bot.

Hook up the device selection to the WebAudio live audio.
WebAudio live audio needs to pass the session_id to the browser process so that Chrome can open the correct input device for unitfied IO.

This CL looks big because it touches quite some interfaces from the render to the browser. But the change is simple and basically adding a session_id/device_id to the classes. All the changes some together and it is very hard to break it down.
It also makes the media output code more similar to the media input code as well, and it will be easier to merge them for the future.


TBR=henrika@chormium.org

BUG=147327
TEST=http://chromium.googlecode.com/svn/trunk/samples/audio/visualizer-live.html
Change the device using the camera icon on the right of the omnibox, then reload. Verify the sound is coming from the correct input device.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=203695

[commit-bot: I haz the power, 2013-06-03 10:49:20]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/xians@chromium.org/15979015/67

[no longer working on chromium, 2013-06-03 11:09:15]

Committed patchset #2 manually as r203695 (presubmit successful).

[Jeffrey Yasskin, 2013-08-07 23:00:47]

https://codereview.chromium.org/15979015/diff/67/media/audio/audio_manager_ba...
File media/audio/audio_manager_base.cc (right):

https://codereview.chromium.org/15979015/diff/67/media/audio/audio_manager_ba...
media/audio/audio_manager_base.cc:260: return new
AudioOutputProxy(dispatcher.get());
This leaks the dispatcher_params. What did you mean to do with that object?

You should also add a direct test of this code path so that we'd find the leak
faster. I saw it in
http://build.chromium.org/p/chromium.memory.fyi/builders/Webkit%20Linux%20%28...,
but it doesn't look reliable there.

[no longer working on chromium, 2013-08-08 08:34:59]

https://codereview.chromium.org/15979015/diff/67/media/audio/audio_manager_ba...
File media/audio/audio_manager_base.cc (right):

https://codereview.chromium.org/15979015/diff/67/media/audio/audio_manager_ba...
media/audio/audio_manager_base.cc:260: return new
AudioOutputProxy(dispatcher.get());
On 2013/08/07 23:00:47, Jeffrey Yasskin wrote:
> This leaks the dispatcher_params. What did you mean to do with that object?
> 
> You should also add a direct test of this code path so that we'd find the leak
> faster. I saw it in
>
http://build.chromium.org/p/chromium.memory.fyi/builders/Webkit%2520Linux%252...,
> but it doesn't look reliable there.

Oh, thanks for finding it out, I am making a CL to fix it now.
We do have tests for this MakeAudioOutputStreamProxy() API, but the reason why
they don't exercise the piece code is that we use AUDIO_FAKE stream on our bots,
in order to avoid opening real hardware devices.