Support more than 8 channels for AudioContext.destination node on OSX

Support more than 8 channels on AudioContext.destination node on OSX. Note that this patch does not fix the issue for Windows/Linux multichannel support.

When an audio hardware with 8 or more outputs used by Chrome, setting AudioContext.destination.channelCount beyond 8 results silence.

context.destination.channelCount = 9; // Result silence.

This is because the current implementation of channel layout with corresponding channel count is incomplete.

This patch resolves the issue by:
- Enforcing CHANNEL_LAYOUT_DISCRETE when user sets the channel count more than 8.
- When the channel layout is 'discrete', use explicit channel count specified by user to create a WebAudioDevice object.

BUG=424795
Committed: https://crrev.com/1d8996d3c518a873d213962990ebb6c755fdb8df
Cr-Commit-Position: refs/heads/master@{#340305}

[hongchan, 2015-01-08 21:48:06]

hongchan@chromium.org changed reviewers:
+ dalecurtis@chromium.org, rtoy@chromium.org

[Raymond Toy, 2015-01-08 21:55:54]

https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
File content/renderer/renderer_blink_platform_impl.cc (left):

https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
content/renderer/renderer_blink_platform_impl.cc:704: // TODO(crogers): WebKit
should give the channel layout instead of the hard
This TODO is still true isn't it?  WebKit does still give the channel count and
not the layout, right?

[hongchan, 2015-01-08 21:58:33]

https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
File content/renderer/renderer_blink_platform_impl.cc (left):

https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
content/renderer/renderer_blink_platform_impl.cc:704: // TODO(crogers): WebKit
should give the channel layout instead of the hard
On 2015/01/08 21:55:54, Raymond Toy wrote:
> This TODO is still true isn't it?  WebKit does still give the channel count
and
> not the layout, right?

Still true for WebKit, but here at Blink we have the implementation of
CHANNEL_LAYOUT_*. I thought it only applies to WebKit and the comment is
outdated - but I don't mind reverting the comment back.

[DaleCurtis, 2015-01-08 22:01:21]

This is fine with me, but I think this will only work on OSX since only the OSX
AudioManager and output driver supports CHANNEL_LAYOUT_DISCRETE.

The other platforms will need more extensive modifications to support this
correctly.

[Raymond Toy, 2015-01-08 22:02:57]

https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
File content/renderer/renderer_blink_platform_impl.cc (left):

https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
content/renderer/renderer_blink_platform_impl.cc:704: // TODO(crogers): WebKit
should give the channel layout instead of the hard
On 2015/01/08 21:58:33, hoch wrote:
> On 2015/01/08 21:55:54, Raymond Toy wrote:
> > This TODO is still true isn't it?  WebKit does still give the channel count
> and
> > not the layout, right?
> 
> Still true for WebKit, but here at Blink we have the implementation of
> CHANNEL_LAYOUT_*. I thought it only applies to WebKit and the comment is
> outdated - but I don't mind reverting the comment back.

But creatAudioDevice still takes unsigned channels, and not some kind of channel
layout enum or object. I think that's what the comment is referring to. 
(Substitute Blink for WebKit).

[DaleCurtis, 2015-01-08 22:03:03]

For this file I don't think there's much unittest you can add if there's not one
already. It'd be a lot of work to set one up here I believe.

I'm also not an owner for this code, so while lgtm, you'll need to find a
content owner.

[hongchan, 2015-01-08 22:09:41]

On 2015/01/08 22:01:21, DaleCurtis wrote:
> This is fine with me, but I think this will only work on OSX since only the
OSX
> AudioManager and output driver supports CHANNEL_LAYOUT_DISCRETE.
> 
> The other platforms will need more extensive modifications to support this
> correctly.

Could you please let me know where to start? Not sure where to look for  audio
devices on other platforms. Currently I am looking at:

src/media/audio/audio_manager.cc
src/media/audio/audio_manager.h
src/media/audio/audio_manager_base.cc
src/media/audio/audio_manager_base.h

With that said, it would be great if we can make this work on OSX at least.

[hongchan, 2015-01-08 22:11:38]

On 2015/01/08 22:02:57, Raymond Toy wrote:
>
https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
> File content/renderer/renderer_blink_platform_impl.cc (left):
> 
>
https://codereview.chromium.org/810763006/diff/1/content/renderer/renderer_bl...
> content/renderer/renderer_blink_platform_impl.cc:704: // TODO(crogers): WebKit
> should give the channel layout instead of the hard
> On 2015/01/08 21:58:33, hoch wrote:
> > On 2015/01/08 21:55:54, Raymond Toy wrote:
> > > This TODO is still true isn't it?  WebKit does still give the channel
count
> > and
> > > not the layout, right?
> > 
> > Still true for WebKit, but here at Blink we have the implementation of
> > CHANNEL_LAYOUT_*. I thought it only applies to WebKit and the comment is
> > outdated - but I don't mind reverting the comment back.
> 
> But creatAudioDevice still takes unsigned channels, and not some kind of
channel
> layout enum or object. I think that's what the comment is referring to. 
> (Substitute Blink for WebKit).

I see. Then I will bring the comment back in the patch set 2. Thanks!

[DaleCurtis, 2015-01-08 22:19:31]

Sorry I don't really have any pointer other than to see what AudioManagerMac is
doing with discrete.  You'll want to look at the PulseAudio, WASAPI, and WaveOut
drivers within media/audio/pulse and media/audio/win to figure out how channel
configurations work.

[hongchan, 2015-01-08 22:19:59]

On 2015/01/08 22:19:31, DaleCurtis wrote:
> Sorry I don't really have any pointer other than to see what AudioManagerMac
is
> doing with discrete.  You'll want to look at the PulseAudio, WASAPI, and
WaveOut
> drivers within media/audio/pulse and media/audio/win to figure out how channel
> configurations work.

Thanks for the pointer!

[hongchan, 2015-01-08 23:07:48]

hongchan@chromium.org changed reviewers:
+ avi@chromium.org
- dalecurtis@chromium.org, rtoy@chromium.org

[Avi (use Gerrit), 2015-01-09 19:02:39]

On 2015/01/08 22:19:59, hoch wrote:
> On 2015/01/08 22:19:31, DaleCurtis wrote:
> > Sorry I don't really have any pointer other than to see what AudioManagerMac
> is
> > doing with discrete.  You'll want to look at the PulseAudio, WASAPI, and
> WaveOut
> > drivers within media/audio/pulse and media/audio/win to figure out how
channel
> > configurations work.
> 
> Thanks for the pointer!

If Dale is happy, so am I.

LGTM

[Raymond Toy, 2015-01-09 19:07:13]

rtoy@chromium.org changed reviewers:
+ rtoy@chromium.org

[Raymond Toy, 2015-01-09 19:07:14]

Please file a bug to say this needs to be implemented for Windows and Linux. 
(And Android too?) And reference this bug in the description.

[hongchan, 2015-01-09 19:19:15]

On 2015/01/09 19:07:14, Raymond Toy wrote:
> Please file a bug to say this needs to be implemented for Windows and Linux. 
> (And Android too?) And reference this bug in the description.

As Dale suggested, Ray and I decided to postpone this CL until we test
Windows/Linux platform. We are ordering a 16 channel audio interface soon, so
we'll be able to run this patch locally. Meanwhile I will try to look at
PulseAudio, WASAPI, and WaveOut drivers as well.

I will keep you all posted. Thanks!

[Avi (use Gerrit), 2015-01-09 19:27:41]

On 2015/01/09 19:19:15, hoch wrote:
> On 2015/01/09 19:07:14, Raymond Toy wrote:
> > Please file a bug to say this needs to be implemented for Windows and Linux.

> > (And Android too?) And reference this bug in the description.
> 
> As Dale suggested, Ray and I decided to postpone this CL until we test
> Windows/Linux platform. We are ordering a 16 channel audio interface soon, so
> we'll be able to run this patch locally. Meanwhile I will try to look at
> PulseAudio, WASAPI, and WaveOut drivers as well.
> 
> I will keep you all posted. Thanks!

CLs that require you to buy new toys with which to test them are the best CLs.
Enjoy! :)

[hongchan, 2015-07-24 17:23:52]

Here is the verification report after L-G-T-M:

Tested PS3 with the recently obtained multichannel audio device (FCA1616):

- On OSX (10.10.4), PS3 successfully could play the audio through 16 individual
channels.
- On Windows 7, I've tested and verified this patch with FCA1616 which
unfortunately supports only up to 8 channels on a single audio device entry.
- On Ubuntu, the test device is not compatible with PulseAudio so I couldn't
verify the multichannel test. However, the patched build didn't crash with the
default stereo device.
xsrf_token: 5b9f5ad1eab75c0703395ac6c38f29dd
send_mail: on
subject: Support more than 8 channels for AudioContext.destination node on OSX

[hongchan, 2015-07-24 17:52:44]

Patchset #3 (id:40001) has been deleted

[hongchan, 2015-07-24 17:52:51]

The CQ bit was checked by hongchan@chromium.org

[hongchan, 2015-07-24 17:52:52]

The patchset sent to the CQ was uploaded after l-g-t-m from avi@chromium.org,
dalecurtis@chromium.org
Link to the patchset: https://codereview.chromium.org/810763006/#ps60001 (title:
"Bring ToT")

[commit-bot: I haz the power, 2015-07-24 17:53:11]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/810763006/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/810763006/60001

[DaleCurtis, 2015-07-24 17:53:12]

dalecurtis@chromium.org changed reviewers:
+ dalecurtis@chromium.org

[DaleCurtis, 2015-07-24 17:53:12]

Still lgtm, thanks for getting the toys and testing :)

[commit-bot: I haz the power, 2015-07-24 19:05:49]

Committed patchset #3 (id:60001)

[commit-bot: I haz the power, 2015-07-24 19:06:36]

Patchset 3 (id:??) landed as
https://crrev.com/1d8996d3c518a873d213962990ebb6c755fdb8df
Cr-Commit-Position: refs/heads/master@{#340305}