Defer changes to the channel count mode to the pre or post-render phase

Defer changes to the channel count mode to the pre or post-render phase

In some situations, the channel count mode is updated just before the
audio rendering starts. During audio processing, the channel count
mode is used to set the number of channels for processing.  However,
for AudioBasicProcessor's, the number of kernels has not yet been
updated to the correct number of channels, so the incorrect number of
channels is accessed.

We delay the setting of the channel count mode until the pre or post
rendering phase of the audio thread where it is safe to change things
since the audio graph is not being pulled.

BUG=407489
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=181573

[Raymond Toy, 2014-09-05 20:00:29]

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

[Raymond Toy, 2014-09-05 20:00:29]

PTAL. 

I'm not sure I got everything right for Oilpan.

[haraken, 2014-09-06 15:29:33]

haraken@chromium.org changed reviewers:
+ oilpan-reviews@chromium.org

[haraken, 2014-09-06 15:29:34]

LGTM

It looks safe to use a hash set of raw AudioNode pointers for
m_deferredCountModeChange, because the AudioNode pointers are removed from the
hash set in AudioNode::dispose(). In other words, it's not possible that the
hash set holds dangling AudioNode pointers.

https://codereview.chromium.org/543073004/diff/1/Source/modules/webaudio/Audi...
File Source/modules/webaudio/AudioContext.cpp (right):

https://codereview.chromium.org/543073004/diff/1/Source/modules/webaudio/Audi...
Source/modules/webaudio/AudioContext.cpp:912:
m_deferredCountModeChange.remove(node);

Shall we add ASSERT(isGraphOwner()) ?

[haraken, 2014-09-06 15:30:50]

Probably can we add a test for this change?

[Raymond Toy, 2014-09-08 17:49:41]

On 2014/09/06 15:30:50, haraken wrote:
> Probably can we add a test for this change?

I had a test based on the example from the bug report but there are issues in
debug mode. I'll explain further in the bug report.

[Raymond Toy, 2014-09-08 18:04:39]

Thanks for the review!

https://codereview.chromium.org/543073004/diff/1/Source/modules/webaudio/Audi...
File Source/modules/webaudio/AudioContext.cpp (right):

https://codereview.chromium.org/543073004/diff/1/Source/modules/webaudio/Audi...
Source/modules/webaudio/AudioContext.cpp:912:
m_deferredCountModeChange.remove(node);
On 2014/09/06 15:29:33, haraken wrote:
> 
> Shall we add ASSERT(isGraphOwner()) ?

Done.

[Raymond Toy, 2014-09-08 18:04:48]

The CQ bit was checked by rtoy@chromium.org

[commit-bot: I haz the power, 2014-09-08 18:05:53]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/rtoy@chromium.org/543073004/20001

[commit-bot: I haz the power, 2014-09-08 19:55:52]

Committed patchset #2 (id:20001) as 181573