Fixed ended event when playing audio on linux.

Fixed ended event when playing audio on linux.

AudioRendererBase::FillBuffers() sends ended event only when playback_delay == 0, but with ALSA output the delay was never set to 0 because it includes hardware delay. Changed ALSA output to include only internal buffers in the delay value: this matches windows and mac versions.

BUG=45074
TEST=<audio> fires ended event when necessary

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=52407

[Sergey Ulanov, 2010-07-13 20:12:10]

Here I removed calls to snd_pcm_delay() from the alsa output and so now the
delay passed to OnMoreData() includes delay of the buffers, but not device
drivers delay and hardware delay. This is not a perfect solution, but this
behavior matches Windows and Mac versions, and on most systems hardware delay is
low enough to be ignored, so this should not cause video and audio to be out of
sync.

[awong, 2010-07-14 19:43:41]

...okay, the specific change makes sense, but I don't know if I agree with
removing the pcm delay check.

Can you add more explanation into the CR description for how this exactly fixes
what it's fixing? That's important because it's put into the changelog, and it's
non-obvious from the change why we're removing this.

As for whether we should remove the snd_pcm_delay, have you tested this on a USB
audio device?  I think that's where we might actually see real delays.

-Albert

[Sergey Ulanov, 2010-07-14 21:26:27]

On 2010/07/14 19:43:41, awong wrote:
> ...okay, the specific change makes sense, but I don't know if I agree with
> removing the pcm delay check.
> 
> Can you add more explanation into the CR description for how this exactly
fixes
> what it's fixing? That's important because it's put into the changelog, and
it's
> non-obvious from the change why we're removing this.
Added more detail on how the bug is fixed.

> As for whether we should remove the snd_pcm_delay, have you tested this on a
USB
> audio device?  I think that's where we might actually see real delays.

I didn't try it, and I agree that it may be a problem with some devices.
Currently this problem exists on all platforms, and needs to be solved on all of
them.

> 
> -Albert

[awong, 2010-07-14 21:32:15]

LGTM

Can you create a new bug about how to deal with device delay?

On Wed, Jul 14, 2010 at 2:26 PM, <sergeyu@chromium.org> wrote:

> On 2010/07/14 19:43:41, awong wrote:
>
>> ...okay, the specific change makes sense, but I don't know if I agree with
>> removing the pcm delay check.
>>
>
>  Can you add more explanation into the CR description for how this exactly
>>
> fixes
>
>> what it's fixing? That's important because it's put into the changelog,
>> and
>>
> it's
>
>> non-obvious from the change why we're removing this.
>>
> Added more detail on how the bug is fixed.
>
>
>  As for whether we should remove the snd_pcm_delay, have you tested this on
>> a
>>
> USB
>
>> audio device?  I think that's where we might actually see real delays.
>>
>
> I didn't try it, and I agree that it may be a problem with some devices.
> Currently this problem exists on all platforms, and needs to be solved on
> all of
> them.
>
>
>
>  -Albert
>>
>
> http://codereview.chromium.org/2978005/show
>