Cancel pending seek in ChunkDemuxer when a new seek occurs

Cancel pending seek in ChunkDemuxer when a new seek occurs

This makes it so playback doesn't stall if one seeks into an unbuffered
region then seeks back into a buffered region.

BUG=132935
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=150489

[vrk (LEFT CHROMIUM), 2012-08-07 19:59:51]

PTAL! The threading bug in the last CL was easily reproducible by adding a
sleep() in Pipeline::SeekTask(), and this CL fixes the problem.

I added a CANCELLED state to ChunkDemuxerStream in which it ignores incoming
Seek() requests and returns null buffers. ChunkDemuxerStream transitions out of
this state when StartWaitingForSeek() is called, which will always happen before
a new Seek() request. Let me know if this is too icky/weird.

[acolwell GONE FROM CHROMIUM, 2012-08-07 20:26:51]

Looks pretty good to me. Tests?

http://codereview.chromium.org/10835044/diff/3001/media/filters/chunk_demuxer.cc
File media/filters/chunk_demuxer.cc (right):

http://codereview.chromium.org/10835044/diff/3001/media/filters/chunk_demuxer...
media/filters/chunk_demuxer.cc:297: ChangeState_Locked(CANCELED);
I think you may need to copy the read_cbs code from StartWaitingForSeek() to
here as well since any pending reads will need to be aborted.

http://codereview.chromium.org/10835044/diff/3001/media/filters/chunk_demuxer...
media/filters/chunk_demuxer.cc:630: audio_->BeginReturningNullBuffers();
nit:s/BeginReturningNullBuffers/CancelPendingSeek/ ?

[vrk (LEFT CHROMIUM), 2012-08-07 22:44:51]

Now with tests!

https://chromiumcodereview.appspot.com/10835044/diff/3001/media/filters/chunk...
File media/filters/chunk_demuxer.cc (right):

https://chromiumcodereview.appspot.com/10835044/diff/3001/media/filters/chunk...
media/filters/chunk_demuxer.cc:297: ChangeState_Locked(CANCELED);
On 2012/08/07 20:26:51, acolwell wrote:
> I think you may need to copy the read_cbs code from StartWaitingForSeek() to
> here as well since any pending reads will need to be aborted.

Done.

https://chromiumcodereview.appspot.com/10835044/diff/3001/media/filters/chunk...
media/filters/chunk_demuxer.cc:630: audio_->BeginReturningNullBuffers();
On 2012/08/07 20:26:51, acolwell wrote:
> nit:s/BeginReturningNullBuffers/CancelPendingSeek/ ?

...yeah that would make more sense :P Done!

[acolwell GONE FROM CHROMIUM, 2012-08-07 23:36:11]

LGTM % nits

https://chromiumcodereview.appspot.com/10835044/diff/1007/media/filters/chunk...
File media/filters/chunk_demuxer.cc (right):

https://chromiumcodereview.appspot.com/10835044/diff/1007/media/filters/chunk...
media/filters/chunk_demuxer.cc:285: // Ignore seek requests when cancelled.
s/cancelled/canceled/

https://chromiumcodereview.appspot.com/10835044/diff/1007/media/filters/chunk...
media/filters/chunk_demuxer.cc:639: audio_->CancelPendingSeek();
I think these should to be inside the locked section.

[vrk (LEFT CHROMIUM), 2012-08-08 00:01:21]

Thanks for reviewing!

https://chromiumcodereview.appspot.com/10835044/diff/1007/media/filters/chunk...
File media/filters/chunk_demuxer.cc (right):

https://chromiumcodereview.appspot.com/10835044/diff/1007/media/filters/chunk...
media/filters/chunk_demuxer.cc:285: // Ignore seek requests when cancelled.
On 2012/08/07 23:36:11, acolwell wrote:
> s/cancelled/canceled/

Done.

https://chromiumcodereview.appspot.com/10835044/diff/1007/media/filters/chunk...
media/filters/chunk_demuxer.cc:639: audio_->CancelPendingSeek();
On 2012/08/07 23:36:11, acolwell wrote:
> I think these should to be inside the locked section.

Oops, yes! Just the callback should be run outside of the lock.

[commit-bot: I haz the power, 2012-08-08 00:01:34]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/vrk@chromium.org/10835044/2003

[commit-bot: I haz the power, 2012-08-08 03:07:28]

Change committed as 150489