Immediately reset for vp8 when there's nothing to flush.

Immediately reset for vp8 when there's nothing to flush.

If we finish playback of a clip up to end-of-stream, the codec is
already Flush()'d so when we later recieve a Reset(), if we try
to start another Flush() there is no event-pump driving the flush
and we'll hang forever waiting for the flush that never was.

The fix is to instead only flush when there are actually buffers in
the decoder. If there are no buffers, there is nothing to flush.

BUG=619967
TEST=looping video plays

Committed: https://crrev.com/eec6e9013286349a66ace2ec0858c99e96132f46
Cr-Commit-Position: refs/heads/master@{#400338}

[DaleCurtis, 2016-06-16 22:45:31]

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

[Tima Vaisburd, 2016-06-16 23:01:34]

https://codereview.chromium.org/2074983002/diff/1/media/gpu/android_video_dec...
File media/gpu/android_video_decode_accelerator.cc (right):

https://codereview.chromium.org/2074983002/diff/1/media/gpu/android_video_dec...
media/gpu/android_video_decode_accelerator.cc:1272:
!bitstream_buffers_in_decoder_.empty()) {
What if the decoder stalled, i.e. no input for some time but waiting for more
input? Did you consider having some initial state that ResetCodecState() would
come to? Right now it is NO_ERROR which is the same as during normal playback...

[DaleCurtis, 2016-06-16 23:20:47]

https://codereview.chromium.org/2074983002/diff/1/media/gpu/android_video_dec...
File media/gpu/android_video_decode_accelerator.cc (right):

https://codereview.chromium.org/2074983002/diff/1/media/gpu/android_video_dec...
media/gpu/android_video_decode_accelerator.cc:1272:
!bitstream_buffers_in_decoder_.empty()) {
On 2016/06/16 at 23:01:34, Tima Vaisburd wrote:
> What if the decoder stalled, i.e. no input for some time but waiting for more
input? Did you consider having some initial state that ResetCodecState() would
come to? Right now it is NO_ERROR which is the same as during normal playback...

Does it matter? I assume if the decoder has nothing in it, we wouldn't hit the
issue this fix is solving.

[Tima Vaisburd, 2016-06-16 23:47:29]

Inital state feels more straightforward.

lgtm

https://codereview.chromium.org/2074983002/diff/1/media/gpu/android_video_dec...
File media/gpu/android_video_decode_accelerator.cc (right):

https://codereview.chromium.org/2074983002/diff/1/media/gpu/android_video_dec...
media/gpu/android_video_decode_accelerator.cc:1272:
!bitstream_buffers_in_decoder_.empty()) {
On 2016/06/16 23:20:47, DaleCurtis wrote:
> On 2016/06/16 at 23:01:34, Tima Vaisburd wrote:
> > What if the decoder stalled, i.e. no input for some time but waiting for
more
> input? Did you consider having some initial state that ResetCodecState() would
> come to? Right now it is NO_ERROR which is the same as during normal
playback...
> 
> Does it matter? I assume if the decoder has nothing in it, we wouldn't hit the
> issue this fix is solving.

If the decoder has nothing inside, we proceed to ResetCodecState() and flush().
I agree that this flush() should not hang...

[DaleCurtis, 2016-06-17 00:16:40]

The CQ bit was checked by dalecurtis@chromium.org

[DaleCurtis, 2016-06-17 00:16:40]

Lets save the state machine changes for MediaCodecLoop.

[commit-bot: I haz the power, 2016-06-17 00:17:11]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2074983002/1

[commit-bot: I haz the power, 2016-06-17 03:16:12]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-06-17 03:19:42]

Description was changed from

==========
Immediately reset for vp8 when there's nothing to flush.

If we finish playback of a clip up to end-of-stream, the codec is
already Flush()'d so when we later recieve a Reset(), if we try
to start another Flush() there is no event-pump driving the flush
and we'll hang forever waiting for the flush that never was.

The fix is to instead only flush when there are actually buffers in
the decoder. If there are no buffers, there is nothing to flush.

BUG=619967
TEST=looping video plays
==========

to

==========
Immediately reset for vp8 when there's nothing to flush.

If we finish playback of a clip up to end-of-stream, the codec is
already Flush()'d so when we later recieve a Reset(), if we try
to start another Flush() there is no event-pump driving the flush
and we'll hang forever waiting for the flush that never was.

The fix is to instead only flush when there are actually buffers in
the decoder. If there are no buffers, there is nothing to flush.

BUG=619967
TEST=looping video plays

Committed: https://crrev.com/eec6e9013286349a66ace2ec0858c99e96132f46
Cr-Commit-Position: refs/heads/master@{#400338}
==========

[commit-bot: I haz the power, 2016-06-17 03:19:43]

Patchset 1 (id:??) landed as
https://crrev.com/eec6e9013286349a66ace2ec0858c99e96132f46
Cr-Commit-Position: refs/heads/master@{#400338}