media: Continue aborting FFMpegDemuxerStream::Read() calls until Seek().

media: Continue aborting FFMpegDemuxerStream::Read() calls until Seek().

For decoders that support multiple in-flight decodes (ie.
GpuVideoDecoder), it was possible to schedule a Read() after
AbortPendingReads() but before Seek() was called. These reads were
satisfied with data that decoders were not expecting (because the
aborted buffers were fully dropped, there were gaps in the input).

Now AbortPendingReads() is defined to continue to abort Read()s until
Seek(), which solves the problem. ChunkDemuxer was already implementing
this behavior, so only FFmpegDemuxer requires changes.

BUG=667210
Committed: https://crrev.com/216002cff0b287035a57cacc510419e15e5217c9
Cr-Commit-Position: refs/heads/master@{#436428}

[sandersd (OOO until July 31), 2016-12-03 00:57:57]

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

[sandersd (OOO until July 31), 2016-12-03 01:06:39]

The CQ bit was checked by sandersd@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-12-03 01:07:14]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[DaleCurtis, 2016-12-03 01:11:08]

https://codereview.chromium.org/2546263002/diff/1/media/filters/ffmpeg_demuxe...
File media/filters/ffmpeg_demuxer.cc (right):

https://codereview.chromium.org/2546263002/diff/1/media/filters/ffmpeg_demuxe...
media/filters/ffmpeg_demuxer.cc:648:
base::ResetAndReturn(&read_cb_).Run(kAborted, nullptr);
return;

[sandersd (OOO until July 31), 2016-12-03 01:15:18]

https://codereview.chromium.org/2546263002/diff/1/media/filters/ffmpeg_demuxe...
File media/filters/ffmpeg_demuxer.cc (right):

https://codereview.chromium.org/2546263002/diff/1/media/filters/ffmpeg_demuxe...
media/filters/ffmpeg_demuxer.cc:648:
base::ResetAndReturn(&read_cb_).Run(kAborted, nullptr);
On 2016/12/03 01:11:08, DaleCurtis wrote:
> return;

Done.

[sandersd (OOO until July 31), 2016-12-03 01:15:28]

The CQ bit was checked by sandersd@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2016-12-03 01:16:19]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[DaleCurtis, 2016-12-03 01:16:20]

lgtm

[commit-bot: I haz the power, 2016-12-03 02:50:58]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-12-03 02:50:59]

Dry run: This issue passed the CQ dry run.

[sandersd (OOO until July 31), 2016-12-05 19:05:30]

The CQ bit was checked by sandersd@chromium.org

[commit-bot: I haz the power, 2016-12-05 19:05:58]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-12-05 22:29:02]

CQ is committing da patch.

Bot data: {"patchset_id": 20001, "attempt_start_ts": 1480964730786190,
"parent_rev": "ad024a47705e7bbdad3eb3c20d0ca0f01a625fc3", "commit_rev":
"b564f655a1da06394ed785babcf216128fc1e16a"}

[commit-bot: I haz the power, 2016-12-05 22:29:33]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-12-05 22:32:51]

Description was changed from

==========
media: Continue aborting FFMpegDemuxerStream::Read() calls until Seek().

For decoders that support multiple in-flight decodes (ie.
GpuVideoDecoder), it was possible to schedule a Read() after
AbortPendingReads() but before Seek() was called. These reads were
satisfied with data that decoders were not expecting (because the
aborted buffers were fully dropped, there were gaps in the input).

Now AbortPendingReads() is defined to continue to abort Read()s until
Seek(), which solves the problem. ChunkDemuxer was already implementing
this behavior, so only FFmpegDemuxer requires changes.

BUG=667210
==========

to

==========
media: Continue aborting FFMpegDemuxerStream::Read() calls until Seek().

For decoders that support multiple in-flight decodes (ie.
GpuVideoDecoder), it was possible to schedule a Read() after
AbortPendingReads() but before Seek() was called. These reads were
satisfied with data that decoders were not expecting (because the
aborted buffers were fully dropped, there were gaps in the input).

Now AbortPendingReads() is defined to continue to abort Read()s until
Seek(), which solves the problem. ChunkDemuxer was already implementing
this behavior, so only FFmpegDemuxer requires changes.

BUG=667210

Committed: https://crrev.com/216002cff0b287035a57cacc510419e15e5217c9
Cr-Commit-Position: refs/heads/master@{#436428}
==========

[commit-bot: I haz the power, 2016-12-05 22:32:52]

Patchset 2 (id:??) landed as
https://crrev.com/216002cff0b287035a57cacc510419e15e5217c9
Cr-Commit-Position: refs/heads/master@{#436428}