FFmpeg fixups for M33 roll. Now with more Opus!

FFmpeg fixups for M33 roll. Now with more Opus!

Besides the DEPs update there's also:
- Remove deprecated usage of avcodec_alloc_frame()
- Remove deprecated usage of pkt->pkt_pts
- Fix end trimming for ogg opus playbacks.
- Remove fragile bytes decoded check for end trimming.

Not all of the Opus issues are fixed yet, this just
pulls in the FFmpeg side of the fixes.  Chrome changes
are in a followup patch set.

BUG=104241, 166752, 168524, 315165
TEST=unittests, asan, ogg opus links from bugs.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=240320

[vignesh, 2013-12-11 20:10:18]

drive by comments.

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
File media/filters/opus_audio_decoder.cc (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
media/filters/opus_audio_decoder.cc:605: statistics.audio_bytes_decoded =
bytes_decoded;
not sure if this change is required. even though we decode bytes_decoded, we
don't give out that many bytes out of the decoder in some cases. so i think it
should be left as it is. acolwell@, thoughts?

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
File media/filters/opus_audio_decoder.h (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
media/filters/opus_audio_decoder.h:91: scoped_ptr<float[]> output_buffer_;
i believe you need to change this too:
https://code.google.com/p/chromium/codesearch#chromium/src/media/ffmpeg/ffmpe...

Also, we do a fixed point build for opus in android [1]. So may be this should
be ifdef'd over !OS_ANDROID.

[1]
https://code.google.com/p/chromium/codesearch#chromium/src/third_party/opus/o...

https://codereview.chromium.org/111153003/diff/40001/media/filters/pipeline_i...
File media/filters/pipeline_integration_test.cc (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/pipeline_i...
media/filters/pipeline_integration_test.cc:1063:
pipeline_->GetBufferedTimeRanges().end(0).InMilliseconds());
Could you please verify manually if this test passes http://crbug.com/178309 (in
case you haven't already) ?

[acolwell GONE FROM CHROMIUM, 2013-12-11 20:22:31]

lgtm % comment

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
File media/filters/opus_audio_decoder.cc (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
media/filters/opus_audio_decoder.cc:605: statistics.audio_bytes_decoded =
bytes_decoded;
On 2013/12/11 20:10:18, vignesh wrote:
> not sure if this change is required. even though we decode bytes_decoded, we
> don't give out that many bytes out of the decoder in some cases. so i think it
> should be left as it is. acolwell@, thoughts?

This is actually still wrong. This stat is supposed to represent the number of
compressed bytes decoded. If I understand this code correctly, it should be the
following since it appears that opus_multistream_decode_float() consumes all of
the input data.

statistics.audio_bytes_decoded = input->data_size();

[DaleCurtis, 2013-12-11 22:19:36]

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
File media/filters/opus_audio_decoder.cc (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
media/filters/opus_audio_decoder.cc:605: statistics.audio_bytes_decoded =
bytes_decoded;
On 2013/12/11 20:22:32, acolwell wrote:
> On 2013/12/11 20:10:18, vignesh wrote:
> > not sure if this change is required. even though we decode bytes_decoded, we
> > don't give out that many bytes out of the decoder in some cases. so i think
it
> > should be left as it is. acolwell@, thoughts?
> 
> This is actually still wrong. This stat is supposed to represent the number of
> compressed bytes decoded. If I understand this code correctly, it should be
the
> following since it appears that opus_multistream_decode_float() consumes all
of
> the input data.
> 
> statistics.audio_bytes_decoded = input->data_size();

Done.

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
File media/filters/opus_audio_decoder.h (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/opus_audio...
media/filters/opus_audio_decoder.h:91: scoped_ptr<float[]> output_buffer_;
On 2013/12/11 20:10:18, vignesh wrote:
> i believe you need to change this too:
>
https://code.google.com/p/chromium/codesearch#chromium/src/media/ffmpeg/ffmpe...
> 
> Also, we do a fixed point build for opus in android [1]. So may be this should
> be ifdef'd over !OS_ANDROID.
> 
> [1]
>
https://code.google.com/p/chromium/codesearch#chromium/src/third_party/opus/o...

I'm actually going to revert the float change out of this patch set in favor of
a bigger cleanup which removes all the extraneous memcpy's. All float changes
now here: https://codereview.chromium.org/100503006/

https://codereview.chromium.org/111153003/diff/40001/media/filters/pipeline_i...
File media/filters/pipeline_integration_test.cc (right):

https://codereview.chromium.org/111153003/diff/40001/media/filters/pipeline_i...
media/filters/pipeline_integration_test.cc:1063:
pipeline_->GetBufferedTimeRanges().end(0).InMilliseconds());
On 2013/12/11 20:10:18, vignesh wrote:
> Could you please verify manually if this test passes http://crbug.com/178309
(in
> case you haven't already) ?

End trimming works, but start trimming appears broken or at least incomplete.

[commit-bot: I haz the power, 2013-12-12 04:07:42]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/111153003/80001

[commit-bot: I haz the power, 2013-12-12 14:20:10]

Change committed as 240320