Fix media code to work with new ffmpeg.

Fix media code to work with new ffmpeg.

Once ffmpeg git, svn are updated to new ffmpeg, will update DEPS
in this CL.  Which when committed, should seal the deal for the
ffmpeg roll.

API Changes:
avutil: SampleFormat->AVSampleFormat
avutil: av_get_bits_per_sample_fmt -> av_get_bytes_per_sample
avcodec: avcodec_open -> avcodec_open2(..., NULL)
avcodec: avcodec_decode_video2(... AVPacket ...) -> const AVPacket.
avformat: av_open_input_file -> avformat_open_input
avformat: av_register_protocol2 -> ffurl_register_protocol
avformat: av_close_input_file -> avformat_close_input(&...)
avformat: av_find_stream_info -> avformat_find_stream_info(..., NULL)

URLContext now has a url_open2 method as well, for now I've set this
to NULL.

Also fixes:
- ffmpeg_unittests change threading to mirror ffmpeg_video_decoder. There's
an issue where threading causes the last frames of a no-audio video to be
clipped. It existed before this ffmpeg roll, but because threading was
disabled by default in ffmpeg, we never noticed it.

- ffmpeg_demuxer_tests: GetBitrate_UnsetInContainer_NoFileSize now passes.

- New ffmpeg_unittests passes:
sync0_ogv/FFmpegTest.Seek_Video/0, where GetParam() = "sync0.ogv"
sync1_ogv/FFmpegTest.Seek_Video/0, where GetParam() = "sync1.ogv"
sync2_ogv/FFmpegTest.Seek_Video/0, where GetParam() = "sync2.ogv"

FFmpeg fixups here, https://chromiumcodereview.appspot.com/9325049/
New git repo here: http://git.chromium.org/gitweb/?p=chromium/third_party/ffmpeg.git;a=summary
Merge+Patches diff: https://chromiumcodereview.appspot.com/9317107

BUG=110776
TEST=unittests, layouttests, etc. Trybots.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=123123

[DaleCurtis, 2012-02-04 03:42:28]

PTAL. These changes are 99% complete, all that's left is to roll DEPS in this CL
once https://chromiumcodereview.appspot.com/9325049/ lands.

[ilja, 2012-02-04 03:57:38]

Double check that you got all the 53->54 transitions. They lurk in unexpected
places. In particular it will probably break a CrOS autotest (will need to
update binaries there as well). I'll keep an eye on it and will let you know
after this lands.

[DaleCurtis, 2012-02-04 04:02:32]

These are all that code search found, if you know of any others please let me
know:

https://cs.corp.google.com/#search/&q=avcodec-53%20package:%5Echrome&type=cs

On 2012/02/04 03:57:38, ilja wrote:
> Double check that you got all the 53->54 transitions. They lurk in unexpected
> places. In particular it will probably break a CrOS autotest (will need to
> update binaries there as well). I'll keep an eye on it and will let you know
> after this lands.

[Ami GONE FROM CHROMIUM, 2012-02-04 17:04:15]

I'll let scherkus review this.

[scherkus (not reviewing), 2012-02-06 21:13:49]

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_common.cc
File media/ffmpeg/ffmpeg_common.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_common.c...
media/ffmpeg/ffmpeg_common.cc:189: bytes_per_channel << 3,
we should also make a similar API change :)

mind filing a TODO in AudioDecoderConfig to switch to bytes?

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_unittest.cc
File media/ffmpeg/ffmpeg_unittest.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_unittest...
media/ffmpeg/ffmpeg_unittest.cc:50: static const int kDecodeThreads = 2;
we should actually consider refactoring this to use FFmpegVideoDecoder directly
***OR*** merging these tests into PipelineIntegrationTests

WDYT?

We can TODO/bug-ify it for now

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_unittest...
media/ffmpeg/ffmpeg_unittest.cc:140: ascii_path.c_str(),
indent

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/file_protocol.cc
File media/ffmpeg/file_protocol.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/file_protocol.c...
media/ffmpeg/file_protocol.cc:87: &Open2Context,
is it acceptable to leave this as NULL?

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.cc
File media/filters/ffmpeg_glue.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.cc...
media/filters/ffmpeg_glue.cc:127: &Open2Context,
ditto for NULL here

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.h
File media/filters/ffmpeg_glue.h (right):

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.h#...
media/filters/ffmpeg_glue.h:7: // avformat_open_input function, which analyzes
the filename given to it and
nit: mind annotating the function calls in comments w/ () at the end of them?

it's a loosely enforced style we use to help distinguish variable names from
functions

http://codereview.chromium.org/9317096/diff/2001/media/test/ffmpeg_tests/ffmp...
File media/test/ffmpeg_tests/ffmpeg_tests.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/test/ffmpeg_tests/ffmp...
media/test/ffmpeg_tests/ffmpeg_tests.cc:130: NULL, NULL);
indent

http://codereview.chromium.org/9317096/diff/2001/media/tools/media_bench/medi...
File media/tools/media_bench/media_bench.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/tools/media_bench/medi...
media/tools/media_bench/media_bench.cc:244: NULL, NULL);
indent

[DaleCurtis, 2012-02-07 19:09:13]

PTAL. Where is the revision # controlled for rolling the Chrome specific
binaries? In DEPS I only see ffmpeg_revision pointing to the public repo.

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_common.cc
File media/ffmpeg/ffmpeg_common.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_common.c...
media/ffmpeg/ffmpeg_common.cc:189: bytes_per_channel << 3,
On 2012/02/06 21:13:49, scherkus wrote:
> we should also make a similar API change :)
> 
> mind filing a TODO in AudioDecoderConfig to switch to bytes?

Done.

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_unittest.cc
File media/ffmpeg/ffmpeg_unittest.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_unittest...
media/ffmpeg/ffmpeg_unittest.cc:50: static const int kDecodeThreads = 2;
Merge w/ PipelineIntegrationTests would probably be better; the less fake
pipelines we have the better.

I'll drop a TODO here for now.

On 2012/02/06 21:13:49, scherkus wrote:
> we should actually consider refactoring this to use FFmpegVideoDecoder
directly
> ***OR*** merging these tests into PipelineIntegrationTests
> 
> WDYT?
> 
> We can TODO/bug-ify it for now

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/ffmpeg_unittest...
media/ffmpeg/ffmpeg_unittest.cc:140: ascii_path.c_str(),
On 2012/02/06 21:13:49, scherkus wrote:
> indent

Done. Argh, cpplint.py is totally useless.

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/file_protocol.cc
File media/ffmpeg/file_protocol.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/ffmpeg/file_protocol.c...
media/ffmpeg/file_protocol.cc:87: &Open2Context,
On 2012/02/06 21:13:49, scherkus wrote:
> is it acceptable to leave this as NULL?

Done. Seemed to work fine.

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.cc
File media/filters/ffmpeg_glue.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.cc...
media/filters/ffmpeg_glue.cc:127: &Open2Context,
On 2012/02/06 21:13:49, scherkus wrote:
> ditto for NULL here

Done.

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.h
File media/filters/ffmpeg_glue.h (right):

http://codereview.chromium.org/9317096/diff/2001/media/filters/ffmpeg_glue.h#...
media/filters/ffmpeg_glue.h:7: // avformat_open_input function, which analyzes
the filename given to it and
On 2012/02/06 21:13:49, scherkus wrote:
> nit: mind annotating the function calls in comments w/ () at the end of them?
> 
> it's a loosely enforced style we use to help distinguish variable names from
> functions

Done.

http://codereview.chromium.org/9317096/diff/2001/media/test/ffmpeg_tests/ffmp...
File media/test/ffmpeg_tests/ffmpeg_tests.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/test/ffmpeg_tests/ffmp...
media/test/ffmpeg_tests/ffmpeg_tests.cc:130: NULL, NULL);
On 2012/02/06 21:13:49, scherkus wrote:
> indent

Done.

http://codereview.chromium.org/9317096/diff/2001/media/tools/media_bench/medi...
File media/tools/media_bench/media_bench.cc (right):

http://codereview.chromium.org/9317096/diff/2001/media/tools/media_bench/medi...
media/tools/media_bench/media_bench.cc:244: NULL, NULL);
On 2012/02/06 21:13:49, scherkus wrote:
> indent

Done.

[scherkus (not reviewing), 2012-02-07 19:21:28]

LGTM w/ one nit

http://codereview.chromium.org/9317096/diff/13002/media/base/audio_decoder_co...
File media/base/audio_decoder_config.h (right):

http://codereview.chromium.org/9317096/diff/13002/media/base/audio_decoder_co...
media/base/audio_decoder_config.h:39: // bits_per_channel, we should match them.
nit: want to move this down by the function itself?

also annotate w/ ()

I'd also say that this TODO isn't about matching FFmpeg per se, just that bits
are more confusing to work with :)

[scherkus (not reviewing), 2012-02-07 19:21:29]

LGTM w/ one nit

http://codereview.chromium.org/9317096/diff/13002/media/base/audio_decoder_co...
File media/base/audio_decoder_config.h (right):

http://codereview.chromium.org/9317096/diff/13002/media/base/audio_decoder_co...
media/base/audio_decoder_config.h:39: // bits_per_channel, we should match them.
nit: want to move this down by the function itself?

also annotate w/ ()

I'd also say that this TODO isn't about matching FFmpeg per se, just that bits
are more confusing to work with :)

[DaleCurtis, 2012-02-07 22:53:55]

http://codereview.chromium.org/9317096/diff/13002/media/base/audio_decoder_co...
File media/base/audio_decoder_config.h (right):

http://codereview.chromium.org/9317096/diff/13002/media/base/audio_decoder_co...
media/base/audio_decoder_config.h:39: // bits_per_channel, we should match them.
On 2012/02/07 19:21:29, scherkus wrote:
> nit: want to move this down by the function itself?
> 
> also annotate w/ ()
> 
> I'd also say that this TODO isn't about matching FFmpeg per se, just that bits
> are more confusing to work with :)

I put it up here since there are several functions and variables which
references bits_per_channel. The () is just for functions right? These are
variable names.

I did clarify the comment a bit. Once the DEPS roll is ready, I'll ping you
again for one last look at those bits.