Possible regression in mpeg4 decoding.
|Reported by:||dalecurtis||Owned by:|
|Blocking:||Reproduced by developer:||yes|
|Analyzed by developer:||no|
Currently, decoding/playing back this video in ffplay:
(If link dies, it's the MP4 version under the "Download" link here: http://www.silverlight.net/learn/overview/what's-new-in-silverlight-5/silverlight-5-release-overview )
Results in the following spew of errors before playback starts:
[mpeg4 @ 0x2249600] hmm, seems the headers are not complete, trying to guess time_increment_bits
[mpeg4 @ 0x2249600] my guess is 9 bits ;)
[mpeg4 @ 0x2249600] Error, header damaged or not MPEG4 header (qscale=0)
[mpeg4 @ 0x2249600] header damaged
However this used to work without error... git bisect reveals commit http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=59e95fa4a8844d2abe7ddd7b8d269ea8d8eea17d as the culprit; specifically the removal of the forced width reinit: avctx->coded_width= 0.
It's worth noting that ffplay only plays back this file because it ignores the return code of avcodec_decode_video2 (which is <0 in this case) and instead relies on got_picture=true.
I've dug around a bit, but am unable to confirm if this is actually a regression or if the file is truly just corrupt. Can someone more knowledgeable take a look?
Original ChromeOS bug report: http://code.google.com/p/chromium-os/issues/detail?id=28560
Thanks in advance,
Change History (6)
comment:3 Changed 4 years ago by cehoyos
- Keywords regression asp added
- Reproduced by developer set
- Status changed from new to open
- Version changed from unspecified to git-master