[FFmpeg-cvslog] r12759 - in trunk/libavcodec: aac_ac3_parser.c aac_ac3_parser.h aac_parser.c ac3_parser.c

[Michael Niedermayer]

On Wed, Apr 09, 2008 at 04:42:53PM +0200, Michael Niedermayer wrote:
> On Wed, Apr 09, 2008 at 10:14:04AM +0200, Bartlomiej Wolowiec wrote:
> > On poniedzia?ek, 7 kwietnia 2008, Mike Melanson wrote:
> > > bwolowiec wrote:
> > > > Author: bwolowiec
> > > > Date: Mon Apr  7 22:54:08 2008
> > > > New Revision: 12759
> > > >
> > > > Log:
> > > > Corrections of errors in aac_ac3_parser
> > >
> > > This didn't fix the regression suite. Hint: type 'make test' to check
> > > for yourself.
> > 
> > hmm.. what else should I do, not to have the tests failed?
> 
> I tell you as soon as i figure it out myself, didnt yet have time to look
> at what exactly is going wrong. Basically wherever the bug is it should be
> fixed, that could be in the parser, demuxer, muxer or the regression
> checksums.

1. the demuxer seems buggy its output with seeking and normal linear playback
differs. A fix for this MUST be in the demuxer.

2. your frame_in_buffer code is buggy, it causes random number of parsed
bytes to be returned.
Hint you cannot change buf_size.

3. why is need_next_header set for AC-3 ?

4. The return value of ff_combine_frame is ignored. This causes at least a
lost packet at EOF.

I suspect that the best course of action is to revert the changes up to the
point where the regression tests pass again.
Next the rm demuxer must be fixed which will likely change the regression
checksums, which need to be updated after confirming that the rm demuxer
works equally well or better.
And last the parser should be commited again. (but of course only if
regressions pass, if they dont the bugs causing the failure must be found
and fixed similar to the rm demxuer _before_ commiting the parser)

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Opposition brings concord. Out of discord comes the fairest harmony.
-- Heraclitus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-cvslog/attachments/20080409/13cfc07b/attachment.pgp>