[FFmpeg-devel] [PATCH] Clip-Wrapped MXF support (attempt #4)

[Reimar Döffinger]

On Sun, Jan 16, 2011 at 02:12:45PM +0200, Maksym Veremeyenko wrote:
> Reimar D?ffinger ???????(??):
> >If your code is correct, you should be able to remove the
> >st->need_parsing = AVSTREAM_PARSE_FULL;
> >and it should still play. But I very much doubt it.
> it will not play... For clip-wrapped files there are one video
> (or/and audio) chunk with data per file, so removing flag
> AVSTREAM_PARSE_FULL will make demuxer to read a full chunk instead
> of block from it...

That is only because your patch is buggy and misuses need_parsing to
check whether it should do the clip-wrapped stuff.
If you patch does not work without setting AVSTREAM_PARSE_FULL for
DV files, that means the patch is wrong.
In general, doing "parsing" twice is wrong. And your patch adds a
kind of "dumb" parsing.
So there never should be both AVSTREAM_PARSE_FULL set and your "dumb",
edit-unit based parsing both enabled.

> >>>Does this code actually fix any case other that DV video? Because for DV it could be
> >>>considered the real bug that the raw DV demuxer was not implemented using a parser
> >>>(if it was, clip-wrapped DV should already work I think).
> >>Hope it does not /fix/ anything other :-))) at least i did not noticed...
> >
> >That makes it likely that your patch _only_ works for DV, and the question is
> it works for AVC-intra too... i did not check against jpeg2000, but
> i suspect it will works too...

Uh, I am almost sure that it will only work for AVC-intra if it already worked
for it before.
We have a parser for AVC, so there should be no reason why your patch should
change how this case is handled.