[FFmpeg-devel] [PATCH]Do not detect mp2 in mxf as mpeg2video
tomas.hardin at codemill.se
Sat Dec 10 17:06:41 CET 2011
On Sat, 2011-12-10 at 15:34 +0100, Reimar Döffinger wrote:
> On Sat, Dec 10, 2011 at 02:16:52PM +0000, Carl Eugen Hoyos wrote:
> > Reimar Döffinger <Reimar.Doeffinger <at> gmx.de> writes:
> > > > Anyway, splitting mxf_essence_container_uls seems like a decent
> > > > solution, if both streams have the same essence_container_ul.
> > >
> > > They do not, the demuxer is only using 14 bytes of the ULs (as
> > > specified) and completely according to the specification the
> > > first 14 bytes of the ULs MPEG-2 video and audio match.
> > So should I also apply the other patch?
> I think Tomas probably knows better.
> In particular I am also unsure that 15 really is correct and it
> shouldn't rather be 16.
> In general I don't get the system behind the match length.
> If it is just meant to be the minimal length to make the match
> unique it would be better to just change the code to look for the
> maximal match (possibly with some threshold and a warning if multiple
> equally good matches exist).
It's a hierarchical system, see RP224 .
In general, since FFmpeg is able to probe lots of codec parameters on
its own, we should try to match as short part of the UL as
possible/reasonable. My guesss is other systems can't figure out what
profile etc. are used on their own, hence the need for more information
than just "this stream is MPEG-2 video, go parse the essence headers".
More information about the ffmpeg-devel