[FFmpeg-cvslog] r14612 - trunk/tests/seek.regression.ref

[Aurelien Jacobs]

Diego Biurrun wrote:

> On Tue, Aug 05, 2008 at 07:52:33PM +0200, Aurelien Jacobs wrote:
> > Diego Biurrun wrote:
> > 
> > > On Tue, Aug 05, 2008 at 02:46:27AM +0200, Aurelien Jacobs wrote:
> > > > 
> > > > Sorry for the mail bombing...
> > > > I wanted to do a major cleanup of matroskadec since a long time and
> > > > I finally got to it.
> > > > I tried hard to keep each change small and self contained, hence the
> > > > huge number of commit.
> > > > 
> > > > Anyway, the cleanup was worth it. Just some numbers:
> > > > 
> > > >  ** before **
> > > > $ wc matroskadec.c
> > > >   3212   8783 107324 matroskadec.c
> > > > $ size matroskadec.o
> > > >    text	   data	    bss	    dec	    hex	filename
> > > >   39334	     64	      0	  39398	   99e6	matroskadec.o
> > > > 
> > > >  ** after **
> > > > $ wc matroskadec.c
> > > >   1621   5140  57271 matroskadec.c
> > > > $ size matroskadec.o
> > > >    text	   data	    bss	    dec	    hex	filename
> > > >   18748	     64	      0	  18812	   497c	matroskadec.o
> > > > 
> > > > => both object size and lines of code divided by 2...
> > > 
> > > Not bad at all.  How does the libavformat demuxer compare to the MPlayer
> > > demuxer you wrote a long time ago?
> > 
> > Lavf's one is now definitely smaller :-)
> > It is also more functional. So I plan to make it the default one
> > in MPlayer.
> 
> I was thinking about this the other day when trying out the FFmpeg AAC
> decoder in MPlayer.  There are a few samples where the libmpdemux
> Matroska demuxer plays AAC audio at the wrong samplerate while the
> libavformat one works correctly.  For some reason, libfaad2 is not
> affected.  If you are interested I can provide more detail, but maybe
> it's not even worth investigating...

I also spotted various samples which have problems with MPlayer
demuxer, but I don't think it's worth spending time on it.

> > But before doing this, I have a few more improvements planed.
> 
> Neat.  What do you have up your sleeve?

 - Try adding support for float PCM.
 - List all EBML IDs used in the wild to avoid printing warning
   about IDs which are know but not used in the demuxer.
 - Add (at least basic) support for Tags (metadata).
 - ...

Aurel