[MPlayer-users] Re: Re: garbled audio through FIFO

[Jeremy Cook]

Thanks for answering everybody. I've been streaming BBCR2->MP3 all night
using the CVS version from 30/11, without any additional patches, and
can confirm that the sound is still garbled n a regular basis. After
several hours of streaming through fifos one of the mplayer instances
that I was running crashed with a "this shouldn't happen" message.

Jeremy


On Wed, 2005-11-30 at 23:42, Bryan Alton wrote:
> Roberto Togni wrote:
> 
> > On Wed, 30 Nov 2005 21:44:06 +0000
> > Bryan Alton <balton at eircom.net> wrote:
> > 
> > [...]
> >> 
> >> You can it on post #76
> >> http://forums.slimdevices.com/showthread.php?t=16917&page=8
> >> 
> >> There was a problem with KQED and a later small mod is listed in post
> >> #108 http://forums.slimdevices.com/showthread.php?t=16917&page=11
> >> 
> > I'm rewriting the real demuxer to move the packet reassembly and
> > reordering from codec to demuxer; that should also solve (or at least
> > reduce) the problems caused by desync, since it restarts with a new
> > matrix every time it finds an audio keyframe. It won't solve the bad
> > sound caused by bad packets anyway. I'll post a patch soon, it's part of a
> > bigger cleanup on real demuxer/codec interface.
> > 
> > Ciao,
> >  Roberto
> 
> >From what you are saying, I shouldn't bother posting my patch.  I think the
> approach you are taking is the right one - packets have to be gathered and
> held and checked until a block with the right number of packets (as
> specified by the flavour) has been assembled before passing to codec.
> 
> After the rewrite make sure you test with live real audio stream from the
> BBC not just the "listen again".    By comparing behaviour (with test
> files), Realplayer replaces missing packets when only a few are lost - if
> you restart the matrix every time just one/two packet are dropped there
> will be lots of gaps in the audio. BTW keyframes are also lost in live
> streams. I did not come across any bad packets, just missing packets as
> determined by the timestamp.
> 
> Good luck, and I can help with testing if you want.
> 
> Bryan
> 
> _______________________________________________
> MPlayer-users mailing list
> MPlayer-users at mplayerhq.hu
> http://mplayerhq.hu/mailman/listinfo/mplayer-users
--