[vdr] Re: softdevice 0.0.7 cvs crash tests...

[Stefan Lucke <stefan@lucke.in-berlin.de>]

On Donnerstag, 12. August 2004 23:25, Nicolas Huillard wrote:
> Stefan Lucke wrote:

[ .. ]

> > So now the bad news from me: I could play your test streams, but I
> > cannot play most of my normal recordings :-(.
> > While fixing this, your "24" stream now segfaults again, the other two don't 
> > and my recordings play fine again ;-) .
> 
> If I can be of any help...

Did you get results from the stream analyser ?
"24" recording: there are at least 2 video stream ids E0 + E6. crashes where with
 E6 stream passed to decoder.

segfaults with your two other streams are unknown (ffmpeg related).

original demux logic was :
change any video stream id to E0 and decode it. change any audio stream
id to C0 and decode it.

cvs now is:
accept any video stream id and pass only E0 to decoder, accept any audio
stream id and pass only C0 to decoder.

So I tried trailing 8 bytes zero fix. But I got _no_ combination so that
"24" doesn't segfault and my normal recordings play. My normal recordings
are with video stream id E4 ;-) .

Remaining question is: Is it possible/allowed that video/audio stream
id change during transmission ? If the answer would be no, than we can
lock on the first stream id we get and drop others. If the answer is yes,
than ...


Stefan Lucke