[MPlayer-dev-eng] PVA demuxer fix for MultiDec files

[Matteo Giani]

Hi all. Great news (at least for me, I don't know how many of you guys 
care about this... :-) ). After about a month of fighting against this 
ugly bug I've finally overcome it.
It was indeed a MultiDec bug, in that it did not correctly set the flags 
in the PVA packet headers as well as the ones inside the PES headers.
To get it simple, in a PVA file the audio PES packets contained in the 
audio stream are aligned to the starts of PVA container packets. There 
is a flag in each PVA packet header that tells whether a new PES packet 
starts there or not, so that one can tell if he has to parse the PES 
header. MultiDec simply did not set it for about half of the PES 
packets, which is why the audio stream ran about twice as fast as the 
video stream.
Moreover, the same PES packets that are not signalled in the PVA headers 
do NOT contain valid PTS information, nonetheless the "PTS" PES header 
flag WAS SET.
I must admit I did have to introduce quite a hack to get those files 
running, and I'll have to tidy up the code a little bit before I submit 
a patch (hopefully tonight or tomorrow).
This message was just to let you know that we can soon remove the 
"BROKEN" from those PVA files... :-).

So, Arpi was right about the part that the audio stream ran twice as 
fast: about half of the data was discarded!!!
ah, I've already bought the 10 liters of cola I deserve to drink for 
those stoooooopid bugs that caused Dominik's problem. Sorry all. It was 
interesting though that his files ran file on my system even _before_ 
the patch, as I had told him. Anybody have an explanation for it ? Could 
it just be the different compiler?

Cheers.
Matteo