[XviD-devel] XviD not thread-safe anymore?

[Christoph Lampert]

On Mon, 28 Jun 2004, Edouard Gomez wrote:
> Christoph Lampert (chl at math.uni-bonn.de) wrote:
> > I had some problem in 2pass-mode when encoding two sessions of XviD
> > in parallel (using Linux / transcode 0.6.13), e.g. wrong bitrates and 
> > quants in one of the files, and just a few minutes ago, a real ugly
> > crash with "Illegal Machine Instruction". 
> > 
> > Is there a (known) threading problem in current CVS XviD? Or maybe in 
> > transcode? Did anyone else have problems lately?
> 
> Not that i know, there uses to be a known problem when MPEG
> streams do end (MPEG code) w/o really ending (the end of file stream isn't
> reached). Usually in this case, the encoding pipeline is
> reset, resulting in a truncated 2pass log, often completly
> corrupted as a close/create race could occur.

Nono, this is completely different. The AVI file was technically okay, but 
wrong quantizer were chosen. 

> But this is not really xvid's fault, as it's caused by the
> upper layer.
> 
> Can you post xvid4.cfg file, and the recipe to trigger the bug?

I'll try, though it's not trivial, because it doesn't always crash or 
anything. Just the filesize was way off...

chl