[XviD-devel] segmentation fault

[Henning Moll]

Hello!

Unfortunately, i have problems using the xvid codec. The whole thing is 
a little bit strange, so i will tell you all details i recognized up to 
now:

The codec will fail with an segmentation fault _mostly_ during the 
second pass of two-pass-encoding. I checked this with several programms 
on different OSes: (mencoder, transcode, gordian knot). All show the 
same problem.

Now the interesting thing: On the same data, with same parameters, the 
segmentation fault appears on different frames. It's kind of random. 
Now you think of an hardware problem, but i don't think it is: I don't 
have any similar behaviour with other software. Every other codec is 
working with all of the programms mentioned above, but not yours... :-(

To weaken that hardware failure theory, i also recompiled transcode and 
xvid without special hardware usage (assembly code, MMX, and so). But 
the same effect there (only hours later ..;-)

The chance of such a segmentation fault is not very likely, but it 
happens often enough to make encoding with xvid nearly impossible.

On different tries it crashes in different functions, but at the moment 
i remember only on symbol name:

transfer_8to16sub_mmx

All the other functions where it happends sound similar.


Hmm, now i need your help. What can i do to track down that problem 
more?

Any suggestions?

I don't know much of CPUs and that stuff, but I thought of something 
like a hardware random number generator. Is there something like that? 
Maybe there exists a model range of my CPU which have a slightly 
different implementation. Maybe that slightly difference is enought to 
let your code segfault? BTW: i have an AMD XP1800+

Are you agree with my theory, that it can not be a hardware _defekt_, 
because other (comparable) programms do not have any problems? If not, 
why?

Kind regards
Henning