[XviD-devel] segmentation fault

[Henning Moll]

Hi!

some feedback:

On Sunday 03 October 2004 17:35, Christoph Lampert wrote:
> Definitely not. Nondeterministic errors are almost always due to
> hardware. 
> [...]  
> Other programs produce other access patterns RAM and 
> instructions, you cannot be sure the problem would pop up somewhere
> else.

Hmm, it's realy hard to believe, that an error just occours on a single 
programm ( in this case a single codec), but at last it seems that you 
are right...


> First thing I would suggest is to run a RAM checker overnight. Second
> thing is to run a CPU-intensive thing (e.g. "cpuburn-in" for a
> while), or possible both together.

Ok, i did that in all possible combinations. Neither memtest nor 
cpuburn-in reported any problem. But interesting: cpuburn-in is not 
worth its name: xvid heated my CPU much more than cpuburn-in. In fact, 
i can use cpuburn-in to cool it down (!) a bit while running the xvid 
encoder. Maybe cpuburn-in is _that_ stress tool, but only for intel 
cpus...? I recommend xvid instead. ;-)
Anyway, that little decrease of heat was not enough to avoid a segfault 
during the encoding process. So i opened the computer case and did some 
heavy external cooling by an ordinary ventilator. That way i was able 
to achieve several encodings without a segfault. Ok, i think that's 
proof enough to confirm a hardware failure.

Now i will start to find the causing component by replace one after each 
other, starting with the cpu of course.

Anyway, many thanks to you for your advice.

Best regards
Henning