Re: Random segfaults (was: Hup,hup Debian AMD64, Hup, hup)

[Lennart Sorensen <lsorense@csclub.uwaterloo.ca>]

On Thu, Jun 22, 2006 at 04:01:36PM +0000, Dimitris Lampridis wrote:
> For the CPU (x2 3800+) i cannot say, although some other OS that unfortunately 
> can offer me Elder Scrolls Oblivion (...) is not having any problems (not of 
> this kind anyway)
> 
> For the memory (4x512MB), i did  have one faulty module, i returned it, got a 
> new one and system was sweet, plus memtest86 doesn't reveal a problem (this 
> doesn't mean that its OK, but its a rather slim case)
> 
> For the power supply its a very robust Tagan, 480W and it was taking the heat 
> fine for many months now. Also, Dutch power company has never caused me any 
> trouble or reason to worry.
> 
> My system is as cool as it can get with stock CPU fan, CPU under 30 Celsius, 
> case under 40 Celsius, so no overheating problem too.
> 
> I'll start checking all those things the best I can (memory is the easiest to 
> check me thinks, at least at a per-module basis), but I'm almost sure its not 
> my HW.

Hmm, well most of the programs you mentioned were c++ programs.  I guess
there could be a c++ library problem, but if that was the cause it
should consistently fail every time.  Intermitent errors, which other
people have not reported, tend to point towards the specific system's
hardware unfortunately.

Random failures tend to make me think the ram has some flacky area, and
depending on where in memory the program is loaded, it fails or doesn't
fail.  memtest86 isn't perfect, and can't catch all errors.

As for windows, I have seen it run "just fine" (as in didn't crash more
than usual) on a system that couldn't even get past the bogomips
calculation on linux.  Replacing the cpu made that system run perfectly.

Len Sorensen