[Techtalk] Segmentation faults

[Maria Blackmore]

On 11 Sep 2002, mc wrote:

> I am having a very bad morning :(  I logged into my RH 7.3 machine and
> tried to open Opera, it segfaulted on me.  It had done this a couple of
> times last night, but I rebooted when I logged off last night, hoping
> that would clear whatever was causing this.  I then tried to reinstall
> opera, tried to update it, tried to remove it, it tells me the darn
> thing is not installed!

hmm, interesting, might be a red herring or it might be connected

> So, I have 2 problems....
> 
> 1) What is causing the segfaults.  This machine has not ever yet been
> the most stable of any linux machine I have installed.  I have seen
> other segmentation faults, mostly when unmounting during shutdowns and
> reboots.  

This really does sound rather like a hardware problem to me, probably a
memory problem, could well be a CPU problem.

You can use memtest86 to test the memory, this is something that starts at
boot time in place of an operating system and will test the whole memory
for errors, it may take some time.  You can either start it from lilo (or 
other boot loader), or you should be able to make it into a boot disk and
test it from there.

In any case, random things segfaulting without apparent reason or
connection sounds very very much like a hardware problem.  Apart from the
memory one of the first things I would check is to see if the CPU
heatsink/fan is still intact and functioning.  Hot CPUs are not happy CPUs
:)

Unless the memory test turns up something concrete this probably means
swapping things until the problem goes away.  A good test of this is to   
try to compile a kernel (using make -j 3 for example).  A kernel compile  
is one of the most common things that will stretch pretty much every 
aspect of your machine, if there's a problem somewhere it is highly likely 
that it will be tickled by this.


> 2) What is the best way to wipe any and all pieces of Opera off the
> machine's rpm DB so I can get it reinstalled?

hmm .. this could be a corrupted RPM database, as for what caused this I'm
not sure.  I would still suspect that hardware failure is the most likely.

As for how to fix it, I'm afraid I don't know, hopefully someone else will
though.

Good luck

Maria