Re. [Gllug] How to prove a computer is overheating.

[Nix]

On Fri, 14 Jan 2005, Al Tomkins prattled cheerily:
> Try recompiling the kernel.
> 
> I had a problem where the box worked fine all the time but recompiling a
> kernel (obviously CPU intensive and therefore heat generating) always
> failed and with error "Signal 11".
> 
> It took me ages to work out that the CPU was overheating and causing the
> problem.

Repeated compilation errors in different places or with different errors
are almost[1] always a sign of hardware failure. GCC does a lot of pointer
chasing, so any one-bit errors anywhere in the huge blocks of memory it
allocates are likely to cause a compiler crash.

[1] I say `almost' because it *is* slightly nondeterministic internally,
    unless you use the -frandom-seed and -fno-guess-branch-probability
    options. However, the nondeterminism doesn't often lead to compiler
    bugs appearing and vanishing (indeed, I can only recall one such case).

-- 
`Blish is clearly in love with language. Unfortunately,
 language dislikes him intensely.' --- Russ Allbery
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug