cpu overheating

[James Wilkinson]

I really shouldn't be joining in this thread again. It's obvious Mike
can't quite get the hang of a simple concept...

Mike McCarty wrote:
> One could argue that only one machine is having a problem, so
> it is extremely unlikely that Linux is the problem, but rather
> the machine. That argument is correct.

But guess what? You're the *only* person in this thread to make that
argument. It's a "straw man" argument. Stop arguing against it.

> One could also argue that
> this unique case is important because is may be the result of
> a defect in Linux which occurs only very infrequently, so it is
> important to find out its cause if so.

If, as appears *highly* probable, the system really was overheating,
*whatever* Linux does it logically cannot be responsible. 

Hardware should not overheat whatever software does, so whatever Linux
is doing, the hardware should still not overheat. If it was happening in
Windows, Windows would not be responsible. It *cannot* be a software
bug!

There is, however, a related argument you could make. That it is
important to find out exactly what is happening so that whoever is
appropriate can stop it happening again. But for this we need to know
such things as:
 * where did the error messages come from;
 * how hot does that processor actually get;
 * are any temperature probes correctly configured.

But these are different questions, and we'd want to gather different
data. The question "what was going on" is relatively unimportant -- we
know the Original Poster was running yum, and we *know* that stresses
the processor.

> If thinking we should take every opportunity to investigate
> unusual behavior for possible defects is being an "idiot", then
> every industry in the world which considers availability and
> reliability in software to be important is full of idiots.
> This includes telecomm, aviation, and power systems at
> least.

These industries have the sense to know that reliable software is
dependent on reliable hardware. If the Original Poster's hardware is not
reliable, crashing software is expected. Indeed, fly-by-wire aeroplanes
are designed around the possibility that the computers will fail, and
usually have multiple redundant "voting" systems to identify and
neutralise rogue systems.

James.

-- 
E-mail:     james@ | And WinCE... there should be BIG RED WARNING STICKERS,
aprilcottage.co.uk | like those IntelInside things. "This product contains
                   | WinCE and is therefore not suitable for the purposes
                   | stated."                               -- Graham Reed