From: John Phillips (john@linux.demon.co.uk)
Date: 08/09/93


From: john@linux.demon.co.uk (John Phillips)
Subject: Re: cc1 got signal 11 ??? 
Date: Mon, 9 Aug 1993 20:17:32 +0000

In article <240o8fINN34s@gap.caltech.edu> ashley@cco.caltech.edu writes:

>While I was trying to re-compile my kernel I got this
>message and the make aborted. What is it? How can I find
>a list of such error messages? Whod decided to make
>such cryptic messages?
>

I used to have this problem but managed to cure it. There follows
my explanation. Yours may be different...

When first running Linux on a new machine I occasionally (once every
couple of days) got a segmentation fault causing a process - usually
one of the daemons - to exit. This would also happen under MS-WINDOWS
to some specific programs. Turbo C++ was very sensitive and some other
programs failed every few hours. I never noticed the problem under DOS.

Running a Linux kernel compile, gcc (cc1 specifically) proved very prone
to fatal signal 11. Three or four errors during a compile was typical.

So I looked inside the box and noticed (who could fail to) that the
CPU (a 486DX50) was so hot it could fry eggs. The cheapskates had not
fitted a heatsink! Taking out a hairdryer (shh!) and setting it up to
blow cool air onto the CPU stopped the problem completely.

Months and one fan-assisted heatsink later, I've not seen one more
fatal signal 11.

Something was (is?) clearly marginal in the CPU, or the CPU's interface
to the outside world. The segment addressing errors were presumably
due to an occasional hardware error caused by a timing fault or a voltage
level failing to meet specification.

The moral? DOS does not exercise the PC hardware to any great extent,
and system designers seem to be able to get away with murder. Unix uses
resources much more extensively and finds out just what has been swept
under the carpet. My problem may have been due to overheating or to
marginal design or possibly both. I was fortunate that keeping the CPU
cool prevented the error. If the marginality were worse, I might still
be having the same problem, even with the coolest CPU in the world.

Regards

John

-- 
John Phillips
john@linux.demon.co.uk