>>>>> "mouse" == der Mouse <mouse@Rodents.Montreal.QC.CA> writes:

    mouse> ... the symptoms are very similar - almost always a
    mouse> coredump from either make or cc1; occasionally I get an
    mouse> "internal compiler error". Sometimes waiting and/or
    mouse> flushing caches (via something like "find / -type f -print
    mouse> > /dev/null" and/or "find / -type f -print | xargs cat >
    mouse> /dev/null") helps. Sometimes nothing helps. Because it
    mouse> happens across two different machines, different
    mouse> architectures even, I'm reluctant to suspect hardware.

I'm not reluctant to suspect hardware. Signal 11 during heavy
compilation (i.e., the kernel) is the single most common symptom of
flaky hardware. If you have two marginal boxes, it's a good bet that
they'll both be giving sig11s. (I tend to accumulate junk computers so
I have quite a bit of experience with this problem.)

The key test is whether the compilation gets further along if you
rerun make. If it does, you can be confident it's a hardware problem.
It's certainly not a problem with cc.

I admit it is possible that we broke something in the virtual memory
system, but I think it's a lot more likely that since we're using egcs
now (which is much more CPU intensive), more flaky hardware is coming
to light. 

I'm surprised that this question isn't in the NetBSD FAQ yet. (FAQ
Maintainer: Feel free to add this answer).

By the way here's the Sig11 FAQ:

       http://www.bitwizard.nl/sig11/

(When reading, tune your mind to 's/linux/netbsd/').

--Ben Wong
ben@spammenot.cowlove.com 
Registered Washington State Resident