[GLLUG] run away gcc

[Jeremy Bowers]

Caleb Cushing wrote:
> yeah... I know I've been killing the right process.... because I never
> get the "no process killed" error... and I tried killing it's parent
> make... then it just became a child of init. I hope init will take care
> of it. fortunately portage runs with gcc niced so it isn't slowing my
> computer down with cpu at 100% I'm going to let it run for a while to
> see it it goes away... or I can come up with a way to grab an image for
> the bug report I filed...

Sometimes a process can get stuck in the kernel. Since it never comes
out, it never gets around to checking the signals. This is when even -9
is ineffective. I used to see it a lot more than I do nowadays.

I wouldn't expect there's any point in waiting for it to come out. gcc
is a particularly well-constrained program, even if it is complex, and
it's not as if it's likely that you found a hidden O(n^4) algorithm in
gcc while compiling what may be the world's most-often compiled code
with the compiler most often used to compile it. :)

One thing to check is that you didn't over-optimize the kernel, since
you mentioned you're running Gentoo. I've also rarely encountered
problems with certain combinations of options, although for something as
intermittent as this probably is you'll probably never work it all out.

In my experience, upgrading to the next point release of the kernel
tends to take care of these things, either because the relevant bug is
fixed (and you'd never recognize it in the Changelog unless you knew
exactly what to look for) or because it simply shakes up the kernel
enough to never again hit that particular corner case. :)