Re: another mysterious segfault

[Lars Damerow]

>From Mark Burgess <mark@iu.hio.no>, Sun, Aug 28, 2005 at 08:17:02AM +0200:
> Interesting. I get a lot of seg fault complaints from Fedora/redhat
> users. I wonder about the C library package it is using. Is there any
> upgrade/patch you could try for the C library? Upgrading has helped
> several others out of mysterious seg faults.

It's happening during kickstart, which I realize we've discussed repeatedly on
the list. The glibc that FC4 kickstart uses is version 2.3.5, and unfortunately
it's relatively difficult to upgrade since it's stashed away in the kickstart
image.

I really wish there were a better way to debug cfengine during kickstart. I
don't see anything magical about the kickstart environment that makes it
unreliable; the only software I've seen that has trouble running during
kickstart is cfengine. That's not a knock against cfengine--we totally depend
on it--but it makes me feel that kickstart isn't necessarily to blame here.

In the past, whenever these mysterious segfaults have come up, I've resorted to
reorganizing my cfengine scripts, changing the order of the sections, and
basically varying my approaches to editing the files until the segfaults stop.
I've tried redirecting the output using maximum debug levels, examing strace
output, removing any Berkeley DB libraries besides the one that cfengine was
built against, forcing gdb sessions during kickstart, and so on, but have never 
been able to find the root cause of these sorts of problems.

Any suggestions are appreciated!

sorry for the ranting,
lars

--
lars damerow
button pusher
pixar animation studios
lars@pixar.com

Strike while the irony's hot!