Paul de Weerd wrote: > On Sun, Sep 21, 2008 at 03:40:09AM +0200, Robert Urban wrote: > | Hi folks, > | > | OS Env: OBSD4.3 running on a dual-PIII (ProLiant 380). > | > | I've got a perl/CGI script doing a bunch of stuff (talking to PostgreSQL, > | writing files, etc) which is dying with SIGSEGV, but only occasionally. I'm > | unable to reproduce the death outside of the chrooted env manually. I did, > | however, manage to get a ktrace of a good and a bad run. (see below). I > | set up the chroot environment by copying all relevent files for perl to > | /var/www. The SIGSEGV's started after I made some minor changes to the > | script, and if I make yet other changes, such as adding debugging code, the > | nasty behaviour stops. I doubt the script (and modules it uses) are > | relevent, but someone thinks they might shed some light on the problem, I > | can make them available. Naturally, I can make the complete ktrace/kdump > | files available as well, if anybody wants to see them. Access to the > | machine is also no problem. > | > | Is there any hope of tracking down this kind of error? > > A shot in the dark here .. but can you verify you have enough memory > available for use ? Check ulimit .. maybe the chrooted env is running > as a user with lower ulimits. Try upping those. > > Other than that, SIGSEGV's are often caused by bad hardware (bad > memory, notably), but I'd be surprised if you only hit that in the > chroot case. You may still want to check your machine though.
thanks for your suggestion. I checked resource limits, and they're ok. The HW is ok too. Before I ever install an OS on intel hardware I always run memtest86+ at least over night. That includes this box, so mem is ok. Could have been, but wasn't :) cheers, Robert Urban