On Fri, Jun 03, 2005 at 09:52:09AM +0200, Christoph Franke wrote: > Dave Jones - Thu, Jun 02 2005 15:01:20 -0400: > > > On Thu, Jun 02, 2005 at 08:55:40PM +0200, Christoph Franke wrote: > > > Dave Jones - Thu, Jun 02 2005 14:25:48 -0400: > > > > > > > Is it repeatable ? Does it behave again if you boot with > > > > exec-shield=0 ? or exec-shield-randomize=0 ? > > > > > > Yes, it is repeatable, teamspeak segfaults on every start. Ntpd > > > segfaults on both build -30 and -31 during boot, but can be restarted > > > afterwards. Will try a boot with exec-shield=0 tomorrow morning and pass > > > through the results. > > > > Ok. > > Oh, staring on teamspeak I didn't instantly see the old fellow came up > again with the -31 build (booted with "exec-shield=0"). > > This occures during compilation of programs as well as on a cronjob > renicing some processes and is always parallel to the memory.c log > entries. The older builds all showed up with this, -30 didn't but -31 > does again. I think that was just by chance than by design. This has dragged on so long, and with no resolution in sight, that I'm actually getting more and more tempted to backport the current FC4 kernel (based on 2.6.12rc5) to FC3. plus sides - I've not seen any reports of the bad pmd bug on 2.6.12rc kernels. I'm pretty confident that this bug is dead there. - Upstream aren't going to devote much more time to tracking down 2.6.11.x bugs with .12 'due soon', so if it is still present in 12rc, it'll likely get more attention. - its actually pretty stable in my experience so far - An increased userbase for the FC4 kernel is going to shake out bugs faster. downsides - lots of code change - the usual potential breakage of existing userspace - .12 isn't 'final' yet. - There are a few things that aren't 'quite right' in the FC4 kernel that will be shipping that I intend to fix in an update, so its by no means a perfect kernel, just trading some bugs for some different ones. (Nothing new there eh?) The first two above we'd have to deal with when .12 is final anyway) I'll think it over some more. The actual backporting of FC4 kernel to FC3 is probably just an afternoons work. Dave
