Comment 95 for bug 107188

@BigPick: Thanks for your analysis of the patch, when I read it I was wondering what change in it might have caused that the problem got fixed. My current theory is that the bug occur more or less random (or with a pattern that is very hard to reproduce) so that even if the patch itself has little effect, a upgrade test may still be successful because the bug is not triggered within this run. I'm currently trying to reproduce the problem with the virtualbox package in ubuntu. My current idea of reproducing it is to install kubuntu-feisty and try various combinations of stock kubuntu feisty install, install with updates, with backport enabled etc. Each time a snapshot is taken before the upgrade.

I hope using this method we will be able to find a way to reliable trigger the bug. Everybody is welcome to help with the hunt :) (my main machine is currently not usable, so I'm a bit slower with those tests as I would otherwise). As for the debugging output, I think that strace is usually a good first step. If we could get a gziped strace log when the bug happend, that would be really cool. I suspect the actual problem lies somewhere under the python layer of the code (in the C/C++ bits), so pdb is probably not that helpful. gdb might, once we know a bit better what to look for.

Cheers,
 Michael