https://bugzilla.novell.com/show_bug.cgi?id=831629 https://bugzilla.novell.com/show_bug.cgi?id=831629#c15 --- Comment #15 from Jan Matejek <jmatejek@suse.com> 2013-10-03 18:01:31 UTC --- python's memory gets corrupted somehow in this test. So far I was unable to track down what exactly happens, but valgrind reports that python starts to step over unitialized memory in its finalizer, and that crashes it in the end. (btw, Don't Try This At Home, you need to have a special build of python to get meaningful results from valgrind [1]) The problematic sequence is this: 1. install a signal handler for SIGUSR1 2. install a different signal handler for SIGUSR1, storing the original one for later use 3. when you get SIGUSR1, restore the original signal handler with `sigaction` and call it with `raise` (that is pretty much the only thing that the override handler needs to do to trigger the problem - i commented out everything but this restore and it still segfaulted.) The original signal handler is executed correctly, AFAICT, and so far I did not find any difference between its behavior in the crashing and non-crashing case. (if you have any good tips on doing "core diffs", i could use some) But something obviously ends up different ;) if you guys have any idea what could be different between this kernel and the previous one regarding these operations, i'd love to hear it. [1] http://svn.python.org/projects/python/trunk/Misc/README.valgrind -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.