Created attachment 121587 [details]
The server crashes on debian unstable,
related bug report there: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=813716
I'm attaching the GDB backtrace and xorg log here as well.
Created attachment 121588 [details]
Created attachment 121610 [details]
gdb backtrace 2
Another crash, different gdb backtrace
Created attachment 121616 [details]
Yet another slightly different backtrace
Something is wrong with the current version in the Debian repos 2:2.99.917+git20160127-1+b1 - I switched back to 2:2.99.917-2~bpo8+1 version from the jessie-backports and could not reproduce a single crash since a whole week, compared to few times a day...
I hope it helps.
The problem is those crashes are from victims and do not show the cause. You would have to run the Xserver under valgrind (and be sure to have the driver compiled with --enable-debug and/or --enable-valgrind).
Alternatively, doing a (git) bisection on the driver may help.
Thanks Chris! I've compiled the driver from the git repo with --enable-debug and --enable-valgrind, also installed valgrind on the machine. Waiting for the next crash with gdb attached.
Should I do anything valgrind specific other than the regular gdb backtrace retrieval?
If it crashes again, then it is worth running under valgrind. You may have to do a couple of tricks, but if you can run
valgrind --log-file=/tmp/vg-xorg.txt Xorg -ac -noreset
then you can connect your normal Xsession (something like running gnome-session as your user should be enough). You should see both a message that xf86-video-intel is running under valgrind on both the stderr (Xorg.log) and the /tmp/vg-xorg.txt. If you don't see that message, check your --enable-valgrind compilation as otherwise the logs will be full of false positives.