Absolutely unexpectedly, X crashes with this message in log: dix: invalid event type 0 I'm attaching a complete logfile. /usr/bin/X (0x8048000+0x4c05d) [0x809405d] corresponds to verify_internal_event in dix/inpututils.c. I have no idea what might be going wrong. I think I have to touch the keyboard/touchpad for it to crash, but I have no idea what kind of touch it is. Is such an event really a fatal problem deserving a server crash? What could go wrong if I comment out the FatalError call and just ignore the event? Is there anything I can do to trace the source of such events? I have xorg-server 1.11.0 from Debian package, I use the intel driver, synaptics 1.5.0 from Debian git for the touchpad, and evdev for keyboard and trackpoint.
Created attachment 52043 [details] xorg log
Is this a regression? If so, what introduced it? Can you try going back in evdev and synaptics to see if the older versions work on 1.11 or go back to 1.10 with the current driver versions?
I'm not sure whether this is or isn't a regression. I replaced my hardware lately, and after doing so, I may have upgraded a few things — I'm pretty sure I did update the intel driver, but I'm not sure about X and input drivers. I may randomly downgrade the packages you mentioned, but it will take months before I can say whether the issue disappeared or not. Bisecting is even more so out of the question, as I really have no idea when this happens. I hope there is a more systematic approach. :-/
fwiw, we saw a similar issue in Fedora a while ago. was fixed by updating mesa, there is a slim chance this may be similar to this bug. https://bugzilla.redhat.com/688693
Tomas, can you please let us know if updating mesa fixes this for you?
Shall I be using something more bleeding edge than 7.11? If not, than it doesn't. I did some experiments during the last few weeks, and I am inclined to blame synaptics. It seems that every time this crash happened, I was touching the right part of the touchpad trying to scroll. Also, I have numerous other problems with it. Versions 1.5.0 and later seem to get the pointer stuck in the top line of the screen several times a day, usually when the touchpad is touched lightly on one edge (I have never been able to reproduce this at will, but whenever it happened, I recalled touching it accidentally from one side). I can usually recover the pointer by randomly touching the touchpad for several dozens of seconds. Version 1.5.99 additionally brought in some acceleration issues — I had to adjust the accel factor to get same behaviour, and the acceleration seemed to depend on the number of fingers touching the touchpad, etc. Now I'm back to 1.4.1, acceleration is fine and pointer doesn't get stuck. I still get the crashes, though, and I plan to add some code that checks whether any events of type 0 get sent by synaptics or something like that to check whether synaptics is indeed the cause.
(In reply to comment #6) > Shall I be using something more bleeding edge than 7.11? If not, than it > doesn't. judging by the fedora bug, some 7.11 had this bug and dave pushed a fix for it. so your version _may_ be affected > still get the crashes, though, and I plan to add some code that checks whether > any events of type 0 get sent by synaptics or something like that to check > whether synaptics is indeed the cause. fwiw, the driver API doesn't allow any event codes. the driver can't submit events of type 0, type handling is all handled in the DDX/DIX.
Just to give you a heads up -- I think it might actually be caused by memory corruption in the kernel: https://lkml.org/lkml/2011/10/29/73 I will give an update once that thing is sorted out. Thanks for your support so far.
Thanks for following up. I'm closing this as "not our bug" ... please reopen if you have reason to believe it's not the referenced kernel bug.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.