On Tue, Jan 28, 2014 at 8:05 AM, Dr. H. Nikolaus Schaller hns@goldelico.com wrote:
Am 28.01.2014 um 04:32 schrieb Michael Mrozek:
Am Tue, 28 Jan 2014 04:07:09 +0200 hat Grazvydas Ignotas grazvydas.ignotas@openpandora.org geschrieben:
I've looked at this and it's stack corruption of some sort in Xorg itself, somewhere in input handling code to be more precise, not in fbdev or some plugin. A function picks up address 0 from stack as return address, and jumps to it. This doesn't even give meaningful backtrace, so is really difficult to debug. It might be something that only shows up on ARM, so is not fixed upstream. Seems to be triggered by input events, not drawing related.
Well, it could be that drawing (moving a window) damages the stack and then the *next* input event fails. Or some race between window manager triggered events and input events from the mouse.
Yeah it's probably something like that. Actually what is crashing is some event dispatcher, so it might actually be not input event but some redraw event that triggers the corruption.
GraÅžvydas