Bug#324717: debugging X crash that requires hardware reset
submitter 324717 zackw@panix.com
thanks
On Tue, 2005-08-23 at 11:13 -0700, Zack Weinberg wrote:
> Michel Dänzer said:
> > On Mon, 2005-08-22 at 20:47 -0700, Zack Weinberg wrote:
> >> [...] since it was happening on a twice-daily basis
> >> using the 6.8.2.dfsg.1-5 packages, and only happens once every few days
> >> since I backed down to -4, that is where my suspicions lie.
> >
> > I don't see any obviously related changes in the changelog between -4
> > and -5 though...
>
> There was the compiler change in there, though. The mis-optimization may
> have been beneficial in some corner case in the Radeon driver. (handwave,
> handwave)
Yeah, that's the most likely change to have such a quantitative impact I
guess.
> >> Can anyone suggest how to debug this sort of crash? I'm willing to try
> >> just about anything.
> >
> > First of all, does disabling the DRI work around the problem?
>
> I'll try that, but as I have no reliable way to reproduce the problem, I
> will only be able to provide a guess, and that only after several days.
> GL wasn't ever in use at the time of the crashes, though.
That doesn't matter too much; with DRI enabled, the X driver uses a
completely different method (DMA ring buffer as opposed to direct
register address) to send commands to the card.
> (I'll go back to -5 at the same time, to increase the odds of provoking
> the bug if it isn't due to DRI.)
Yeah, that's what I meant.
> > PS: This should probably be a bug report instead of only a list thread.
>
> It did not seem useful to submit a bug report with this level of vagueness
> about the cause. If you would prefer a bug report I can certainly file
> one.
Looks like David Martínez Moreno beat you to it. Trying to set you as
the submitter of bug #324717.
--
Earthling Michel Dänzer | Debian (powerpc), X and DRI developer
Libre software enthusiast | http://svcs.affero.net/rm.php?r=daenzer
Reply to: