The AIGLX patches (enabled by setting the "aiglx" use flag for x11-base/xorg-server-1.1.1-r1) do give a warning that they don't work well with EXA, but they don't imply that X may crash if you try to use, say, beryl or compiz with AIGLX! The cause of the crash is rather simple: the patch 05-offscreen-pixmaps.patch makes it use XAAEvictPixmaps() in GL/glx/glxdri.c; this is fine and dandy if you use XAA where the patch adds in such a function, but with EXA? X fails to find the symbol and crashes. I'm not a brave soul to go hacking the innards of X, but a simple patch solves the problem for me... or at least the fact that I'm reporting this bug while using beryl, EXA, and AIGLX via this one patch seems to prove such to me. (As an aside, the performance of AIGLX+beryl is much, much better for me under EXA than XAA, and less flickery)
Created attachment 99855 [details, diff] Makes an XAAEvictPixmaps symbol while using EXA Okay, it's hacky, but it's the best I could do, I'm afraid.
Bah, after doing a bit MORE searching, I've found that this could be a little bit of a stopgap solution for an issue touched on in bug #147841 although not the ultimate solution (obviously). I hope that helps....
Instead of your patch, try adding a line to the Module section of your xorg.conf: Load "xaa"
(In reply to comment #3) > Instead of your patch, try adding a line to the Module section of your > xorg.conf: > Load "xaa" > Unfortunately, that just makes X crash in the same manner as it did originally. Except I don't see anything about a missing symbol in my gdm log. Could just be cause I'm looking in the wrong log, but it doesn't work.
Grr, I keep making more comments than necessary but bear with me please.... Here's what my log says when it crashes this time: -------------------------------- (II) XAA: Evicting pixmaps Backtrace: 0: /usr/bin/X(xf86SigHandler+0x85) [0x80cc168] Fatal server error: Caught signal 11. Server aborting -------------------------------- So it's a different kind of error, but it's activated by the same conditions and has the same end result. (and yes, this is with xorg-server recompiled without my patch. Also, just for reference, I'm using an R200 chip with an IGP 345M for which EXA performs far better than XAA, YMMV etc. etc.)
Is this still an issue with xorg-server-1.3? Please reopen when you reply.