Created attachment 23243 [details]
Bug detailed description:
start X,then run mesa xdemo glthreads ,sometimes it will run abort (about 1/10)with the message :
glthreads: xcb_io.c:378: _XAllocID: Assertion `ret != inval_id' failed.
following is the backtrace from gdb:
#0 0x00000034ca830ec5 in raise () from /lib64/libc.so.6
#1 0x00000034ca832970 in abort () from /lib64/libc.so.6
#2 0x00000034ca82a11f in __assert_fail () from /lib64/libc.so.6
#3 0x00007f9551f552fc in _XAllocID (dpy=0x10ab070) at xcb_io.c:378
#4 0x00007f9551aef732 in XFixesCreateRegion (dpy=0x10ab070,
rectangles=0x465ba0e0, nrectangles=1) at Region.c:41
#5 0x00007f95522b636e in dri2CopySubBuffer (pdraw=0x5ef0d90,
x=<value optimized out>, y=<value optimized out>, width=-1, height=0)
#6 0x0000000000402a08 in thread_function (p=<value optimized out>)
#7 0x00000034cb406407 in start_thread () from /lib64/libpthread.so.0
#8 0x00000034ca8d4b0d in clone () from /lib64/libc.so.6
btw this issue only happens on g45 and q45,it works on g33,915gm ,945gm and gm965.
Created attachment 23244 [details]
xorg conf file
This seems strange.
Eric, can you share any explanation for this problem?
I'm also running into this assertion at xcb_io.c:378 with my own application that does OpenGL rendering from a separate thread. I'm running with NVidia's libGL and driver so I don't think this has anything to do with Mesa or the Intel driver, but rather libX11.
I've had it happen with libx11 version 188.8.131.52. I haven't yet reproduced it with the just-released version 1.2, but as it is very rare, that might not mean anything.
the issue has gone against:
it still exists with uxa+ums and come out very often ,so reopen it to trace this issue. but sometimes we will another assetion like below:
XlibInt.c:599: _XPrivSyncFunction: Assertion `dpy->synchandler == _XPrivSyncFunction' failed.
It works well on G45 but failed on 945GME. The same issue happens on 945GME with following commits:
It works well on 945GME with current driver. Verified.