[cairo] extend-reflect test hangs system

Jeff Muizelaar jeff at infidigm.net
Tue Mar 20 07:04:36 PDT 2007

On Tue, Mar 20, 2007 at 06:40:23AM +0100, Mathias Hasselmann wrote:
> Am Dienstag, den 20.03.2007, 00:52 -0400 schrieb Jeff Muizelaar:
> > This is a known problem. Nobody has come up with a solution yet.
> Do we have a theory about the cause of the problem? Or do we know at
> least, arround which time that bug appeared? Could investigate the
> problem, but would like to operate with some kind of plan: Endless
> reboot-debug cycles are no fun...

IIRC the xservers ends up looping in fbCompositeSrc_8888x8888mmx() when try to
composite the surface we create in the 'if (attr->extend ==
_cairo_pattern_acquire_surface_for_surface(). The current hypothesis is
that the problem is caused by fbComposite() giving
fbCompositeSrc_8888x8888mmx() bad parameters or data of some sort.

Debugging it shouldn't be too bad if you have debugging symbols or can
build your own xserver and can debug using ssh. (Note: when debugging
the xserver using gdb I've found it a good idea to do 'handle SIGARLM
ignore' to keep from being interrupted).


More information about the cairo mailing list