Brice Goglin a écrit :
retitle 431746 Xorg/sparc: Xinerama active, but <= 0 screens?
thank you
You're welcome.
BERTRAND Joël wrote:
I have tested some configurations to isolate parameters :
1/ sparc32/smp SS20 + CG14 -> bug (I haven't tested with CG6 because
my test workstation has no empty Sbus). I have a ZX framebuffer (leo)
in a SS5, but ZX framebuffer support is totaly broken in 2.6.x kernels;
2/ sparc64/smp U60 (UPA/PCI) + Creator3D -> bug;
3/ sparc64/smp U2 (UPA/Sbus) + Creator3D -> bug;
4/ i386 + prosavage8 -> no bug;
5/ i386 + Radeon Mobility 9200 (AGP) -> no bug.
All configurations are up to date (debian testing).
I don't know if this bug is sparc32/64 specific, but I only see
this bug on sparc32 and sparc64, never on i386.
Unfortunately, I don't have access to any machine like this.
If you want, I can open an account on a U2 (dual UltraSPARC II/300 MHz
with 2 GB and a Creator3D fb).
You said earlier:
"All program that test Xinerama crash because XineramaIsActive returns
TRUE _and_ XineramaQueryScreens returns a NULL pointer and a number of
screens <=0. Maybe the bug is in XineramaQueryScreens ?"
So I guess the next step would be to check the return values of these
functions on your i386 machines?
Maybe, but I don't have decent i386 to rebuild xorg from scratch (only
K6-III/400 with 256 MB, other i386 are laptop, and I cannot use it to test).
Then, it will probably be a good time
to forward the bug upstream (unless you want to rebuild the server with
some debugging printf to find out what's going on in the internal
functions).
I can, if that can help you.
By the way, when you say "<= 0", do you mean "< 0" ?
No, "<= 0". This message comes from /var/log/syslog and is returned by
gdm.
Regards,
JKB
