Bug 103963

Summary: switching screens on Matrox G400 corrupts colors of the X desktop
Product: [openSUSE] SUSE LINUX 10.0 Reporter: Hans-Peter Jansen <hpj>
Component: SaX2Assignee: Stefan Dirsch <sndirsch>
Status: RESOLVED FIXED QA Contact: E-mail List <qa-bugs>
Severity: Normal    
Priority: P2 - High CC: hpj
Version: Beta 1   
Target Milestone: ---   
Hardware: PC   
OS: All   
Whiteboard:
Found By: Beta-Customer Services Priority:
Business Priority: Blocker: ---
Marketing QA Status: --- IT Deployment: ---
Attachments: xorg.conf
Xorg.0.log
hwinfo --gfx
screenshot of the damage

Description Hans-Peter Jansen 2005-08-10 20:22:37 UTC
After switching between console and graphic screen (Ctrl-Alt-F{1,7}), the 
colors of the X desktop are corrupted. I remember a related problem in the 
early days of the MGA driver some years ago, and from the early days of MGA 
G550 support (not so long ago).. 
 
Also, I had one occurence of https://bugzilla.novell.com/show_bug.cgi?id=71981 
on this system with B1 (from about 15 login attempts). I don't remember that 
on a G400 any time before.
Comment 1 Stefan Dirsch 2005-08-10 21:10:34 UTC
Please provide more informations: 
 
http://www.opensuse.org/index.php/Bugs:X 
Comment 2 Hans-Peter Jansen 2005-08-10 21:25:20 UTC
Created attachment 45631 [details]
xorg.conf
Comment 3 Hans-Peter Jansen 2005-08-10 21:25:45 UTC
Created attachment 45632 [details]
Xorg.0.log
Comment 4 Hans-Peter Jansen 2005-08-10 21:26:29 UTC
Created attachment 45633 [details]
hwinfo --gfx
Comment 5 Hans-Peter Jansen 2005-08-10 21:27:03 UTC
Let me know, what else you need.. 
Comment 6 Stefan Dirsch 2005-08-11 21:20:15 UTC
There has been *zero* changes in mga driver since SuSE 9.3 ... 
Comment 7 Hans-Peter Jansen 2005-08-12 16:51:34 UTC
Hmm, than the culprit must be somewhere else. Surely the kernel is involved in 
console switching, and X. Both would lead to others suffering from it, too, 
don't it? 
 
 
Comment 8 Hans-Peter Jansen 2005-08-12 16:59:27 UTC
Created attachment 45940 [details]
screenshot of the damage
Comment 9 Stefan Dirsch 2005-08-12 21:50:51 UTC
BTW, you use a dualhead setup with different color depths on both screens. 
Please test whether it helps to use either 16 or 24 bit color depth for both 
screens. 
Comment 10 Hans-Peter Jansen 2005-08-13 09:50:43 UTC
It wasn't me, who setup dualhead. Blame sax2..    
   
Either adjusting the second port to 24 bit or deactivating it comletely fixed  
it! The typical color corruption is visible for half a second in the upper 
half of the screen after switching back to X. Looks like the screen content is 
copied first, and the color tables(?!) are adjusted later (after vbi?). 
A question remains: could this be prevented without user intervention?  
  
BTW: while I got https://bugzilla.novell.com/show_bug.cgi?id=71981 a few times 
before the fix, it never happened thereafter (restarted X about 20 times, 
switched screens back and forth many times). 
Comment 11 Stefan Dirsch 2005-08-13 10:59:11 UTC
Please describe in detail, what you did during installation and configuration  
of X11. 
Comment 12 Hans-Peter Jansen 2005-08-13 11:16:09 UTC
Well, that's easy: as I happily noticed the new buttons to adjust the  
resolution in YaST, I just reduced the screen resolution from 1280x1024 to   
1152x864, since the default is a bit to much for this poor 17'' CRT.  
  
Nothing else. I can kick another install on this system to support this claim,  
if you like. In real installations with Matrox graphic cards, which I happen  
to have quite a bunch of (about 30 in all flavors from early G200 with 2MB up 
to the G550), I tend to deactivate the second screen, reduce the max. vert. 
refresh to 85Hz (if CRT), and activate 3D acceleration. 
Comment 13 Marcus Schaefer 2005-08-15 08:13:06 UTC
Stefan I'm fed up with that matrox graphics driver it's really a piece 
of broken code. I would suggest to remove every profile entry within the CDB 
using the Matrox_* profile files. 
 
In that case only single head is supported which means there is a small 
chance for the driver to work correctly 
 
Thanks 
Comment 14 Stefan Dirsch 2005-08-15 08:42:23 UTC
I agree. We should finally do this step. 
Comment 15 Stefan Dirsch 2005-08-15 14:07:30 UTC
Matrox_* profile entries are now removed in CDB. Fixed for Beta2.