Downstream openSUSE bug: https://bugzilla.novell.com/show_bug.cgi?id=384022 Example of the type of corruption: https://bugzilla.novell.com/attachment.cgi?id=210633 Sometimes if I do a VT switch or if I suspend and resume (which does the same) the fonts appear corrupted when I return to X. The fonts appear extremely blocky. Using any one of the following appears to work around the issue in some way: Option "AccelMethod" "XAA" Option "EXANoComposite" "Yes" Option "MigrationHeuristic" "greedy"
I can not reproduce it on openSUSE11.0-GM following your configuration. I have tried for many times. Stefan, Andrew Would you like to try this on openSUSE11.0-GM to make sure it can be reproduced? Maybe, openSUSE11.0-GM has resolved this problem. Thanks
Andrew, please try to reproduce on openSUSE 11.0-GM. The machine, on which I've seen this issue is no longer available. It was an Alpha test machine, which has been returned back to Intel
The bug doesn't appear to exist because openSUSE changed the driver to default to XAA rather than EXA. Force EXA without any of the workarounds mentioned to see the problem.
(In reply to comment #3) > The bug doesn't appear to exist because openSUSE changed the driver to > default to XAA rather than EXA. Force EXA without any of the workarounds > mentioned to see the problem. Intel is aware of this. Still they can't reproduce it.
I'm taking over this bug now. I'll wait to hear a confirmation that openSUSE11.0-GM still has the bug before I try to reproduce it myself. Thanks, -Carl
I installed the opensuse11 GM and xf86-video-intel 2.3.1 stable release on 965GM (Thinkpad T61), it uses EXA acceleration. I can't reproduce the fonts corruption issue on VT switch or suspend/resume.
I have to admit I can no longer reproduce this bug either. I turned on EXA and explicitly turned off all the workarounds, I even setup a 1440x900 framebuffer on the console and I cannot reproduce the bug. So, thanks for fixing this! openSUSE should default to EXA again unless there are other bugs holding this back. Thanks!
Andrew, which packages are you currently using? openSUSE released an update (xorg-x11-driver-video/xorg-x11-server) this monday, which made "EXA" (for "intel") and "MigrationHeuristic" "greedy" the default. No other changes than these. So with these changes in place you would need to test with Option "MigrationHeuristic" "always" to reproduce the bug again.
I have tried (In reply to comment #8) > Andrew, which packages are you currently using? openSUSE released an update > (xorg-x11-driver-video/xorg-x11-server) this monday, which made "EXA" (for > "intel") and "MigrationHeuristic" "greedy" the default. No other changes than > these. > > So with these changes in place you would need to test with > > Option "MigrationHeuristic" "always" > > to reproduce the bug again. > I have tried this. Seems OK. However I don't use the latest xserver. Just set the options. Stefan, Have you tried it? Also I have got the information from upstream, it will be better to use the default value of option "MigrationHeuristic". Are there some special reasons to set like this? Also if there are such bugs about this, upstream developer will not spend time to fix this because this option may be deleted in the future for drm-gem.
I double checked that I was setting the MigrationHeuristic properly and I was but I discovered that for some reason compositing is being disabled by the driver: (II) intel(0): EXA: Disabling Composite operation (RENDER acceleration) Since disabling EXA composite is one of the workarounds that would explain why the bug is hard to reproduce. I'll try a few things to see if one of my settings has caused it to be disabled.
> Stefan, Have you tried it? No, the machine, where I've ever seen no longer exists. > Also I have got the information from upstream, it will be better to use the > default value of option "MigrationHeuristic". Are there some special reasons > to set like this? Sure. You get much better performance than with the default "always". Still not comparable with XAA, but at least the system is again somewhat usable. We already got bugreports from users after switching to EXA, which think different here, and went straight ahead to XAA again. > Also if there are such bugs about this, upstream developer > will not spend time to fix this because this option may be deleted in the > future for drm-gem. I don't see GEM in the current kernel. Do you? For openSUSE 11.1 we're trying it now with the default "always" for "MigrationHeuristic". Maybe performance issues are resolved in time.
ok. I'll mark this bug as fixed per above comments. if there are new issues, please open new bugs. thanks.
Michael, I don't think we can consider this bug fixed. We have not been able to see if EXA, including compositing + MigrationHeuristic=always still results in font corruption or not.
Okay, I've been able to check w/ all options and this bug does appear to be fixed. Thanks guys!
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.