Created attachment 14531 [details] XAA.png On my r200 (ATI Technologies Inc Radeon R200 QL [Radeon 8500 LE]), I've noticed a rendering error with EXA that is not there with XAA. Specifically, in the GTK+ progress bars in the clearlooks GTK+ theme. I've noticed the problem on at least xf86-video-ati-6.7.197 and xf86-video-ati-6.8.0 but possibly earlier too. I'll attach screen shots, look in the lower right corner. XAA.png shows correct rendering, EXA.png shows incorrect.
Created attachment 14532 [details] EXA.png
Please attach the full xorg.conf and Xorg.0.log files. In particular though, which version of the X server are you using?
Created attachment 14537 [details] xorg.conf and Xorg.0.log Hmm, that explains why I'd thought that the new progress bar for the default Clearlooks theme looked a bit crap. Testing, it happens with always and smart migration heuristics but not greedy. Disabling page flip and color tiling don't make a difference. This is on xserver 1.4.0.90 (with Gentoo patches none of which touch EXA or radeon) and xf86-video-ati GIT from yesterday. Xorg.0.log and xorg.conf attached
(In reply to comment #3) > Testing, it happens with always and smart migration heuristics but not greedy. So it's probably pixmap migration related (greedy doesn't migrate to offscreen memory as aggressively as smart, let alone always). Does it also happen without Option "AccelDFS"? If yes, can somebody try the xserver Git master branch?
Disabling AccelDFS doesn't make a difference. Odd that I don't see any other obvious corruption though. Will try updating to current xserver Git.
*** Bug 13797 has been marked as a duplicate of this bug. ***
Pretty sure this is fixed in xserver 1.5 / Git. Please reopen if you find otherwise.
Unfortunately, I'm still having this bug. I'm using xserver 1.4.99.905 and xserver-xorg-video-radeon 6.9.0.
Created attachment 17682 [details] Xorg.0.log
Hmm, I wonder if this could actually be an R200 RENDER acceleration bug... Does Option "MigrationHeuristic" "greedy" or "smart" still work around the problem? What about Option "RenderAccel" "off"?
"greedy" and "RenderAccel" "off" work around the problem, it's still broken with "smart".
So I think this probably is an R200 RENDER acceleration bug, as I'm not seeing it on R300.
Created attachment 17752 [details] [review] possible fix This patch (against ati git master) reworks the fragment program setup on r200 to match r300. Does it fix the bug for you?
No, it didn't make any difference.
does git master help? specifically 783cdb7374941bb1d2b63eea375fbf2f1b808cc3
I no longer own any r200 hardware, but perhaps Colin or Alan can confirm the bug?
Just updated to git (shockingly my last update was 30-09-2008) and can confirm the problem is resolved on R200.
I don't have access to a r200 any more. Glad to see it fixed up though!
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.