I have xf86-video-intel-18.104.22.1683, and I get rendering glitches when running full screen SDL apps (specifically dosbox). This did not happen with 901, but did happen with 902.
I'm running metacity with the compositor on, so some form of compositing window manager may be necessary to reproduce.
X.Org version: 22.214.171.124
Pls provide your xorg.conf and X log.
Does git master have your problem too?
Created attachment 16009 [details]
Created attachment 16010 [details]
Here's the xorg.conf and log file. I'll try to test git master as soon as I can, but it probably won't be before the weekend.
Could you describe what's the failure rendering result? I'm running metacity with composite on and dosbox in fullscreen which looks fine at least with simple commands running (like 'intro', etc.).
And what's your mesa version? Is it same for 901 and 903 build tests?
If I run dosbox full-screen, I get short, blinking, horizontal white lines around the edges and down the middle of the screen. I don't have to run any commands in dosbox; merely making it full screen is enough.
If I run a game in dosbox (Master of Orion 2 is what I'm plyaing right now), then the lines are in different parts of the screen depending on what screen I'm displaying, but they're never gone entirely. Some screens are almost entirely covered by them. The effect does seem to depend somewhat on what other windows I have open (although, oddly enough, not necessarily on which windows are on the same desktop as the dosbox window).
I'm running mesa 7.0.2, and it's the only version I've had this install.
I have libsdl 1.2.13 with opengl and xv support enabled.
I have dosbox 0.72.
I've put a picture here:
I can attach it if you'd prefer, but it's fairly big. The glitch is really not bad at the moment, but you can see the line down the middle.
I can't produce on 915G here too.
Is this issue gone with 126.96.36.1991 and how about 903? Better to try mesa 7.0.3 also.
That picture was with 903. It also happens with 902, but not with 901. I'll try mesa 7.0.3
Although still can't produce on another hp nx6310 945GM, could you try to revert this commit between 901 and 902 which might be relate to this and test again?
Author: Keith Packard <firstname.lastname@example.org>
Date: Fri Mar 21 03:08:24 2008 -0700
Transformed coord computed using floats. Don't move bilinear composite dst.
Reverting that commit fixes the problem for me.
I'm more than willing to test any patches you want me to test.
Better to try split that commit into two parts, for is_nearest and float matrix compute. You can split that patch and try to apply one by one and see which one is actually causing your problem.
Okay, false positive, sorry. I accidentally picked up another patch that fixed my problem. It's called xf86-video-i810-188.8.131.523-fix-panel-resize-on-i8xx.patch, and Remi can probably explain which one it is.
Sorry for following the wrong track...
Could you try bumping locally to 2.3.0 without any patches? (I'll push it to portage when I get the chance). That patch I made for .903 is a backport of 4 or 5 patches that sit in between .903 and 2.3.0.
2.3.0 is fine.