Bug 33918

Summary: radeon 6.14.0: OpenGL bad rendering in some applications
Product: xorg Reporter: Alexandre Demers <alexandre.f.demers>
Component: Driver/RadeonAssignee: xf86-video-ati maintainers <xorg-driver-ati>
Status: RESOLVED INVALID QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium    
Version: git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
result with 6.13.2
none
result with 6.14.0
none
dmesg
none
xorg.log none

Description Alexandre Demers 2011-02-04 13:46:34 UTC
I'm not sure if the problem comes from mesa, kernel or the radeon driver itself. However, when installing radeon driver 6.14.0 with kernel 2.6.37 (vanilla) and using latest mesa git version, some applications don't render correctly anymore.

I use RendererFeatTest almost everyday to test the latest mesa version. However, when I installed radeon 6.14.0, the application slowed down and some elements are not rendered correctly anymore. Reverting to 6.13.2 fixes it. So the problem doesn't come from mesa nor kernel itself.

I'm running Ubuntu 10.10 64bit, using XServer 1.9 and vanilla kernel 2.6.37.

I could test with latest 2.6.38-rc to see if this combination works.

I'll add the rendering screenshoots from RendererFeatTest with 6.13.2 and 6.14.0 driver.
Comment 1 Alexandre Demers 2011-02-04 13:56:41 UTC
Created attachment 42945 [details]
result with 6.13.2
Comment 2 Alexandre Demers 2011-02-04 14:09:09 UTC
Created attachment 42949 [details]
result with 6.14.0
Comment 3 Alexandre Demers 2011-02-04 14:14:59 UTC
As you can see, there are differences mostly under test07 and test14 where the shadows are rendered on everything instead of being rendered under the objects. There is also a difference with test10, but I think that's either related or of no importance for now (since the test completly fail to render anyway with prior version).
Comment 4 Alexandre Demers 2011-02-04 14:16:00 UTC
I should also add that I'm using the latest libdrm from git (which is mostly like 2.4.23 for radeon).
Comment 5 Alex Deucher 2011-02-04 14:25:59 UTC
Any chance you can bisect xf86-video-ati and track down what caused this?  Also, what chip and 3D driver are you using?  Can you attach your dmesg output and xorg logs?
Comment 6 Alexandre Demers 2011-02-04 14:33:26 UTC
Onboard chip, RS780, which means it's an HD3200. When testing, I used the r600g driver.

I'll attach dmesg and xorg.log in a few minutes.

For the bisect, is there a way I could test without having to remove my deb package? Maybe, compiling, installing in a different folder and changing a path somewhere? Any suggestion?
Comment 7 Alexandre Demers 2011-02-04 14:40:44 UTC
Created attachment 42950 [details]
dmesg
Comment 8 Alexandre Demers 2011-02-04 14:41:02 UTC
Created attachment 42951 [details]
xorg.log
Comment 9 Xavier Bestel 2011-02-04 15:13:36 UTC
What I do is copy the *_drv.so files I compiled over the system ones (without removing the .deb), and when I'm finished testing I simply reinstall the deb to have everything restored.
Comment 10 Tormod Volden 2011-02-04 17:42:12 UTC
As commented in bug 33738 I had similar corruption with xterm and eog, caused by commit ef9bfb262db7004bef3704e5d914687e50d3fca4.
Comment 11 Alexandre Demers 2011-02-04 18:24:43 UTC
Hmmm, strange. I can't seem to reproduce it with git repository and it might be related to one of two other patches introduced by Ubuntu packaging. I used the debian folder from previous 6.13.2 to base my own packaging. There are two patches included. I'll try to disable them one at a time and I'll come back soon to let you know if that helped.

By the way, thanks for sharing your procedure for testing git versions. That's what I thought was the way to go and it works great.
Comment 12 Alexandre Demers 2011-02-04 20:17:52 UTC
I removed both patches since Ubuntu removed one of them for Natty and the other was ment to be used to select gallium VS classic driver. For some reason, that last patch was causing the problem.

So the problem doesn't come from the radeon driver. We can close this bug. Sorry for taking your time and thank you for your help.
Comment 13 Alex Deucher 2011-02-05 10:13:32 UTC
Just out of curiosity, can you point me to what patch was causing the problem?
Comment 14 Alexandre Demers 2011-02-05 10:25:21 UTC
Of course. Patch named 101_select_between_classic_and_gallium_dri.patch. Maybe it just needs to be modified to work properly. I was using the specified option (ForceGallium) in my xorg.conf, so it shouldn't have had any impact on the rendering when I was testing the r600g driver from git. But in fact, it was playing with both the rendering and the performances (the tests were running choppy).
Comment 15 Michel Dänzer 2011-02-09 01:20:27 UTC
(In reply to comment #14)
> Of course. Patch named 101_select_between_classic_and_gallium_dri.patch.

Please report this to Ubuntu. They took this approach despite my recommendation against it, and they should know it's causing problems.
Comment 16 Alexandre Demers 2011-02-09 10:16:17 UTC
(In reply to comment #15)
> (In reply to comment #14)
> > Of course. Patch named 101_select_between_classic_and_gallium_dri.patch.
> 
> Please report this to Ubuntu. They took this approach despite my recommendation
> against it, and they should know it's causing problems.

(In reply to comment #15)
> (In reply to comment #14)
> > Of course. Patch named 101_select_between_classic_and_gallium_dri.patch.
> 
> Please report this to Ubuntu. They took this approach despite my recommendation
> against it, and they should know it's causing problems.

Reported: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/715939

Thanks for everything

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.