Bug 25031 - rendering and color corruption since 14109abf
Summary: rendering and color corruption since 14109abf
Status: RESOLVED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: 7.4 (2008.09)
Hardware: Other All
: medium major
Assignee: Chris Wilson
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-11-11 04:25 UTC by Petar Velkovski
Modified: 2009-12-01 10:28 UTC (History)
4 users (show)

See Also:
i915 platform:
i915 features:


Attachments
Xorg.0.log (66.50 KB, text/x-log)
2009-11-11 04:25 UTC, Petar Velkovski
no flags Details
dmesg output (39.78 KB, text/plain)
2009-11-11 04:25 UTC, Petar Velkovski
no flags Details
Screenshot of the desktop badly rendered (692.90 KB, image/png)
2009-11-11 04:26 UTC, Petar Velkovski
no flags Details
Screenshot with the desktop rendered with intel 2.9.0+git20091125.917f9bb2 (481.04 KB, image/png)
2009-11-29 05:39 UTC, Petar Velkovski
no flags Details
Desktop rendered with 2.9.0+git20091111.dbb68168 and all is normal :) (486.97 KB, image/png)
2009-11-29 05:48 UTC, Petar Velkovski
no flags Details
Screenshot with the desktop rendered with intel 2.9.0+git20091129.c10850c6 (490.62 KB, image/png)
2009-11-30 02:43 UTC, Petar Velkovski
no flags Details
screenshot after 5e04ded (7.90 KB, image/png)
2009-11-30 04:04 UTC, Martin Pitt
no flags Details

Description Petar Velkovski 2009-11-11 04:25:08 UTC
Created attachment 31109 [details]
Xorg.0.log

After making this upgrade:

xserver-xorg-video-intel (2:2.9.0+git20091106.dbb68168-0ubuntu0tormod) to 2:2.9.0+git20091110.14109abf-0ubuntu0tormod

a drawing corruption appeared on my system.

I am sending Xorg.0.log and dmesg output as advised by the ubuntu packager. Also I am sending a screenshot so that you can see the kind of rendering coruption that appeared.

Explanation of the screenshot:
1. The system is using compiz
2. There is problem rendering of alpha channel noticeable around the icons on the desktop
3. Alpha still functions properly for some things (Guake terminal window with alpha set still renders properly.
4. The text colour is changed to white everywhere, that's why gedit text editor in the screenshot only shows the part of the text document that I selected, but the rest of the text is invisible (white text on white background)

According to the package builder you should be able to make a conclusion when was this bug introduced (dbb68168 worked just fine, 14109abf is broken)
Comment 1 Petar Velkovski 2009-11-11 04:25:57 UTC
Created attachment 31110 [details]
dmesg output
Comment 2 Petar Velkovski 2009-11-11 04:26:49 UTC
Created attachment 31111 [details]
Screenshot of the desktop badly rendered
Comment 3 Petar Velkovski 2009-11-25 21:10:52 UTC
Today I made upgrade to xserver-xorg-video-intel 2:2.9.0+git20091125.917f9bb2-0ubuntu0tormod~karmic only to see that this problem is still present! This bug needs more love :)
Comment 4 Tormod Volden 2009-11-26 03:24:02 UTC
The difference between the two versions are the five commits on 2009-11-10, all by Chris Wilson:

i915: Fix texture sampling coordinates.
i915: Use the color channels to pass along solid sources and masks.
i915: Check for overflow before overflowing.
Check that batch buffers are atomic.
uxa: Force alpha bits to fill remaining bits
Comment 5 Martin Pitt 2009-11-26 06:52:52 UTC
Confirming here:
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)
Comment 6 Chris Wilson 2009-11-29 03:28:49 UTC
Petar, do you mind retesting with the latest bits? I know not everything is resolved but I want to see what issues remain.

Also can you post a screenshot from with the 2.9.1 drivers just so I can be sure which bits of the desktop are actually regressions...
Comment 7 Petar Velkovski 2009-11-29 03:40:22 UTC
As soon as Tormod builds the latest package from git, I'll test it. I'll write him an email. 
Comment 8 Petar Velkovski 2009-11-29 03:41:55 UTC
bugzilla-daemon@freedesktop.org wrote:

Tormod they asked me to test with the latest build of intel driver. Can 
you please build it, so I can send my results?
> http://bugs.freedesktop.org/show_bug.cgi?id=25031
>
>
>
>
>
> --- Comment #4 from Tormod Volden <bugzi07.fdo.tormod@xoxy.net>  2009-11-26 03:24:02 PST ---
> The difference between the two versions are the five commits on 2009-11-10, all
> by Chris Wilson:
>
> i915: Fix texture sampling coordinates.
> i915: Use the color channels to pass along solid sources and masks.
> i915: Check for overflow before overflowing.
> Check that batch buffers are atomic.
> uxa: Force alpha bits to fill remaining bits
>
>
>   
Comment 9 Tormod Volden 2009-11-29 04:57:54 UTC
When you tested the 2:2.9.0+git20091125.917f9bb2-0ubuntu0tormod~karmic in comment 3 you already had the latest bits, there is nothing newer.
Comment 10 Chris Wilson 2009-11-29 05:03:19 UTC
(In reply to comment #9)
> When you tested the 2:2.9.0+git20091125.917f9bb2-0ubuntu0tormod~karmic in
> comment 3 you already had the latest bits, there is nothing newer.
> 

Ok, I was just misled by comment 4 - I didn't realise that Petar had tested with the couple of bug fixes since those commits.

Petar, can you please grab a screenshot with 2.9.1?
Comment 11 Petar Velkovski 2009-11-29 05:39:23 UTC
Created attachment 31548 [details]
Screenshot with the desktop rendered with intel 2.9.0+git20091125.917f9bb2

Ok this is a screenshot with intel 2.9.0+git20091125.917f9bb2. The only difference I can can notice is that the text in gedit is rendered correctly. Other than that the other rendering and colour corruption remain. i hope this is what you wonted me to try, because I don't have intel 2.9.1 in ubuntu.We used to have it in jaunty (Ubuntu 9.04), but not in Karmic (Ubuntu 9.10). As far as i can remember it worked just fine.  If I should test with 2.9.1 than Tormod should build it for Karmic too. :)
Comment 12 Petar Velkovski 2009-11-29 05:48:57 UTC
Created attachment 31550 [details]
Desktop rendered with 2.9.0+git20091111.dbb68168 and all is normal :)

This is how the desktop should be rendered normaly. This is rendered using xserver-xorg-video-intel_2.9.0+git20091111.dbb68168
Comment 13 Chris Wilson 2009-11-29 07:38:18 UTC
Petar, thanks for the extra screenshots.

Knowing that we are incorrectly blending the transparent pixels in the images should be a good clue as to what the bug is. I didn't expect that code path to be affected at all...
Comment 14 Eric Appleman 2009-11-29 13:11:51 UTC
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)

Linux kingfisher 2.6.32-5-generic #6-Ubuntu SMP Mon Nov 23 11:49:07 UTC 2009 i686 GNU/Linux

Confirming with latest bits.

If you guys need any hard data from me in the way of logs or configs, I'll be glad to provide.
Comment 15 Petar Velkovski 2009-11-30 02:43:53 UTC
Created attachment 31576 [details]
Screenshot with the desktop rendered with intel 2.9.0+git20091129.c10850c6

No visible progress with intel 2.9.0+git20091129.c10850c6
Comment 16 Chris Wilson 2009-11-30 03:17:25 UTC
Ok, I've found one bug. rendercheck is just a noise of errors so I can't determine if everything is now correct...

commit 5e04ded2bce4c135b57d391f5f4e24e030103e61
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Nov 30 10:57:04 2009 +0000

    i915: Fix missing texture offset for mask.
    
    In commit e581ceb, I modified the shader generation to accommodate mixed
    textures and solids but missed applying the new computed sampler for the
    mask.
    
    References:
    
      Bug 23803 [bisected i915] gnome characters disappear
      http://bugs.freedesktop.org/show_bug.cgi?id=23803
    
      Bug 25031 rendering and color corruption since 14109abf
      http://bugs.freedesktop.org/show_bug.cgi?id=25031
    
      Bug 25047 [945GM bisected] rendercheck/repeat/triangles regressed
      http://bugs.freedesktop.org/show_bug.cgi?id=25047
  
Comment 17 Petar Velkovski 2009-11-30 03:39:00 UTC
I'll wait for Tormod to build a new package and I'll give you a report.
Comment 18 Martin Pitt 2009-11-30 04:04:45 UTC
Created attachment 31582 [details]
screenshot after 5e04ded

Icons are back to normal, thanks! So most of the desktop now looks fine.

However, it seems that this caused a regression in painting menus, radio buttons, etc. Their upper half is fine, but their lower half is blending into the background color now.
Comment 19 Eric Appleman 2009-11-30 08:57:17 UTC
b118a52c seems to have also broken Flash for me. Whenever I try to watch a Flash video on that revision, X will crash and restart.
Comment 20 Petar Velkovski 2009-11-30 15:28:26 UTC
Chris I made this upgrade today:
xserver-xorg-video-intel (2:2.9.0+git20091111.dbb68168-0ubuntu0sarvatt) to 2:2.9.0+git20091130.2.6729b508-0ubuntu0tormod~karmic
And everything is NORMAL AGAIN :). If this build contains your fixes, congratulations, you nailed it :)

Now I would like to ask for a favour :). If you can have a look at bug http://bugs.freedesktop.org/show_bug.cgi?id=25345 and see if it makes sense to you at all. Also it it makes sense could you advertise it a little bit among other intel driver developers because it seams like a serious bug to me.

After you read this message you can close this bug. :) 
Comment 21 Martin Pitt 2009-12-01 01:18:00 UTC
(In reply to comment #18)
> However, it seems that this caused a regression in painting menus, radio
> buttons, etc. Their upper half is fine, but their lower half is blending into
> the background color now.

With latest git this is fixed again, too. So all is well for me, too. Thanks!
Comment 22 Chris Wilson 2009-12-01 10:28:27 UTC
Eric, I can't think how the commit you pin-point would cause Flash to crash the system, if the problem reoccurs please open a new bug report.

All, I've reverted the centre-point sampling for the next release and thus fix text rendering and apparently the gradient/solid on the menu -- but the underlying issues remain unresolved. Thanks for your patience.


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.