Bug 60049 - corruption using qt apps and opera in gnome
Summary: corruption using qt apps and opera in gnome
Status: RESOLVED WORKSFORME
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium normal
Assignee: Chris Wilson
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-01-30 01:17 UTC by Conner
Modified: 2013-05-21 13:02 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
corruption shown (622.04 KB, image/png)
2013-01-30 01:17 UTC, Conner
no flags Details
corruption shown in picture (622.04 KB, image/png)
2013-01-30 01:20 UTC, Conner
no flags Details
xorg.0.log (26.23 KB, text/plain)
2013-01-30 01:45 UTC, Conner
no flags Details
corruption in opera (76.63 KB, image/png)
2013-01-30 22:00 UTC, Conner
no flags Details

Description Conner 2013-01-30 01:17:29 UTC
Created attachment 73893 [details]
corruption shown

I have noticed a lot of corruption using qt applications in gnome with last few versions of intel and even more using the Opera browser. I posted a screenshot of Minitube (qt app) showing some corruption.
Comment 1 Conner 2013-01-30 01:20:40 UTC
Created attachment 73894 [details]
corruption shown in picture

Picture attached as plain text accidentally.
Comment 2 Chris Wilson 2013-01-30 01:31:38 UTC
Can you please attach your Xorg.0.log?
Comment 3 Conner 2013-01-30 01:39:36 UTC
Sorry for the sloppiness. Here are some more details.

Archlinux x64

first gen intel i5, gateway nv79c

xf86-video-intel 2.20.19

mesa 9.0.2

xorg-server 1.13.2

qt 4.8.

linux 3.7.5

20-intel.conf:

Section "Device"
    Identifier  "Intel Graphics"
    Driver      "intel"
    Option      "AccelMethod"  "sna"
    Option      "Backlight"    "intel_backlight"
    Option      "TearFree" "true"
EndSection
Comment 4 Conner 2013-01-30 01:45:43 UTC
Created attachment 73895 [details]
xorg.0.log
Comment 5 Chris Wilson 2013-01-30 08:56:21 UTC
Can you please test without TearFree? Just in case that is affecting the damage tracking.
Comment 6 Chris Wilson 2013-01-30 08:57:07 UTC
Out of curiosity, which backlight does SNA detect by default for you?
Comment 7 Chris Wilson 2013-01-30 09:33:31 UTC
Doesn't immediately reproduce here, perhaps a few details on the qt version and themes?
Comment 8 Conner 2013-01-30 19:14:58 UTC
This is what I get without intel_backlight specified: 
found backlight control interface acer-wmi (type 'platform') 

UXA detects intel_backlight correctly. To get either sna or uxa to work with the backlight I have to use acpi_backlight=vendor regardless and even more strangely  is that adding i915 to mkinitcpio (initramfs creation) causes the backlight not to work.

To reproduce the corruption in minitube, use gnome and play a video. I don't get corruption if I don't play a video. 

To reproduce the corruption in opera, use gnome and open the preferences (ctrl-f12) window in opera and click the tabs in the windows.
Comment 9 Conner 2013-01-30 19:16:02 UTC
Still get same corruption with tearfree disabled.
Comment 10 Chris Wilson 2013-01-30 19:23:29 UTC
(In reply to comment #8)
> This is what I get without intel_backlight specified: 
> found backlight control interface acer-wmi (type 'platform') 
> 
> UXA detects intel_backlight correctly.

I think you mean "incorrectly". :-p

Your system is specified to use the acer-wmi *platform* backlight interface. The intel_backlight is one of last resort that is often not hooked up to anything at all. So you need to file a bug with the acer-wmi acpi guys for the broken backlight controller.

Let's see what happens in the opera's preference window...
Comment 11 Chris Wilson 2013-01-30 20:25:41 UTC
Hmm, do you have an example handy of the opera corruption? Haven't seen anything amiss, the suggestion from an earlier bug I fixed with Opera suggests that I check for missing migrations...

Anyway seeing the style of corruption should help me pinpoint the likely cause.
Comment 12 Conner 2013-01-30 22:00:08 UTC
Created attachment 73960 [details]
corruption in opera

The same colourful corruption on minitube shows up on the cancel button in the opera preferences' window but that corruption disappears with gnome-screenshot.
Comment 13 Chris Wilson 2013-02-01 11:50:11 UTC
The error does look like corruption inside the temporary buffers used for trapezoids. Can you please build from xf86-video-intel.git and ./configure with --enable-debug?
Comment 14 Conner 2013-02-01 18:39:04 UTC
Yeah, I built xf86-video-intel with debug from git but was unable to start x (probably my error in building). Anyway, I instead built 2.21.0 with debug.
Comment 15 Chris Wilson 2013-02-12 08:59:48 UTC
Ok... Is the error still reproducible? Have you tried different compiler flags (thinking back to earlier similar bugs)?
Comment 16 Conner 2013-02-16 16:18:16 UTC
No longer reproducible in 2.21.2. What did you do to fix it?
Comment 17 Chris Wilson 2013-02-16 21:19:37 UTC
Nothing explicitly targeting this bug, but plenty of candidates for accidental fixes. If it stays fixed in the next release, I'm just going to pretend it never happens and hope it never repeats.
Comment 18 Chris Wilson 2013-02-20 13:55:40 UTC
I've tagged new release. Can you double check that the bug hasn't resurfaced?
Comment 19 Conner 2013-02-23 13:30:01 UTC
It is still reproducible in qt apps and opera. It's hard to get a screenshot but the corruption for example is minitube slider is milder and doesn't flicker when watching videos and the corruption in the opera preferences window is limited to very small grey boxes in the tabs of the preferences window. I am using 2.21.3. The situation is definitely improved.
Comment 20 Chris Wilson 2013-02-25 11:04:24 UTC
Your description still reminds me of earlier compiler artifacts.
Comment 21 Chris Wilson 2013-03-11 19:26:50 UTC
Might as well check with 2.21.4 in case this turns out to be a read-read bug instead.
Comment 22 Conner 2013-03-12 03:23:56 UTC
Still reproducible. What do you mean by read-read bug?
Comment 23 Chris Wilson 2013-03-12 08:38:37 UTC
read-read: bug 61628
Comment 24 Chris Wilson 2013-04-04 18:23:15 UTC
It still appears that you are the only person encountering this issue. Can you try enabling a few debug options and changing compiler options to identify the cause?
Comment 25 Chris Wilson 2013-05-21 13:02:42 UTC
Been a while, hopefully things are better. Or at least not worse.


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.