Bug 59351 - [Zaphod] Artefacts on rotated display
Summary: [Zaphod] Artefacts on rotated display
Status: RESOLVED DUPLICATE of bug 55500
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: 2013-01-14 10:20 UTC by Till Matthiesen
Modified: 2013-02-04 15:28 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
artefacts acrobat rotated display (1.16 MB, image/jpeg)
2013-01-14 10:22 UTC, Till Matthiesen
no flags Details
xorg.conf (1.27 KB, text/plain)
2013-01-14 13:56 UTC, Till Matthiesen
no flags Details
corrupted glyphs on rotated screen (1.10 MB, text/plain)
2013-01-17 09:36 UTC, Till Matthiesen
no flags Details
MIME type corrected (1.10 MB, image/png)
2013-01-17 09:41 UTC, Till Matthiesen
no flags Details
artefacts acrobat rotated display [video] (2.38 MB, application/octet-stream)
2013-02-04 10:39 UTC, Till Matthiesen
no flags Details

Description Till Matthiesen 2013-01-14 10:20:47 UTC
Hi Chris,

I run two monitors in zaphod mode. The left one is rotated.
Acrobat reader shows artefacts along the TOC bar typically when I switch workspaces several times. A photo is attached.

It was not possible to take a screenshot directly as the artefacts vanished on the dump. Also, as I move the mouse pointer to the other screen, the artefacts vanish immediately.

The IGP is a Q965.

xorg-server-1.13.1
xf86-video-intel-2.20.17
libdrm-2.4.40

Thanks for your excellent work!

~Till
Comment 1 Till Matthiesen 2013-01-14 10:22:27 UTC
Created attachment 72997 [details]
artefacts acrobat rotated display

The artefacts are not present when I start Acrobat on the other screen with normal orientation.
Comment 2 Till Matthiesen 2013-01-14 10:46:49 UTC
This issue is not reproducable with xrandr/xinerama.
Comment 3 Till Matthiesen 2013-01-14 12:01:10 UTC
I just realised the description is not correct.

The artefacts do not vanish due to the pointer crossing the screen.
They vanish when the pointer triggers the focus of any other window on whatever screen.
Comment 4 Till Matthiesen 2013-01-14 12:48:25 UTC
Building the driver from git didn't improve the situation.
Comment 5 Chris Wilson 2013-01-14 13:41:42 UTC
I suspect it is akin to bug #55500. Can you describe how you configure your zaphod displays so I can try and replicate your setup?
Comment 6 Till Matthiesen 2013-01-14 13:56:25 UTC
Created attachment 73009 [details]
xorg.conf

The window manager is "awesome wm (3.4.13)".
Acrobat Reader is version 9.5.1

Take for example this document and open the bookmark panel:
https://01.org/linuxgraphics/sites/default/files/documentation/965_g35_vol_2_3d_media_web_updated_1.pdf
I don't know how to reproduce the issue with other window managers but I can trigger the artefacts by switching to another workspace and back.

If you need further information, let me know.

Thanks for looking into this!
Comment 7 Till Matthiesen 2013-01-17 09:36:35 UTC
Created attachment 73180 [details]
corrupted glyphs on rotated screen

Hi Chris,

I saw you added some patches and, thus, tried the current git version.
The artefacts are still there (I guess it's expected) but now there are additional glyph corruptions (see attachments). Again, they only appear on the rotated screen.

Are you aware of them or should I file a new bug report?
Comment 8 Till Matthiesen 2013-01-17 09:41:02 UTC
Created attachment 73181 [details]
MIME type corrected

[The artefacts are not caught by the screendump]
Comment 9 Till Matthiesen 2013-02-04 10:39:46 UTC
Created attachment 74167 [details]
artefacts acrobat rotated display [video]

Hi Chris,

I just gave the current git a try.

The glyph corruption I reported in #7 is gone.
However, the artefacts are not.
While I know that you have other priorities right now,
I just wanted to know if you could reproduce the problem.
For me it is easy to trigger the artefacts by changing size
of the bookmark bar (see attached video).
Comment 10 Chris Wilson 2013-02-04 10:58:33 UTC
That video definitely looks like bug #55500. Looks like you have a situation where it consistently occurs. :(

*** This bug has been marked as a duplicate of bug 55500 ***
Comment 11 Till Matthiesen 2013-02-04 14:18:12 UTC
I just did a bisection and came up with this commit:

commit 1aee8acacfe5869a072d9f20f3b8290b16683260
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Nov 30 12:17:25 2012 +0000

    sna: Unify gen4 acceleration again
    
    After disabling render-to-Y, 965g seems just as happy with the new code
    paths as g4x.
    
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Comment 12 Till Matthiesen 2013-02-04 15:07:14 UTC
Judging from the date of the bisected commit, it does not *seem* to be a duplicate of #55500.
Comment 13 Chris Wilson 2013-02-04 15:28:20 UTC
Yes, it is. It is exactly the failure of the GPU in handling VUE that is causing it to generate spurious fragments for the pixel shader. That is the underlying gen4 bug that is behind this and bug 55500.


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.