Bug 64767

Summary: Resizing java window leaves artifacts on windows behind
Product: xorg Reporter: Clemens Eisserer <linuxhippy>
Component: Driver/intelAssignee: Chris Wilson <chris>
Status: RESOLVED FIXED QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: medium    
Version: git   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
artifacts on background window
none
window of the application with some artifacts on the left
none
artifacts on background window
none
window of the application with some artifacts on the left none

Description Clemens Eisserer 2013-05-19 18:36:19 UTC 
Created attachment 79537 [details]
artifacts on background window

When resizing the "policytool"-application quickly with the xrender java2d backend enabled (will be default for Java8), some artications of the application's window are left on the window behind it (probably some missing/buggy clipping somewhere).

Steps to reproduce:
- disable composition manager
- download the java8 runtime environment from: https://jdk8.java.net/download.html 
- untar the archive and start bin/policytool
- open some simple application (e.g. terminal)
- resize the application window on top of the terminal quickly.

What happens:
- Some elements of the application window sometimes appear on the window of the background application. Most often this is the first entry of the file-menu "File" as well as some dots, as shown in the screenshot.
Comment 1 Clemens Eisserer 2013-05-19 18:37:15 UTC 
Created attachment 79538 [details]
window of the application with some artifacts on the left
Comment 2 Clemens Eisserer 2013-05-19 18:37:42 UTC 
Created attachment 79539 [details]
artifacts on background window
Comment 3 Clemens Eisserer 2013-05-19 18:44:24 UTC 
Created attachment 79540 [details]
window of the application with some artifacts on the left
Comment 4 Clemens Eisserer 2013-05-19 19:11:59 UTC 
Forgot to mention, happens on SNB with SNA enabled
Comment 5 Chris Wilson 2013-05-20 11:08:05 UTC 
Currently testing on ILK, so far I have not been able to trigger it. Using a mixture of firefox, chromium and terminals in order to test against various types of rendering.
Comment 6 Chris Wilson 2013-05-20 11:43:49 UTC 
Repeated the tests with IVB and xfce4 (no compositing). Lots of redraw flicker, but no visible corruption.
Comment 7 Chris Wilson 2013-05-20 15:44:47 UTC 
How can I be sure that bin/policytool is using the bundled java?
Comment 8 Chris Wilson 2013-05-24 14:12:27 UTC 
Which version of the ddx? i.e. might this be bug 64841 as well?
Comment 9 Clemens Eisserer 2013-06-07 15:25:37 UTC 
I can reproduce this issue on my i5 540M system as well, running a fresh installation of the default XFCE spin of F19-Beta with the latest git ddx.

I couln't resist: http://www.youtube.com/watch?v=dObEIBMoOsE&feature=youtu.be ;)
Comment 10 Clemens Eisserer 2013-06-07 15:27:03 UTC 
you can also use the openjdk version (1.7 and higher), starting it with:

policytool -J-Dsun.java2d.xrender=True

It will print out a confirmation that its really using the xrender backend with this option enabled.
Comment 11 Chris Wilson 2013-06-12 12:28:37 UTC 
A couple of recent coherency bug fixes have landed, which could well be related to this effect. Can you please try with -intel.git?
Comment 12 Clemens Eisserer 2013-06-14 10:04:39 UTC 
Unfortunately I can still reproduce it with 2.21.9-25-g8f340f9 on core i5 540M.
Comment 13 Chris Wilson 2013-07-01 09:32:31 UTC 
Apologies if I keep sending you on a wild goose chase, but you can you test -intel-2.21.11 with the nonblocking wait fixed kernel?
Comment 14 Clemens Eisserer 2013-07-04 07:03:52 UTC 
Unfortunately I can still reproduce it with 2.21.11 + kernel-patch
Comment 15 Chris Wilson 2013-07-21 14:09:18 UTC 
How about with xf86-video-intel.git? I wonder if the changes to exposes in sna_do_copy() were the culprit.
Comment 16 Chris Wilson 2013-10-01 14:18:26 UTC 
Ping? Clemens, can we close this as unreproducible? :)
Comment 17 Clemens Eisserer 2013-10-03 16:09:30 UTC 
still there with todays git and 3.11.1
Comment 18 Chris Wilson 2013-10-04 08:33:03 UTC 
Can I blame Java? Please let me blame Java! :)

My only guess at this point is that one of the clipped render commands to the GPU is missing the enable-clip bit.
Comment 19 Chris Wilson 2013-10-30 10:26:10 UTC 
Still? (Ever the optimist that this bug will just quietly whither away.)
Comment 20 Clemens Eisserer 2013-10-30 19:15:31 UTC 
Indeed, it is fixed :)
Comment 21 Chris Wilson 2013-10-30 21:54:18 UTC 
Let's claim victory whilst we still can!

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.