Bug 25497

Summary: Scrolling in various applications causes artifacts
Product: xorg Reporter: Clemens Eisserer <linuxhippy>
Component: Server/GeneralAssignee: Xorg Project Team <xorg-team>
Status: RESOLVED MOVED QA Contact: Xorg Project Team <xorg-team>
Severity: major    
Priority: highest CC: aros, auxsvr, jan, jbeh, keithp, virtuousfox
Version: git   
Hardware: All   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
screenshot
none
eclipse problems
none
garbage caused by KDE's startup animation, when scrolling was done.
none
Screenshot with NVIDIA binary drivers
none
Scrolling with Intel drivers 2.12/kernel 2.6.34
none
Scrolling in Xephyr (X.org server 1.8)
none
Adobe Reader 9 scrolling artifacts
none
r300g & 2.6.37 - akregator scrolling issue.jpg none

Description Clemens Eisserer 2009-12-07 12:00:00 UTC
Created attachment 31825 [details]
screenshot

When scrolling in firefox, with another window on top of it, firefox doesn't seem to repaint the areas correctly that have been covered by the window on top.
Please see the attached screenshot, it illustrates the problem quite well.

I don't use a composition manager.

My System:
- i945GM
- Fedora 12 (+updates)
Comment 1 Clemens Eisserer 2009-12-07 12:04:20 UTC
I also experience sometimes unpainted areas with the code-autocompletion feature in eclipse - I guess its the same problem.
Comment 2 Chris Wilson 2009-12-09 03:30:49 UTC
Hmm, I always though this was a firefox issue not generating GraphicsExpose when it handles scrolling from obscured areas. At the very least I'm not sure if it is even a driver issue. See XChat for another example of how to handle scrolling badly.
Comment 3 Clemens Eisserer 2009-12-09 10:08:03 UTC
I can't reproduce both problems (firefox's and eclipe's) when using the vesa driver. I also don't see the problem with Xorg-server-1.6 + intel.
Comment 4 Clemens Eisserer 2009-12-09 10:08:41 UTC
Created attachment 31889 [details]
eclipse problems
Comment 5 Clemens Eisserer 2009-12-13 12:18:04 UTC
Another affected application is Adobe Adcorbat Reader 9.2.

Even KDE's animated application-is-starting-up "cursor" is leading to garbage on screen, while scrolling in firefox/acroread while its visible.
Comment 6 Clemens Eisserer 2009-12-13 12:19:17 UTC
Created attachment 32044 [details]
garbage caused by KDE's startup animation, when scrolling was done.
Comment 7 Carl Worth 2010-02-17 09:36:36 UTC
Chris seems to have at least seen this issue before, so reassigning...

-Carl
Comment 8 Clemens Eisserer 2010-03-03 08:14:41 UTC
still happens with 2.10.901 + xorg-1.7.5

As far as I can tell this started to appear with xorg-1.7, I remember using the same driver-version with xorg-1.6 caused no problems, but with 1.7 the repaint problems started to appear.

However, when using the vesa driver everything repaints fine. Maybe its just timing related, who knows :/
Comment 9 Artem S. Tashkinov 2010-04-29 03:50:26 UTC
I have the exact same problem and it's causing a real headache.
Comment 10 Artem S. Tashkinov 2010-04-29 03:57:18 UTC
I'm now running Fedora 13 i686

xorg-x11-server-Xorg-1.8.0-8.fc13.i686
xorg-x11-drv-intel-2.11.0-3.fc13.i686

Firefox 3.6.3 from mozilla.org

No xorg.conf

Intel Clarkdale i5 GPU

All hardware is running at its stock speeds.
Comment 11 Clemens Eisserer 2010-05-24 15:21:35 UTC
Ok, quite the same report has been filed against Radeon.

The bug seems to be XServer dependent. I just played with an older system I own - with xorg-xserver-1.6 everything works as expected, but as soon as I upgraded to xorg-xserver-1.7 (everything else left untouched) the corruptions started to appear. 

To some degree I guess this is a timing related problem, as I don't see it on an Pentium-M 1.6ghz + Radeon-Mobility M9. I'll play arround with my dual-core laptop to see which impact disabling one core or forcing lower clocks has.

I guess this problem will never see a fix :/
Comment 12 Clemens Eisserer 2010-05-24 15:23:04 UTC
Radeon Report: Bug 28235
Comment 13 Artem S. Tashkinov 2010-05-29 05:03:16 UTC
This problem doesn't occur with proprietary NVIDIA and ATI drivers, so I believe it's still Intel driver bug - it's not an X server bug, though a combination of a new X server and new open source drivers may have this problem.

Also it doesn't happen with VESA driver on any hardware.
Comment 14 Chris Wilson 2010-05-29 05:26:04 UTC
Note the two drivers with artefacts are the dri2, the ones you say disqualify this as being a core bug are not. So it could be a similar bug in EXA and UXA, or it may be a race in dri2 and damage. Or it could just be that drivers that hit fallbacks rendering firefox (VESA, nvidia, fglx) are too slow for this bug... Speculation is rather pointless, what is needed are reproducible test cases and traces.
Comment 15 Artem S. Tashkinov 2010-05-29 07:27:34 UTC
(In reply to comment #14)
> Note the two drivers with artefacts are the dri2, the ones you say disqualify
> this as being a core bug are not. So it could be a similar bug in EXA and UXA,
> or it may be a race in dri2 and damage. Or it could just be that drivers that
> hit fallbacks rendering firefox (VESA, nvidia, fglx) are too slow for this
> bug... Speculation is rather pointless, what is needed are reproducible test
> cases and traces.

A reproducible testcase would be the following.

Install Fedora 13 on a system with any Intel GPU, download and unpack the official Firefox distribution (at the time of writing that will be http://releases.mozilla.org/pub/mozilla.org/firefox/releases/3.6.3/linux-i686/en-US/firefox-3.6.3.tar.bz2 ), run this Firefox, run any other application on top of it and try scrolling any webpage.

The artifacts will be immediately visible.

I can post screenshots of a bug or any debugging information you may want.
Comment 16 Artem S. Tashkinov 2010-05-30 08:39:31 UTC
Created attachment 35956 [details]
Screenshot with NVIDIA binary drivers

I was *wrong*.

This bug is also reproducible with NVIDIA binary drivers.

See the attached screenshot.

Please, change the component to X.org server itself.
Comment 17 Clemens Eisserer 2010-06-15 12:51:16 UTC
Strange, seems really to be a race somewhere.
With intel-2.11.901 I can't reproduce this anymore, probably because rendering is now "fast enough" to not trigger the issue.
Comment 18 Artem S. Tashkinov 2010-06-26 06:34:04 UTC
Created attachment 36518 [details]
Scrolling with Intel drivers 2.12/kernel 2.6.34

(In reply to comment #17)
> Strange, seems really to be a race somewhere.
> With intel-2.11.901 I can't reproduce this anymore, probably because rendering
> is now "fast enough" to not trigger the issue.

You cannot, I can.

It is a bug in *X server itself*. I wonder why no one even cares to change the component and attach/mark as a duplicate a similar bug which happens under Radeon open source driver.
Comment 19 Clemens Eisserer 2010-06-26 06:45:39 UTC
Yes I know its a bug in X itself, thats why I said there seems to be a race somewhere. Well it happend between Xorg-Server-1.6 and 1.7 - so I guess someone will have to bisect.
Comment 20 Artem S. Tashkinov 2010-06-26 07:47:13 UTC
Created attachment 36519 [details]
Scrolling in Xephyr (X.org server 1.8)

(In reply to comment #19)
> Yes I know its a bug in X itself, thats why I said there seems to be a race
> somewhere. Well it happend between Xorg-Server-1.6 and 1.7 - so I guess someone
> will have to bisect.

Scrolling artifacts accumulate *even* in Xephyr, so probably it will not be that difficult to bisect, if someone has enough vigor and patience.
Comment 21 Artem S. Tashkinov 2010-06-26 07:51:58 UTC
Chris!

Please, change the component to X.org server itself, and "Assigned To" to maybe xorg-dev-team or Keith Packard.
Comment 22 Artem S. Tashkinov 2010-07-07 06:07:08 UTC
(In reply to comment #17)
> Strange, seems really to be a race somewhere.
> With intel-2.11.901 I can't reproduce this anymore, probably because rendering
> is now "fast enough" to not trigger the issue.

I'm now running 2.12 with kernel 2.6.34.1 and I still see this bug.

And mind that I have the fastest Intel GPU available (on CPU Intel Core i5 HD graphics).
Comment 23 Clemens Eisserer 2010-07-07 07:20:27 UTC
Sure you still see this bug - why should it have been fixed if nobody is willing to go the route of bisecting.
Going postal here won't make things better.
Comment 24 Clemens Eisserer 2010-07-07 07:24:29 UTC
*** Bug 28235 has been marked as a duplicate of this bug. ***
Comment 25 Artem S. Tashkinov 2010-07-08 02:33:47 UTC
I could have bisect this issue but I don't know if there's a bug free X org server.
Comment 26 Artem S. Tashkinov 2010-09-30 01:07:52 UTC
This bug is now almost a year old, and it's still reproducible. Sigh.

X.org devs are too lazy to even change its component:


(In reply to comment #21)
> Chris!
> 
> Please, change the component to X.org server itself, and "Assigned To" to maybe
> xorg-dev-team or Keith Packard.
Comment 27 Clemens Eisserer 2010-09-30 02:31:13 UTC
@Artem: Please don't get personal.
Comment 28 Artem S. Tashkinov 2010-10-01 02:25:15 UTC
Created attachment 39097 [details]
Adobe Reader 9 scrolling artifacts

Adobe Reader 9 also exhibits this problem.

Please, change this bug title to something more general, like "[X.org server regression] Scrolling artifacts in different applications".
Comment 29 Michel Dänzer 2010-10-01 08:37:19 UTC
(In reply to comment #28)
> Please, change this bug title to something more general, like "[X.org server
> regression] Scrolling artifacts in different applications".

Feel free to do it yourself, rather than abusing others for not doing it.
Comment 30 Artem S. Tashkinov 2010-10-01 09:12:33 UTC
Only a bug owner or bugzilla power users/administrators can do that.
Comment 31 Julien Cristau 2010-10-01 09:31:26 UTC
> --- Comment #30 from Artem S. Tashkinov <t.artem@mailcity.com> 2010-10-01 09:12:33 PDT ---
> Only a bug owner or bugzilla power users/administrators can do that.
> 
Pretty sure that's not true.
Comment 32 Artem S. Tashkinov 2010-10-10 12:15:38 UTC
(In reply to comment #31)
> > --- Comment #30 from Artem S. Tashkinov <t.artem@mailcity.com> 2010-10-01 09:12:33 PDT ---
> > Only a bug owner or bugzilla power users/administrators can do that.
> > 
> Pretty sure that's not true.

Maybe with this bugzilla permissions are more relaxed but on others my statement certainly holds true.

BTW, I've found out that applications running under Windows emulator Wine are also affected by this bug.
Comment 33 Sergey Kondakov 2011-01-15 18:10:22 UTC
Created attachment 42091 [details]
r300g & 2.6.37 - akregator scrolling issue.jpg

this and worse happens while i casually scroll pages in akregator.
ATI Technologies Inc Mobility Radeon X2300/Gallium 0.4 on ATI RV515; kernel is 2.6.37; libdrm, xf86-video-ati, mesa from git; xorg-server is 1.9.3

i don't remember if there were a time without this bug.
Comment 34 Andriy Gapon 2011-03-21 01:56:36 UTC
Just in case, can this bug be related to Bug 27627?
Comment 35 Clemens Eisserer 2011-03-21 02:00:25 UTC
No, unfourtunatly not.
Comment 36 debguy 2014-11-28 01:42:18 UTC
$ select x in cache libs ; do echo fixit ; done

1) try video card options, look at cache lines, that was needed on an intel card i had at one time for same thing.  try 'vesa' driver.  is that one ok?  if so, also check it detects ammount of memory correctly.

2) your libs are installed wrong

-------------

3) you compiled yourself, were having trouble, and failed to

a) install to fresh /usr/local (fresh, nothing in it, new tree)

b) you didn't make clean and make everything, so some .o were stale but used

4) it may not be Xorg but broken gnome libs / firefox stuff.  try re-installing those from scratch.  make sure you REAALLY remove all depends and reinstall.  (ie, possibly upto all libs firefox depends on: which happends to be everythign nearly)
Comment 37 debguy 2014-11-28 05:20:26 UTC
obvioulsy remove every trace of the firefox, including reverse deps, and reinstall

if you can scroll in gedit(1) or gv(1) or xpdf(1) then you have lib / install issues.

re-install at least firefox
Comment 38 Artem S. Tashkinov 2014-11-28 11:54:58 UTC
(In reply to debguy from comment #36)
> $ select x in cache libs ; do echo fixit ; done
> 
> ...
>

What an inane comment. This bug is perfectly reproducible in Fedora 20 stable/21 beta/latest Ubuntu under all video drivers (intel, nvidia, ati). VESA cannot even be used under modern kernels - KMS is enabled on boot.
Comment 39 Artem S. Tashkinov 2018-01-30 12:04:27 UTC
*** Bug 103753 has been marked as a duplicate of this bug. ***
Comment 40 GitLab Migration User 2018-12-13 22:22:13 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/xorg/xserver/issues/387.

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.