Screen refresh problems with nvidia cards

I also have this problem, running Intrepid with latest updates. For me, the problem is worst in Firefox. Under Metacity, everything looks fine.

I also confirm this problem. Setting status to confirmed and importance to medium.

Is everyone here using the nvidia-glx-177 drivers? Does this problem occur when using the nvidia-glx-173 drivers?

It sounds like a problem with Compiz's damage system, but it'd be worth ruling out the drivers. I don't see this problem myself.

Reverting to 173 seems to have solved the problem here. I could only test it for a few minutes so I'll update if it shows up again, but it seems like it's gone so far.

For me, reverting to -173 does not help.

Reverting to 173 solved the problem for me.

I'm having the same problem with an Intel GM965. I am using the Vesa driver, so that may be the problem. It seems like it's happening less in the last few days, not sure what has changed though.

I take back my earlier comment. -173 fixes the problem for me too.

Yes, I also confirm, this bug only happens with nvidia-graphics-drivers-177

I'm currently using 177, but when i used 173, I remeber having the same problem.

Of note, Metacity's compositor does not have any problems.

Yes Alex, previous comments already noted that Metacity is fine...

It's important to know that Metacity's *COMPOSITOR* (enable it in gconf) does not have any problems. That's the difference.

Posibly related: https://bugs.launchpad.net/ubuntu/+source/openoffice.org/+bug/271103

I ran the nvidia-bug-report tool, to see if there were something strange with my installation.
That tool scanned /var/log/messages for NVRM messages.
Several API-mismatch messeges were reported, wonder why 173.14.12 is still present in my system.

Sep 15 19:43:22 martin-laptop kernel: [ 280.996545] NVRM: API mismatch: the client has the version 173.14.12, but
Sep 15 19:43:22 martin-laptop kernel: [ 280.996546] NVRM: this kernel module has the version 177.70. Please
Sep 15 19:43:22 martin-laptop kernel: [ 280.996547] NVRM: make sure that this kernel module and all NVIDIA driver
Sep 15 19:43:22 martin-laptop kernel: [ 280.996548] NVRM: components have the same version.
Sep 15 19:45:05 martin-laptop kernel: [ 15.172429] NVRM: loading NVIDIA UNIX x86 Kernel Module 173.14.12 Thu Jul 17 18:11:36 PDT 2008

I'll attach the output from var/log/messages

I can reproduce this frequently just by switching between images in GQView that have different widths: the sides of the wider image are often left behind next to the narrower image (see attached screenshot. The nvidia driver put the black bars in on the side).

This doesn't happen with the nvidia 173.14.12 driver, only with the 177.70 driver. I'm running 2.6.27-3 amd64 and GQView 2.0.4-3.

It looks like the 177.76 update fixed the issue for me. I haven't been able to recreate the rendering problems after I updated.

I still get the corruption problems in GQView with the 177.76 driver.

Also, if I make a blank document in GIMP, make the mouse the paintbrush and move the mouse around, bits of the mouse cursor still get left randomly on the screen image with this driver.

Neither problem happens in the 173.14.12 driver.

I still get the problem with 177.76, albeit less frequently.

I can confirm that the gimp-problem that Rocko describes is still there for me, using 177.76.

I had this problem on kernel 2.6.27.3 but after the update to 2.6.27.4 it seems to have gone away. I'm using the 177 nvidia drivers on a nvidia mobile 8400gs

I'm still getting the problem, though I think it's happening less often than before.

Most GTK+ things are redrawing fine now with the exception of Firefox and webkit loading images (sometimes the image loads as black until I either scroll it off the page and back on again, or force a whole-screen refresh using Compiz' desktop cube effect)

I'm still getting a lot of issues with Qt apps - the most notable being Amarok. It's easily reproducable here by selecting a menu and skimming over all the entries. Some are left selected when they shouldn't be.

That's with a 8800GTS, 177.76 and 2.6.27-4-generic (2.6.27.4.4)

I am using the latest updates as of 12:51pm Oct-04-2008. Nvidia 177.78 Driver, Nvidia Quadro 135m and Kernel 2.6.27-4 32bit. I receive the problem with 177.70, 177.76 and 177.78. Sits at Searching for Driver and will not proceed. Once i choose cancel compiz effects are present. But screen refresh problems exist in many tabbed applications such as Firefox or Avant Window Manager Preferences.

JKing

I can confirm this problem with 177.80.
Nvidia Quadro 140m, 2.6.27-6 64 bit.
The system is fully updated at the moment.
I notice it primarily with firefox since I am not using awn or gimp at the moment.

Confirming this still happens with 177.80. In fact it seems to have gotten worse.

I agree with Nick B. above me, it is a LOT worse, actually un-useable to some extent so I will be going to 173 until I hear something is fixed with this.

Dell D630

As I posted in ubuntu forums:
I have been using intrepid since sunday and had no problems, but with the new version, every time I go to a screen in firefox that has any pictures, if I Page down, it looks as if nothing has changed, but then if I click the down button from there the screen will appear (or if I roll over any links that change colour with the mouse, they will appear).

For me this bug is present in 177.80 in exactly the same way as with the previous 177.76 drivers. (Dell XPS 1530 laptop, NVidia mobile 8600 card.)

The easiest way is to confirm it is to run GIMP, create a blank drawing, select the drawing tool so you get a circular cursor, and move the cursor around a blank drawing. Bits of the cursor remain behind on the blank drawing.

GCView also has the same problem as I reported before - it sometimes only redraws the new image, leaving parts of the old behind.

Occasionally I get other problems like Firefox not refreshing properly but they are relatively rare.

It seems to have gotten worse on my hardware. I didnt even notice it until i switched to 177.80.

Confirm it on a Dell XPS M1330 with the Nvidia 177 drivers.

If I enable 'scale' plugin in compiz, and then 'scale' by moving the mouse pointer to the edge of the screen, the firefox (and possibly other) windows refresh.

This sounds a lot like this problem, which is unfortunately a fundamental problem in the way the Damage extension is currently designed: http://lists.freedesktop.org/archives/xorg/2008-August/038022.html

The quick summary is that the X server sends damage events to compiz before sending the rendering commands to the hardware, and compiz can respond to those events before the damage has actually occurred. Some solutions have been proposed, but nothing has been implemented yet.

I see different refreshing problems with intrepid:
- Everytime a new window or menu opens, for a short time I see artifacts of the background image or an old window.
- When viewing pictures with gthumb or gqview in fullscreen mode, the old image ist left behind when drawing a new one (even appears sometimes later when opening a new window, see above).
- In Ooo and Evolution while writing text or a new email, when I manually move the cursor with the arrow keys it leaves black bars on every position. The screen is refreshed correctly when i point and click somewhere else with the mousepointer. Especially this is really annoying!

I noticed one interesting thing: when I start compiz manually on startup, using FusionIcon, these problems do not occur. When compiz is started by ubuntu i.e. the visual effects settings, i get refresh problems. After disabling them by visual effect settings and reactivating them again, I still experience artifacts. But when i use FusionIcon, all refreshing problems are gone. I could not figure out, what the difference between starting compiz via FusionIcon or by visual effects settings is.

I confirm this on Dell XPS M1330 with nvidia 177.80. Running Intrepid with latest updates.

I confirm this on Asus Z53S with nvidia 177.80 driver installed with envy. I try the driver proposed by ubuntu, same result. 173 driver too....
I use Intrepid with all update.
please help :(

Same problem with Dell XPS M1330 in intrepid.
I have problem with Firefox and emesene.

This seems to have gotten a lot worse with the latest set of nvidia updates, Firefox now doesn't update properly whenever I open a tab, this used to only happen occasionally.

Same problem for me : https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/281065
Ubuntu Intrepid 64 bits beta + nvidia 177

With an Nvidia mobile 8400GS (on a Dell XPS M1330 running Intrepid 64-bit), it happens for me with both the 177 and 173 drivers.

I have almost a 100% reproducibility rate in CompizConfig Settings Manager (clicking a plugin to show its settings), but the worst thing is Firefox, since I use it more often...

As I said in bug 281065, disabling --loose-bindings in compiz avoids the problem. But this is not a solution, it slows down the perfs...

I am also experiencing this problem.
Dell XPS M1330, geforce 8400gs

Especially using firefox on ubuntu is extremely annoying, but _all_ applications are affected.
I think the importance of this bug is higher than Medium.

I can confirm that running without --loose-bindings fixes the problem. (A web site I'm working on with lots of mouse-over/hover actions seems to have 100% reproducibility of this problem.)

I disabled --loose-bindings by commenting out lines 340-344 in /usr/bin/compiz (which really is a shell script). This is not recommended unless you know what you are doing...

Unlike ®om, I don't notice any significant slowdown, though. Then again, I run compiz with a quite small number of plugins and animations, I mostly just use it to get the GPU to do stuff it does better than the CPU (like moving windows), and as a bonus, I get more eye candy in the window decorations (using emerald). ;)

A more simpler fix in /usr/bin/compiz is to change line 80 from
  INDIRECT="no"
to
  INDIRECT="yes"
(This change also disables --loose-bindings but at the same time enables --indirect-rendering)

For me, this seems to give the same performance as the change in my last comment.

For the moment, I use your first solution (commenting lines 340-344), because --indirect-rendering causes compiz to not use nvidia-settings config (anisotropic filter) even if you launch 'nvidia-settings -l' before.

Both changes work fine for me. I don't notice any drop in performance in normal use, though. The only measure I can find that shows a slight drop is glxgears, where the drop is maybe 5% (compared to an increase of 40% if I don't run compiz at all).

Changing Line 80 to INDIRECT="yes" helps for me, too.

INDIRECT="yes" is equivalent to remove --loose-bindings and to add --indirect-rendering

removing --loose-bindings avoids the problem
adding --indirect-rendering seems useless, and add another problem

To see the problem, launch nvidia-settings, enable anisotropic filter, then compiz --replace, and see the quality when looking at a cube face from 70° (no filter at all is applyed).
Do the same test without --indirect-rendering : it's far better :)

I would like to add it worked fine since 2~3 days ago, a daily update (intrepid beta) broke something.

That would most likely be when you updated to the Nvidia 177.80 driver.

I confirm what ®om said, it was temporary working properly before recent updates.

OK, if it's not resolved, I hope nvidia drivers will be downgraded for final release?

Does everyone use Compiz here ?

I have the refresh problem on firefox while I'm NOT using compiz.
I use only the NVidia binary driver.

You should try to see if it changes something when desactivating the desktop effects.

If the problem remains then it would mean something is very wrong in the latests drivers.

Corsaire : desktop effects are compiz, so desktop effects enabled = compiz enabled

Sorry you can ignore my last comment.

I have seen it happen without Compiz enabled. But it very rarely happens compared to with Compiz enabled. It seems Compiz amplifies the bug much more.

®om: I don't think that downgrading a stable driver to a beta release would be such a good idea.

Maybe downgrading to 173?

I don't think it's a good idea to release a "stable" version of Ubuntu with such a bug... which has an importance more than medium... the system is unusable out-of-the-box (once desktop effects enabled).

Removing --loose-bindings avoids in majority the problem, but this morning, even without --loose-bindings I had the problem once.

Downgrading to 173 isn't a solution either. I get no noticeable differences at all between 173 and 177 on my 8400GS.

®om: Ubuntu doesn't ship with the NVIDIA driver preinstalled therefore it's not true that the system is unusable out-of-the-box. Furthermore you can choose between 173 and 177 and the problem doesn't affect all NVIDIA cards. According to this report Geforce 8xxx are affected. Maybe we could simply blacklist the identifiers of the cards which have this problem.

Can all the users who can experience this problem post the following details?
1) your Graphics card model
2) the output of this command:
lspci -n | grep 300

$ lspci -n | grep 300
01:00.0 0300: 10de:0425 (rev a1)

I use Ubuntu 64 bits, but I think the problem occurs even in 32 bits, could you confirm?

$ lspci -n | grep 300
01:00.0 0300: 10de:040c (rev a1)

Running a Quadro 570m

I had the problem as seen in the bug #269904 using a GeForce 8800GT, and I still have it now that I have a GTX260.

$ lspci -n | grep 300
01:00.0 0300: 10de:05e2 (rev a1)

I use Ubuntu 8.04 64bit with vanilla NVIDIA 177.13 drivers (not some .deb pre-made packages), but at the time I filled the bug #269904 it was much more like a 173.* release so the problem is fairly old.

@ Saïvann Carignan : Yes I'm perfectly aware of this. This is not the issue here.

$ lspci -n | grep 300
01:00.0 0300: 10de:0160 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 6500 (rev a1)

Problems for me are mainly compiz-ised window title bars becoming obviously visually corrupt. Problems occur with both 177 and 173 drivers (did not test earlier ones).

$ lspci -n | grep 300
02:00.0 0300: 10de:0402 (rev a1)

$ lspci | grep VGA
02:00.0 VGA compatible controller: nVidia Corporation GeForce 8600 GT (rev a1)

$ lspci -n | grep 300
01:00.0 0300: 10de:0407 (rev a1)

$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8600M GT (rev a1)

$ lspci -n | grep 300
01:00.0 0300: 10de:0429 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation Quadro NVS 140M (rev
a1)

$ lspci -n | grep 300
01:00.0 0300: 10de:0427 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M GS (rev a1)

I've been playing around with the settings and I noticed that if you *do not* set this:

nvidia-settings -a InitialPixmapPlacement=2 -a GlyphCache=1

it greatly reduces the problem. It doesn't eliminate the problem but at least things are usable.
Has everyone else having this problem been using that setting?

This problem seems to be acknowledged by Nvidia
"Please note: it has been reported that using a compositing manager in conjunction with this option will cause newly-created windows to briefly display random contents instead of black before being first drawn. We are aware of this non-critical issue and planning to fix it in a future NVIDIA Linux graphics driver release."
It doesn't mention "refresh" problems specifically but it seems to be in the same category as this.

lspci -n | grep 300
01:00.0 0300: 10de:0427 (rev a1)
lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M GS (rev a1)

$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 9600M GS (rev a1)
$ lspci -n | grep 300
01:00.0 0300: 10de:0648 (rev a1)

$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8600M GS (rev a1)

$ lspci | grep 300
01:00.0 0300: 10de:0425 (rev a1)

I switched to from nvidia 177 to nvidia 173, no problem now :)

At least a viable solution for me :)

$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation Quadro NVS 140M (rev a1)
$ lspci -n | grep 300
01:00.0 0300: 10de:0429 (rev a1)

177 was completely unusable
173 is much better, but i sometimes still get the problem

I decided to try -173 once again and, strange enough, it seems to work all of a sudden... (Yes, I did reboot after my previous failures.)

Although, this time, I did an 'apt-get remove --purge compiz-core compiz-wrapper' and also noticed that the USE_EMERALD variable had moved (?) to /usr/bin/compiz-decorator. Until now, I had started emerald by setting '/usr/bin/emerald --replace' as window decorator in ccsm (which means that gtk-window-decorator starts first, then, before you'll have had time to start any windows, emerald immediately takes over...

Will keep my eyes open for weirdness.

I modified compiz to use XDamageReportNonEmpty instead of XDamageReportRawRectangles. It's nowhere near a complete fix for the problem I mentioned earlier, but for my laptop at least, it's enough to work around the problem.

I figured out how to put together a PPA build: https://launchpad.net/~aplattner/+archive

Olá Alberto e a todos.

On Sunday 12 October 2008 10:26:37 Alberto Milone wrote:
> Can all the users who can experience this problem post the following details?
> 1) your Graphics card model

NVIDIA FeForce 8400m G

> 2) the output of this command:
> lspci -n | grep 300
01:00.0 0300: 10de:0428 (rev a1)

--
BUGabundo :o)
(``-_-´´) http://LinuxNoDEI.BUGabundo.net
Linux user #443786 GPG key 1024D/A1784EBB
My new micro-blog @ http://BUGabundo.net

Olá Aaron e a todos.

On Monday 13 October 2008 06:46:53 Aaron Plattner wrote:
> I figured out how to put together a PPA build:
> https://launchpad.net/~aplattner/+archive

I have Version: 1:0.7.8-0ubuntu2 and your is compiz - 1:0.7.4-0ubuntu8~ppa1 so it wont auto-install.
Can you bump it?

--
BUGabundo :o)
(``-_-´´) http://LinuxNoDEI.BUGabundo.net
Linux user #443786 GPG key 1024D/A1784EBB
My new micro-blog @ http://BUGabundo.net

I've noticed, that selection "Reload Window Manager" option in Compiz
Fusion Icon fixes problem for me.

Effectively, this problem happens with nvidia 173 too, but quite rarely (while with 177 it's very very often).

I used to have this problem too, but, strictly speaking, I can no longer confirm it because I had to switch Compiz off after the latest updates because it no longer draws window decorations.

Anyway, here is the requested information for my card:

tawmas@tylke:~$ lspci -n | grep 300
01:00.0 0300: 10de:0421 (rev a1)

tawmas@tylke:~$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8500 GT (rev a1)

I came here from a bug duplicate, because I had only noticed the problem in firefox. I dont seem to have the same issues in gedit, (those are about the only tools I use!) and it certainly doesn't affect my winXP running on virtualboxOSE.
This has only started since yesterday, (iv'e been doing regular updates) and it's really annoying, because firebug is next to useless now.

8.10 intrepid
2.6.27-7
nvidia 177 driver

:~$ lspci -n | grep 300
01:00.0 0300: 10de:0427 (rev a1)
:~$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M GS (rev a1)

Just read through the comments here in a bit more detail, and I reckon I have actualy had this problem for a while. Bits of cursors hanging around, etc. But it has got ALOT worse with the latest upgrades. Firefox is getting really annoying.

BUGabundo wrote:
> I have Version: 1:0.7.8-0ubuntu2 and your is compiz - 1:0.7.4-0ubuntu8~ppa1
> so it wont auto-install. Can you bump it?

Yeah. Intrepid builds uploaded.

$ lspci -n |grep 300
01:00.0 0300: 10de:0427 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M GS (rev a1)

a little note:
i have this problem with compiz
and under metacity
i think it's a problem of driver, if i haven't it it's all ok..

$ lspci -n |grep 300
01:00.0 0300: 10de:0425 (rev a1)

$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8600M GS (rev a1)

I have this problem as well: Firefox can get pretty irritating.
I think it deserves an importance of more than "Medium".

$ lspci -n |grep 300
01:00.0 0300: 10de:042b (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation Quadro NVS 135M (rev a1)

I had this problem, most easily reproducible switching between firefox tabs.

After trying the fix at:

https://launchpad.net/~aplattner/+archive

things seem to be working well now. As BUGabundo pointed out it would help to up the package version there, since trying this fix otherwise requires removing/reinstalling the older compiz packages, and adding a pin-priority entry in /etc/apt/preferences to do it the sources.list way.

kiomava wrote:
> things seem to be working well now. As BUGabundo pointed out it would
> help to up the package version there, since trying this fix otherwise
> requires removing/reinstalling the older compiz packages, and adding a
> pin-priority entry in /etc/apt/preferences to do it the sources.list
> way.

It was right as of yesterday. It looks like 1:0.7.8-0ubuntu3 was released
earlier today. I'll build an updated version soon.

-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------

Aaron Plattner wrote:
> [legalese boilerplate inserted automatically by mail server]

D'oh. Please disregard that.

$ lspci -n |grep 300
01:00.0 0300: 10de:0611 (rev a2)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8800 GT (rev a2)

I've noticed this happening more with QT3/4 (KDE) applications, but on occasions I have a problem with Firefox not refreshing web pages or refreshing text boxes (particularly annoying when you are editing a Wikipedia article and you don't know where you are).

Currently using 177.80 drivers with Compiz 0.7.8.

$ lspci -n | grep 300
01:00.0 0300: 10de:0427 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M GS (rev a1)

Same here.

$ lspci -n | grep 300
01:00.0 0300: 10de:0407 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8600M GT (rev a1)

I don't think anyone should blacklist any nvidia cards - that happened to ATI cards just before the Hardy release and it was awful - compiz just stopped working without any explanation. And that was because in a small number of cases it froze the kernel. This bug is not like that one at all, it's just annoying and there are workarounds that several people have confirmed do work that can be done (like disabling compiz, installing the 173 driver, modifying the /usr/bin/compiz script).

Instead, if it can't be fixed before the final release, I think it would be much better if compiz warned there might be problems when you try to enable it for affected nvidia cards.

Rocko: I suggested that before Aaron came up with a fix.

Aarons packages seems to solve the problem better than -173 for me. With -173 I had some corruption when switching context manager tabs in Amarok (KDE based). I fail to reproduce those problems now.

$ lspci -n | grep 300
01:00.0 0300: 10de:0428 (rev a1)
$ lspci | grep VGA
01:00.0 VGA compatible controller: nVidia Corporation GeForce 8400M G (rev a1)

Thanks for posting your ids (with lspci) but this is not required any more.

I think I spoke too soon. I experienced some quite bad corruption with a maximised gnome-terminal running VIM (over SSH on a Debian machine, with syntax highlighting editing a PHP file).

However, again reverting to -173 but still using the fixed compiz, I only experienced very minor corruption, which occurs very seldom while quickly switching tabs in Amarok. Well enough fix to live with, anyway.

@Alberto and all users

- Some descriptions sounds like "RGBA trashing", maybe to disable Murrine and test ? or another version ?
Using Murrine SVN myself and some apps trashes..

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

- Dell users, BIOS version or maybe warranty cases. Also HP got this problem with weak materials.

http://direct2dell.com/one2one/archive/2008/07/25/nvidia-gpu-update-for-dell-laptop-owners.aspx

http://direct2dell.com/one2one/archive/2008/08/18/nvidia-gpu-update-dell-to-offer-warranty-enhancement-to-all-affected-customers-worldwide.aspx

Every model must be checked. Just to visit Dells support !

No, I would say this is different from RGBA trashing. This happens regardless if you're using a GTK engine that supports RGBA. Also not all apps support RGBA yet anyways.

I've actually been able to minimize the problem by 80% or so by not setting the InitialPixmapPlacement and GlyphCache settings.

I tried the commenting out of the 5 lines in /usr/bin/compiz and that seemed to work for a while but it does not completely fix the problem for me. It happens about 3 times a day now instead of once every 5 minutes, but it still happens. I have not tried any of the other fixes.

Also, when my screen does not refresh, I can click anywhere on the screen and it will refresh everything and draw everything as it should be at that time.

As suggested by (R)om, remove --loose-binding in /usr/bin/compiz seems to be a
workaround for me that solves the problem. I haven't got any refresh issues
since I changed that.

@tseliot, what action do you think is needed in order for this bug to close? Anything I could assist with?

Bryce: I think that the only thing we can do is adopt Aaron's patch for Compiz. Here's his PPA:
https://launchpad.net/~aplattner/+archive

Maybe either you or Michael (mvo) can review his patch (see the attachment)?

Alberto, thanks for pointing to the patch and the quick reply.

I've reviewed the patch, although I'm going to defer to Michael on whether it's acceptable. I don't see anything wrong with the patch, although I'm not so familiar with compiz internals, and since the patch involve alterations to a realloc() and introduction of some arrays, it's tough for me to be certain just from eyeballing it. Michael, your thoughts?

Not to be pessimistic, but has anyone tried the patched compiz on an ATI card?

Removing the "loose binding" option has slightly decreased the frequency of this issue, but it is still here...

I enabled "loose binding" a few days ago, and back then it seemed to reduce the amount of artifacts left on the screen.
But now, its about the same as it was before.

Alberto should it be enable or not?

This bug was fixed in the package compiz - 1:0.7.8-0ubuntu4

---------------
compiz (1:0.7.8-0ubuntu4) intrepid; urgency=low

  * debian/patches/049-damage-report-non-empty.patch:
    - add workaround patch by Aaron Plattner to fix
      problems with the damage extension and refresh
      (LP: #269904)

 -- Michael Vogt <email address hidden> Fri, 17 Oct 2008 09:29:01 +0200

Since we have a workaround in compiz now, the nvidia bug (or the Xserver one, whereever the root cause is) is not intrepid RC any more. Thank you!

I'm afraid the workaround is ineffective, for me.

I have also installed the packages from the PPA and still have problems.
(nVidia Quadro NVS 140M)

Since the last updates in intrepid repositories, I reactivated the drivers 177 (I switched to 173 due to this problem), and it seems to work... for the moment.

This bug has not disappeared completely, however the reproducibility factor has gone down to <10%. My setup is a Dell M1330 GF 9400M GS. Disabling loose bindings did more or less the same result for me. Now i'm running with: compiz.real --ignore-desktop-hints --replace --loose-binding core ccp

However, if I scroll for past commands (arrow-up) in gnome-terminal the refresh bug shows by failing to update the text field when vsync is enabled. Can someone confirm this?

Otherwise this fix has worked flawlessly. Great work! :)

I've been using the aplattner-patched compiz version for a few days now, no compiz tweaks (no messing with loose bindings or indirect rendering or similar), updating to the latest intrepid at least daily. I also haven't messed with nvidia-settings, as I have done in the past with Hardy. I have apt set up with a pin priority to avoid accidentally updating to an unpatched compiz. I use terminal all the time, and firefox, among other things, and haven't seen any of the refresh problems I saw before. I did see very frequent problems in terminal before the patch, especially it seemed when scrolling/paging and with the cursor at the bottom of the window.

But no refresh/damage-related problems at all after that patch, and I've been trying to push it the last few days. More details about my setup:

Nvidia 177.80 driver, installed using administration/hardware drivers thing (not directly from nvidia installer or envy)
2.6.27-7-generic #1 SMP Tue Oct 14 18:40:44 UTC 2008 i686 GNU/Linux
Dell D830
nVidia Corporation Quadro NVS 135M (rev a1)

nvidia-settings InitialPixmapPlacement is 1 (the default)
nvidia-settings GlyphCache is 0 (also default)

Also someone mentioned about nvidia hardware problems being important to consider, but since I boot to XP and Hardy and don't see problems I don't think that's a factor for me.

since the last update and restart it appears to be fixed for me.
Quadro 140M, Lenovo T61.
tamoneya@lithium:~$ lspci -n | grep 300
01:00.0 0300: 10de:0429 (rev a1)
driver 177.80 installed through hardware drivers in ubuntu.
I dont see any rendering problems although compiz appears a little sluggish when rotating the cube. Barely noticeable though.

I've been updating at least twice a day recently because this problem really has been a big pain. If I'd known what using intrepid entailed, I probably would have stuck with hardy.
AAaanyway, the latest update included compiz 1:0.7.8-0ubuntu4
Kudos to Aaron and everyone who's been involved, It seemed to work, for about 30 seconds, and then firefox crashed. I hope this was a one off, all seems OK now, certainly better.
Firefox was certainly the worst affected for me (this got loads worse before it got better) Only been testing for a few minutes, but it seems a really screwed up computer has become a usable one.
Thanks

yup the latest updates fixed it for me too!!!

The latest compiz updates fixed it for me. But I've gone back to the 173.14.12 driver because the difference in fps in glxgears is large - I only get 3000 fps with 177.80 versus 4400 with 173 (and in Hardy I get even better- 4800 with 173), and I assume that translates to slower game speed with 177.

have the same problem on my 5200 AMD64/x2 with Geforce 8800gt

Especially with Eclipse (JAVA-IDE)

--------------------------------------------------------
nVidia Quadro NVS 135M (on a Dell D630)
01:00.0 0300: 10de:042b (rev a1)
--------------------------------------------------------

I'm still having this problem on Intrepid Beta; Firefox is incredibly annoying. On pages with flash, the flash will show up overlaying the previous tab I was viewing... until I click the mouse somewhere within the Firefox window (including the titlebar). I'm using gtk-window-decorator with the latest compiz from the above-mentioned PPA (latest as of 20 Oct 08).

It also happens extremely frequently with Konsole -- I can't change directories or ssh or scroll (or many other things) without getting my view clobbered. Again, I have to click somewhere in order to refresh the window. Konsole is almost unusably annoying.

I'm willing to help in any way.

Unfortunately I am very much still experiencing these problems with the 173 drivers and all the latest updates. The problem is most visible in Lyx and gvim.

Hey, JoeLinux, which compiz versions do you have exactly? We have identical cards and nearly identical laptops D830 vs D630, and I used to see the firefox problems you describe but not any more.

$ dpkg -l |grep compiz
ii compiz-core 1:0.7.8-0ubuntu4~ppa1 OpenGL window and compositing manager
ii compiz-fusion-plugins-extra 0.7.8-0ubuntu2 Collection of extra plugins from OpenComposi
ii compiz-fusion-plugins-main 0.7.8-0ubuntu2 Collection of plugins from OpenCompositing f
ii compiz-gnome 1:0.7.8-0ubuntu4~ppa1 OpenGL window and compositing manager - GNOM
ii compiz-plugins 1:0.7.8-0ubuntu4~ppa1 OpenGL window and compositing manager - plug
ii compiz-wrapper 1:0.7.8-0ubuntu4~ppa1 OpenGL window and compositing manager, wrapp
ii compizconfig-backend-gconf 0.7.8-0ubuntu1 Settings library for plugins - OpenCompositi
ii compizconfig-settings-manager 0.7.8-0ubuntu3 Compiz configuration settings manager
ii libcompizconfig0 0.7.8-0ubuntu2 Settings library for plugins - OpenCompositi
ii python-compizconfig 0.7.8-0ubuntu1 Compiz configuration system bindings

$ uname -a
Linux kiomava-d830 2.6.27-7-generic #1 SMP Fri Oct 17 22:24:21 UTC 2008 i686 GNU/Linux

$ dpkg --list | grep compiz | awk '{print $1" "$2" "$3}'
ii compiz 1:0.7.8-0ubuntu4~ppa1
ii compiz-core 1:0.7.8-0ubuntu4~ppa1
ii compiz-fusion-plugins-extra 0.7.8-0ubuntu2
ii compiz-fusion-plugins-main 0.7.8-0ubuntu2
ii compiz-fusion-plugins-unsupported 0.7.6-1
ii compiz-gnome 1:0.7.8-0ubuntu4~ppa1
ii compiz-plugins 1:0.7.8-0ubuntu4~ppa1
ii compiz-wrapper 1:0.7.8-0ubuntu4~ppa1
ii compizconfig-backend-gconf 0.7.8-0ubuntu1
ii compizconfig-settings-manager 0.7.8-0ubuntu3
ii emerald 0.7.2-0ubuntu2
ii libcompizconfig0 0.7.8-0ubuntu2
ii libemeraldengine0 0.7.2-0ubuntu2
ii python-compizconfig 0.7.8-0ubuntu1

Unfortunately, it still isn't working for me, as Konsole continues to appear very "affected" in this respect. I'm using gnome-terminal right now, as the effects on it are less intense than Konsole (almost non-existent). Is the Aaron's PPA supposed to be seen as a downgrade by the package manager? Because I had to manually download the packages and install them myself instead of using the repo, and it complained that I was downgrading the packages. I'd previously had the Compiz PPA packages (deb http://ppa.launchpad.net/compiz/ubuntu intrepid main) installed.

@kiomava: Looks like we have the same package versions... except I have emerald installed, not that it should matter (I am assuming, of course).

JoeLinux:

I turned on emerald, using GrayDarkIce theme, and it doesn't seem to be any different. No terminal/tab switching artifacts. So now our package versions match exactly. xorg.conf maybe? I'll attach mine. These might be relevant:

 Load "glx"
 Disable "dri2"

Are you using KDE? Maybe that's the difference, I'm using Gnome.

I'm also using Gnome... so there shouldn't be a difference there. Either way, I think Konsole would be using the KDE libs instead of the Gnome libs, right?

Anyway, I'll attach my xorg.conf.

I have a D630:

$ dpkg --list | grep compiz | awk '{print $1" "$2" "$3}'
ii compiz 1:0.7.8-0ubuntu4
ii compiz-core 1:0.7.8-0ubuntu4
ii compiz-fusion-plugins-extra 0.7.8-0ubuntu2
ii compiz-fusion-plugins-main 0.7.8-0ubuntu2
ii compiz-gnome 1:0.7.8-0ubuntu4
ii compiz-plugins 1:0.7.8-0ubuntu4
ii compiz-wrapper 1:0.7.8-0ubuntu4
ii compizconfig-backend-gconf 0.7.8-0ubuntu1
ii compizconfig-settings-manager 0.7.8-0ubuntu3
ii libcompizconfig0 0.7.8-0ubuntu2
ii python-compizconfig 0.7.8-0ubuntu1

And I am experiencing the artifacts as well. This is extremely frustrating - gnome-terminal, firefox and editor (komodo) are all causing me great suffering.

I'm pretty sure it's not the nvidia bug plauging us - I just had my mobo replaced and the symptoms of the failed board where MUCH more pronounced than these - and very, distinctly, different.

So I did end up seeing the refresh problems while using the aplattner-patched compiz and Emerald. After turning Emerald back off, no refresh problem so far. I wonder if this is some timing thing and I am just lucky when I have emerald turned off.

fimbulvetr, did you try the aplattner-patched compiz? It looks like you don't have it installed.

@kiomava: aplattner's patched compiz made it into the main repositories (see https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/269904/comments/109). fimbulvetr's version does contain the patches. The only differences between ...ubuntu4 and ...ubuntu4~ppa1 should be the build date if I'm not mistaken.

Anyway, since I installed the patched compiz _and_ nvidia-glx-173, I haven't seen any problems at all (using an GeForce 8400M GS) for about a week with daily use. :)

I tried downgrading to 173 and while it appeared to work alright, I couldn't stay on the version because it appeared to have issues with nvidia's twinview settings, which rendered it unusable for me.

I'm still seeing this using metacity and 177

Finally, I still have some refresh problems (3 times since this morning), in xchat, in firefox and in systray.

Yes, I still have some refresh problems, too, although they are less frequent than before the latest update.

I have a multimonitor configuration with nVidia Quadro140 (Dell Latitude D830) and have solved the problem setting the same vertical resolution for the monitor, using TwinView option.
This is just a workaround, I had to force the notebook panel to work @1024 while its real vertical resolution is 1050.

the problem it's solved for me, but the refresh is very very very too bad ..... :(

I haven't noticed the bug for a while, so i would say its fixed for me.

On Thu, Oct 23, 2008 at 2:00 PM, enubuntu <email address hidden> wrote:

> the problem it's solved for me, but the refresh is very very very too
> bad ..... :(
>
> --
> Screen refresh problems with nvidia on intrepid
> https://bugs.launchpad.net/bugs/269904
> You received this bug notification because you are a direct subscriber
> of a duplicate bug.
>
> Status in "compiz" source package in Ubuntu: Fix Released
> Status in "nvidia-graphics-drivers-177" source package in Ubuntu: Confirmed
> Status in compiz in Ubuntu Intrepid: Fix Released
> Status in nvidia-graphics-drivers-177 in Ubuntu Intrepid: Won't Fix
>
> Bug description:
> Binary package hint: compiz
>
> I'm using latest compiz on intrepid.
>
> Since jockey-gtk does not work on my box (I already filed a bug), I
> manually installed nvidia-glx-177 and added the "driver → nvidia" line to
> xorg.conf.
>
> Now I seem to be having some screen refresh problems when using compiz:
> parts of the screen are refreshed, but others aren't. That means nautilus
> doesn't show files or folders until I move the cursor over them, firefox
> doesn't show webpages properly until I scroll down, and so on.
>
> Please, tell me if there is any additional info I can provide to find out
> the cause.
>

It was totally fixed for me from 17 october (my last message in this bug report) until today. Today, I experienced several refresh problems. It far better than before the "fix", but I think there was a regression in the last upgrade.

The problem is not resolved at all for kde apps (kile for example), many refresh problems occur (running under gnome).

It's confirmed that the problem is still not fixed in nvidia drivers, so it's not necessary to have more than two confirmations. This will help the bug report to stay productive. Thanks for your implication with ubuntu!

Hi

I've encountered same problems with Hardy !
In xterm, it's a pain...

nvidia drivers 177.80
compiz 1:0.7.4-0ubuntu7

VGA compatible controller: nVidia Corporation Quadro NVS 135M

I'll comment lines 340-344 in /usr/bin/compiz and come back to give a feedback.

Having this problem too
Dell XPS 1710 nVidia 7950GTX 177.80drivers

Intrepid with latest git build of compiz.

Downgraded to 173 drivers and less frequent issues on Konsole, but same issues in firefox.

It happens without Compiz at all.. Well, maybe not 100% same, but just turning on Metacity's lazy window selecting produces the same behavior..

Sorry - didn't realize intrepid was enabling compiz on it's own now.. That kinda sucks - it has not been stable yet...

Got the issue here after upgrading from Hardy to Intrepid. Weird corruption of window titlebars when moving mouse over them. More specifically, seems to trigger when moving over window control icons at top right corner of windows (minimize etc). Have not seen the issue exhibit elsewhere in the system, but only upgraded today so early days.

Video card is FX5500 256mb. Driver is NVIDIA Accelerated Graphics Driver v.173

I am not on that computer, but came here looking for solace that I am not crazy... that I did not jack my system. I am running a 7600GS and I am seeing the following problems:

  - Window title bars not fully skinning, then when I move the mouse over the controls (Min, Max, Close) it instantly re-renders properly. Sometimes the title bars do not show at all.

  - Firefox opens with no window decorations at all and full screen (almost kiosk mode except it still has the launch and address bars). Pressing F11 twice forces it to render correctly.

  - AWN (ppa version), OMG what a mess! I get this odd white box over the desktop switcher, the icons do not always render correctly, and every so often, I get the scrambled box. All the issues disappear by moving the mouse over the affected area except the white box. The white box does go behind the switcher, and in front of maximized programs obscuring the status bar at the bottom of many programs (i.e. the Firefox loading bar is 1/2 obscured).

This problem does not disappear with 173. Both 177 and 173 seem equally bad, but I did not do any type of actual testing to compare the two, this is just my empirical experience.

OOPS!!

This problem DOES occur with 173!

Also should have mentioned, the only computer which exhibits this behavior is the one with the nVidia card. I also run a Dell D620 with the Intel chipset with no display issues (network yes, but that is a different bug, but no display issues). Also, I run a Sony Vaio BX540B also with an Intel 915.

Hi,
this is my configuration:
* GeForce 8400 GS (rev a1)
* Kubuntu intrepid amd 64
* kernel 2.6.27-7-generic
* Driver: 177.80

After reading https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-177/+bug/289964 I've changed to the nv driver in xorg.org and firefox-3.0 now renders ok.

Anyway the problem now is that the screen shows a fast flick every 10 seconds. Not so annoying as being unable to use firefox, but...

Further to my report, I now have an easy fix for my occurrence...

To recap: with Intrepid and my nVidia FX5500/256mb card, I was experiencing window titlebar corruption when mousing over the righthand controls (minimize, close, etc.). Moving the mouse cursor over the controls would corrupt half the titlebar; moving the mouse again over them would restore a correctly rendered titlebar. This is the full extent of the window-drawing corruption for me: just in the window titlebars.

I experimented with switching to different Window Borders...

SYSTEM > PREFERENCES > APPEARANCE > THEME > CUSTOMIZE > WINDOW BORDER

Sure enough, when I switched away from the default Window Border of HUMAN, e.g. to CLEARLOOKS, the titlebar rendering problem ceased. From my tests, it looks like the default HUMAN Window Border is the only border theme that triggers the rendering fault (for my setup, anyway).

Anyone else able to solve their issue doing the same??

Cheers,
Belzecue.

I've noticed nvidia released a new version of the driver yesterday (177.82) and seems to address that problem, at least there's "Fixed an image corruption issue seen in FireFox 3" in the description. I didn't try that - I'm waiting to have it from ubuntu repos. If anyone had a chance to see if it fixes the problem - I'd be glad to hear some news.

Regards,
Roman

Roman, I tried the new nvidia driver (177.82) and cannot reproduce the error anymore. It seems to have fixed the issue for me.

Cool,

Now, the $64,000,000.00 Question... When will this make its way into the Intrepid Ibex repositories. I have stopped using my home computer except in case of me forgetting to bring home my laptop because it is so screwed up (especially firefox)

Kevin

I've installed the new driver (177.82) as well and it didn't fix the problem for me (my problem concerns almost exclusively Konsole - changing tabs, ls -l, and basically every other action requires a mouse click to be seen; with Firefox i've observed this problem maybe 3 times).

same for me here. With a current ibex installation the refresh problem is still there. Any upstream package that I could try? Has this error been confirmed to be really an nvidia bug?

This *was* an issue for me, but now it's fixed. This report is clearly a metabug for more than one issue, then. (Unsubscribing.)

@ Belzecue wrote on 2008-11-05:

I do not remember having any other (eg. Firefox) problems but changing the window border scheme solved the window title bar rendering error for me, so I can confirm this part. AMD64 w/32-bit Ubuntu, nVidia Corporation C51PV [GeForce 6150] (rev a2)

Thanks, Nathalie

The problem *was* fixed, but I have the problem again since the last upgrade today :(

Is anyone already fixing the bug? Here, the refresh issues mainly affect firefox and konsole. Any information we could provide to help the maintainer to fix the bug? Does it affect other distributions than ubuntu?

Where's the fix? Sorry for the question.

My last comment refers to the "fix released" state as shown above. Where is the supposed fix?

Still encountering the problem (firefox scrolling refreshing with compiz enabled) with intrepid and latest 177er nvidia driver (177.82-0ubuntu0.1). Seems to be twin-view specific here (thinkpad t61with nVidia Corporation Quadro NVS 140M).

I have no error since I installed the last driver from Nvidia's ftp. It's
seems the official page shows up to 177.xx version but the ftp stores the
180.16 version.

Anyway I have the same bugs only when I use a second screen from my laptop.

ftp://download.nvidia.com/XFree86/Linux-x86/

2008/12/14 Soenke <email address hidden>

> My last comment refers to the "fix released" state as shown above. Where
> is the supposed fix?
>
> Still encountering the problem (firefox scrolling refreshing with compiz
> enabled) with intrepid and latest 177er nvidia driver
> (177.82-0ubuntu0.1). Seems to be twin-view specific here (thinkpad
> t61with nVidia Corporation Quadro NVS 140M).
>
> --
> Screen refresh problems with nvidia on intrepid
> https://bugs.launchpad.net/bugs/269904
> You received this bug notification because you are a direct subscriber
> of the bug.
>

--
-= http://www.legadodekain.net =-

I had very little problems using the 173.xx version in intrepid, and after installing the 180.11 driver from jaunty (still running intrepid), the situation seems to be even better =)

To do this (ONLY do this if you are sure about what you are doing, since jaunty is still in heavy development and things like this might break your installation, the packages that work for me might be replaced by packages that don't work when you try):
0. READ through ALL of these steps to make sure you understand everything, and can revert the steps if something would go wrong.
1. Download the individual packages from http://packages.ubuntu.com/search?keywords=nvidia+180 (you need all of them except nvidia-glx-180-dev).
2. Uninstall your current versions of nvidia-XXX-kernel-source and nvidia-glx-XXX. (XXX should be either 173 or 177 in intrepid.)
3. Install all three downloaded packages at once using sudo dpkg -i in a terminal.

Yes this bug is still alive and definitely specific to compiz, nvidia, mozilla applications (maybe all cairo based programs) and twin-view. How can we help to get it fixed?

funny, i got the same symptoms, not on moz-apps, also any gnome app is working fine, but on other apps,
like openoffice, skype and all apps that are qt.
this on ubuntu intrepid, gnome, nvidia 8400m gt.
cheers

As for me I have it on all aps and effects, using Xubuntu. Also with nvidia + twin-view. When it happens (regularly, every few days) I simply don't have any window refreshing itself unless I move the window or another window above it forcing a refresh. Even xmms stops updating itself. Effects are broken when it happens, for example I can switch between workspaces but the effect don't play completely. Thankfully I can use compiz-fusion-icon to reload the window manager which fixes the problem.

Hello,
think I have a similar problem when compiz is enabled.

Please confirm if you are talking here about display errors like this.
In my case it is gnome-terminal not being updated sometimes.

See the attached video.

I am using intrepid amd64.
NVIDIA 180.06 drivers.
NVIDIA Quadro NVS 140M

@ Christoph Korn:
That is just the kind of display errors I get, albeit not that often...

@ All:
Installing 180.11 drivers makes the situation better for a while, but I did experience some errors when compiz had been running for a long time (and only on my external screen (my computer is a laptop), which had a higher frequency of errors earlier). Restarting compiz (alt+f2 -> compiz --reload, I need to do that twice compiz crashes the first time and fallbacks to metacity) fixed the problem.

http://www.nvnews.net/vbulletin/showthread.php?t=124059

The changelog mentions that bugs related to glyphcache were fixed.

Maybe that fixes my problem.

180.11 did not change anything.

Bug is still present.

$ lspci -n |grep 300
00:0d.0 0300: 10de:03d0 (rev a2)

$ lspci |grep VGA
00:0d.0 VGA compatible controller: nVidia Corporation GeForce 6150SE nForce 430 (rev a2)

173.xx solved a minor screen flickering problem. 177 still doesn't fix it.

01:00.0 0300: 10de:042d (rev a1)
01:00.0 VGA compatible controller: nVidia Corporation Quadro FX 360M (rev a1)
nvidia-180-kernel-source:
  Installed: 180.11-0ubuntu1

None of the suggested workarounds helped avoid the damage bug. Loose binding does not help. I see no difference at all between the 173, 177, and 180 versions of the driver.

Of the applications I use, these three are affected:

  1. (seldom) Firefox: damaged and non-damaged areas overlap and become unreadable while scrolling with the mouse or with window scrollbars. Paging with the keyboard works correctly.

         Remedy: transfer focus to another window and back or click anywhere on the window boroder.

         Sometimes I can work for days without seeing it. But once it starts happening (usually in one of many open firefox windows, it tends to continue until I restart X11). But sometimes it can return to normal without me consciously doing anything.

  2. (often) Gimp: bits of paintbrush cursor hang on, leaving a messy track.

         Remedy: zoom in and out.

  3. (almost always) xterm: typically, sending a large amount of text (more than a few lines), leaves it with only the top part of the window rendered correctly when the process completes; the bottom part remains blank. Sometimes just one line will fail to render, as in the case of 'date'; it will look as if 'date' is still running, while in fact it will have finished instantly. I can't get used to it: I often find myself waiting for the commands to return because their output is not visible.

         Remedy: click into the window or transfer focus to it.

         xterm had its own rendering bugs since about 1994, the most obnoxious being the incorrect command line wrapping in wide windows; the other prominent bug is falure to erase the old text on update. I associate that (not knowing much about it) with the transition from termcap to whatever other system we use today. If I recall correctly, I first encountered it in the first RedHat release, and it hasn't been fixed since then. I think it affects other readline-based terminals besides xterm. This, combined with the compiz damage bug, makes xterm close to unusable. But I can't make myself use anything else. I'm getting old.

I see no ill effects reported by some users in the window decorations. Nautilus always works fine (although I don't use it much). Have never seen problems in emacs-gtk. Overall, I am not sure whether I'd rate this problem as mild or severe; I think I have enough patience to carry on like this, but I "severely" desire to have it fixed one day.

And I forgot to say: it did not change with transition from Hardy to Intrepid.

selkovjr, I second everything you've written regarding xterm. I'm experiencing the same problems to the letter. Blank lines at the bottom after a large amount of text is sent to the terminal. Thanks for describing it so well (and saving me the pain of trying the 180 driver.

I never had this problem with hardy, however.

It might be worth trying the very latest driver, 180.18. I had occasional problems with the 173 and 177 series of drivers, then 180.08 was pretty good but not perfect, then 180.11 and 180.16 were progressively worse than 180.08 for refresh corruption. But since I installed 180.18 on the 23rd Dec, I've had no refresh problems at all. I installed using the NVIDIA-Linux-x86_64-180.18-pkg2.run file that I got from nvidia's ftp site, ie at ftp://download.nvidia.com/XFree86/

I can confirm that commenting out lines 341-344 of the /usr/bin/compiz script has completely resolved my rendering issues. I haven't experienced any problems since making the change. Thanks to those that suggested it.

Just to clarify, these lines in /usr/bin/compiz enable the loose-binding option for all nvidia cards. Commenting out just causes compiz to run without the loose-binding option (which I never found worked very well in previous versions anyway). Here are the lines in their commented-out form:

# if check_nvidia; then
# if [ "x$INDIRECT" != "xyes" ]; then
# COMPIZ_OPTIONS="$COMPIZ_OPTIONS --loose-binding"
# fi
# fi

I have done both things -- installed 180.18 and disabled loose binding.

Looks like the same bug has been kicked around from one spot to another.

I can now scroll in firefox without mangling the readable text. The problem I have now is with table cell backgrounds in firefox (see attached screenshot). It does not significantly affect readability, but given how bizarre it is, I have little hope that it will be fixed soon. This is no longer a scrolling bug; this is a draw-on-visibility bug (if that's really different). I have this page showing a tree, rendered as an html table with the dotted lines representing the tree branches rendered as background images in table cells. If I cover this window with another window, then uncover it, I see 2-3 of those "tree" cells filled with images from whatever window on the screen that was updated last before this firefox window was uncovered. I have pasted part of my gnome menu to show where the cell background was copied from. The wicd icon updates itself every second or so, showing the strength of the wireless signal. The next time, it may be a digit from the clock, or a small segment of it, or just a blank portion of the gnome menu.

This is not entirely unlike the pattern I saw in one of the earlier versions, where any new window would be initially painted with chopped and randomly tiled fragments of the root window. That artifact is now gone (or almost gone).

Text refresh in xterm has not been fixed. I have a feeling that it doesn't occure as often -- let's say, it works more often than it doesn't, but the problem did not go away. It is exactly the same.

And this one is new. I'm seeing hideous refresh problems in Terminal Screenlet. The pattern is not new, though. Someone in this thread pasted a screenshot a while ago (I can't find it -- the thread has grown so big), showing the same artifact in some graphical program (or was that Nautilus?). What happens is that only a narrow vertical strip on the left-hand side is refreshed, at some odd offset. The next refresh will take place in a strip that will be wider or narrower, and probably placed at a different offset. Since Terminal Screenlet refreshes its client area on every keystroke, it instantly becomes an unreadable mess, removed only by hiding/showing the entire widget layer.

Worst of all, the system was unable to wake up from suspend 3 times during the 3 days that I'm running 180.18. It wakes up with a black screen and dead keyboard. I didn't know whether the kernel was alive or not, so I had to reboot it on all 3 occasions.

Overall, no improvement. I don't feel I can trade suspend for scrolling. Suspend on this machine (Dell Precision M4300) was very unreliable in Hardy (failed exactly every other time), but I haven't seen a single failure since I installed Intrepid. 180.18 obviously interferes with it in some way.

I just noticed another artifact -- a very insignifcant one, but it may be a symptom of a problem. I use the Emerald Aqua theme from Mac4Lin, with the rounded blue buttons and scroolbars. The buttons look fine, but all scrollbars show a gray rectangular outline (see the attached image).

Hi,

I can confirm the behavior that selkovjr describes with 180.06 driver (T61 laptop, nVidia Corporation Quadro NVS 140M (rev a1)):

- vertical scrolling refresh problems disappear
- after suspend computer does not wakeup
- changing to console does not work anymore -> switches to x-server immediately again
- enabling second monitor with nvidia-settings: sometimes a big black rectangle occures (because of different solutions of screens)

So 180.xx is really *beta* and not usable.

I will try the other fix with loose-binding right now.

-soenke

That suspend-wakeup problem is a segfault. I have a suspicion that the server is distributed with an incompatible kernel module. I just made the 177 version do the same, when I installed it without properly cleaning up the 180. I thought the ubuntu package would simply put the right things in all the right places, but it didn't. For that, the nvidia installer must be run with the --uninstall option. Interestingly, it didn't log any errors before it crashed. It read the config file, the screen blinked a couple times, then segfault.

If it runs under gdm, the latter will attempt to restrat it (in the absence of verbose errors), so this situation is a dead-end. The segfault apparently clobbers the console driver and likely something else. In one instance, it opened the console in the vga mode, with all sorts of random colored ascii characters filling the screen. It writes everything it wants everywhere it wants. I don't know how much of the system survives the crash; the only function I was able to use was the power button -- a single touch on the button brings up the Ubuntu shutdown splash with the progress bar, and then it shuts down.

So I'm back to 177. Still running without loose binding (to ascertain that it makes any difference in scrolling). The only effect so far is that without loose binding, compiz's window fading became rough and jerky, so I had to disable it.

Yeah problem seems to disappear here without loose binding, too. But it makes 2.1GHz CoreDuo feel like 80486 :-/

wow, long list here...
you can add me to the loose-bind-junks. qt-apps look bad, refresh problems, openoffice affected too, googleearth: black space is transparent,
qt-app-icons on systray: u do not wanna look at them, AND very interesting: the pop-up-menus from qt-apps-icons on systray do not have shadows, seems like there is a great conflict in my head or in the computer. thanks.

Today the nvidia drivers 180.22 were released.
http://www.nvnews.net/vbulletin/showthread.php?t=125891

The changelog mentions a regression related to compiz is now fixed:
Fixed a regression that could result in window decoration corruption when running Compiz using Geforce 6 and 7 series GPUs.

Nice -- they fix it in their unusable beta drivers and not in the stable release :-(((

180.22 is a stable release.

Ah ok, it's *the* stable release :) Hope it's in intrepid-backports soon.

I find 180.22 to be MUCH worse than 180.18 for this compiz corruption issue. In only two hours of use, I've already seen significant screen corruption five times when scrolling in Firefox. With 180.18, I only saw corruption once in over two weeks of use.

I'm using an Nvidia GeForce 8600M GT card.

Indeed, it is not fixed.

But I cannot say that it happens more often.

It happens much more often with 180.18. It is almost unusable.

Same with with 180.22.

Not a 'real' solution to the refresh issue but the following has brought sanity back to my life:

System -> Preferences -> Appearance
 click on Visual Effects tab
 select None

Same problem on Debian Lenny/Sid with drivers 177.80 and 180.22 from repository and from Nvidia's site.
Acer Aspire 5920G, Nvidia GeForce 8600GS

Peter Draksler, Visual Effects = None means compiz is disabled. Bug appears only with compiz enabled.

I'm also experiencing this problem with nvidia drivers after 173, but only in the terminal. Doesn't matter which one. I tried gnome-terminal, xterm, konsole. But I noticed two things which may be useful to track it down:

1. For me it only occurs in single-head mode. As soon as I start my X-Server with a second display attached I have no screen update problems.

2. I noticed that the problem does not occur when a compiz animation is played. So enabling the benchmark window fixes the problem. Currently I use the Paint Fire feature of Compiz to workaround the bug when I'm in single-head mode. Usually I put the fire into the trash bin. ;-)

Please everybody, click the "This bug doesn't affect me (change)" link at the top of the bug instead of posting a "me too" comment. The problem here is understood and additional confirmation comments only add clutter to the bug.

I think problem = Nvidia powermizer

Powermizer work 3 performance level
0 - 169Mhz
1 - 275Mhz
3 - 475 Mhz

My laptop default performance level = 0
NV Clock = 169Mhz (with screen refresh problem)

I run this script in gnome-terminal:
while true; do
    if on_ac_power; then
        nice /usr/bin/nvidia-settings -q all > /dev/null
    fi
    sleep 25;
done

performance level automatically switch to 2
NV Clock = 475Mhz

And no screen refresh problem!!!

Sorry very poor english....

this would explain, why some problems disappear, like openoffice issues or qt-apps, when connecting second monitor and running twinview.
but the scrambled qt-app-icons on the panel-taskbar and the missing shadows on their drop-down menu remain, (since driver 177.XX)

I've been running the script that AcCEsS mentioned for several months now (upon login), but I still had screen refresh problems anyway.

I just upgraded (about 2 days ago) to the latest Jaunty Alpha on the only computer I have with an nVidia card, so I don't know how much help I can be to the Hardy and Intrepid users, since I don't have a stable version.

I have 180.22, and nVidia Quadro NVS 135M.

Not trying to add another "this happens to me too" note, just saying that I think it has to be more than PowerMizer, considering that mine is always on the highest performance level, and it doesn't seem to make a difference.

One situation I can always count on to mess up is Amarok 2, when I collapse an artist branch to view his/her songs, and then close that branch back up again. I have to move my mouse over the area EVERY SINGLE TIME, so that's easily repeatable for me at least, in case someone wants a somewhat "dependable" test.

[This is an automated message]

In Jaunty (9.04), we are dropping the nvidia-graphics-drivers-177 package, which
is now superseded by then nvidia-graphics-drivers-180 package.

  http://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers-180

To transition your bug into the new package, we need your help. Please do the following:

 a. Verify the bug occurs in Jaunty with the -180 driver
     (ISOs: http://cdimage.ubuntu.com/releases/jaunty/)
 b. If you haven't already, please include in the bug:
     * Your /var/log/Xorg.0.log
     * Your /etc/X11/xorg.conf
     * The output of `lspci -vvnn`
     * Steps to reproduce the issue
 c. Under Affects, click the down arrow to the left of 'nvidia-graphics-drivers-177'
 d. Edit the Package to change 'nvidia-graphics-drivers-177' to 'nvidia-graphics-drivers-180'
 e. Click Save Changes

Thank you!

[We'll expire the remaining -177 bugs in a month or so.]

I have all the updates on Jaunty.
I use gnome.
When Compiz is enabled, kmail will still not refresh the all window, and I can see part of the screen divided vertically.
It is happening much less then before (even less today, then yesterday (not sure if any packages changed)).
I'll try to take a screenshot

$ apt-cache policy compiz
compiz:
  Installed: 1:0.7.8-0ubuntu9
  Candidate: 1:0.7.8-0ubuntu9
  Version table:
 *** 1:0.7.8-0ubuntu9 0
        500 ftp://darkstar.ist.utl.pt jaunty/main Packages
        500 ftp://archive.ubuntu.com jaunty/main Packages
        100 /var/lib/dpkg/status

$ apt-cache policy kmail
kmail:
  Installed: 4:4.2.0-0ubuntu7
  Candidate: 4:4.2.0-0ubuntu7
  Version table:
 *** 4:4.2.0-0ubuntu7 0
        500 ftp://darkstar.ist.utl.pt jaunty/main Packages
        500 ftp://archive.ubuntu.com jaunty/main Packages
        100 /var/lib/dpkg/status

$ apt-cache policy nvidia-glx-180
nvidia-glx-180:
  Installed: 180.27-0ubuntu1
  Candidate: 180.27-0ubuntu1
  Version table:
 *** 180.27-0ubuntu1 0
        500 ftp://darkstar.ist.utl.pt jaunty/restricted Packages
        500 ftp://archive.ubuntu.com jaunty/restricted Packages
        100 /var/lib/dpkg/status

 affects ubuntu/nvidia-graphics-drivers-180

I've been trying to track down the cause of this problem for ages, and I think the comment from another user at bug #270377 has just helped me find it.

Apparently the redraw problems described here and in the other bug reports are caused by a race condition, that can occur with compiz in combination with nvidia driver versions above 177.76. According to nvidia developers at the nvnews forums: "[the cause of the problem] is a race condition between the X server and compiz that's the result of a fundamental design problem in the way the Damage extension works." and "[fixing it] will require more changes to the Damage and GLX extensions, the X server, Compiz, and the NVIDIA X driver."

Because it is apparently caused by a design problem, it will probably take some time before the problem is fixed. The only real way to ensure it doesn't happen at the moment is not to use compiz. But on some systems it helps to change settings that affect the timing of pixmap uploads to the GPU. That's why it may help to disable the compiz "--loose-binding" option, or to play with the powermizer settings.

The fact that the occurrence of race-conditions may be affected by timing also explains the fact that everyone seems to experience this bug (and its many duplicates) differently, depending on hardware-configuration, which programs are running, the number of open windows, X.org and driver settings, etc.

For me, restarting compiz without loose binding immediately solved the refresh problems, albeit at the cost of compiz rendering performance. I just had a look at the wrapper script that launches compiz in Ubuntu. It enables --loose-binding on nvidia hardware (which in turn works around a performance issue with texture binding on nVidia cards), but it only seems to do this when indirect rendering is disabled (the default setting). So in theory this should disable --loose-binding without having to edit the wrapper script or (re)start compiz by hand every time:

mkdir -p ~/.config/compiz; echo INDIRECT=0 >> ~/.config/compiz/compiz-manager

It just adds "INDIRECT=0" to the configuration file ~/.config/compiz/compiz-manager, to force indirect rendering. This will of course degrade performance (e.g. with desktop cube) and will also disable some effect plugins (e.g. water). But I much prefer it to having the current redraw problems or disabling desktop effects all together.

I do hope this issue gets properly fixed though...

See also:
http://www.nvnews.net/vbulletin/showthread.php?t=120892
http://www.nvnews.net/vbulletin/showthread.php?t=124470

Oops, sorry for posting duplicate information about the race condition and the loose binding thing. I guess I should've read the entire thread before pasting the comment. Still, apparently this bug is hard to find through Google or the search option in Launchpad, because it has so many manifestations (and "solutions"). So I hope it's still helpful for people like me, who get directed here from one of the 15 duplicates and don't scroll all the way up... :-P

Actually, as mentioned in previous posts the fix should be:

mkdir -p ~/.config/compiz; echo INDIRECT=yes >> ~/.config/compiz/compiz-manager

(note the INDIRECT=yes instead of INDIRECT=0). This launches compiz as:

  /usr/bin/compiz.real --ignore-desktop-hints --replace --indirect-rendering "" core ccp

instead of the normal (or with the INDIRECT=0 option):

 /usr/bin/compiz.real --ignore-desktop-hints --replace --loose-binding "" core ccp

Carrick

The bug still occurs on intrepid with 180.27 drivers on a geforce 9600M.

And please stop calling the workarounds that only work sometimes 'fixes'.
They just change timings a bit and stop the race condition kicking in all the time. Still with all
suggestions tried out, that I found around the web, compiz is still unusable (esp. with texmaker
running).

Metacity is a fine windowmanager but I still hope this will be fixed in jaunty.

Actually, there _is_ a workaround which works all the time at the cost of a bit cpu time (at least for me). Just go to the compiz settings manager (ccsm). Activate the 'draw fire' plugin, reduce all the settings to mininum, except for background brightness, which should be set to 100. Now when you draw fire on the desktop, it should be invisible, but for me it solves the screen refresh problems completely! So just draw a bit of fire after every reboot, and you're fine.
Note: If you use the workaround with activating --indirect-rendering and disabling --loose-binding, it will also work, but the cpu time needed for the fire will be much greater (25% compared to 5% if you leave the settings to standard). Since the fire solves the problem all the time, you should leave INDIRECT=no.

Hope this helps.
Cheers, Jan

Are you sure drawing the fire effect actually helps?I haven't tried by I
think by doing that you are forcing the GPU to set to higher clock speed.
That is why the screen refresh problem does not occur. I'm running Hardy and
I fixed my GPU clock speed (geforce 8400m) to level 2. The problem does not
occur except in openoffice 2.4 (sometimes) and adobe reader 8.

On Fri, Feb 20, 2009 at 11:13 PM, Jan Krieg <email address hidden> wrote:

> Actually, there _is_ a workaround which works all the time at the cost of a
> bit cpu time (at least for me). Just go to the compiz settings manager
> (ccsm). Activate the 'draw fire' plugin, reduce all the settings to mininum,
> except for background brightness, which should be set to 100. Now when you
> draw fire on the desktop, it should be invisible, but for me it solves the
> screen refresh problems completely! So just draw a bit of fire after every
> reboot, and you're fine.
> Note: If you use the workaround with activating --indirect-rendering and
> disabling --loose-binding, it will also work, but the cpu time needed for
> the fire will be much greater (25% compared to 5% if you leave the settings
> to standard). Since the fire solves the problem all the time, you should
> leave INDIRECT=no.
>
> Hope this helps.
> Cheers, Jan
>
> --
> Screen refresh problems with nvidia on intrepid
> https://bugs.launchpad.net/bugs/269904
> You received this bug notification because you are a direct subscriber
> of the bug.
>

I'm igniting my "fire-place" every evening. At work the problem does not occur (dual-head configuration) but at home (single-head) I always have the problem in the gnome-terminal. And everytime I start a small fire in the corner and then I have no problems anymore. So I can reproduce the bug AND the workaround every day and it definitely helps.

@aptgetmoo: The higher clock speed could be the reason for why the fire helps, but my assumption was that by drawing the fire I force the X-server(?) to continuously refresh the screen. To be honest, I don't know much about the internals, so be free to correct me :). By doing this, _all_ screen refresh problems disappear, so you could give it a try when it occurs again with oo or adobe.

On Fri, 20 Feb 2009, Jan Krieg wrote:

> Actually, there _is_ a workaround which works all the time at the cost
> of a bit cpu time (at least for me). Just go to the compiz settings
> manager (ccsm). Activate the 'draw fire' plugin

It was of no help to me with any settings. The only thing that really
works, sort of, is manually "refreshing" the screen. I don't need any
extra gadgets to keep it under control, because I run a widget layer,
which is a big part of why I want compiz. So when anything on my screen
goes bad, I clean it up with two keystrokes -- activate and deactivate the
widget layer.

Those of us who remember working on remote machines through a VT100 will
not find it terribly annoying. Remember how we used to send Ctrl-L
whenever something wrote garbage on the screen? Same thing here. It's
wired in my brain. Much I as want to be rid of the bug, I can manage it
this way.

i have this problem in jaunty
see the attachment for image

i've open a new bug for jaunty:
https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/335405

I am not using Ubuntu anymore, but the issue has just been solved on openSUSE with the latest 180.29 driver.
No need to deactivate loose binding anymore. Hurrah !
Unfortunately I am not able to tell you what solved the issue, because it seems that some people with the same driver still have the problem.
But the good news is there is a solution ! Good luck !

Jean-Christophe, could you link the OpenSUSE bug report using https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/269904/+distrotask , so it would be easier to follow what they are doing ? :-)

help for this bug in jaunty with nvidia 8400M G on a laptop with nvidia 180.29 driver

so i've install the update of today and now i have nvidia 180.35 driver, but no change...
see the screenshot

without compiz i haven't this problem..
please help

Just to add to this (I think someone already mentioned), this happened to me with 177, 180 and 180.25 from jaunty. Reverting to 173 stopped the problems althougether. Haven't tried any of the workaround becaused they aren't proper fixes.

The following "fix" helped me to get rid of the problem on Skype, Quassel and other Qt based applications, making them usable:

https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/269904/comments/215

I was finally able to completely circumvent this problem by enabling *both* indirect rendering and loose bindings for compiz (by modifying the appropriate line in /usr/bin/compiz). Compiz runs really fast now (a lot (!) faster than with indirect-rendering only and even more faster than without both).
I'm running an unsupported combination of software versions however, so could someone check whether this works with plain Intrepid, too?

I am running:
Ubuntu 8.04 with stock compiz 0.7.4
Nvidia driver 180.29, installed manually
on a Quadro FX 370 (dual monitor setup)

Without loose-bindings I especially had very bad performance when running OpenGL applications with compiz enabled.

I now have in my /usr/bin/compiz:
build_args()
{
        [...]
        if check_nvidia; then
                if [ "x$INDIRECT" != "xyes" ]; then
                        COMPIZ_OPTIONS="$COMPIZ_OPTIONS --loose-binding --indirect-rendering "
                fi
        fi
}

apparently I have no more problem since I use nvidia 180.11 on 8.10 (amongst other updates)

in this thread

http://www.nvnews.net/vbulletin/showthread.php?t=129711

you can read about a new compiz workaround to force X-to-GLX synchronization
that has been implemented.

http://gitweb.compiz-fusion.org/?p=fusion/plugins/workarounds;a=commit;h=46960f12a9d213e5f0e841557e2ed2f7ea18cc79

i have not checked it myself but it seems like it circumvents the
race condition for those users.
please include that commit in jaunty!

Here is a debdiff for compiz-fusion-plugins-main to add that workaround. The new package is available in my PPA:
https://launchpad.net/~anders-kaseorg/+archive/ppa

To enable the workaround, after installing the patched package, you must run System → Preferences → CompizConfig Settings Manager, open Workarounds in the Utility section, and turn on “Force synchronization between X and GLX”. Then restart compiz.

Will this patch be included in jaunty?

I haven’t seen the problem in a week with my patch installed and the workaround enabled. I’ve subscribed ubuntu-main-sponsors.

If anyone else has also tried my patched compiz-fusion-plugins-main and enabled the new “Force synchronization between X and GLX” workaround, it would be helpful for you to comment whether it works for you.

@Anders:

I made my own intrepid package with your patch and haven't experienced any (or only very minor) issues with the patch enabled (using original /usr/bin/compiz, I.E. _with_ --loose-bindings and _without_ --indirect-rendering).

I'm using the 180.37-0ubuntu1 nvidia driver from jaunty, and the situation is even better than with my old solution (using the 173 driver). No mentionable slowness of X or Compiz either.

Finally a working workaround! :)

Anders Kaseorg, I added your package, but when I apt-get upgrade, it seems there is a syntax error somewhere in your repository (or package) :

$ sudo apt-get upgrade
Lecture des listes de paquets... Fait
Construction de l'arbre des dépendances
Lecture des informations d'état... Fait
Les paquets suivants seront mis à jour :
  compiz-fusion-plugins-main cpp-4.3 deluge deluge-common deluge-core deluge-torrent gcc-4.3 gcc-4.3-base git-core lib32gcc1 lib32stdc++6 libgcc1 libgfortran3
  libgomp1 libstdc++6
15 mis à jour, 0 nouvellement installés, 0 à enlever et 0 non mis à jour.
Il est nécessaire de prendre 0o/17,9Mo dans les archives.
Après cette opération, 3846ko d'espace disque supplémentaires seront utilisés.
Souhaitez-vous continuer [O/n] ? O
dpkg: erreur d'analyse syntaxique, dans le fichier « /var/lib/dpkg/available » près de la ligne 24315 paquet « openjdk-6-jre »:
 champ « Depends », référence à « libasound2 » : version contenant « ) »
E: Sub-process /usr/bin/dpkg returned an error code (2)

(We solved ®om’s problem over IRC with `dpkg --clear-avail; apt-get update`. Obviously, it’s unrelated to this bug.)

Joakim Andersson: could you make that intrepid package available for for the public? :)

@Karl Frisk:

Sure, but I only have it for amd64... It's attached, anyway.

(It's labeled ~andersk1 since he made the original patch, I just applied it as is, without any modification)

that seems to work beautifully. no problems at all in the scenarios where this bug bugged me the most. :)
this patch should be in jaunty by default.

thanks to everyone involved.

I added a patched compiz-fusion-plugins-main for Intrepid to my PPA:
<https://launchpad.net/~anders-kaseorg/+archive/ppa>

(I’d prefer not to see ~andersk labels attached to binaries that someone else compiled.)

@Anders Kaseorg,

Unfortunately, the patch doesn't work at all for me (before and after a fresh reinstall of jaunty alpha 6).
The menu doesn't appear in ccsm, so I enabled the option in gconf-editor (/apps/compiz/plugins/workarounds/allscreen/options/force_glx_sync), but I have still the refresh problem in openoffice and in digikam : they are near unusable.

I finally installed the patch as explained here : http://www.nvnews.net/vbulletin/showpost.php?p=1957219&postcount=18
(I exactly followed the steps)

And it totally resolves the problem ! No more refresh problems in openoffice, even if powermizer is at the lowest frequency !

Please make it available in jaunty (and enabled by default this workaround).

®om: The directions in that post should give you a workaround plugin that is very similar to the one in my package (despite that post’s author’s misunderstanding of Git; `git revert` means undo a diff, not “revert to” a commit).

> The menu doesn't appear in ccsm,

I suspect you have something weird in your home directory that you copied over across the fresh reinstall. Did you have a ~/.compiz/plugins/libworkarounds.so and ~/.compiz/metadata/workarounds.xml that would have been overriding the global ones? (You certainly do now that you’ve followed those directions; but if you had ever done any manual installation of compiz plugins in the past as well, that would obviously have prevented my package from having any effect.)

I recommend removing your ~/.compiz/plugins and ~/.compiz/metadata directories (or even entirely wiping out your compiz settings with rm ~/.compiz -rf; gconftool-2 --recursive-unset /apps/compiz), then reinstalling all compiz packages from apt. Or you could try again after creating a fresh new account and logging into it.

Another vote for Anders' fix. I was suffering from the refresh problems in Jaunty's OpenOffice 3 with both nvidia 185.37 and 185.13, but they work fine now. (I assume the fix implements the 'Force synchronization between X and GLX' by default? Using the workarounds package also fixes the OO refresh issues.)

@Anders Kaseorg:
I installed your package from a fresh reinstall of Jaunty alpha 6, with a totally new /home, without any manual installation (I tried your package before the steps described with git).

As many people affirm that your package works, I suspect a localization problem (my system is in French, and in the French version, "workaround" is translated by the word "solutions").

I will let you know if I have the files you asked me (I am at work).

I can confirm that Anders' workaround works indeed, please include it by default.

I want to join this bug to the compiz I reported some time ago:

http://bugs.opencompositing.org/show_bug.cgi?id=1044