At some point, there seems to have been a regression. Most visibly, when changing virtual desktops in WindowMaker, the fadeout-animated notification text ("Workspace #") is occasionally rendered as a corrupted mess. This does not happen all the time, however. Additionally, sometimes, albeit extremely rarely, images in Firefox (version 20.0 beta) are now corrupted. Both of these issues seem to be very similar to what I have experienced and reported in past, just reintroduced again. The pixmap corruptions in Firefox are semi-persistent, but may disappear after switching focus in and out a few times (probably triggering a re-render of the page) or scrolling. I will try to bisect this issue later, if it does not turn out to be something blatantly obvious. -- Window manager: WindowMaker 0.95.3 -- chipset: 00:02.0 VGA compatible controller: Intel Corporation 82G33/G31 Express Integrated Graphics Controller (rev 10) -- system architecture: i686 / 32bit -- xf86-video-intel: GIT a0a2faefdefbea63669dfeb49f7e701196ab5631 -- xserver: X.Org X Server 1.12.4 from Debian testing -- mesa: 8.0.5-3 from Debian testing -- libpixman: 0.26.0 -- libdrm version: 2.4.40-1 -- kernel version: 3.8.0 (vanilla from kernel.org) -- Linux distribution: Debian Testing aka Wheezy -- Machine or mobo model: Asus P5KPL-CM -- Display connector: VGA
Sounds like a damage bug is back, the form these take nowadays is that it switches to using an inplace mapping that turns out to not be coherent (i.e. we only update the information on the CPU and not on the GPU as we expected). I don't think I have the assertions in place to catch that. I'll try and write those. In the meantime, can you try and do a quick hunt for the last known good version? And I'll setup a machine here to reproduce.
Out of curiousity, have you found a more reliable trigger? As you may have noticed, there is a definite bug around read-read synchronisation, but I have not been able to reliably trigger it myself.
I've pushed a patch to disable read-read synchronisation optimisations which probably affects this issue. Can you please test with master?
(In reply to comment #2) > Out of curiousity, have you found a more reliable trigger? As you may have > noticed, there is a definite bug around read-read synchronisation, but I > have not been able to reliably trigger it myself. Unfortunately not, and even though I can sometimes hit the issue, bisecting has been dog-slow due to it happening so rarely. Will try latest git asap.
It would seem (after over two days of testing) that this issue is fixed now, (or at the very least hiding itself extremely well). I have not been able to even randomly reproduce the corruptions during this time period.
A little spring cleaning, might as well now consolidate this bug. *** This bug has been marked as a duplicate of bug 61628 ***
Scratch that .. still getting corruption in FF with driver GIT 93ecd5606eaf2e1db2b844f39bb04d77fa53cf8d, and kernel 3.8.2. Though extremely rarely and still no sign of those WMaker corruptions anymore.
Created attachment 76406 [details] FF corruption with driver git 93ecd5606eaf2e1db2b844f39bb04d77fa53cf8d and kernel 3.8.2
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.