Description of problem: With kernel 3.14.2, kwin crashes at login and the desktop effects are disables and cannot be enabled. Version-Release number of selected component (if applicable): kernel-3.14.2-200 How reproducible: Boot with new kernel, log in, behold Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info: Booting with previous 3.13.10-200 or any previous kernel works fine.
Please attach the dmesg output from both 3.13 and 3.14 kernels.
FWIW, I am currently running [egreshko@meimei ~]$ uname -r 3.14.2-200.fc20.x86_64 With desktop effects enabled with about 20 of them "checked". No crashes. I am, however, running the nVidia drivers.
Created attachment 892208 [details] dmesg output from the working 3.13 kernel
Created attachment 892209 [details] dmesg output from the not-working 3.14 kernel
I am experiencing the same problem as the OP. Attaching the dmesg outputs. On tuesday (6. May) I'll get access to my second machine and I will be able to attach more dmesg outputs, if needed. HTH, :-) Marko
Created attachment 892232 [details] dmesg output from working 3.13.10 kernel
I am using Intel graphics. I haven't tried updating the laptop since this happened, but it also uses Intel graphics. If required, I could update it later. I am not running the 3.14.2 kernel presently. dmesg appears to only store the most recent boot. Is the information in the journal somewhere? Do you still need it?
I discovered a wonderful command: journalctl -b -1 that is supposed to give me only the output from the previous boot. Hopefully this will satisfy your needs :-) Yes?
Created attachment 892233 [details] dmesg output from not-working 3.14.2 kernel
I got curious and updated my laptop. The desktop has Intel 965G graphics and it will not run kde desktop effects with kernel 3.14.2. The laptop has older Intel 915GM graphics and it does run kde desktop effects with kernel 3.14.2. However, with the laptop and the 3.13 kernels, I was getting mce errors that were apparently unreportable, but with the 3.14 kernel, I have not yet experienced any such error. The machine shows no evidence of malfunction.
My machine also has Intel GM965 chip. The other one (which I'll be able to access on Tuesday to provide details) has nVidia graphics, running with nVidia binary drivers from rpmfusion. Both of these machines experience problems in my case. HTH, :-) Marko
(In reply to Peter Gückel from comment #8) > I discovered a wonderful command: > > journalctl -b -1 that is supposed to give me only the output from the > previous boot. Hopefully this will satisfy your needs :-) Yes? Yes, that's great. It shows kwin crashing as you said. It doesn't immediately show any kernel problems. The only thing I saw that might be related is that powerdevil (whatever that is) tried to talk to dbus to find the backlight interface and couldn't: May 03 19:21:37 desk.home dbus-daemon[547]: dbus[547]: [system] Activating service name='org.kde.powerdevil.backlighthelper' (using servicehelper) May 03 19:21:37 desk.home dbus[547]: [system] Activating service name='org.kde.powerdevil.backlighthelper' (using servicehelper) May 03 19:21:37 desk.home dbus-daemon[547]: no kernel backlight interface found May 03 19:21:37 desk.home dbus-daemon[547]: dbus[547]: [system] Successfully activated service 'org.kde.powerdevil.backlighthelper' May 03 19:21:37 desk.home dbus[547]: [system] Successfully activated service 'org.kde.powerdevil.backlighthelper' Then later kwin crashes with an assertion. That crash is a problem in kwin, even if it's due to not being able to find a kernel interface. It would be interesting to get the same journalctl -b -1 output for the "working" kernel to see if it similarly fails to find a backlight interface. Hans has been working on some backlight issues recently and wrote a great blog post about debugging those issues. Please read it over and see if it results in any progress on your side: hansdegoede.livejournal.com/13889.html
It's a Mesa bug: https://bugs.launchpad.net/ubuntu/+source/mesa/+bug/1299499
I have definitely been having backlight problems for a couple of years and have reported them once and again over that period of time. Curiously, the backlight problems (the screen going about 95% black with just a 5% hint of what is there after a few minutes of inactivity and no way to make it come to full strength short of a reboot) are almost always on the laptop, which works with the 3.14 kernel, and only rarely with the desktop, which doesn't work with the 3.14 kernel. Since this is a mesa bug, I will have a look at Hans' blog, but I don't think I'm going to spend a lot of time 'verstelling' (misconfiguring) anything until the mesa updates land.
Created attachment 892299 [details] dmesg, really journalctl, output of working 3.13.10 kernel I didn't feel like rebooting, so this is the currently running session that began last night, got suspended while I slept, and was just resumed. I hope this is untainted enough to make sense of.
Aside re: Comment 10 I wish I could find out what is causing the mce errors on my laptop. I just got the kernel error message, even with the 3.14.2 kernel, that had appeared not to give me them. Something is triggering them, but what? How can I find out?
I have updated to today's mesa updates 10.1.1.2. The problem persists on the desktop (intel 965 graphics) with the 3.14 kernel.
The last 2 boots, I got no more kernel mce error messages. I'm weirded out. Hopefully that's gone.
I'm experiencing the same problem. After upgrading to kernel-3.14.2 kwin crashes when "Enable desktop effects at startup" is checked (for "Composition type" OpenGL; it doesn't crash when this is set to XRender but most of the effects don't work with this setting). Today I've upgraded KDE to the one from updates-testing (4.12.5) and kwin still crashes. ps. the laptop I'm experiencing problems has Intel GPU, mesa-libGL-10.1.1-2.20140419.fc20.x86_64.
I've just upgraded to kernel 3.14.3 from "testing" and the problem still persists.
As another data point, the 3.14 kernel broke video playback for me (with Intel graphics): See bug 1094519. I don't know if it is the same problem or not, but it only happens on my system with Intel graphics and only happens on the 3.14 kernel.
kernel-13.4.3 is defective, too. The last kernel that worked was 3.13.10. Curiously, it works fine on intel 915 graphics, but breaks compositing on intel 965 graphics. ]As an aside (not sure if it is part of the comorbidity of this problem), the mce machine error message (that appeared only on the intel 915 graphics machine) has no longer appeared for a number of days.]
(In reply to Peter Gückel from comment #22) > kernel-13.4.3 is defective, too. Correction: 3.14.3
It seems this is a duplicate of https://bugzilla.redhat.com/show_bug.cgi?id=1095375 I upgraded to the latest mesa from updates-testing (10.1.3-1.20140509.fc20.x86_64) and changed "Compositing type" from OpenGL 3.0 to OpenGL 2.0 and kwin seems to be working without a crash. Cheers. ps. kwin still crashes on "Compositing type" OpenGL 3.0
(In reply to Martin from comment #24) > It seems this is a duplicate of > https://bugzilla.redhat.com/show_bug.cgi?id=1095375 Possibly. It is similar, but the crash is not on that mobile system, but on a desktop computer running Intel 965 graphics. The Intel 915 graphics laptop works fine. > I upgraded to the latest mesa from updates-testing > (10.1.3-1.20140509.fc20.x86_64) I am not using updates-testing, so I have not tried this. I will see about it in a bit. > and changed "Compositing type" from OpenGL > 3.0 to OpenGL 2.0 and kwin seems to be working without a crash. Been there, done that (about a year ago). I learned then not to use OpenGL 3.0 and I have NOT changed the default settings. I have OpenGL 2, raster graphic system set, which is the default.
(In reply to Martin from comment #24) > I upgraded to the latest mesa from updates-testing > (10.1.3-1.20140509.fc20.x86_64) and... kwin seems > to be working without a crash. Confirmed. mesa-10.1.3-1 from updates-testing has solved the problem with the desktop i965 graphics compositing problem.
yeah, this is a dup *** This bug has been marked as a duplicate of bug 1095375 ***