FS#47801 - [qt] plasma does not start after upgrade to qt5*5.5.1-9

Attached to Project: Arch Linux
Opened by Sebastian Frohn (sebafroh) - Sunday, 17 January 2016, 08:14 GMT
Last edited by Antonio Rojas (arojas) - Thursday, 04 February 2016, 18:31 GMT
Task Type Bug Report
Category Packages: Extra
Status Closed
Assigned To Antonio Rojas (arojas)
Felix Yan (felixonmars)
Architecture All
Severity High
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 0
Private No

Details

Description:

After upgrade to qt5*5.5.1-9 packages and plasma-desktop 5.5.3-2 kde does not start any more. After login in SDDM graphical interface is terminated und sddm restarts without any further inrformation.

Downgrade of plasma-desktop to 5.5.3-1 does not solve the problem.

Additional info:

[2016-01-17 08:11] [PACMAN] Running 'pacman -U plasma-desktop-5.5.3-1-x86_64.pkg.tar.xz'
[2016-01-17 08:11] [ALPM] transaction started
[2016-01-17 08:11] [ALPM] downgraded plasma-desktop (5.5.3-2 -> 5.5.3-1)
[2016-01-17 08:11] [ALPM] transaction completed
[2016-01-17 08:12] [PACMAN] Running 'pacman -U qt5-base-5.5.1-8-x86_64.pkg.tar.xz qt5-declarative-5.5.1-8-x86_64.pkg.tar.xz qt5-graphicaleffects-5.5.1-8-x86_64.pkg.tar.xz qt5-imageformats-5.5.1-8-x86_64.pkg.tar.xz qt5-location-5.5.1-8-x86_64.pkg.tar.xz qt5-multimedia-5.5.1-8-x86_64.pkg.tar.xz qt5-quick1-5.5.1-8-x86_64.pkg.tar.xz qt5-quickcontrols-5.5.1-8-x86_64.pkg.tar.xz qt5-script-5.5.1-8-x86_64.pkg.tar.xz qt5-sensors-5.5.1-8-x86_64.pkg.tar.xz qt5-svg-5.5.1-8-x86_64.pkg.tar.xz qt5-tools-5.5.1-8-x86_64.pkg.tar.xz qt5-translations-5.5.1-8-x86_64.pkg.tar.xz qt5-webchannel-5.5.1-8-x86_64.pkg.tar.xz qt5-webkit-5.5.1-8-x86_64.pkg.tar.xz qt5-x11extras-5.5.1-8-x86_64.pkg.tar.xz qt5-xmlpatterns-5.5.1-8-x86_64.pkg.tar.xz'
[2016-01-17 08:12] [ALPM] transaction started
[2016-01-17 08:12] [ALPM] downgraded qt5-base (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-xmlpatterns (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-declarative (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-graphicaleffects (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-imageformats (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-location (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-multimedia (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-sensors (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-webchannel (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-webkit (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-script (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-quick1 (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-quickcontrols (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-svg (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-translations (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-tools (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] downgraded qt5-x11extras (5.5.1-9 -> 5.5.1-8)
[2016-01-17 08:12] [ALPM] transaction completed

Steps to reproduce:
This task depends upon

Closed by  Antonio Rojas (arojas)
Thursday, 04 February 2016, 18:31 GMT
Reason for closing:  Duplicate
Additional comments about closing:   FS#48015 
Comment by Antonio Rojas (arojas) - Sunday, 17 January 2016, 08:34 GMT
Which graphics card and driver?
Comment by Hector Mtz-Seara Monne (hseara) - Sunday, 17 January 2016, 11:24 GMT
I have the same behavior, but not only with plasma but also gnome (except wayland version).
My computer has an intel graphics card (Intel Corporation Broadwell-U Integrated Graphics (rev 09)
and I'm using the "intel" driver from xf86-video-intel.

I have found that when using the default "uxa" Acceleration method the reported problem is shown.
When I change the acceleration method to "sna" then plasma starts properly. This is at least
a temporary fix. I guess then that the problem is related to the intel driver.
Comment by Sebastian Frohn (sebafroh) - Sunday, 17 January 2016, 12:30 GMT
Intel graphics card, xf86-video-intel driver, and OpenGl 3.1 with GLX (with EGL screen freezes some times). I searched in logfiles from X, but the startup-process is interupted without any warnig or errormessages.

But journalctl says:

an 17 08:03:51 archaeopteryx systemd[434]: Started D-Bus User Message Bus.
Jan 17 08:03:52 archaeopteryx ksplashqml[483]: QXcbConnection: Could not connect to display :0
Jan 17 08:03:52 archaeopteryx systemd-coredump[484]: Process 483 (ksplashqml) of user 1000 dumped core.

Stack trace of thread 483:
#0 0x00007fda01cf95f8 raise (libc.so.6)
#1 0x00007fda01cfaa7a abort (libc.so.6)
#2 0x00007fda0247f1e1 _ZNK14QMessageLogger5fatalEPKcz (libQt5Core.so.5)
#3 0x00007fd9f48cf064 _ZN14QXcbConnectionC1EP19QXcbNativeInterfacebjPKc (libQt5XcbQpa.so.5)
#4 0x00007fd9f48d483d _ZN15QXcbIntegrationC1ERK11QStringListRiPPc (libQt5XcbQpa.so.5)
#5 0x00007fda03a9c3ad n/a (/usr/lib/qt/plugins/platforms/libqxcb.so)
#6 0x00007fda029a0e62 _ZN27QPlatformIntegrationFactory6createERK7QStringRK11QStringListRiPPcS2_ (libQt5Gui.so.5)
#7 0x00007fda029ad052 _ZN22QGuiApplicationPrivate25createPlatformIntegrationEv (libQt5Gui.so.5)
#8 0x00007fda029adf8d _ZN22QGuiApplicationPrivate21createEventDispatcherEv (libQt5Gui.so.5)
#9 0x00007fda02677126 _ZN16QCoreApplication4initEv (libQt5Core.so.5)
#10 0x00007fda02677196 _ZN16QCoreApplicationC1ER23QCoreApplicationPrivate (libQt5Core.so.5)
#11 0x00007fda029afd2d _ZN15QGuiApplicationC2ERiPPci (libQt5Gui.so.5)
#12 0x0000000000406221 _ZN9SplashAppC2ERiPPc (ksplashqml)
#13 0x00000000004055d8 main (ksplashqml)
#14 0x00007fda01ce6610 __libc_start_main (libc.so.6)
#15 0x0000000000405639 _start (ksplashqml)
Jan 17 08:03:52 archaeopteryx kdeinit5[506]: QXcbConnection: Could not connect to display :0
Comment by Antonio Rojas (arojas) - Sunday, 17 January 2016, 12:32 GMT
Have you restarted your system after updating Qt (not just the Plasma session)?. Also, is this reproducible with a clean user?
Comment by Sebastian Frohn (sebafroh) - Sunday, 17 January 2016, 12:37 GMT
System rebootet after upgrade, downgrade of plasma desktop and after the qt5*-packages, too.
Comment by Sebastian Frohn (sebafroh) - Sunday, 17 January 2016, 12:40 GMT
Stack trace continues:

Jan 17 08:03:52 archaeopteryx systemd-coredump[507]: Process 506 (klauncher) of user 1000 dumped core.

Stack trace of thread 506:
#0 0x00007fd4231e95f8 raise (libc.so.6)
#1 0x00007fd4231eaa7a abort (libc.so.6)
#2 0x00007fd4235ed1e1 _ZNK14QMessageLogger5fatalEPKcz (libQt5Core.so.5)
#3 0x00007fd411f5a064 _ZN14QXcbConnectionC1EP19QXcbNativeInterfacebjPKc (libQt5XcbQpa.so.5)
#4 0x00007fd411f5f83d _ZN15QXcbIntegrationC1ERK11QStringListRiPPc (libQt5XcbQpa.so.5)
#5 0x00007fd424b5f3ad n/a (/usr/lib/qt/plugins/platforms/libqxcb.so)
#6 0x00007fd423b0ee62 _ZN27QPlatformIntegrationFactory6createERK7QStringRK11QStringListRiPPcS2_ (libQt5Gui.so.5)
#7 0x00007fd423b1b052 _ZN22QGuiApplicationPrivate25createPlatformIntegrationEv (libQt5Gui.so.5)
#8 0x00007fd423b1bf8d _ZN22QGuiApplicationPrivate21createEventDispatcherEv (libQt5Gui.so.5)
#9 0x00007fd4237e5126 _ZN16QCoreApplication4initEv (libQt5Core.so.5)
#10 0x00007fd4237e5196 _ZN16QCoreApplicationC1ER23QCoreApplicationPrivate (libQt5Core.so.5)
#11 0x00007fd423b1dd2d _ZN15QGuiApplicationC2ERiPPci (libQt5Gui.so.5)
#12 0x00007fd41205540a kdemain (libkdeinit5_klauncher.so)
#13 0x0000000000408560 n/a (kdeinit5)
#14 0x000000000040584c main (kdeinit5)
#15 0x00007fd4231d6610 __libc_start_main (libc.so.6)
#16 0x0000000000405d39 _start (kdeinit5)
Jan 17 08:03:52 archaeopteryx ksmserver[508]: Cannot connect to the X server
Jan 17 08:03:52 archaeopteryx ksmserver[508]: QXcbConnection: Could not connect to display :0
Jan 17 08:03:52 archaeopteryx systemd-coredump[509]: Process 508 (ksmserver) of user 1000 dumped core.

Stack trace of thread 508:
#0 0x00007f6ce2d515f8 raise (libc.so.6)
#1 0x00007f6ce2d52a7a abort (libc.so.6)
#2 0x00007f6cde5281e1 _ZNK14QMessageLogger5fatalEPKcz (libQt5Core.so.5)
#3 0x00007f6cce585064 _ZN14QXcbConnectionC1EP19QXcbNativeInterfacebjPKc (libQt5XcbQpa.so.5)
#4 0x00007f6cce58a83d _ZN15QXcbIntegrationC1ERK11QStringListRiPPc (libQt5XcbQpa.so.5)
#5 0x00007f6ce35273ad n/a (/usr/lib/qt/plugins/platforms/libqxcb.so)
#6 0x00007f6cdea49e62 _ZN27QPlatformIntegrationFactory6createERK7QStringRK11QStringListRiPPcS2_ (libQt5Gui.so.5)
#7 0x00007f6cdea56052 _ZN22QGuiApplicationPrivate25createPlatformIntegrationEv (libQt5Gui.so.5)
#8 0x00007f6cdea56f8d _ZN22QGuiApplicationPrivate21createEventDispatcherEv (libQt5Gui.so.5)
#9 0x00007f6cde720126 _ZN16QCoreApplication4initEv (libQt5Core.so.5)
#10 0x00007f6cde720196 _ZN16QCoreApplicationC1ER23QCoreApplicationPrivate (libQt5Core.so.5)
#11 0x00007f6cdea58d79 _ZN15QGuiApplicationC1ER22QGuiApplicationPrivate (libQt5Gui.so.5)
#12 0x00007f6cdf01151d _ZN12QApplicationC2ERiPPci (libQt5Widgets.so.5)
#13 0x00007f6ce30d236a kdemain (libkdeinit5_ksmserver.so)
#14 0x00007f6ce2d3e610 __libc_start_main (libc.so.6)
#15 0x00000000004007b9 _start (ksmserver)
Jan 17 08:03:52 archaeopteryx systemd[434]: Stopping D-Bus User Message Bus...
Comment by Antonio Rojas (arojas) - Sunday, 17 January 2016, 12:42 GMT
Please test with 5.5.1-7
Comment by Stanislaw Pitucha (viraptor) - Monday, 18 January 2016, 04:29 GMT
Confirming. Getting the same coredump on standard kernel.
Grsec kernel discovers the corrupted memory way earlier and doesn't even allow sddm to start up properly.

If you try enough times, you can log in actually, but the plasma loading progress bar is stuck for a long time at 100%.
(on 5.5.1-9 and -8, I'll follow up with -7 when it downloads)
Comment by Stanislaw Pitucha (viraptor) - Monday, 18 January 2016, 05:19 GMT
Same crash on 5.5.1-7, -8, -9. Which is weird, because everything worked just fine even with grsec on -8. This is a new machine, I never even had -7 on it.
Crash started only after upgrading -8 -> -9, so it looks like some cached data triggers it.

> "Also, is this reproducible with a clean user?"

The actual bug happens before the login. grsec/pax kills the sddm process before it even displays anything. On stock kernel it just writes in some random memory, continues to run, and crashes on login.
Comment by Sebastian Frohn (sebafroh) - Monday, 18 January 2016, 06:55 GMT
On -7 & -8 there have never been any problems - only with -9.
Comment by Antonio Rojas (arojas) - Monday, 18 January 2016, 10:41 GMT
I'm at a loss here. Felix - any idea what may be causing this? is there any difference at all between -7 and -9?
Comment by Antonio Rojas (arojas) - Monday, 18 January 2016, 10:42 GMT
Everybody: please remove *both* /var/lib/sddm/.cache and ~/.cache and see if it improves things.
Comment by Stanislaw Pitucha (viraptor) - Monday, 18 January 2016, 11:35 GMT
I started with just /var/lib/sddm/.cache for a start and it solved the issue for me. Unfortunately I didn't think to keep a copy for debugging :(
Multiple login/logout works without issues. Restarting sddm many times also works just fine. grsec kernel is also happy again.

So just to summarise, I reapplied all the updates (went to 5.5.1-9), got crashes, removed /var/lib/sddm/.cache, can't reproduce crashes anymore.
Comment by Felix Yan (felixonmars) - Monday, 18 January 2016, 12:06 GMT
The -7 and -9 PKGBUILD is exactly the same except pkgrel. The dependencies could be at different versions though.

So Stanislaw's issue seems really about sddm cache. @Sebastian could you confirm this?
Comment by Péter Károly JUHÁSZ (stone) - Monday, 18 January 2016, 15:18 GMT
Deleting the caches doesn't help for me, even with a clean user. Downgraded to -7 and it works now.

pacman -U qt5-base-5.5.1-7-x86_64.pkg.tar.xz qt5-declarative-5.5.1-7-x86_64.pkg.tar.xz qt5-graphicaleffects-5.5.1-7-x86_64.pkg.tar.xz qt5-location-5.5.1-7-x86_64.pkg.tar.xz qt5-multimedia-5.5.1-7-x86_64.pkg.tar.xz qt5-quickcontrols-5.5.1-7-x86_64.pkg.tar.xz qt5-script-5.5.1-7-x86_64.pkg.tar.xz qt5-sensors-5.5.1-7-x86_64.pkg.tar.xz qt5-svg-5.5.1-7-x86_64.pkg.tar.xz qt5-tools-5.5.1-7-x86_64.pkg.tar.xz qt5-translations-5.5.1-7-x86_64.pkg.tar.xz qt5-webchannel-5.5.1-7-x86_64.pkg.tar.xz qt5-webkit-5.5.1-7-x86_64.pkg.tar.xz qt5-x11extras-5.5.1-7-x86_64.pkg.tar.xz qt5-xmlpatterns-5.5.1-7-x86_64.pkg.tar.xz
Comment by Sebastian Frohn (sebafroh) - Monday, 18 January 2016, 16:48 GMT
No, removing the cache did'nt help (in -9). Downgraded to -8 and login works again.
Comment by Antonio Rojas (arojas) - Monday, 18 January 2016, 17:13 GMT
Just to make sure: have you restarted sddm (systemctl restart sddm) after cleaning the cache?
Comment by Péter Károly JUHÁSZ (stone) - Monday, 18 January 2016, 17:54 GMT
Sure, restarted.
Comment by Sebastian Frohn (sebafroh) - Monday, 18 January 2016, 19:17 GMT
Cleared, sddm restartet, cleared, rebooted...
Comment by Doug Newgard (Scimmia) - Monday, 18 January 2016, 19:29 GMT
Try the maui theme.
Comment by Sebastian Frohn (sebafroh) - Friday, 22 January 2016, 05:50 GMT
Always when there is a new plasma-framwork-package installed (and overwrites the cache of ?), KDE starts exactly one time.

By the way, there are some problems with the Screen, so a back border flickering at the left screen side while starting and sometimes later. This is only with -8, with -9 or -7 the display ist correct.
Comment by Dmitry (funbringer) - Sunday, 24 January 2016, 19:44 GMT
I've had exactly the same problem. Removing /etc/sddm.conf resolved this issue.
Comment by Péter Károly JUHÁSZ (stone) - Monday, 01 February 2016, 08:03 GMT
I can confirm that removing the /etc/sddm.conf solves the problem since it resets the theme, but if I switch back to Breeze the problem returns.
Comment by Antonio Rojas (arojas) - Wednesday, 03 February 2016, 20:10 GMT
With intel driver 2.99.917+544 it doesn't crash anymore for me, but the display stays frozen in the SDDM login screen until I switch to tty2 and back. Can someone confirm?
Comment by Sebastian Frohn (sebafroh) - Thursday, 04 February 2016, 07:19 GMT
It seems to depend really on Breeze-Theme for sddm - with other themes login works normally. And, if I failed to login with "sddm/breeze-Theme/actual package stack" and change session type from "plasma desktop" to "plasma media center" (which is not installed) or back, then login to desktop works for one (or up to four) times.

Loading...