Description of problem: In a clean cauldron net install the "Sorry but graphical server could not be started...." message that ends in "Good Luck :) " appears and then after a short delay (~4 secs) the sddm login screen appears and nothing is wrong. This is disconcerting and needs to be fixed. Not sure how to get the required logs from journalctl magic :\ Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3.
(In reply to Barry Jackson from comment #0) > > Not sure how to get the required logs from journalctl magic :\ > just do, as root, after having logged in with sddm: journalctl -ab > journal.txt and attach journal.txt to this bug report.
Keywords: (none) => NEEDINFOCC: (none) => marja11
Created attachment 8070 [details] journal.txt
Display server stopped 4 times: Jun 25 09:43:21 localhost sddm[1490]: Display server stopped. Jun 25 09:43:23 localhost sddm[1490]: Display server stopped. Jun 25 09:43:28 localhost sddm[3329]: Display server stopped. Jun 25 09:43:32 localhost sddm[3571]: Display server stopped. And there are sddm segfaults: Jun 25 09:43:30 localhost kernel: sddm[3329]: segfault at 0 ip 00007f0f71847e73 sp 00007ffc32c1ef10 error 4 in libQt5Core.so.5.6.0[7f0f715a9000+4a9000] and Jun 25 09:43:33 localhost kernel: sddm[3571]: segfault at 0 ip 00007f68dd085e73 sp 00007ffda679ab20 error 4 in libQt5Core.so.5.6.0[7f68dcde7000+4a9000] Assigning to sddm & its maintainer
Keywords: NEEDINFO => (none)Assignee: bugsquad => mageiaSource RPM: (none) => sddm
although it's a qt5 library that segfaults :-/ ...
Source RPM: sddm => sddm, qt5
I got similar issues on my old cauldron before I reinstalled it a few days ago. It happened only with kernel 4.6.x, but not with kernel 4.6.0 RC6 that I had been mistakenly using for a while. Now I can no longer reproduce it in a freshly installed cauldron, but I'll keep my eyes open :)
I just made a clean net-install of plasma to test for this (and other bugs) and this is still there.
duplicate of bug 17653 (or vice versa)?
Created attachment 8241 [details] journal.txt #2 FYI, I see exactly the same errors as in comment 3 on real hardware, with Mga6-sta1 and newer. Maybe is the GUI trying to start before another service is initialized, which would explain why the GUI finally appears after a few seconds?
CC: (none) => LpSolit
Please check you have a fully updated system, so theese packages are installed: initscripts-9.55-21.mga6 plymouth-0.9.2-5.mga6 sddm-0.13.0-6.mga6 xinitrc-2.4.21-17.mga6
CC: (none) => tmb
(In reply to Thomas Backlund from comment #9) > Please check you have a fully updated system, so theese packages are > installed: > > initscripts-9.55-21.mga6 > plymouth-0.9.2-5.mga6 > sddm-0.13.0-6.mga6 > xinitrc-2.4.21-17.mga6 Yes, fully updated, and that's the versions I have.
Yes still same here: [baz@jackodesktop ~]$ rpm -q initscripts plymouth sddm xinitrc initscripts-9.55-21.mga6 plymouth-0.9.2-5.mga6 sddm-0.13.0-6.mga6 xinitrc-2.4.21-17.mga6
Created attachment 8243 [details] journal output 24-7-2016 journal output 24/7/2016 fully updated.
(In reply to Barry Jackson from comment #12) > Created attachment 8243 [details] > journal output 24-7-2016 > > journal output 24/7/2016 fully updated. Jul 24 00:44:23 jackodesktop kernel: sddm[3596]: segfault at 80000004 ip 0000000080000004 sp 00007fff3aed17b8 error 14 Jul 24 00:44:16 jackodesktop kernel: sddm[1603]: segfault at 0 ip 00007fddd0c3ec43 sp 00007ffe2ef8e330 error 4 in libQt5Core.so.5.6.1[7fddd099e000+4ab000] Maybe I was wrong to think that the "error 14 in sddm" segfaults are a different bug than the "error 4 in libQt5Core.so.5.6.1" ones, so that this would be a dup of bug 17653, after all?
Summary: Sorry but the graphical server failed to start (or words to that effect) then it starts after another 4 seconds => SDDM segfaults - error 4 in libQt5Core.so.5.6.0 [was: Sorry but the graphical server failed to start (or words to that effect) then it starts after another 4 seconds]
I wonder if this bug has already been fixed in Qt 5.7 as such problems have already been reported in Qt 5.6, see e.g. https://bugs.kde.org/show_bug.cgi?id=348194#c20.
As said in bug 18822 comment 60, switching to LXDM or LightDM (using drakdm) and then switching back to SDDM fixes this issue. Barry, can you confirm?
Yes, I confirm that installing the lxdm package (it was not installed), switching to it in MCC and using it once fixes this. Removing the lxdm package (and theme) then makes no difference, SDDM continues to work without the segfaults. My installation was a net-install later than the isos and I have another with this issue that was installed much later than the last isos.
I did a clean net-install of plasma, installed lxdm but did not yet use it. I then made a copy of /etc as /etc_bak. I then set LXDM in MCC as DM and used it once. I then set DM back to SDDM and rebooted. I then took a diff of /etc against the original bakup. [baz@localhost ~]$ su Password: [root@localhost baz]# diff -ur /etc /etc_bak Only in /etc: .pwd.lock Only in /etc/sysconfig: desktop Binary files /etc/sysconfig/harddrake2/previous_hw and /etc_bak/sysconfig/harddrake2/previous_hw differ Only in /etc: .updated [root@localhost baz]# cd /etc [root@localhost etc]# cat .updated This file was created by systemd-update-done. Its only purpose is to hold a timestamp of the time this directory was updated. See systemd-update-done.service(8). [root@localhost etc]# cat /etc/sysconfig/desktop DISPLAYMANAGER=SDDM [root@localhost etc]# /etc/sysconfig/desktop had DISPLAYMANAGER=LXDM after changing to LXDM and =SDDM after changing back, but /etc/sysconfig/desktop did not exist from fresh install. I will try removing /etc/sysconfig/desktop to see if issue re-appears.
Bingo! \o/ Yes removing /etc/sysconfig/desktop restores the issue. So the problem seems to be that a clean install of plasma with SDDM does not create /etc/sysconfig/desktop with DISPLAYMANAGER=SDDM
You rock Barry :D It's crazy that a missing line like that lets SDDM work nevertheless but with segfaults... Assigning to Thierry, as I guess the installer needs to be fixed to write /etc/sysconfig/desktop.
CC: (none) => mageiaAssignee: mageia => thierry.vignaud
Priority: Normal => release_blocker
Are you kidding? The segfault must be fixed! As for autologin configuration, we current do not support sddm yet (which is covered by bug #17913 -- work in progress) so we rely on the generic autologin package for now.
Status: NEW => ASSIGNEDAssignee: thierry.vignaud => mageia
I am obviously not kidding, but sorry if I offended you by my proposal to fix the issue as drakdm fixes it already. If we could fix the segfault instead (or more likely do both), that would of course be very welcome.
i confirm that i had the "ooops" error and adding the desktop file fixed it for me too.
So should this be split into two bugs? 1. SDDM segfaults when /etc/sysconfig/desktop is missing 2. Plasma default installation is missing /etc/sysconfig/desktop
Assignee: mageia => kde
Still happens in nonfree net-install today.
QA Contact: (none) => stormi
@ Nicolas, any clue how to debug this sddm segfault now that we know how to reproduce it? @ Thierry: Would the part about the missing /etc/sysconfig/desktop on SDDM installs be fixed already, or should it be split into a new bug report?
CC: (none) => thierry.vignaud
/etc/sysconfig/desktop is a prefdm file generated by drakedm It may or may not exist. This shouldn't affect sddm. If sddm segfaults when it doesn't exist, that's an sddm issue.
I think we should get rid of prefdm, this can fix this issue too. Where is prefdm configurer in the installer ?
prefdm relies on /etc/sysconfig/desktop
yes but we need to get rid of prefdm ( it had been removed from systemd 230. We are one of the latest distribution to use it :/
Is fixed since latest systemd-unit update that made prefdm not conflict with plymouth. Please reopen if I was wrong.
Status: ASSIGNED => RESOLVEDResolution: (none) => FIXED