Summary: | Plasma crash in ShellCorona::createWaitingPanels() after logging on | ||
---|---|---|---|
Product: | [Plasma] plasmashell | Reporter: | a.hoj |
Component: | general | Assignee: | Noah Davis <noahadvs> |
Status: | RESOLVED FIXED | ||
Severity: | crash | CC: | angelatwiggs123, arojas, bugzilla.rogelio, dashonwwIII, gjunk, gyy, kishore96, leviatan1, liubomirwm, mike.cloaked, nate, plasma-bugs, tckarthikeyan2211, tittigol |
Priority: | VHI | Keywords: | drkonqi, regression |
Version: | 5.23.0 | ||
Target Milestone: | 1.0 | ||
Platform: | Neon | ||
OS: | Linux | ||
See Also: | https://bugs.kde.org/show_bug.cgi?id=477941 | ||
Latest Commit: | https://invent.kde.org/plasma/plasma-desktop/commit/a6e732f35f4fcbfeeb0cbf41609e62c040326ab7 | Version Fixed In: | 5.23.1 |
Sentry Crash Report: | |||
Attachments: |
stack trace
minimal config test case New crash information added by DrKonqi |
Description
a.hoj
2021-10-14 13:13:39 UTC
If I type "kstart5 plasmashell" to Konsole, wallpaper show for a while, but early turns black again. Main panel never appeared. Created attachment 142431 [details]
stack trace
Distro: I'm running on Arch Linux fully updated with testing repo. My comment text got lost - here it is again: I too see plasmashell core dump after I upgraded from 5.22.5 5o 5.23.0 Downgrading back to 5.22.5 makes all fine again This occurred on intel NUC10i5 with CPU / graphics card: Intel(R) Core(TM) i5-10210U CPU VGA compatible controller: Intel Corporation CometLake-U GT2 [UHD Graphics] (rev 02). I also tested on a very old laptop and that did not crash if that helps. That has : Intel(R) Core(TM) i5-6260U 00:02.0 VGA compatible controller: Intel Corporation Iris Graphics 540 (rev 0a) Versions: kernel 5.14.12 using modesetting driver on Xorg - i have not tried wayland. qt 5.15.2 glibc 2.23 kwin, plasma 5.23.0 *** This bug has been marked as a duplicate of bug 435715 *** Are you sure this is a duplicate of bug 435715? The backtrace looks completely different to me, and this is a regression in 5.23 Yeah I guess you're right. Created attachment 142474 [details]
minimal config test case
This is a regression from 47697dce9bc9d443bb5b79869e434ca93d829830 Attached is a minimal config that reproduces it *** Bug 443738 has been marked as a duplicate of this bug. *** *** Bug 443780 has been marked as a duplicate of this bug. *** Noah, could you take a look? Thanks! Created attachment 142485 [details]
New crash information added by DrKonqi
plasmashell (5.23.0) using Qt 5.15.2
- What I was doing when the application crashed:
Upgraded from Plasma 5.22 to Plasma 5.23, logged out and then logged in again and i was greeted with a black screen and two crash reporters with the same backtrace. Tried to launch plasmashell from Konsole but it again crashed with this backtrace.
- Unusual behavior I noticed:
Black screen on login.
-- Backtrace (Reduced):
#4 0x00007f69a6e9e86b in QQmlObjectCreator::requiredProperties (this=<optimized out>) at ../../include/QtQml/5.15.2/QtQml/private/../../../../../src/qml/qml/qqmlobjectcreator_p.h:136
#5 QQmlComponentPrivate::requiredProperties (this=this@entry=0x55601e53f6e0) at qml/qqmlcomponent.cpp:344
#6 0x00007f69a6ea1748 in QQmlComponentPrivate::completeCreate (this=0x55601e53f6e0) at qml/qqmlcomponent.cpp:1071
#7 0x00007f69a6ea196d in QQmlComponent::completeCreate (this=this@entry=0x7ffe8951a700) at qml/qqmlcomponent.cpp:1066
#8 0x00007f69a6e92595 in QQmlEnginePrivate::singletonInstance<QJSValue> (this=this@entry=0x55601ce258e0, type=...) at qml/qqmlengine.cpp:2526
I'm guessing it's choking on some kind of interaction between a required property, the singleton and something from KDeclarative, but the KDeclarative thing could also be completely unrelated. The Kicker DragHelper is not in KDeclarative, but I wonder if I replaced it with a Qt Quick Drag attached property the crash would go away. After all, my patch to re-add drag and drop was apparently responsible for this crash. *** Bug 443838 has been marked as a duplicate of this bug. *** A possibly relevant merge request was started @ https://invent.kde.org/plasma/plasma-desktop/-/merge_requests/605 I have updated to 5.23.0-2 on three different machines. Two are machines which had new installs and new users added in the past year - they are fine after the update. A third machine has a user who has been running kde and then plasma for some years. On that machine although sddm greeter presents the user for login - once the password has been entered, the login hangs with a screen showing the splash but does not proceed to the graphical desktop. Nothing useful appeared in the logs. As a means to achieve a login, the .config user directory was renamed to .config.prev and then the login completes - but of course the desktop needs to be customised from defaults. It would be very useful to know which particular file or files in the .config/ directory are no longer compatible with 5.23 so that a minimal edit will achieve a login in these cases. I have another computer with long established user areas and they will need to be treated the same way. If the underlying issue can be diagnosed it would save some users quite a lot of work to fix their desktops once they update to 5.23 @MikeC as for me removing [Compositing] section from .config/kwinrc was enough to fix booting with plasma 5.23 so you may try that. For the record I had following settings in this file: [Compositing] GLColorCorrection=false GLCore=true GLPlatformInterface=egl GLPreferBufferSwap=e GLTextureFilter=1 HiddenPreviews=5 OpenGLIsUnsafe=false UnredirectFullscreen=false WindowsBlockCompositing=true XRenderSmoothScale=false @Ardith Metz thank you. I will certainly move that file aside on another machine when I update that in a couple of days - it will save quite a bit of work if that is the only change needed. By the way on the two machines where there was no problem logging in after the update that section of the kwinrc file has only the lines: [Compositing] GLCore=true OpenGLIsUnsafe=false So perhaps removing all other lines in that section, but leaving those two will be enough. When I have had a chance to do that in a couple of days I will report back to say whether that alone is enough to allow normal login to the graphical desktop again. *** Bug 443933 has been marked as a duplicate of this bug. *** I can now confirm that on two user accounts where there were additional lines in the [compositing] section of kwinrc changing the lines to include only two lines as per: [Compositing] GLCore=true OpenGLIsUnsafe=false and then logging in via sddm gave no repeat of the failure to login to the plasma desktop. The original lines in one case were: [Compositing] Backend=OpenGL Enabled=true GLColorCorrection=false GLCore=true GLPlatformInterface=egl GLPreferBufferSwap=a GLTextureFilter=2 HiddenPreviews=5 OpenGLIsUnsafe=false UnredirectFullscreen=false XRenderSmoothScale=false Clearly one or more of the lines in that section of kwinrc that had been placed in the file in the past is not compatible with 5.23, but it is useful to know that this simple change prevents the bug where login fails. Please open a separate report, that's unrelated to the original crash tracked here. Git commit d754291b64d7a19520bcd571c9f57848a7eb6a69 by Noah Davis. Committed on 18/10/2021 at 17:44. Pushed by ndavis into branch 'master'. kickoff: Replace Kicker DragHelper with Qt Quick Drag Originally, I thought it wasn't possible to use Qt Quick Drag here, but then I found out I could use Item::grabToImage()'s function callback parameter to set Drag.imageSource. M +10 -16 applets/kickoff/package/contents/ui/KickoffItemDelegate.qml M +0 -11 applets/kickoff/package/contents/ui/KickoffSingleton.qml https://invent.kde.org/plasma/plasma-desktop/commit/d754291b64d7a19520bcd571c9f57848a7eb6a69 Git commit a6e732f35f4fcbfeeb0cbf41609e62c040326ab7 by Noah Davis. Committed on 18/10/2021 at 17:46. Pushed by ndavis into branch 'Plasma/5.23'. kickoff: Replace Kicker DragHelper with Qt Quick Drag Originally, I thought it wasn't possible to use Qt Quick Drag here, but then I found out I could use Item::grabToImage()'s function callback parameter to set Drag.imageSource. (cherry picked from commit d754291b64d7a19520bcd571c9f57848a7eb6a69) M +10 -16 applets/kickoff/package/contents/ui/KickoffItemDelegate.qml M +0 -11 applets/kickoff/package/contents/ui/KickoffSingleton.qml https://invent.kde.org/plasma/plasma-desktop/commit/a6e732f35f4fcbfeeb0cbf41609e62c040326ab7 (In reply to Antonio Rojas from comment #23) > Please open a separate report, that's unrelated to the original crash > tracked here. Does your commit at https://github.com/archlinux/svntogit-packages/commits/packages/kwin/trunk for "Properly unredirect windows if compositing is not possible" entered earlier today now fix the compositor issue, which means an additional report is no longer necessary? The github commit I referred to is only for the arch package kwin - so I guess is not an upstream plasma fix - so should that be referred to a separate kde bug report to see if it can go in at source? *** Bug 444053 has been marked as a duplicate of this bug. *** *** Bug 444196 has been marked as a duplicate of this bug. *** |