SUMMARY Discover seems to be endlessly looping when clicking the "About" dialog. STEPS TO REPRODUCE 1. Open Discover 2. Click the "About" dialog OBSERVED RESULT Hangs spamming this two messages in the console: Qt Quick Layouts: Polish loop detected. Aborting after two iterations. file:///usr/lib/qt/qml/org/kde/kirigami.2/AboutItem.qml:135:5: QML FormLayout: Binding loop detected for property "implicitHeight" file:///usr/lib/qt/qml/org/kde/kirigami.2/AboutItem.qml:135:5: QML FormLayout: Binding loop detected for property "implicitHeight" Qt Quick Layouts: Polish loop detected. Aborting after two iterations. file:///usr/lib/qt/qml/org/kde/kirigami.2/AboutItem.qml:135:5: QML FormLayout: Binding loop detected for property "implicitHeight" file:///usr/lib/qt/qml/org/kde/kirigami.2/AboutItem.qml:135:5: QML FormLayout: Binding loop detected for property "implicitHeight" Qt Quick Layouts: Polish loop detected. Aborting after two iterations. file:///usr/lib/qt/qml/org/kde/kirigami.2/AboutItem.qml:135:5: QML FormLayout: Binding loop detected for property "implicitHeight" file:///usr/lib/qt/qml/org/kde/kirigami.2/AboutItem.qml:135:5: QML FormLayout: Binding loop detected for property "implicitHeight" EXPECTED RESULT Actually, not hanging :-). SOFTWARE/OS VERSIONS Linux/KDE Plasma: Linux 5.15.12-arch KDE Plasma Version: 5.23.5 KDE Frameworks Version: 5.89.0 (within that, kirigami2 5.89.0) Qt Version: 5.15.2
One thing I realized: This does not happen when the Discover window is maximized, only when it's in a non-maximized window. May be a Kirigami bug?
Cannot reproduce with current git master of everything. Does the same thing happen in System Monitor? Or Just Discover?
Created attachment 145382 [details] Screenshot of bug occuring/terminate dialog Hi Nate, thanks for looking into this. I attached a screenshot what it looks like when it crashes. No, this does not happen in the System Monitor (Pressing the three "lines" and "About System Monitor". Method used (discover git): $ cmake . $ make DESTDIR=/home/XXX/devel/deploy/discover install $ export LD_LIBRARY_PATH=/home/XXX/devel/deploy/discover/usr/local/lib/:/home/XXX/devel/deploy/discover/usr/local/lib/plasma-discover/:/home/XXX/devel/deploy/discover/usr/local/lib/libexec/:/home/XXX/devel/deploy/discover/usr/local/lib/plugins/discover/:/home/XXX/devel/deploy/discover/usr/local/lib/plugins/discover-notifier/:/home/XXX/devel/deploy/discover/usr/local/lib/plugins/plasma/ $ cd /home/XXX/devel/deploy/discover/usr/local/bin $ ./plasma-discover That's enough to reproduce it, though, ofc, it does not properly find plugins etc, but the "about" dialog shows the same issue as if I run discover from archlinux. I'm using two 1920x1080 monitors where the primary monitor is not the "primary" set in KDE; and they are switched sides (means the one detected left is the right one and the right one the left one, so I switched them in displays). If that may cause an issue. Is there anything I can do to help debug the issue? Best regards, G
PS: I replaced my username with XXX ofc ;-).
Thanks for the info. Still cannot reproduce with my git master version.
Hi Nate, that's somewhat ... strange. Though, I actually succeeded reproducing this on KDE Neon (which I installed for that purpose), so it probably really has to do with my setup and being a corner case, and seriously - who presses the about dialog but me :D. This will be difficult and strange to find out. I'd like to tackle this though. I have no former experience debugging kirigami, since looking at the code of discover, I think it hardly is connected to that, since you just use KF5 and within that kirigami. But I'll give it a shot and see if I can come up with better information. //G
@nate: I tried and could reproduce this on X11 KDE Neon User Edition with Discover window size of about 830x650, or 810x630 in other case in VM. (Maybe depends on how big the UI items are because of a language; the latter case was in English language Discover) Try if a size in that ballpark hangs the git master version. ( I used xdotool search --name "About – Discover" windowsize 810 630 )
That's an interesting finding - I'm on the english language too, but I doubt it's the issue. I did some more testing. If I delete my ~/.config/discoverrc and start discover it crashes in the About dialog. If I resize the window (and if just for a few pixels) it does not crash. It does not matter if i make the window bigger or smaller, after resizing it it never crashes again.
I just realized: This does no always happen, but for me about 90 % of the time in KDE Neon. Switching between the options "Installed", "Settings", "About" and "Up to date" reproduces it pretty well if I did not resize the window before. If I resize the window before - this does not happen.
I believe I'm hitting a very similar, if not the same, bug. If I click on many app entries, on some rare occasions, the interface freezes until I resize the window. I captured the terminal log and right when it freezes it outputs: Qt Quick Layouts: Polish loop detected. Aborting after two iterations. The same line is in this bug report (and this is how I found it too). If it's indeed the same bug, it means it's not only the About page that can trigger it, but app pages as well. Googling about this error points to this page: https://bugreports.qt.io/browse/QTBUG-88339 They describe that that output can cause apps to freeze, just like Discover is doing here. There's a comment also saying that >See https://codereview.qt-project.org/c/qt/qtdeclarative/+/299581 : a polish loop is when polish() is called within updatePolish(). If you do too much of that, maybe polishing never completes and your UI will hang; so we have a conservative warning and abort the polishing early to avoid people reporting bugs that it's our fault that their UI hung. If you see this warning, the application probably needs work. ------------------- (In reply to Georg Grabler from comment #9) > I just realized: This does no always happen, but for me about 90 % of the > time in KDE Neon. > > Switching between the options "Installed", "Settings", "About" and "Up to > date" reproduces it pretty well if I did not resize the window before. > > If I resize the window before - this does not happen. I managed to reproduce this issue by doing what you said, and the freeze looks exactly the same as when it happens when navigating applications. If you resize or maximize/unmaximize the window, it unfreezes.
The ~/.config/discoverrc which I can reproduce this every time (this is on a fresh fedora install for testing purposes): [Window] geometry=1920,29,810,540 visibility=2 where I do not see this issue if i minimally resized the window (note, it's just 1px resized): [Window] geometry=1920,29,810,541 visibility=2 It does not matter if the discoverrc exists or not, it's definitely an issue with the exact size (which seems to be the default size on my computer). I couldn't find what could make the difference in kirigami, kcoreaddons or discover why this would trigger polish() almost permanently (which seems to be the case). Seems like polish() is called too often, cancelled and then re-called by kirigami, but I couldn't find what could cause this in KAboutDialog.
This is a Fedora Linux 36 final blocker: https://bugzilla.redhat.com/show_bug.cgi?id=2057563 I can reproduce this every time with Discover 5.24.3, except when the window is maximized.
99% sure this is another manifestation of bug 448784. *** This bug has been marked as a duplicate of bug 448784 ***
Git commit 57187b7e6a1c6396c95b9450ff4c8acb8e70f431 by Aleix Pol. Committed on 08/04/2022 at 15:02. Pushed by apol into branch 'master'. AboutPage: Prevent infinite loop Do not use the width of the item but the width of the page to decide if we are on wide mode in the FormLayout. For some reason our default height hits a perfect storm of: - growing past the page height - adds a scrollbar - this narrows the view and we enable the compact mode - this makes it fit again - goto 0 M +2 -0 src/controls/AboutItem.qml M +1 -0 src/controls/AboutPage.qml https://invent.kde.org/frameworks/kirigami/commit/57187b7e6a1c6396c95b9450ff4c8acb8e70f431
*** Bug 452326 has been marked as a duplicate of this bug. ***
*** Bug 454915 has been marked as a duplicate of this bug. ***
*** Bug 456451 has been marked as a duplicate of this bug. ***
*** Bug 457675 has been marked as a duplicate of this bug. ***
*** Bug 458733 has been marked as a duplicate of this bug. ***