Created attachment 508206 [details] emerge --info since the last qt update (from 5.9.2 to 5.9.3), clicking the calendar plasmoid causes the system memory to immediately fill up. I'm using Kontact with akonadi and have the calendar events module activated in the calendar plasmoid. how to reproduce: 1. open htop/ top on a tty in background 2. click on the clock plasmoid in your plasma control bar to open the calendar plasmoid (in my case with akonadi integration) 3. immediately switch to the tty and observe the plasmashell process eating up all the memory (in my case >8GiB) installed package versions: kde-plasma/plasma-workspace-5.10.5-r1 kde-apps/akonadi-17.08.3 dev-qt/qtcore-5.9.3
judging by the emerge.log this issue appeared after updating qt from 5.9.2 to 5.9.3 while neither akonadi nor plasma-workspaces were rebuilt.
Which akonadi backend did you choose?
I'm using the default MariaDB-10.1.26-r1 backend.
I can confirm this however with different versions: plasma-workspace-5.10.5 akonadi-17.08.3 qtcore-5.7.1-r3 Recompiling all akonadi/plasma/calendar related stuff solved the problem for me: emerge -va1 plasma-workspace plasma plasma-integration plasma-desktop PyQt5 kdepim-apps-libs libkdepim kdepim-addons akonadi akonadi-calendar Although I don't know if PyQt rebuild is really necessary. The downside of this brute force solution is that I cannot tell which package it was at the end. But I can life with that. My bet would be either plasma-integration, akonadi or plasma.
Recompiling any of plasma-workspace plasma plasma-integration plasma-desktop PyQt5 kdepim-apps-libs libkdepim kdepim-addons akonadi akonadi-calendar qtdeclarative didn't fix the problem for me. I recompiled each package separately and restarted plasmashell afterwards, the problem always remained. I ran plasmashell under valgrind, though, the log looks quite bad: https://seafile.orlives.de/f/9263703be348406e8c24/?dl=1 I'm currently running qtdeclarative-5.9.3, plasma-5.11.5, akonadi-17.12.1, libkdepim-17.12.1
(just for clarification: I killed plasmashell with SIGTERM once it started eating up memory, that's where the signal 15 comes from)
Problem occurred again after update to the latest amd64 stable versions of related packages. Rebuilding the aforementioned packages didn't solve the issue. However I realised that this error only occurred on special calendars....well I could ping it down to TODO events with an empty start time or a start time before the end time on todos (with all-day eventually created by other clients). Correcting/resolving those todos fixed it for me. t.schmittlauch can you confirm? If so, I'll fill a bug report upstream.
Several iterations of the packages in question later, does this still happen?
Tested it again for a week in daily use and also tried to reproduce the issue with the start time. Could not reproduce the issue on AMD64 stable with: plasma-workspace-5.13.5 akonadi-18.04.3 qtcore-5.11.1-r1
Thanks! Closing then.