https://bugs.kde.org/show_bug.cgi?id=306352
--- Comment #17 from Timo Gurr <timo.g...@gmail.com> --- (In reply to Robin Bankhead from comment #16) > Just to update on my earlier, I haven't found a way of re-triggering the > high CPU usage after killing and relaunching kglobalaccel5 at the start of > my VNC session, and it has not recurred again during such a session. (It's > possible it has happened without me noticing though.) > > This is now under kglobalaccel-5.99.0, plasma-desktop-5.26.2. I have a relatively reliable way to reproduce this issue, first of course as already known it happends on every login / plasma session start at least on X11 and requires a manual restart of the plasma-kglobalaccel user service to behave again: $ systemctl --user restart plasma-kglobalaccel.service This happens on another hardware machine with completely different graphical setup where I use Intel Iris mesa driver contrary to the NVIDIA proprietary one on my desktop, and I'm also able to reproduce this on VMware where my guest uses the SVGA3D driver so I don't think it has anything to do with the graphical stack per se. We also got reports from other users of our distribution experiencing the same. Now for being able to reproduce it in a running plasma session, it always happens to me when using Steam, not when playing regular games on the desktop itself but when using the Steam Link Android app on for example my TV (but should probably also work on a device like a phone to reproduce - not tested this myself yet though) to stream the games, when I come back to my desktop I can see the kglobalaccel5 process churning away with 100% again. To make it a little bit less annoying I've now setup KDE Connect to issue the above mentioned command so I'm able to remotly restart the service but it's still annoying as hell especially as I have no idea when it kicks in, it could also happend during playing and/or when restarting a game and influence the performance. I would still like to have some guidance on how to be able to debug the issue. Maybe for yet another workaround I'll gonna try adding something like e.g. CPUQuota=1% to the plasma-kglobalaccel.service [Service] section. Still, I'd would love to see this but get some attention from the maintainers who have some deeper knowledge what could be the root cause and how this issue materializes. Also I have no clue why it doesn't happend on e.g. a Fedora 36 KDE Plasma install next to me having more or less the identical hardware setup and also using an X11 session. Afaiks this issue started when the change over to systemd user units was made with a plasma update, but this is just a vague guess as I first thought it might be a issue on my machine and kept on killing the process manually until having time to search for bug reports and found this one. However since we're recycling an original issue from 2012 here it would've probably been better to create a new bugreport even if it's about the same outcome, the issue is most probably not the same as the original report here. -- You are receiving this mail because: You are watching all bug changes.
