https://bugs.kde.org/show_bug.cgi?id=521757
Thomas <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] --- Comment #9 from Thomas <[email protected]> --- I can reproduce what appears to be the same KActivities/Plasma CPU regression on CachyOS with Plasma 6.7.5. Environment Hardware: Minisforum UM890, AMD Ryzen 9 PRO 8945HS, Radeon 780M Distribution: CachyOS (Wayland) Kernel: 7.2.9-2-cachyos plasma-activities: 6.7.5-1.1 plasma-activities-stats: 6.7.5-1.1 plasma-workspace: 6.7.5-3.1 Observed behavior With plasma-kactivitymanagerd.service running, both KDE processes consume CPU continuously: plasmashell: approximately 120–125% CPU kactivitymanagerd: approximately 55–60% CPU AMD-reported PPT: approximately 30–35 W CPU temperature: approximately 71–77°C Stopping Activity Manager immediately reduces plasmashell CPU usage to approximately 0–1%, with AMD PPT falling to roughly 6 W and CPU temperature dropping to approximately 35–45°C. Restarting Activity Manager initially reproduced the high-CPU condition consistently. Database findings The Activity Manager SQLite database passed PRAGMA integrity_check. I identified resource names containing literal backslashes: ResourceScoreCache: 11 records ResourceEvent: 17 records ResourceInfo: 12 records I created a verified SQLite backup and removed these 40 records from a test copy. After replacing the active database with the cleaned copy, I restarted Activity Manager without restarting Plasma Shell. The CPU loop returned immediately, even though a subsequent database query confirmed that all three tables still contained zero backslash-containing records. Successful recovery I subsequently: Set KDE System Settings → Recent Files → Remember opened documents to Do not remember. Restarted plasma-plasmashell.service while Activity Manager was stopped. Unmasked and started plasma-kactivitymanagerd.service. Both processes then remained near 0% CPU, and AMD PPT returned to approximately 6 W. After a full reboot, Activity Manager remained active, Plasma remained near idle, and average reported PPT remained below 8 W. Additional observations The earlier database-only test did not restart Plasma Shell. Therefore, it does not rule out stale resource entries retained in Plasma memory. Likewise, because both recent-file tracking and Plasma's process state changed before the successful test, I cannot determine which change was essential. I have preserved the original SQLite database backup for further investigation but have not attached it because it may contain private file paths and activity history. This appears consistent with the activity-resource feedback loop described in this issue. The power-consumption impact was particularly significant on my system. I would be happy to perform further targeted diagnostics or test a proposed patch, provided the tests can be conducted without unnecessarily modifying personal resource history. -- You are receiving this mail because: You are watching all bug changes.
