https://bugs.kde.org/show_bug.cgi?id=398908
--- Comment #119 from Alexander Meshcheryakov <alexander....@gmail.com> --- I have noticed that memory leaks are happening not gradually but in sharp spikes. During such spikes Dolphin consumes about 70% of a single CPU core. I have not figured out the trigger of such moments. Sometimes leaks happen even while user session is locked, and as far as I can tell there are no updates in content of directories which Dolphin displays. But it seems leaks happen more often when I interact with my PC. The patch from #c116 does not remove these CPU usage spikes. But with patched solid instead of leaking ~200 MB at a time, Dolphin leaks just 5-20 MB. Here is some data: Obligatory heaptrack of Dolphin running for 3 days with solid patched from #c116 http://45.86.162.4/files/2022/heaptrack.dolphin.solid_patch_116.19.zst CPU and RSS consumption graphs for Dolphin with patched solid http://45.86.162.4/images/2022/dolphin_patched_solid_cpu.png http://45.86.162.4/images/2022/dolphin_patched_solid_rss.png Here are the graphs for Dolphin with unpatched solid http://45.86.162.4/images/2022/dolphin_cpu_2.png http://45.86.162.4/images/2022/dolphin_rss_2.png Note how memory leaks are simultaneous with CPU spikes. >From all these observations I am rather confident that the core issue is not an inefficiency of StorageAccess::clearTextPath which is fixed by Méven Car patch, but rather mysterious spurious ~2 min intensive process related to KDirWatch. I suppose this process does not do anything useful and should be disabled/optimized. Also this process seems to be related to filesystem tags (user.xdg.tags in xattrs). If I remove /home/self/.local/share/user-places.xbel at the moment it gets populated with tags:/ again Dolphin leaks. -- You are receiving this mail because: You are watching all bug changes.
