I had previously mentioned how my system was freezing and had seen the cgroup memory continually climb…
The freezing was attributable to the ancient Haswell iGPU that is problematic, and only happened when using mpv with hardware acceleration.
But the issue of plasmashell (or something to do with it) taking up a crazy amount of Shmem, and the usage plummeting once I restart plasmashell, still exists.
It looks like heaptrack is not the tool to use to sniff that out because what I’ve got is not looking bad in terms of leakage in that image.
Before restarting plasmashell, /proc/meminfo showed Shmem: 1806844 kB.
And when I closed all windows, it only dropped to Shmem: 1426688 kB.
It’s currently at 434924 kB with most windows re-opened after restarting plasmashell.
If not heaptrack, what should I be using to sniff out the Shmem leavings?
Is there any hw acceleration used for plasmashell that I can disable as a test? It looks like kcmshell6 qtquicksettings set to “Software” might be the way. I thought that used to be in regular settings. I think I want to find out what other options exist for that command that I may be missing out on…
Again, I think this may be GEM-related: Making sure you're not a bot!
Searches relevant: How to diagnose a vRAM leak in plasmashell?