https://bugs.kde.org/show_bug.cgi?id=520190
--- Comment #12 from [email protected] --- Yeah, kudos for the deep dive and finding a fix for at least one path to that leak! I was wondering why some people can't reproduce this on i915 and found out one more reason while testing on more iGPU models: large values of scaling (>= 200%, fractional or not doesn't seem to matter) may also be needed to trigger. Just to be clear, the GPU mem leaked from plasmashell does stay leaked until the process exits, I'm not talking about the normal fluctuating gem count and VRAM usage (which can be pretty substantial, easily a few hundred M when displaying tooltip previews for multiple windows in a collapsed group) - that bounces back very quickly. -- You are receiving this mail because: You are watching all bug changes.
