https://bugs.kde.org/show_bug.cgi?id=489862
Tom Dzmelyk <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Assignee|[email protected] |[email protected] Target Milestone|1.0 |--- Product|plasmashell |kwin Component|general |general --- Comment #16 from Tom Dzmelyk <[email protected]> --- I believe I've chased this down. The super noticeable stuck on edge behavior was a regression introduced by this commit: https://invent.kde.org/plasma/kwin/-/commit/31e0e256ad534c146bd42a6ad3c22bdc16d56685 . That is fixed now, but only in 6.4 by https://invent.kde.org/plasma/kwin/-/commit/64eb6cab7e9c88940a6743cd476640b5b3008b9b . Backporting that commit to 6.3 fixed 99.9 percent of the issue for me but I could still produced stickiness on slow movements. I easily reproduced this by doing steep diagonal inputs. The last 0.1% is that Workspace::outputAt() assumes a that the screen layout is contiguous, but for certain resolution and render scales Kwin isn't generating a contiguous layout. At 1440p and 120% resolution scale for dual monitors side-by-side it creates QRectF(0,0 2133x1200) and QRectF(2134,0 2133x1200), which leaves a one pixel gap between outputs. At 125% render scale it generates QRectF(0,0 2048x1152) and QRectF(2048,0 2048x1152) which is contiguous. Some logging I hacked to Workspace::outputAt() confirmed this. pos QPointF(2133,904.333) Geometry: QRectF(0,0 2133x1200) Output Distance: 1 Geometry: QRectF(2134,0 2133x1200) Output Distance: 1 Which explains the behavior I was seeing, whichever monitor is iterated over first wins. Moving from plasmashell to kwin so the right people get their eyes on it. -- You are receiving this mail because: You are watching all bug changes.
