https://bugs.kde.org/show_bug.cgi?id=468430
Marco Martin <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Latest Commit|https://invent.kde.org/plas |https://invent.kde.org/plas |ma/plasma-workspace/-/commi |ma/plasma-workspace/-/commi |t/5aec35f6be56aee6c32c3cdf7 |t/54a9dba9ec549cc06bdefaa72 |13825edfa30b6fa |7b2fcc52874c65e --- Comment #41 from Marco Martin <[email protected]> --- Git commit 54a9dba9ec549cc06bdefaa727b2fcc52874c65e by Marco Martin. Committed on 18/05/2026 at 15:29. Pushed by mart into branch 'master'. ShellCorona: Remove The invariant timer as is not deterministic enough Looking at bug 468430 and at sentry issues https://crash-reports.kde.org/organizations/kde/issues/267303?project=9 https://crash-reports.kde.org/organizations/kde/issues/401445/?project=9 It seems that sometimes it can happen that the invariants after a screen remove can be executed by the timer before handleOutputOrderChanged. when this happens, it causes the same backtrace as the issues. Now, the assumption is that every single time that a screen is added or removed, an output order change will happen eventually, so we check for the screen data integrity only after handleOutputOrderChanged, where we're guaranteed we are in a settled state. If this doesn't, that would be a bug of KSCreen or KWin. Adds a test that checks the internal state of shellcorona inbetween an screen removal and an output order change. M +58 -0 shell/autotests/shelltest.cpp M +7 -15 shell/shellcorona.cpp M +1 -3 shell/shellcorona.h https://invent.kde.org/plasma/plasma-workspace/-/commit/54a9dba9ec549cc06bdefaa727b2fcc52874c65e -- You are receiving this mail because: You are watching all bug changes.
