https://bugs.kde.org/show_bug.cgi?id=474768
--- Comment #17 from Harald Sitter <sit...@kde.org> --- I have good news and bad news: the good news is I know why this crashes and it should no longer happen with Plasma 6. the bad news is this isn't reliably fixable for Plasma 5. What happens is a fairly unfortunate set of circumstances. The applets in question all appear to use XHR [1] to access the local file system (deprecated in Qt 5 and actually raising a warning). XHR is implemented using QNAM [2] and that in turn is backed by an implementation of QNAM in KIO. So far so complicated. What happens before the crash is that plasma receives a new window event, feeds it into the grouping system and that issues a beginRemoveRows event. This then triggers QML binding re-evaluation. This causes the aforementioned XHR. Finally things explode because the XHR opens another event loop because of a well known defect [4] where KIO needs to run KJob::exec. During the nested loop another (similar) change event is received by Plasma and processed, changing internal states. When the call chain eventually returns to the original event processing the internal states are not what they should be and things explode. I'll put some bandaids in place for Plasma 5 but ultimately this problem can manifest in an endless number of ways. All we can do is hope that the bandaids will hold things together well enough. [1] https://developer.mozilla.org/en-US/docs/Web/API/XMLHttpRequest [2] https://doc.qt.io/qt-5/qnetworkaccessmanager.html [3] https://invent.kde.org/frameworks/kio/-/blob/kf5/src/widgets/accessmanager.cpp [4] https://invent.kde.org/frameworks/kio/-/blob/kf5/src/widgets/accessmanager.cpp#L303 -- You are receiving this mail because: You are watching all bug changes.
