hallas added a comment.
In D19170#415760 <https://phabricator.kde.org/D19170#415760>, @dfaure wrote: > If some code is deleting this job from a slot connected to it, that code needs to be fixed. This hack isn't a fix, it will only create more problems.. > > E.g. if some other bad code runs the event loop in the slot connected to finished, it will now delete the job, and we'll get the exact same crash but on `if (emitResult)` this time. > > This is a big no from me. The emitter of a signal MUST NOT be deleted in slots connected to it. > > The bug needs proper analysis. Hi David, thanks for the reply :) I completely agree with you that this is a hack and shouldn't go in, it was merely meant as an aid in doing proper analysis of the bug. I hope one of the people who can actually reproduce the problem can try with this patch enabled and see if it goes away. If it does then I think we can be pretty certain that someone is deleting jobs that they shouldn't. Otherwise I don't have much to go on. I have tried to reproduce the problem in various ways, copying to/from an NTFS drive and also copying to/from SMB but I can't get it to crash. I have been looking some more into the code and I still cannot figure out how the KJob instance can get invalidated. Have you looked at the backtraces? Do you have any ideas or pointers to where to look? Appreciate the feedback REPOSITORY R244 KCoreAddons REVISION DETAIL https://phabricator.kde.org/D19170 To: hallas, #frameworks, elvisangelaccio, dfaure Cc: dhaumann, kde-frameworks-devel, michaelh, ngraham, bruns
