Let's find out :-)
Great analysis, thanks! Now this makes a lot more sense ;-)
Ok, this seems to work now here, too ;) Will merge it, thanks again.
Yeah, I just missed to recheck the sorting logic. Before the change we just needed the best match so the partial sort was completely sufficient. Now we need a complete sorted vector, so stable sort is the way to go ;)
I think it would be nicer to have it directly as enum in the config, otherwise in one year, we will be all confused how these two bools interact.
Thanks for taking a look again ;=)
We could do that, I just didn't not to break currently existing systems
That a nested event loop of an error dialog caused the crashes makes sense, because they were reported for alien drives (NTFS) or transfers with permission problems. What actual error dialogs did you get and are they reproducible?
Thanks for cleaning up the tests after me!
Implement a proper fix
I think I have finally found the root cause of this. The following events occurs:
No problem, I'll take care of this for you!
@apol I didn't applied for KDE Developer Account yet (I'm new here) so I can't, sorry.
Cool! :D welcome to KDE!
Can you land the patch or should we do it for you?
I think this looks good now, but I would prefer an OK from another developer, as this has become quite large.
These two conversion functions should just be deprecated, as they are not expecially useful.
- fix MTimeDB::get loop
Landed, don't know why this didn't update.
Remove unrelated debug info
- review warning fixes
Ah sorry completely missed that part. Unfortunately I'm out of office already, I'll take a look as soon as I'm back tomorrow.
Actually, do we need a partial sort? Should we not just stable_sort the complete vector by prio? That should keep stuff with same prio untouched, or?
I think the + 1 is the only real issue, the other issue stems from my local setting that I have no compiled in QRC.
I think one error is the sort of the empty vector with the begin() + 1
But even then, make test fails, after fixing this.
Please take a look and run the tests.
The partial sort is bad ;=()
Beside that this fails ;=)
Yes, this is nice, nice trick with value(0) to avoid adding an element to the vector but just letting the API default construct it.
That should implement all the latest suggestions/comments