https://bugs.kde.org/show_bug.cgi?id=506919

Harald Sitter <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |WAITINGFORINFO
             Status|CONFIRMED                   |NEEDSINFO

--- Comment #13 from Harald Sitter <[email protected]> ---
Lifetime management is super convoluted here and the theory would be that we
are simply tapping into an already deleted object, depending on timings (that
we may well only hit on fedora for some reason).

We have a deleteLater, a manual delete, a setTransaction which relinquishes
control of the UpdateTransaction but also taking control of an outside created
transaction (albeit in a strangely different way than when the
UpdateTransaction constructs -- why are we even re-managing signal
connections?), we also have non-trivial connections to the destroyed signals
(huge risk of blowing up on partially destructed object trees which I think we
are seeing in at least some of the sentry events with wildly different traces).

All that said, since the sentry issue is an amalgam of at least 6 different
crashes I've seen so far, I cannot dig for context or find anything.

Can someone who can reproduce this problem maybe run discover through valgrind?

And maybe also build it with a patch from source so we can add some checks?

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to