D3977: Fix memleak in KDynamicJobTracker, KWidgetJobTracker needs QApplication

[Friedrich W. H. Kossebau]

kossebau added a comment.


  In https://phabricator.kde.org/D3977#155721, @elvisangelaccio wrote:
  
  > @kossebau 
  >  After this change we get the
  >
  >      Tried to unregister a kio job that hasn't been registered.
  >   
  >   warning with a `KCompositeJob` in Ark (batchextract.cpp).
  >   
  >
  > The job is now automatically unregistered and the `unregisterJob(this)` call in the job destructor is causing this warning.
  >  Are KCompositeJobs still supposed to unregister themselves?
  
  
  KJob related knowledge sadly completely swapped out of my brain, would need to investigate myself. No idea if there should be done anything special for KCompositeJobs.
  
  On a quick look the automatic unregistering based on the finished signal seems to make sense for any KJob subclass. Perhaps the change in this commit conflicts with some older code in Ark to work around the old behaviour and which does some additional manual unregistration? Or perhaps some slot in accidentally invoked 2x so the unregistration happens more than once, with the second try than failing? Wild guesses is all I have to offer, sorry. I cannot remember anything questionable about this patch, so no pointers into it, instead my initial reaction would be to look at the consumer side.

REPOSITORY
  R241 KIO

REVISION DETAIL
  https://phabricator.kde.org/D3977

To: kossebau, #frameworks, kfunk
Cc: elvisangelaccio, kfunk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20171015/5a610a05/attachment.html>