Page MenuHomePhabricator

Selecting a surface crashes app
Closed, ResolvedPublic

Description

The problem occurs in the vessel segmentation view, but seems to be related to the datamanager and/or cherry selection adapter (see call stack attached as file).

To reproduce perform segmentation on an image using vessel segmentation view, then perform boosting and cropping. Turning the surface to not visible leads to the crash.

Event Timeline

Removing dependency to vesselSegmentationView. This is a general problem of the propertyListView.

This seems to be fixed. I tried to reproduce the crash, but everything works as expected. Can this bug be closed?

This bug still exists and can be reproduced, but it only appears if the "Property List View" is activated.

Using only the default perspectives everything seems to work fine.

To reproduce:

Load an image (BronchialtreeWin32.pic.gz).
Change to VesselSegmenation, run a bronchialtree segmentation and confirm.
Turn the surface off.
Boost segmentation.
Turn the surface off.
Crop a branch.
Turn the surface off.
boost again.
Turn the surface off.

and so on...
at one of these steps while turning the surface to not visible, the app crashes. (See call stack attached)

Removing 3M3 keyword. This is not relevant as the property list is not delivered with 3M3.

(In reply to comment #4)

Removing 3M3 keyword. This is not relevant as the property list is not
delivered with 3M3.

Actually, it is delivered: Window->Show View->PropertyList
It just isn't part of any perspective. Maybe we should remove it from the installer?

When doing the workflow described in comment 3 but selecting the surface between the steps instead of turning visibility on and off, the same behaviour is observed. Suspicion: There is a orphan pointer which is held by some gui class and therefore a corrupted state of the property list is created by the respecting classes.

[SVN revision 20912]
FIX (#2978): member vars m_DeleteObserverTag and m_ModifiedObserverTag are now reverted back to -1 in OnObjectDelete to allow new observers to be registered for itk::DeleteEvent objects

shall I merge this in the 0.14 branch?

Looks like an important fix to me.

+1

Fixed by Michael Mueller. Thank you!