Pushed new branch T26219-Crash_by_selection_widget_after_close_project.

Found it.

OK, I have narrowed it further down.

Pushed new branch T25779-StpFiles2.

I experienced a similar problem, and I opened an issue in CMake: https://gitlab.kitware.com/cmake/cmake/issues/19132 .

Pushed new branch T25826-RadiomicsGUIUpdate-V2.

Pushed new branch T26234-TSE.

Another place where inverse matrix calculation has been introduced:

As I see, the inverse matrix calculation came in with this commit of mine:

@kalali @kislinsk , added you two has you have some experience with at least either rendering/mapper or selection concept. May be you have additional ideas.

The problem occured in context of selection widgets and may be could be patched for selection widgets in context of T26218.

This assumption was incorrect :(. Even if the widget properly release the node if it is deleted the problem still occures?!?

The callstack as additional information:

Was too fast? Thought you might meant what I added. If you had something else in mind shall I reopen the task?

Pushed an implementation for option 2. Widgets are now synced and remove selections that are not part of their datastorage any more.

Pushed new branch T26218-Sync_node_selection_widget_with_datastorage_node_delete.

Uh oh, that was fast. ;)
I hadn't even made up my mind on that matter. My first purpose was to open up the discussion on that topic...

I just realized that it is actually checked in SetIndexToWorldTransform :-D

While I changed the matrix check, I left it at the current position in in the PlaneGeometry.
We should discuss if it is sensible to move this ckec and also it's applications to a different location, e.g. to SetIndexToWorldTransform of BaseGeometry.

Pushed new branch T25682-DominantAxesError.

Pushed new branch T26217-FfRefBval.

Does anyone know why the inverse is always calculated and not simply set to the transpose? After all, we are talking about rotation matrices.
In this context, the EnsurePerpendicularNormal method (https://phabricator.mitk.org/source/mitk/browse/master/Modules/Core/src/DataManagement/mitkPlaneGeometry.cpp$37) seems rather complex and also does not catch all cases since it only checks if axis 3 is perpendicular to 1&2 but not if 1&2 are perpendicular to each other. I would prefer to simply check if the matrix R is a rotation matrix by checking

1. det(R) == 1?
2. R*R^T = ID?

Pushed new branch T26228-VisualStudio2019.

The meaning of the dominant axes is also documented here:

Pushed new branch T26188-fixes.

It's apparently an issue of the "dominantAxis" calculation. In some cases this does not make sense and yields the same index twice. Not sure what this "dominantAxis" is supposed to be anyway. Supposedly related to T22254.

What other aspects do you mean?