Yes, I've seen this too. I'm assigning this over to Nicole since I believe it is an issue with the Markups themselves and not with their use in this module.

Nicole, let me know what you think. This is one of the things I emailed about earlier.

After a first pass at debugging:
Disabling the picking manager for the interactors solves this problem, but that puts us back to having two old bugs reappear.
A workaround could be to disable the picking manager when you go into the landmarks module and reenable when you exit. I'd want to push the check/enable up from the fiducial displayable manager though, as I put it in a few places to make sure it stuck. Putting the picking manager enable in the init method of qMRMLSliceView would make it universal for Slicer.

@Nicole: Thanks for investigating. Does that mean there is a bug with the picking manager ? I don't think it make sense to disable/enable picking to pick a fiducial.

May be you could create a simple VTK example to illustrate the problem ?

As Nicole noticed, turning off the vtkPickingManager solves this bug, but brings back 0003319.

Translating the view allow to select the fiducial.

Interesting note: Translating the view allows you to move the fiducial, but it fails again when the fiducial is back at the exact center of the view!

This is reproducible with the Markups module: add a couple fiducials. In the Markups module enable Click to Jump Slices, then pick the Centered option. Some of the fiducials will be unmovable when centered.

I will try to make a pure-VTK example of this for testing.

Okay - I figured out what is wrong. It is an issue I will be able to reproduce in pure VTK. I have a work-in-progress patch to VTK that we can put in Slicer/VTK for use with 4.4 until something adequate is available in VTK proper.

The essence of the issue is between that way vtkWidgetHandles do picking vs the way the vtkPickingManager expects them to operate. That is even a widget that is not visible can be consider 'in front' for the purposes of the picking manager code.

This is simple to reproduce:

• create two markup fiducial lists with one fiducial in each one. Make the scale big and make them different colors for convenience.

• in the threeD view, rotate the camera so that one is in front of the other. Notice that with VTKv6, the one closer to you is always picked, which is due to the vtkPickingManager and is an improvement over VTKv5.

• now turn off visibility of the nearer fiducial using the eye icon of the markup module. Now the remaining fiducial is not pickable!

The reason is that the vtkPickingManager performs a Pick operation on all pickers in the scene and finds that invisible widget. The widget handles have custom internal actors that are the 'real' objects in the scene so they are too clever to work with the vtkPickingManager as currently implemented.

@jcfr I'd like to review this all with you and come up with a strategy so we can cut the release with a working VTK version for slicer, while putting in motion the correct fix for inclusion in a future VTK release.

@Steve: Great investigative work ! :) I was able to reproduce the problem. We can definitively push a fix/hack in our Slicer/VTK fork while we are maturing a more suitable solution in Gerrit. That way, the 4.4 release won't be on hold waiting for the fix to be integrated into upstream VTK.

Thanks @jcfr!

See the attached sscce.

I believe this is all fixed now - the details are in the pull request here:

https://github.com/Slicer/Slicer/pull/179

Thanks @ Steve. Great work on figuring our the problem and improving VTK infrastructure.

Have a good weekend.

Commit r23648 should fix that problem.
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&revision=23648

A fix was merged into VTK master that should address this: https://gitlab.kitware.com/vtk/vtk/commit/8e4f2f7c3a2e87e4567e5d1ff5fe93a32da8f1f0

@johan.andruejol

The fix merged into VTK master seems to have caused this exact same regression to reappear: https://issues.slicer.org/view.php?id=4634

Any thoughts?

My understanding is that the fix integrated in VTK addresses the problem for 3D, I think test for 2D viewer should be added in VTK.