Pushed new branch T25759-ContextMenuCrashInMultilabelSegmentation.

This happens if the source image is set to the original image and the segmentation is set to the converted image. It does not happen if the source image is set to the pseudo segmentation image (the image that is converted to the segmentation). The pixel value in the crashing function is set to -1 then but there is no label with this pixel value leading to a nullptr access.

no unfortunately not. A guy from the radiology department asked and he claimed that he downloaded the 2018-04 release. I'll ask him again

Dark mode is default on all platforms and it works for me on our macbook and mac minis. Do you have more information and are you really sure that you didn't use the 2016-11 release? 🍺 😄

I mixed the Segmentation View and the Multilabel Segmentation View. It turned out that the latter one does not even have a protection against non-reinited image/segmentation pairs.

Pushed new branch T25750-SegmentationViewsResetOpacityOfSegmentations.

My initial thought was that I accidentally fixed a bug with the new context menu actions so looking closer into the original opacity-slider action in the old data manager might reveal the flaw that is responsible for this behavior.
Something like - not correctly setting the opacity using the opacity value from the data node when opening the context menu.

@kalali Thanks for looking into this. The bug that the Multilabel segmentation view resets the opacity to 1 is probably easy for me to fix for the upcoming bugfix release but I do not want to backport a big feature like the new context menu, which is out of the scope of a bugfix release. I may also try to sync the opacity slider to the opacity property in a seperate commit that we do not merge back to the master.

May it be that whenever the tools are deactivated that there is a message above like "please perform a reinit on the segmentation image"? I can reproduce your issue just by loading a scene file with your example image and a segmentation. Sometimes the tools are deactivated, and sometimes not (race condition of some kind). But the message is always shown when the tools are deactivated. The actual issue for me is, that the segmentation is not reinitialized even though the tools are activated right from the beginning, leading to very strange effects when using the tools.

As briefly discussed in the hall way:
I load two files into the object manager. One file is the MR image, the other file holds the segmentation for the MR image. The file holding the segmentation is then converted into a segmentation (right click -> convert to segmentation). This leaves me with 3 files in the data manager.

What do you mean by "the converted segmentation from this file"? So there's two or three files in the Data Manager?

Pushed new branch T25746-FixSegmentationToolsFor4DImages2.

The LiveWire tool stores a reference to a PlaneGeometry as raw pointer. Of course this doesn't go well for long and the pointer is invalid when scrolling through the image slices while drawing a contour. I replaced the raw pointer by a smart pointer. I also cleaned up the code.

Pushed new branch T25756-FixCrashInLiveWireTool.

I can reproduce your bug.
However, when using the latest master the problem is solved. I changed the context menu (and decoupled it from the data manager) and now everything works fine. This fix was done after the release.
However, one could discuss if reopening the multilabel segmentation plugin should actually reset the opacity to 100% or leave it as it is.