Only tests with a scaling != 1 fail. All tests which fail render Pic3D in either sagittal or coronal view.

The valid images show half pixels at bottom and top of the image. I guess that is incorrect and we have to reproduce them. However, we still dont know the correct size of the rendered image. Either Mac/Linux are rendering wrong or Windows is.

If you load an image in the Step2 tutorial, everything works fine. So both results, windows and mac/linux, are incorrect.

Qt must somehow interfere and takes care about the correct zooming and size of the image.

I figured out that a resize() call on the mitkVtkPropRenderer solves this issue, because inside a resize of the BaseRenderer is called and there the display geometry is adapted to fill out the complete render window.

I can push a quick on Wednesday or after the release. But we have to decide if this is desired behavior for a rendering pipeline.

This bug was solved in the context of T10161