Halo computation in downsample function does not work correctly

[constantinpape]

fortunately it is only necessary for the interpolate mode, for the others the 'natural' halo that results from upscaling the bounding box is sufficient.

[martinschorb]

Do you think this is due to the same iterpolation?

That's what happens when setting the EM sources to occluding.

[martinschorb]

During my conversion, I call make_bdv without any specific downscale_mode.
Since when is the default nearest? I guess that should be what was used here...
So maybe it has instead something to do with the viewer dealing with downsampled occluding sources. Probably if the image size does not match the downscaling factors...

[martinschorb]

just talked to @tischi In this case it is happening on the viewer side.