If the opacity of a segment is reduced from the default 1.0, either for the whole segmentation or for the selected segment, the 3D representation does not persist. It appears again if set to 1.0.
This is seen in Linux builds on 2 machines (5.10, today’s preview, self-built), all of them using Qt5.
It does not seem to have been reported yet, though it is observed in 5.10 stable.
May be some users might report if it’s a systemic issue.
Thank you.
This video in 5.10 is demonstrative.
I’m also unable to replicate using Slicer 5.10.0 on the Windows platform.
Therefore, this may be Linux specific? @chir.set Did this suddenly start happening or do you think it has been an issue for awhile? Do you know a version when it did work before? Just wondering if it is a graphics issue specific to your machine, or something more generally observed on the Linux platform you are using.
I noticed that as from today. I can’t say if it existed a few days ago since I don’t tweak opacity of segments every day.
Both machines run Arch, regularly updated, and have a Ryzen GPU. AFAIU, the 3D representation of segments is not GPU rendered like Volume Rendering, the latter having no malfunction.
I’ll be attentive to this and report if it gets back to normal, magically.
Thanks for the video. I tried the exact same thing on a Ubuntu machine with an nvidia card and did not see the behavior (all worked as I expected). Maybe you can try on a different machine or boot a different OS, since it may be GPU or driver related.
5.10 is widely used and nobody has reported this so far, so most likely this is something specific to your system, such as a recent regression in a video card driver.
SLICER_OPENGL_PROFILE environment variable to no, core, or compatibility?Yes, with the builds from download.slicer.org.
The 3D opacity is rightly applied in Ubuntu 24.04 booted as an ISO, and in a third Arch machine that has been updated about 3 weeks ago.
Else, it’s on or off with any tweak : SLICER_OPENGL_PROFILE, 5.10, latest preview, Wayland, X11, older kernel, disabling slicerrc.py, creating a new user… I could not test depth peeling as I don’t find it in settings.
Thanks all for your input, wait and see…
The solution found here is to use:
export MESA_LOADER_DRIVER_OVERRIDE=zink
The problem occurs similarly for Models and for surfaces rendered in the ExtraMarkups extension, i.e, for any transparent polydata surface.
The default driver for AMD GPUs is radeonsi, which seems to be in cause. But maybe VTK has some additional things to consider here.
The fallback llvmpipe driver works also.
The fundamental fix has been applied in the mesa package.
To be complete, the zink driver (Vulkan) resolves the transparency problem, but is found bad for Volume rendering.
@chir.set Hi, after looking into your issue, I noticed an interesting similarity with a problem I’m experiencing on Linux.
When I set the segmentation opacity to anything below 1.0, the rendering also becomes incorrect. In your case, the segmentation disappears completely, while in my case it becomes fragmented/corrupted.
The symptoms are different, but both problems seem to start when the opacity goes below 1.0. Could you take a look and see if you can guess what might be happening in both cases, and whether they could be related somehow?
Link to my issue: Color display error on 3D Volume
That’s a hard exercise 
Look for any messages in the application log and in the python console. Start from the command line and see if there are unusual output. Try to determine the changes in your machine since the last normal result. Tweak the application settings relevant to display and segmentation. …