The following SOPs take reference bounding geometry:

• OpenVDB Fill
• OpenVDB Clip
• OpenVDB Points Group
• OpenVDB Rasterize Points

and make calls to geo->computeQuickBounds(). This seems to evaluate "incorrectly" for VDBs as it's based on the VDB transform (and therefore overall table) rather than the VDB activity (which is what I'd expect).

This is easy enough to replicate, see the attached example. The hip file contains an example with the native clip and group sop. the image is with the aswf fill sop where it's easier to see the result.

JeffLait, I appreciate this may not actually be a bug, but I wanted to clarify if this was intended and whether or not there might be a better way to compute the bounds of incoming VDBs through the detail API. I'd expect the bounds to be equal to feeding a VDB through a bounds SOP.