System information

• SolveSpace version: master @ 5671941 (3.0)
• Operating system: macOS 11.5.1 (Big Sur)

Expected behavior

Opened turc-nurbs-1.slvs.zip
Expected no NURBS issues

Actual behavior

The 'board-1-inset' group shows "The Boolean operation failed", there are angry red lines around some of the geometry from this group and Analyze->Naked Edges reports "The mesh has naked edges (NOT okay, invalid). 10 problematic edges, bad." On the console I get:

failed: I=20, avoid=7
print 4 edges
failed to assemble polygon to trim nurbs surface in uv space

Additional information

The problematic model has a couple of holes through a solid board (called "top-plate" in the model), modelled as a difference extrude ("board-1-holes" being the extrude of the "board-1" sketch). These holes are countersunk by adding a second sketch of slightly-larger holes and a second difference extrude that doesn't go all the way through the board ("board-1-inset-sketch" and "board-1-inset"). I've simplified board-1-inset-sketch to reduce the number of holes, and only countersink the top-left hole, to hopefully make the test case easier. When I tried removing some more construction geometry from "board-1" the problem went away and I'm not sure why.

The overall project that this model is part of quickly degenerated into having to tick "Force NURBS surfaces to triangle mesh" early on and is pretty complex and incredibly slow to recalculate (Generate::ALL takes 30s). I read a comment on another issue suggesting that the triangle meshing might be contributing to the performance issues, hence why I'm raising this to see if I can get the whole thing to work in NURBS.