3D Print Toolbox Intersection Check Problems

Question

The intersection checking function in 3D Print Toolbox seems to give a lot of false positives at the moment.

An easily replicable example (assuming it's a broad problem with the add-on, and not just my setup):

1. Open up a new scene.
2. Add a subdivision surface modifier to the default cube.
3. Set it to 'Simple' mode and the number of subdivisions to 5.
4. Apply the modifier.
5. Run the intersection check in 3D Print Toolbox.

This gives me a result of 112 intersections, despite this being a simple flat cube. The intersections seem to be random faces on the sides of the cube.

Interestingly, the following process gives different results:

1. Open up a new scene.
2. Add a subdivision surface modifier to the default cube.
3. Set it to 'Simple' mode and the number of subdivisions to 1.
4. Apply the modifier.
5. Repeat steps 2-4 four more times.
6. Run the intersection check in 3D Print Toolbox.

For me, this gives a result of 0 intersections, despite the second cube being seemingly identical to the first.

As you might imagine, on more complex models the intersection check seems to be largely useless at the moment. It seems to have a lot more trouble with geometric shapes than organic, sculptural geometry, though. Complex flat N-gons that have been triangulated (a common scenario in 3D printing) seem to be particularly problematic, perhaps because this often produces a lot of long, thin triangles?

So if anyone knows whether this is likely to be a problem with some setting, a problem with the add-on in general at the moment (I feel like it used to give better results in earlier versions), or just something obvious that I'm missing, then I would very much appreciate the assistance!

I'm using Blender 2.90.1

Thanks.

Answer 1

This is a real regression indeed, compared e.g. to version 2.83.4. I see you've also asked here: https://blenderartists.org/t/strange-results-using-3d-print-toolbox-to-check-intersections/1258126 (pasting to ensure others can find an answer if it's posted there).

The only solution for now is to keep the 2.8x version around and use it primarily for running the 3D analysis, then moving up to 2.90.1. I have no other solution for now.

Is the issue reported on the blender bug tracker? Given importance of 3D printing and such, I would think Blender can afford an auto test ensuring such regressions does not slip between stable versions.

Cheers.