Kenny

I still think TAU should be exported from constants, NOT added to the global Math object. I don't like modifying the global math object. It breaks encapsulation of the jscad...

Yep this looks like an issue. Here is a simplified SVG that shows the same problem: ```svg ```

If you look at the imported SVG in 2D you see the correct "shape":  But when extruded it's not treating it like a hole:  The problem is that...

Actually after some more investigation I realized I was wrong. The real problem is that the deserializer produces two separate `geom2`'s instead of one. If you "merge" the two objects...

Intersect is not the correct solution, in general. It works specifically in this case because the hole essentially defines "all the plane except this hole" and the outer polygon defines...

I used `git bisect` to track the regression specifically to commit [b29993a8](https://github.com/jscad/OpenJSCAD.org/commit/b29993a8fd3da3bd43a5f871edae4c60d7f56fc7) from PR #928 It appears there was previous discussion of lighting bugs between @z3dev and @hrgdavor in #928...

Should the transforms be applied before it gets to the shader? Or should the transforms be passed in and applied inside the shader? Is [vColorShaders.js](https://github.com/jscad/OpenJSCAD.org/blob/master/packages/utils/regl-renderer/src/rendering/commands/drawMesh/vColorShaders.js) the relevant code?

> @platypii are you still looking to make these changes? V2 or V3? I still strongly believe this is a good change. It makes the designs more robust, for cases...

@z3dev I re-rebased it onto V2 master branch again. Please review :-)

@z3dev good call. I added similar logic and tests for zero size for rounded primitives. Star has a confusing relationship of inner and outer radius, so I left it alone.