Extrusion produces wrong indexing -> visual artifacts

[martinRenou]

Similar to #335 but for extrusion result

[trungleduc]

hmm, could you test it with a previous version, maybe before #353 ?

[martinRenou]

hmm, could you test it with a previous version, maybe before #353 ?

I also reproduce the same-ish issue before that PR

[martinRenou]

It looks like it was working in 2.0.0a0 and stopped working in 2.0.0a1

[trungleduc]

wow, there are only 3 commits between a0 and a1, and two are meaningless. The last one does not touch the rendering part

[martinRenou]

It doesn't make any sense right.

Also we hard pin the open-cascade build so nothing could have changed there.

[trungleduc]

actually, we are wrongly pinning the jupytercad-* package in the meta package. Installing jupytercad always gets the latest jupytercad-* ones. This issue happens between a2 and a4, which contains the 3D refactoring pass. I will take a look at it

[martinRenou]

Good catch!

[trungleduc]

I managed to track it down into 2 issues:

• the front face and back face create additional meshes in the scene.

  JupyterCAD/packages/base/src/3dview/helpers.ts

  Line 208 in 71e093c

  const backFaces = new THREE.Mesh(geometry, mat0);

It looks like these faces are related to the clipping panel, they are introduced in #333

• if I remove it, I get a better one

it still has the face normal points to the inside, but I'm not sure if it's an issue or it's the behavior of OpenCascade, since if I extrude in the opposite direction (LengthFwd=-10), I get the face normal points to the outside.

[martinRenou]

The clipping indeed relies on the normals, so if they are not correct, it results in visual artifacts.

if I extrude in the opposite direction (LengthFwd=-10), I get the face normal points to the outside.

Interesting! Should we... reverse the normal depending on the sign of LengthFwd?

[martinRenou]

What's weird though is that it seems to be coming from the 3D refactoring pass? You were not only changing the code structure right? You also fixed the normals for STL files #353, could this impact the extrusion?

[trungleduc]

What's weird though is that it seems to be coming from the 3D refactoring pass? You were not only changing the code structure right? You also fixed the normals for STL files #353, could this impact the extrusion?

I don't think it's related to the 3D pass, here is the result at 234f9bf, it works correctly

Here is the result one commit after, at 85c0dde

[trungleduc]

Interesting! Should we... reverse the normal depending on the sign of LengthFwd?

Or just using double-sided material?

[trungleduc]

It looks like the additional front and back meshes are made for the equivalent 3D solid tube since there are meshes for the top and bottom circles.

[martinRenou]

I'll have a look. It seems the visibility toggle does not hide all the meshes created for a specific object (there are multiple ones to achieve the clipping).

I'm not sure we can use the double-sided material without breaking clipping. Using double-sided material would only workaround the wrong normals but not fix it, and wrong normals would result in broken clipping.

[trungleduc]

I'll have a look. It seems the visibility toggle does not hide all the meshes created for a specific object (there are multiple ones to achieve the clipping).

but we should not have the meshes at the top and bottom of the extruded shape ?

I'm not sure we can use the double-sided material without breaking clipping. Using double-sided material would only workaround the wrong normals but not fix it, and wrong normals would result in broken clipping.

For the extrusion case, I'm not sure what is the definition of correct normals, it's like when you push for one direction you will get the outside of the circle, but with the opposite direction you will get the inside

[martinRenou]

but we should not have the meshes at the top and bottom of the extruded shape ?

I'm not sure what you mean by that?

I'm not sure what is the definition of correct normals

Always pointing outside?

Arf this is annoying... Clipping is what makes it difficult here because it needs proper normals, without this I would agree using double-sided materials and call it a day.

[martinRenou]

As discussed offline, this issue comes from the clipping logic.

Clipping involves some logic to fill the gap left by the cut, this only makes sense for solid objects. However extrusion produces a non-solid object by default and that logic gets broken in that case, resulting in visual artifacts.

We should disable that logic for non-solid objects.