@Wannes Malfait (Wannes), I believe part of this problem is because BKE_mesh_new_nomain is used instead of BKE_mesh_new_nomain_from_template, which copies the materials/slots.

Another is probably that face attributes are transferred to vertices, but that isn't possible with material indices. I guess there's probably some way to transfer face attributes to new faces? Though maybe that exposes some arbitrary decisions in the edge cases.

In T93663#1342061, @Hans Goudey (HooglyBoogly) wrote:

@Wannes Malfait (Wannes), I believe part of this problem is because BKE_mesh_new_nomain is used instead of BKE_mesh_new_nomain_from_template, which copies the materials/slots.

Another is probably that face attributes are transferred to vertices, but that isn't possible with material indices. I guess there's probably some way to transfer face attributes to new faces? Though maybe that exposes some arbitrary decisions in the edge cases.

If BKE_mesh_new_nomain_from_template fixes the problem of not being able to assign materials correctly then I think just making that change should be good enough.

There is no "good" way to copy material indices like you said. I don't think it's possible to to any better than if you used the Set Material Index node with the to point domain converted Material Index attribute. Since it's possible to do this with the current node, I don't think it's worth the added effort and complexity to add this to the node itself. I think in general you would want to assign materials after all the geometry processing is done anyway.

That makes sense, thanks. I guess we can just do the first fix then.