Sounds like a quite expensive operation to do.
Would think that allocating a single array of integers with a size of number of edges, and then iterate over mloop and keep counts of edges is cheaper.

is_manifold.

Also comment actually belongs to a function, not a variable.

I think it could be quite a bit faster in some cases (but never slower) to just iterate over all loops, instead of having this nested loop.

Also, doesn't this function fit better in mesh_validate.c?

If you think than an extra bool check will be that much slower, I can simply set poly_index to totpoly to end the loop if you wish.

We discussed moving this to mesh_validate.c. IIRC we decided to wait until something else needed this too and write a some more general functions for this then.

Nah, that would be quite hacky.

The two benefits of just iterating over all loops are:

1. The code is more simple. Only one loop is required and you could just break out of it.
2. It's faster. Not mainly because of the extra condition (which is highly predictable), but because the memory access pattern is more predictable. By iterating over the polygons in an outer loop, you might have to jump around in the loops array, resulting in a worse cache hit rate.

Note, I don't care too much about whether this is changed or not. I just happened to see this function and thought about how it could be made better.

Ah sorry, I misunderstood you.

I can try to just have one for loop instead that loops over all mloops!

I do really appreciate that you go through the changes and propose changes like these.
Thanks!

This is now pushed to master:
rBfc9e921495312ace23af11a69e738a8a7adbaeed

Thanks again for pointing this out.