Differential D16505
Fix T84078: Improve UV unwrapping for quads with an internal reflex angle. Authored by Chris Blackbourn (chrisbblend) on Nov 15 2022, 1:18 AM. Tags Subscribers None
Details When triangulating meshes, the UV unwrapper was previously using a heuristic to split quads into triangles. If one of the internal angles is greater than 180degrees, ("reflex angle") the heuristic was giving a poor choice of split. Instead of using a special case for quads, this change routes everything through the generic ngon polyfill_beautify method instead.
Diff Detail
Event TimelineComment Actions This doesn't seem correct as as the check for existing triangles uses a fan-fill which may not be the topology used once BLI_polyfill_calc_arena has run. A simple fix could be to handle the case of quads (where p_quad_split_direction takes existing faces into account), but that doesn't help for ngons. BLI_polyfill_beautify could optionally take existing triangles not to use ... but this seems like a fairly involved workaround which would be better spent on more general support for n-gons.
Comment Actions kk, I'll split the differential up into 3 parts.
Comment Actions From what I remember, the reason I didn't use p_add_ngon also for quads is that lead asymmetric triangulation, and in turn asymmetric UVs. Would be good if we could try to keep symmetry somehow. Comment Actions Is this an instance of T78101 or is something else going on? Do you have an example .blend file which illustrates the asymmetry? Comment Actions I tried to create a model to illustrate the asymmetry problem: Here's unwrap without change:  And after applying D16505 in it's current form:  Comment Actions With git blame I traced it back to T32198: UV sphere unwrap is an elongated ellipse. It seems that at least in the test file there, the n-gon splitting also preserves the symmetry, and in a few other files it also seems to work ok. | ||||||