Trevor Laughlin wondered (2018-04-27 ...
sourceforge.net/p/netgen-mesher/discussi...306/thread/360df8b7/
) if there is a chance to get his PR into Netgen
github.com/NGSolve/netgen/pull/10
? (He seems to be the author of that patch.) He says that it is a patch from Salome that seems to fix an issue with meshing a face with internal edges.
Now this is a huge patch with basically no explanation for the changes. The only part of this patch which I understand are the modifications to libsrc/meshing/improve3.cpp, because I need to patch exactly the same issue.
However, even so I now also converted my "throw" into a continue, I still write some text to cout. This is because the topology of the mesh is broken, if we get there. (Not sure how netgen sometimes manages to recover from that broken mesh topology.) In my experiments it was always SwapImprove itself who broke the topology, however not necessarily in the current run. One mechanism by which SwapImprove can break the topology is by reverting the orientation of a tet during a swap operation, i.e. by creating "negative" tets. This happens if the 1e24 penalty for this fatal error is smaller than the badness of the alternative. The very far away auxiliary points added by Delaunay1 can cause those extremely bad alternatives, even if the mesh itself is not too bad.
Currently github says "This branch has no conflicts with the base branch". Is there any plan to ever apply that Salome patch in it current form, or does it have to be broken down into a series of smaller patches with explanations anyway?
