Segfault when smoothing on generated spherical meshes

[Grégory Operto]

Hello everybody

After generating spherical meshes using AimsMeshGenerate, running AimsMeshDecimation or AimsMeshSmoothing (or AimsMesh2Sphere) will end by an error (segmentation fault). It seems that smoothing along the mesh using <aims/primalsketch/finiteElementSmoother_d.h> (which is my primary interest here) does not work as well, as the "node ordering" step loops infinitely.
One kind of output associated to the crash (with AimsMeshDecimation and AimsMeshSmoothing) is :

Code: Select all

[grg@localhost ~]$ AimsMeshGenerate -i test.minf -o atlasout.mesh
[grg@localhost ~]$ AimsMeshDecimation -i atlasout.mesh  -o atlasout2.mesh

reading triangulation   : done
decimating mesh         : building facets from mesh :      0Warning: non cyclic neighborhood ... trying to recover
     1Warning: non cyclic neighborhood ... trying to recover
     2Warning: non cyclic neighborhood ... trying to recover
     3Warning: non cyclic neighborhood ... trying to recover
     4Warning: non cyclic neighborhood ... trying to recover
     5Warning: non cyclic neighborhood ... trying to recover
     6Warning: non cyclic neighborhood ... trying to recover
     7Warning: non cyclic neighborhood ... trying to recover
     8Warning: non cyclic neighborhood ... trying to recover
     9Warning: non cyclic neighborhood ... trying to recover
    10Warning: non cyclic neighborhood ... trying to recover
    11Warning: non cyclic neighborhood ... trying to recover
    12Warning: non cyclic neighborhood ... trying to recover
    13Warning: non cyclic neighborhood ... trying to recover
decimating Erreur de segmentation
[grg@localhost ~]$ 

Could someone please give a hint on this ? Thanks a lot in advance !

Grégory

[riviere]

Hi Grégory,

The main problem here is that the gui who wrote the decimation algorithm has left the lab long ago, and the algorithm has not been maintained so nobody exactly knows how it works...
I guess the decimation algorithm expects a given structure in meshes local neighbourhoods (triangles linked to a given vertex) which is always satisfied by the mesher algorithm (written by the same guy), but not satisfied by meshes generated by AimsMeshGenerate. It's possible that generated meshes (or at least spheres) have some illnesses like duplicated vertices on the initial meridian or at poles, or invisible "holes", which may cause the decimation algorithm to fail. We have to check that...
Denis