CDT_3 - DT3 assertion

[janetournois]

At commit 16c57e8 in PR #8186.

Running example
conforming_constrained_Delaunay_triangulation_3_fpmap.exe Data/data/meshes/snowflake_50mm.off

runs into an assertion in Delaunay_triangulation_3 (side_of_oriented_sphere() would return NEGATIVE)

   CGAL error: assertion violation!
   Expression : false
   File       : C:\Jane\CGAL\cgal-wt\Triangulation_3\include\CGAL/Delaunay_triangulation_3.h
   Line       : 1417

Note that without using params::face_patch_map, it works

[janetournois]

Same issue with the remeshed version of snowflake surface... snowflake_50mm_remeshed.zip
(I tried remeshing to have better edge lengths in input)

[janetournois]

Same issue with Base.zip

[sloriot]

"Note that without using params::face_patch_map, it works" Do you mean if you really simplify coplanar patches? or simply using the triangle mesh?

[janetournois]

I mean using the triangle mesh, without face_patch_map

[sloriot]

OK then I guess the next thing to try is simply the mesh using remesh_almost_planar_faces and see if it crashes with no face map

[janetournois]

using this code

  PMP::remesh_planar_patches(mesh, mesh2,
    params::edge_is_constrained_map(eif).face_patch_map(fpmap).cosine_of_maximum_angle(1.),
    params::face_patch_map(fpmap2).do_not_triangulate_faces(true));
  auto ccdt = CGAL::make_conforming_constrained_Delaunay_triangulation_3(mesh2,
                      CGAL::parameters::face_patch_map(fpmap2));

with the np face_patch_map, the crash happens, and without it, it works fine

[lrineau]

fixed by: dccd25e

[lrineau]

At commit 16c57e8 in PR #8186.

Running example conforming_constrained_Delaunay_triangulation_3_fpmap.exe Data/data/meshes/snowflake_50mm.off

runs into an assertion in Delaunay_triangulation_3 (side_of_oriented_sphere() would return NEGATIVE)

The issue was that the snowflake (once the "patches" are detected) has patches with internal boundaries. And the code dealing with multiple polylines for a given patch was buggy (see the tiny patch at dccd25e).

Now it works as expected:

Screencast_20250402_210640.webm

(Note that it was recorded with non-optimized Debug version of CGAL Lab, with even ASAN and UBSAN.)

[lrineau]

Same issue with the remeshed version of snowflake surface... snowflake_50mm_remeshed.zip (I tried remeshing to have better edge lengths in input)

Same issue with Base.zip

Those are fixed as well. The issue was the faces with holes: all the boundaries of those faces were taken as the boundary of a face without hole. And then there were overlapping faces. The commit dccd25e fixed that.