Skip to content

SLS2: Improve robustness of skeleton extrusion - #9842

Open
MaelRL wants to merge 49 commits into
CGAL:6.2.x-branchfrom
MaelRL:SLS2-Extrusion_robustness-GF
Open

MaelRL wants to merge 49 commits into
CGAL:6.2.x-branchfrom
MaelRL:SLS2-Extrusion_robustness-GF

Conversation

@MaelRL

@MaelRL MaelRL commented Nov 9, 2026 •

Copy link
Copy Markdown
Member

Summary of Changes

  • Add fallback strategies when CDT2+normal fails to triangulate a face
  • Enable using different kernels for skeleton and for extrusion
  • Various bug fixes

TODO:

Release Management

  • Affected package(s): Straight_skeleton_extrusion_2
  • Issue(s) solved (if any): -
  • Feature/Small Feature (if any): -
  • License and copyright ownership: no change

MaelRL added 11 commits November 6, 2026 15:05
Although we will ignore that point in the builder of the offset polyhedron,
some visitor such as the one used in skeleton extrusion expects
to have an offset point for all skeleton halfedges that are
intersected by the offset isoline.
The correct values are to be recomputed from the converted contours
And default it to SK, because conversion to exact kernel is very costly...
The core idea is to always produce a mesh.
The mesh may be self-intersecting due to inexact constructions,
but we get a mesh.
The former approach simply computed for each face (+bottom & top)
the contributions, and let PMP do some heavy work stitching things
together.
Now, we do things combinatorially, so by construction things
are stitchable since the skeleton is combinatorially valid.
If the triangulation fail due to inexact constructions, we
use two fallbacks:
- first one is to use a different projection traits
- second one is to rebuild the face with a converted-to-exact
  skeleton and thus the boundary of the face is necessarily
  embedded (note that the sqrt error is absorbed at normalization;
  there can be no embedding issue with epeck)

In the current state, the snapping was not adapted and is broken
and disabled. It will be fixed in an upcoming commit.
With (0,0,0) as the origin, we can get some numerical issues
for meshes with a very large bbox (1e9+)
@MaelRL
MaelRL force-pushed the SLS2-Extrusion_robustness-GF branch from d195a20 to 1df4cad Compare November 20, 2026 14:02
@lrineau

lrineau commented Nov 20, 2026

Copy link
Copy Markdown
Member

The CI failure is about dependencies:

Differences in Surface_mesh_parameterization:
Random_numbers
are new and not committed.
Differences in Straight_skeleton_extrusion_2:
Polygon_mesh_processing
Union_find
have disappeared.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants