Surface remeshing crash fix

[Angela-Oo]

Description

Performing SurfaceRemeshing Adaptive with Features can lead to crashes.

To reproduce the crash do the following

• Start the remeshing project with a mesh file that has degenerated triangles and kinks (if you need a concrete file for reproduction I can provide you one)
  The remeshing project does not work in the latest commit as already reported in Issue
  Inheriting from MeshViewer on Windows crashes due to the imgui context being null Inheriting from MeshViewer on Windows crashes due to the imgui context being null #87.
  The latest working commit is as far as I know SHA-1: 3145a49
• Click Detect Features with 70
• Execute Adaptive remeshing
  The program will crash because mesh_->halfedge(f) is called with an invalid Face f

The reason for this crash is, that for some degenerated triangles solving minimize_squared_areas lead to an vector containing nan, so one solution is to check at the end of minimize_squared_areas if the solved vector contains invalid values.

By doing so the crash is prevented but it takes a really long time to solve (might be an endless loop).
One solution for this might be to change the function is_too_long, so that an edge is never to long if it is smaller than min_edge_length_.

Motivation

Adaptive surface remeshing with features should not crash

Benefits

no crash

Drawbacks

I am not sure if there is a reason that an edge is currently allowed to be smaller than min_edge_length_, but everything seems to work with the change.

Maybe it would be better to abort the calculation if it is not possible to calculate the minimize_squared_areas.

[dsieger]

Hi @Angela-Oo

I'm currently looking into remeshing robustness issues in relation to #90. I've got some changes in the pipeline and would like to test them with some more models. If you could share any models that crashed for you, I'd be glad to have a look.

Thanks,
Daniel

[dsieger]

You can also have a look at the remeshing-robustness branch if you want to try for yourself.

[Angela-Oo]

Hi @dsieger,

Here is one example mesh where the adaptive remeshing crashed.
surface_remeshing_crash.zip

Thank you too,
Angela =)

[dsieger]

The good news is that it doesn't crash anymore. Yay!

The bad news is that the resulting mesh might not be satisfactory, but that's a different issue. :-)

Here's what I observed: The initial mesh quality is too bad to detect proper feature edges. By proper I mean single connected line segments. I think the kinks also lead to extremely high curvature values and therefore to lots of refinement in otherwise flat / low curvature regions. Some smoothing and/or manual cleanup of the feature edges could help.

[Angela-Oo]

Hi @dsieger,

Thank you for the fix =)
I don't expect the library to detect the correct feature edges, as these are already known in my input data.

Kind regards,
Angela