A degenerate triangle has any point on it

[stla]

Issue Details

When t is a degenerate Triangle_3, then t.has_on(pt) is true for any Point_3 pt.

Source Code

Triangle_3 t(p1, p1, p2);
t.has_on(any_point);

[afabri]

Hi @stla, I confirm what you observe. The thing is that we somehow prefer adding the precondition that the triangle is not degenerate. The function do_intersect(Triangle_3, Point_3) has this precondition (and there the "bug" is that this is not documented). We think that it makes little sense to make low level functions work with degenerate input as it means additional tests that probably will be called again and again by higher level functions. It would probably even make sense to add the precondition of not being degenerate to the constructor of Triangle_3.

[MaelRL]

Constructing degenerate kernel objects must be possible for inexact constructions or degenerate inputs (e.g. mesh) to still work. However, it is just for the user's responsibility to detect and handle degenerate kernel objects if they think there might be some (using Kernel::Is_degenerate_23). So we need to add consistently preconditions in the code and in the doc.

We were thinking about adding a helper function K::O2 get_non_degenerate_representation_of_kernel_object(K::O1), but we would have to find a solution to the fact that we do not have a kernel object to represent degenerate tetrahedra that are (planar) quads.

Related: #4654

[lrineau]

Maybe this has_on member function should have a precondition.