CGAL users discussion list

[Andre Massing <>]

Hi,


Eric Berberich wrote:
> Andre Massing wrote:
> > Hi again,
> >
> > sorry for bothering again the cgal-ml, but this time I came up with a 
> > rather simple example, where the Simple_cartesian<double> kernel 
>
> See my very recent emails on this list concerning 'double arithmetic' 
> and Andreas Fabri's trick with the
> CGAL::Exact_predicates_exact_constructions_kernel

Yes, I know that doubles are evil and not suited for excact predicates 
because of round-off errors etc.
My question is rather why in  aforementioned implementation in the 
Triangle_3_Tetrahedron_3_intersection.h file the 
k.bounded_side_3_object() is  *not* used for the intersection detection: 
If it is detected that one vertex lies either on the boundary in the 
inside of the tetrahedron, why it is not stated as an intersection 
between  the triangle and the tetrahedron? (of course, it might or might 
not be a real intersection in the SimpleCartesian Kernel case, but it is 
anyway outside the scope of the kernel to verify it).
At least it would still work fine in the 
CGAL::Exact_predicates_exact_constructions_kernel
but additionally catch some "corner cases" which are not detected by the
Triangle_3 - Triangle_3 intersection in the SimpleCartesian.

Kind regards,
Andre