CGAL users discussion list

[Eric Berberich <>]

Panagiotis Foteinos wrote:
> Hello all.
>
> I am using the 3D Delaunay Triangulation package with a custom 
> refinement algorithm. In my program, there are almost flat elements 
> which I want to identify as flat (by checking the volume for example), 
> and leave them alone without performing any calculations on them any 
> more. The problem is that the calculation of the circumcenter or the 
> radius (and the volume as a consequence) of an almost flat element gives 
> a "floating point" exception (in release mode) and terminations follows.
>
> A workaround is to hard-code a couple of lines inside CGAL package's 
> code that throw an exception if there is division with zero , and then 
> catch it from my code. If the catch statement is activated that means 
> that I am possibly dealing with an almost element and I should just 
> overpass it.
>
> Is there any better way to do this?

Using an exact number type instead of a floating point type.

Please see http://www.cgal.org/philosophy.html for details and this page
http://www.mpi-inf.mpg.de/departments/d1/ClassroomExamples/ for problems 
of floating-point numbers in geometric reasoning.


eriC