CGAL users discussion list

["Laurent Rineau (GeometryFactory)" <>]

Le dimanche 27 septembre 2009 16:23:55, Ben Supnik a écrit :
> Hi Hichem,
> 
> The order of computation is different - the first (failing) case is I
> think more efficient because the code can use a merge-sort-like sequence.
> 
> You say that your polygons "seem" to be correct...did you use the CGAL
> tests for self-intersection, etc. or a different unrelated set of code
> to check the input data?
> 
> Also, I'm not sure what the MSVC version of -frounding-math is...if the
> low level rounding in CGAL is incorrect, polygon ops can become
> intermittently unreliable...I saw that when I had compiler options
> wrong, and the symptoms were similar.

With recent MSVC, it seems /fp:strict is needed to have reliable IEEE754 
floating point arithmetic.

-- 
Laurent Rineau, PhD
Release Manager of the CGAL Project
http://www.cgal.org/
R&D Engineer at GeometryFactory
http://www.geometryfactory.com/