CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

Without reading the paper, the two first sentences in the FAQ are enough:

"While it may well be a bug in CGAL, with such number types you are more 
likely to suffer from a problem with robustness. In particular, using 
the kernel CGAL::Cartesian with number types double or float is likely 
to give you wrong results."

So if the result is correct with a kernel with exact constructions, then
the error is due to floating point calculation.

Sebastien.

On 04/25/2012 04:02 PM, gupta61 wrote:
> Hi Sebastian,
>
> Thanks for the reply. Actually I did follow your advice and read the
> article. Also if you look in my post I mentioned that if I am using exact
> constructions, then we get the right result. Also as you have mentioned, if
> I use conversion from one kernel to another it gives me correct
> intersection.
>
> I just wanted to confirm that this is a problem in "inexact constructions"
> and not another bug as we have previously found. My only worry is that the
> result is so wrong and different for "segment" and "opposite_segment". I
> encountered this kind of error before in BBox and segment intersection and
> we realized that the error is because "segment" was considered as a "ray".
>
> If you plot this opposite_segment and triangle, and consider both the things
> constructed from rays, you will get the intersection as current output. This
> particular assumption still doesn't explain the intersection computed for
> segment and triangle.
>
> Regards,
> Piyush Gupta
>
>
> Sebastien Loriot (GeometryFactory) wrote
> > You did not follow my advice and did not read this:
> > http://www.cgal.org/FAQ.html#inexact_NT
> > :)
> >
> > Note that you can use Cartesian_converter to convert from a kernel
> > with exact constructions to another kernel.
> >
> > http://www.cgal.org/Manual/latest/doc_html/cgal_manual/Kernel_23_ref/Class_Cartesian_converter.html
> >
> > Sebastien.
> >
> > On 04/24/2012 09:39 PM, gupta61 wrote:
> > > Hi,
> > >
> > > We just encountered another bug in CGAL in segment triangle intersection.
> > > I
> > > am trying to find out intersection between the segment and a triangle.
> > >
> > > The segment is defined by
> > >
> > > CGALSegment3 segment_query( CGALPoint3(2.1, 2, 0.5),
> > >                               CGALPoint3(3.23135, 2, 0.5) );
> > >
> > > and the triangle by
> > >
> > > CGALTriangle3 triangle( CGALPoint3(2.18182, 1.81818, 0.666667) /* vertex
> > > 0*/,
> > >                           CGALPoint3(2.18182, 2, 0.5) /* vertex 2*/,
> > >                           CGALPoint3(2.09091, 1.90909, 0.5833335) /* vertex 
> > > 3*/
> > >
> > > We are using exact predicates and inexact constructions.
> > >
> > > The output we are getting is as follows:
> > >
> > > Segment query: 2.1 2 0.5 3.23135 2 0.5
> > >
> > > Triangle: 2.18182 1.81818 0.666667 2.18182 2 0.5 2.09091 1.90909
> > > 0.5833335
> > >
> > > CGAL::do_intersect (0/1) = 1
> > >
> > > Encountered intersection object is a point (for the segment)
> > >            2.1 2 0.5
> > >
> > > Encountered intersection object_oppos is a point (for segment.opposite())
> > > 1.89801666666667 2 0.5
> > >
> > >
> > >
> > > If we use exact constructions we get the right result. Let me know if I
> > > am
> > > missing anything here.
> > >
> > > BTW, the correct result is
> > >
> > > Segment query: 2.1 2 0.5 3.23135 2 0.5
> > >
> > > Triangle: 2.18182 1.81818 0.666667 2.18182 2 0.5 2.09091 1.90909
> > > 0.5833335
> > >
> > > CGAL::do_intersect (0/1) = 1
> > >
> > > Encountered intersection object is a point
> > >        2.18182 2 0.5
> > >
> > > Encountered intersection object_oppos is a point
> > >        2.18182 2 0.5
> > >
> > >
> > > -----
> > > Piyush Gupta
> > > UIUC
> > > --
> > > View this message in context:
> > > http://cgal-discuss.949826.n4.nabble.com/Bug-in-Segment-Triangle-intersection-tp4584602p4584602.html
> > > Sent from the cgal-discuss mailing list archive at Nabble.com.
> >
> >
> > --
> > You are currently subscribed to cgal-discuss.
> > To unsubscribe or access the archives, go to
> > https://lists-sop.inria.fr/wws/info/cgal-discuss
>
>
> -----
> Piyush Gupta
> UIUC
> --
> View this message in context: 
> http://cgal-discuss.949826.n4.nabble.com/Bug-in-Segment-Triangle-intersection-tp4584602p4586683.html
> Sent from the cgal-discuss mailing list archive at Nabble.com.