CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

Bernd Gaertner wrote:
> My guess is that since the decimal value 0.1 is not representable as a 
> double value under the binary IEEE standard, Lazy_exact_NT gives you a 
> surrounding interval that is more than just one point. If so, this 
> particular problem should go away if you use Point p(1,2).

The problem you have comes from the fact that x() and y() don't return
references. Due to the type_equality_wrapper (the mechanism making 
Kernel::Point_2 and Point_2<kernel> being compatible), a reference to
a temporary is returned at some point.

You should write:
        const MP_FT& x() const {return *pos;}
        const MP_FT& y() const {return *(pos+1);}


S.

> Best,
> Bernd.
>
> Yue Biek wrote:
> > I first write a simple demo to test the validity of C_Point2.
> > int main()
> > {
> > Point p(0.1, 0.2);
> > MP_FT x = p.x();
> > std::cout << p.x() << std::endl;
> > }
> >
> > However, the problem issued when p.x() executes. I found the error is 
> > caused by a test down in to_double() method:
> > // Interval_nt.h
> > bool is_point() const
> >  {
> >    return sup() == inf();
> >  }
> >
> > It seems that it complains the cartesion component of C_Point has 
> > different supremum and infimum. If my guess is right, how could this 
> > happen?
> > Really need some pointers here.