CGAL users discussion list

[David Keller <>]

Hello,

Some problems seem to have survived the provided patch. I would like to 
append these to the above discussion on assertion violations even though 
the behavior is a slightly different one. The problem can be reproduced 
by calling the Bezier_curves example with the attached input file 
Bezier.dat. The problem manifests in the following output:

---
CGAL error: precondition violation!
Expr: t_res != EQUAL
File: /opt/CGAL-3.3.1/include/CGAL/Arr_traits_2/Bezier_x_monotone_2.h
Line: 676
Explanation:
terminate called after throwing an instance of 
'CGAL::Precondition_exception'
 what():  CGAL ERROR: precondition violation!
Expr: t_res != EQUAL
File: /opt/CGAL-3.3.1/include/CGAL/Arr_traits_2/Bezier_x_monotone_2.h
Line: 676
Aborted
---

Apart from this issue I would like to suggest an additional constructor 
in Bezier_point_2.h:
_Bezier_point_2 (const Rational& x, const Rational& y).

Would this conflict with the existing constructor _Bezier_point_2 (const 
Algebraic& x, const Algebraic& y)  ? (Rational is convertible to Algebraic).
The use of the function
template <class Traits, class Dcel, class PointLocation>
typename Arrangement_2<Traits,Dcel>::Vertex_handle
insert_point (Arrangement_2<Traits,Dcel>& arr,
             const typename Traits::Point_2& p,
             const PointLocation& pl)
is now limited to the insertion of algebraic-points.
Algebraic-points can be constructed from  rational-points and inserted 
by the above function of course. However, if one decides first to insert 
rational points which may later coincide with source- or target-points 
of Bezier-curves may encounter difficulties to do so with the current 
implementation. Or is there another preferred way to carry out such an 
operation?

Thanks and kind regards,
David Keller

Efraim Fogel wrote:
> Hi David,
>
> Ron Wein, has provided a fix. Attached is a patch you can use to apply 
> the fix. Apply it from the include/CGAL directory ('patch < 
> Bezier.diff').
>
> It suppose to affect the following files:
>
>    Arr_traits_2/Bezier_cache.h
>    Arr_traits_2/Bezier_curve_2.h
>    Arr_traits_2/Bezier_x_monotone_2.h
>    Arr_traits_2/Bezier_point_2.h
>    Arr_traits_2/Bezier_bounding_rational_traits.h
>    Arr_Bezier_curve_traits_2.h
>
> The problem in test case #1 can simply be avoided if you comment out 
> the assertion on line 247 in Bezier_bounding_rational_traits.h
>
> The problem in test case #2 happens when there exists a vertical 
> tangency point that occurs exactly at the middle of the curve. As you 
> can see from the size of the patch, it was a bit more difficult to 
> fix. We hope that we eliminated the bugs or at least reduced their 
> number...
>
> Naturally, the fix will make it into the next release.

Attachment: Bezier.dat
Description: MOPAC data