CGAL users discussion list

[Thorben Kröger <>]

On Wednesday 14 October 2009 09:06:48 Thorben Kröger wrote:
> On Wednesday 14 October 2009 07:52:02 Marc Pouget wrote:
> > One thing is surprising on this picture:
> >
> > all the vertices on a "line" of the mesh should have the same
> > computed differential quantities, cause the mesh is regular and the
> > patches symmetric, so the estimations are only done on translates. So
> > i think there may be a bug in the way neighbors are retrieved, you
> > may want to check and compare the patches for vertices along the same
> > "line". Note that if this is the problem, it is only related to the
> > example code and not the core of the package.
> 
> I have modified my plane example to output the values computed for another
> point translated in x-direction along the plane, it should receive the same
> normal vector.
> 
> in plane_01.png, I have added the point to the right (it's located where
>  the arrow originate) and it's neighors, again painted in blue '*'. The top
>  view plane_03.png shows that the neighborhood is not completely the same,
>  but close. I would have thought that using such a large neighborhood would
>  be robust to such slight changes?

On closer inspection, the neighborhoods are completely identical... This is 
then even more strange.

Also it's interesting that the sign of the angle respective to the true plane 
normal flips... 


> The side view is plane_02.png, the two red vectors are the ones estimated
>  by CGAL for the two neighborhoods.
> 
> Any thoughts?
> 
> Cheers,
> 
> Thorben
>