CGAL users discussion list

["J.L.M." <>]

Sebastien,

Thank you for your reply. I have been on holiday, or I would have 
responded sooner. I hope you will be able to provide a fix to the 
boundary function, as this is essential to my research. If there is 
anything I can do to help, please let me know.

Here are a few responses to your statements. The call to boundary is 
necessary for my work. What I am computing is the feasible region that 
contains all points in which the object can be placed and be wholly 
contained in the container (stored in result). If you would like, I have 
a paper I can provide to you that shows how this works.

Second, I am wondering about your statement that I need to make a copy 
of container. I see that the minkowski_sum_3 modifies the input, and in 
my example code the input is container_boundary, and object. 
container_boundary should be a copy of container (well just the boundary 
of course).

You are correct that the result is empty in this example. The correct 
algorithm uses the interior of the object (actually the inverted 
interior) in the Minkowski sum. I removed that part while simplifying 
the code to one that shows the problem.

Thank you for pointing out the non-standard OFF files. I am creating my 
shapes in Blender, and I have to post-process them. First, Blender gives 
me coordinates that are something like 1.00001 or 0.99999, and I need 
them to be exact whole numbers for accurate computations. Also, Blender 
has Z increase up, and Y increase into the scene, which is different 
than I am used to with OpenGL, so I swapped those coordinates. It must 
be in my swapping of those that I have switched the orientation of the 
facets.

Thank you again for your time.

On 12/20/2010 08:26 AM, Sebastien Loriot (GeometryFactory) wrote:
> J.L.M. wrote:
> > I have been writing some programs that use the Nef_polyhedron_3 and
> > Minkowski_sum_3 portions of the CGAL library. I have discovered a
> > number of assertions with various input shapes. The attached backtrace
> > shows one example problem. I have also attached a small example
> > program, and the input shapes. The l-shape is shown in the attached
> > jpeg file.
> >
> > This problem shows up in both CGAL 3.7 and 3.6.1. I am using GCC 4.5.1
> > on Arch Linux with kernel version 2.6.36.
> >
> > I think this is a bug, since it does not occur with other input
> > shapes. I would be very grateful if someone can help me fix it since
> > the Minkowski_sum_3 is essential to my research.
> >
> > Thank you for your time.
> >
> > Jason M'Sadoques
> >
> >
> >
> > ------------------------------------------------------------------------
> It seems their is a problem in boundary function.
>
> If I replace:
> Nef_polyhedron_3 container_boundary(container.boundary());
> by
> Nef_polyhedron_3 container_boundary(container);
>
> Your program runs fine (with your data, the call to boundary is
> superfluous).
>
> Note that as you are then reusing "container", you need to make a copy
> of it
> (see
> http://www.cgal.org/Manual/latest/doc_html/cgal_manual/Minkowski_sum_3/Chapter_main.html)
>
>
> Note also that the difference you are computing will be empty as the
> Minkowski sum contains "container".
>
> Finally, note that facets of your cube are clockwise oriented while the
> usual convention is to use counterclockwise oriented faces in off files.
>
> I'll try to find a fix to boundary function and will post it here.
>
> S.