CGAL users discussion list

["Amit Pendharkar" <>]

Hi,

I tried the test case posted on 1st April and I could not recreate the problem. Everything looks fine.

I am using CGAL-3.3.1 that comes with gmp-4.1.4 on Fedora core 6, 32 bit.

Just few inputs for people who are looking into the case

Thanks & Regards,
Amit Pendharkar

On Tue, Apr 8, 2008 at 10:49 AM, Andreas Fabri <> wrote:

Hello,

I just looked into it (sorry for having given it low priority).
I tried it with CGAL-3.3.1   in release and debug mode
and on an XP64/VC8 as well as on XP32/VC7, and couldn't
reproduce the error behaviour.


Which CGAL version do you use? Did you install the tar.gz
or the Windows installer with precompiled libraries?
Which compiler do you use?

Is GMP installed on your machine ?


Finally, can you provide a stack trace, so that I see where
the exception gets thrown.

andreas

Ryan Henry wrote:

The test-case posted to this mailing list by Louis Feng on April 1st:

   #include <CGAL/basic.h>
   #include <CGAL/Exact_predicates_exact_constructions_kernel.h>
   #include <CGAL/Delaunay_triangulation_2.h>
   #include <iostream>
   #include <CGAL/FPU.h>
   struct K : CGAL::Exact_predicates_exact_constructions_kernel {};
   typedef CGAL::Delaunay_triangulation_2<K> Triangulation;
   typedef Triangulation::Point Point;


   int main()
   {
     CGAL::force_ieee_double_precision();
     Triangulation T;
           Point p = Point(6.68896e-005, 1);
     T.insert(p);
     p = Point(6.68852e-005, 2);
     T.insert(p);
     p = Point(6.68807e-005, 0);
     T.insert(p);
     p = Point(6.68762e-005, 1);
     T.insert(p);
     p = Point(6.68717e-005, 2);
     T.insert(p);
     p = Point(6.68673e-005, 0);
     T.insert(p);
     /*
     for (long i=14950; i<15000; i++) {
       Point p = Point(1.0f/i, i%3);
       //Point p = Point(1.0f/i, i%100);
       std::cout << p << std::endl;
       T.insert(p);
     }
     */
     return 0;
   }

There are also two reproducible test-cases in my original message on the subject, although they are not so simple. If these are somehow insufficient, I can look for more.

-Ryan


Sylvain Pion wrote:

Ryan,

Your best bet is to make a small (smallest possible) reproducible test-case.


Ryan Henry a écrit :

Sorry to be a bother! However, we are still unable to proceed with our CGAL backed Delaunay implementation due to the bug described earlier. Is there any more suggestions out there and, if not, are there any developers in particular who work in this portion of code that I could perhaps contact? Thanks in advance.

-Ryan




Ryan Henry wrote:

I tried adding the CGAL::force_ieee_double_precision(); call to my code, but with no luck. Certain data sets still make the Delaunay triangulation crash. It might be worth noting that we are only adding points to the triangulation that have distance greater than or equal to 0.001 from any other point currently in the triangulation. therefore, it does not appear to be a problem with points being "too close" together. Any other thoughts?

-Ryan


Olivier Devillers wrote:

You may try :

#include <CGAL/FPU.h>

and

 CGAL::force_ieee_double_precision();


at the beginning of your code.




it should at least help to have the same behavior
if your ar in debug mode or not.

Olivier

--
You are currently subscribed to cgal-discuss.
To unsubscribe or access the archives, go to
https://lists-sop.inria.fr/wws/info/cgal-discuss