[vtkusers] vtkKDTree FindClosestPoint bug?

[David Doria]

On Tue, Sep 15, 2009 at 10:17 PM, Kevin H. Hobbs <hobbsk at ohiou.edu> wrote:

> On 09/15/2009 03:33 PM, David Doria wrote:
> >
> >
> > Has anyone else had this problem?
>
> Yes, but that was long ago, and a test was added at the end of
> /Filtering/Testing/Cxx/TestPointLocators.cxx that was supposed to check
> for this.
>
> Please confirm that you are using a reasonably new VTK.
>
> I can't remember if it was VTK or ITK where the last remaining issues
> with the KdTree seemed to only occur when the two points were almost
> exactly the same distance from the trial point, so close that the error
> could be from calculations being made in float verus double precision.
>
> Your test looks reasonable but I've been awake for way too long and
> can't give it a thorough look.
>
>
I am using the CVS VTK. I see the test in TestPointLocators.cxx, but
apparently it is not testing the right thing! haha

These points are not any where near the same distance. The squared distance
is ~9 for the correct one and ~3722 for the incorrect one!

Maybe someone can go through this example and see if this actually looks
like a bug.

Thanks,

David
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20090916/b1a53f5d/attachment.htm>