[vtkusers] still a bug in vtkImageStencil class in v5.2.0

[Mathieu Malaterre]

On Thu, Oct 16, 2008 at 1:52 PM, Michael Knopke <Michael.Knopke at gmx.de> wrote:
> Hello,
>
>
>
> I reported this some time ago, but wasn't sure about the consistency of my
> vtkBuild. Now after upgrading to 5.2.0 I'm convinced that the improved
> vtkImageStencil class (or one of the prerequesits) are faulty.
> My tests where done in two different environments:
>       -     Borland6 compiled vtk5.2.0
>
> -          Visual studio compiled vtk5.2.0 both static
>
>
>
> The test pipeline is simple. VtkImageData loaded as volume. Generating a
> tube that intersects the volume similar to example: ImageTracerWidget (I
> visualized the tube, it nicely intersects the volume completely with the
> ending protruding the volume).
> Now converting the polyDataToImageStencil (with setting OutputSpacing,
> OutputExtent and OutputOrigin as in the sample).
>
> Results: depending on the rotation of the tube it crashes in _vector.h at
> function: push_back (from stl). Sometimes especially when the tube is axial
> aligned it works, sometimes it produces wrong results.
> Using vtk 5.1 (old cvs build) the same approach (leaving out the spacing,
> extent and orgin settings as those functions did not exist) it never crashes
> and never produces wrong results.
> I can send a sample application that reproduces the problem.
>
> As I'd really like to port my application to vtk 5.2.0 I tried to exchange
> the following files but without success: vtkPolyDataToImageStencil,
> vtkImageStencil, vtkImageStencilData, vtkImageStencilSource with the ones
> from my working cvs build.
> It seems the stencil mechanism doesn't fit in the (modified?) pipeline
> algorithm of vtk5.2.0. Are there other files I could try to replace?

Do you have a small example to reproduce the bug ?

Thx
-- 
Mathieu