[vtkusers] Possible bug in CVS version of vtkBandedPolyDataContourFilter?

[Charles Boivin]

Hello everyone,

I have recently switched to the CVS version of VTK for a couple of
reasons. In my application, I am using the
vtkBandedPolyDataContourFilter module. I have noticed today that this
stopped working the way it was before. I expect to get bands of cells
that are all colored the same out of this filter; this certainly was
what I was getting before with VTK 4.4.

Now, the contours are generated properly but they are colored based on
the input scalar value, not using a scalar value that is uniform within
the contour. Hence I get a very smooth output in terms of color -- no
specific contours can be distinguished. If I put the output in wireframe
mode, I can verify that the contours are indeed generated, just not
colored properly.

I checked the differences between version 4.4 of the C++ file, and the
version from CVS (updated a few minutes ago). There are some minor
changes, but one caught my eye. At the beginning of the Execute() member
function, there is this:

>From the CVS version:
  vtkPointData *outPD = output->GetPointData();

Now, in VTK 4.4, this was:
  vtkPointData *outPD = input->GetPointData();

Certainly, the CVS version seems correct, at least based on variable
name. I tried changing "output" to "input" back, as it was in v4.4, and
this actually fixed my problem -- it created bands of uniformly colored
cells. 

It seems like the CVS version *should* be right (in terms of that
declaration, at least), but it does not work for me... Am I possibly
using this module wrong? Or is there a problem in the CVS version of
this object?

Thank you for your answers,

Charles Boivin