[vtkusers] 64bit indexing problem with vtkImageData

[David Gobbi]

Hi Jim,

Fortunately, the "git blame" command can answer questions like these
fairly quickly.  In the vtkImageData method that printed the error, just
after line 2117, I found this:

11b22c6c Filtering/vtkImageData.cxx (2009-09-11 2122) vtkIdType idx;

This tells me than in a 2009 bug fix, an important variable was
changed to vtkIdType.  Your best bet is to upgrade to VTK 5.6 or
VTK 5.8, both of which include this bug fix.

 - David


On Fri, Sep 16, 2011 at 1:04 PM, Jim Geach <jimgeach at physics.mcgill.ca> wrote:
>
> Dear list,
>
> I'm trying to volume render data cubes with typical sizes of >(3000x3000x3000) (these are outputs from a cosmological simulation).
>
> Currently I'm using VTK v5.4.2 (built with the 64bit indexing turned to "ON" in the CMake config). Everything seems to be going okay until it comes to the volume itself, or rather, the ImageData - I'm getting a crash with error like:
>
> ERROR: In /install/OCF/software/vtk/5.4.2/VTK/Filtering/vtkImageData.cxx, line 2117
> vtkImageData (0x72cacd0): Coordinate (1279, 1279, 719) out side of array (max = 4194303999
>
> I read somewhere about 64-bit indexing not being properly implemented in vtkImageData, even in relatively recent VTK versions -- is this the case for version 5.4.2, and if so is it fixed in the latest version?
>
> On the same note, I should note that I'm using VTK via the Python binding, and converting numpy arrays into vtkStructuredPoints. Will VTK just figure-out that 64-bit indexing is required when i do this?
>
> Thanks, Jim
>
> ---
> Dr James Geach
> McGill University
>
> Ernest Rutherford Building,
> 3600 Rue University, Montreal, QC
>
> jimgeach at physics.mcgill.ca
> 514-398-1449
> http://www.physics.mcgill.ca/~jimgeach