[vtk-developers] VTK-Python testing: update.

[Berk Geveci]

I would go with command line arguments. In Tcl, it is like:
test -D path/to/data/dir -V relative/path/to/data/dir/filename
(I think Bill also added support for -B path/to/baselin/root).
This gives more flexibility to the test writer in that the
filename does not have to be hardcoded. Multiple files are
differentiated by a number trailed by underscore (volProt.png,
volProt_1.png...). The images are compared until the first match
is found. If all comparisons fail, the difference with minimum 
error is reported (and the corresponding images are used by
Dart).

-Berk

 
On Wed, 2002-09-11 at 13:01, Prabhu Ramachandran wrote:
> Hi Dan,
> 
> Thanks for the feedback.
> 
> >>>>> "DJB" == Daniel J Blezek  writes:
> 
>     DJB> Prabhu, I would not try to wrap my brain around
>     DJB> rtImageTest.tcl too much, it's pan..., err, I mean it is a
>     DJB> bit tricky.  It was originally written to tease out the
>     DJB> proper image from the VTK Tcl example code that was already
>     DJB> in place.  As such, it really is a bit hairy, even for a Tcl
> 
> Oh ok. :) 
> 
>     DJB> guru.  Rather I think going with a unit test package for
>     DJB> Python is the best approach.  Multiple images per test are a
>     DJB> fine idea.  In rtImageTest.tcl, multiple images per test
>     DJB> pertain to multi-platform regression images, rather than
>     DJB> multiple outputs.
> 
> Ok.  So I'll allow support for multiple images (since no one has
> objected thus far).  I'll basically configure it so that the argument
> to the script will be 
> 
>   -V relative/path/to/img_dir/
> 
> and its the scripts responsibility to know the name of the image(s)
> that it creates in that particular directory.
> 
> cheers,
> prabhu
> _______________________________________________
> vtk-developers mailing list
> vtk-developers at public.kitware.com
> http://public.kitware.com/mailman/listinfo/vtk-developers