[vtk-developers] VTK/Python and Sun new discoveries

[David Gobbi]

Andy, Prabhu,

I agree that including a vtkpython binary (I much prefer calling
it 'vtkpython' rather than 'vtkPython') is a good idea.  It is
trivial to do and it solves a lot of problems.  For the SGI I
have to re-link the python executable with the C++ compiler
to make it work, I'm not terribly surprised that it is the
same under Solaris.

Perhaps the 'vtkpython' executable could also automatically load
libvtkCommonPython et al as well?

And when VTK/python is built, both VTK and python should definitely
be built shared.  Otherwise it is guaranteed that the
vtkTkRenderWidget will not work, since when static linking is used
both VTK and Python will each have their own copy of the Tk code
and the Tk/Tcl global variables.

There are ways of building VTK/python static, for example by
linking the vtkTkRenderWidget.so against the _tkinter.so module
that comes with python instead of linking it against libtk.so.
But unless someone wants to work through all the details, add them
to the cmake files, and maintain them it is better just to tell
everyone to build shared.  This is going to be a problem with
python and ITK, since there is no 'shared' option with ITK.

Prabhu is correct that I am busy right now, I'm reading the list
and recompiling VTK every so often to make sure that all my stuff
works with the cvs version of VTK, but I probably won't be
committing much of anything in the next few months.

 - David



On 24 May 2002, Andy Cedilnik wrote:

> Prabhu,
>
> I agree with you. I wish we could use python executable for vtk/python
> and tclsh (or wish) for vtk/tcl. Unfortunately that will not work on Sun
> using Sun's compiler and probably on HP,SGI,.... Please correct me if I
> am wrong. The problem is that C++ code does not get initialized
> correctly when the binary is compiled using C compiler and not C++ one.
>
> As a test, please try this code:
> from vtk.tk.vtkTkRenderWidget import *
> root = Tk()
> pane = vtkTkRenderWidget(root)
>
> This will crash on all platforms where classes do not get initialized
> properly. The funny thing is that all VTK python tests will pass. This
> however is not the case with gcc, which does not have any problem with
> this.
>
> So, I think we should add vtkPython, which is just a wrapper around the
> actual python library and does not include any code at all.
>
> 				Andy
>
> On Fri, 2002-05-24 at 10:37, Prabhu Ramachandran wrote:
> > Well, I am not sure that this is the right way to go.  Making a new
> > binary seems unnecessary.  Also I dont understand clearly what the
> > issues are with the current setup and why you need to add a special
> > vtkPython binary.  Maybe you could explain it a little more?  I
> > understand that if you want specific features like initializing MPI or
> > something like that you'd need a special interpreter but doing it just
> > for VTK is a no-no IMHO.
> >
> > I dont know what David has to say.  I guess he is very busy these
> > days.
> >
> >     AC> Note to python developers: Please add some tests to vtk that
> >     AC> will use tk render widget.
> >
> > I'll keep that in mind when I get some time to do the test stuff.  The
> > trouble is that I'm not too familiar with how the testing exactly
> > works and don't have too much time to check it out either.
> >
> > prabhu
> > _______________________________________________
> > vtk-developers mailing list
> > vtk-developers at public.kitware.com
> > http://public.kitware.com/mailman/listinfo/vtk-developers
> >
>
>
> _______________________________________________
> vtk-developers mailing list
> vtk-developers at public.kitware.com
> http://public.kitware.com/mailman/listinfo/vtk-developers
>