[Pkg-scicomp-devel] paraview status

[Christophe Prud'homme]

Ondrej,


> Excellent work with the shared libraries. It builds both on i386 and
> amd64 in pbuilder. However, during install:
>
> $ wajig install paraview_3.2.1-1_amd64.deb
> Selecting previously deselected package paraview.
> (Reading database ... 181424 files and directories currently installed.)
> Unpacking paraview (from paraview_3.2.1-1_amd64.deb) ...
> dpkg: error processing paraview_3.2.1-1_amd64.deb (--install):
>  trying to overwrite `/usr/bin/vtkWrapPythonInit', which is also in
> package python-vtk
> dpkg-deb: subprocess paste killed by signal (Broken pipe)
> Errors were encountered while processing:
>  paraview_3.2.1-1_amd64.deb
>
>
> So we should do either of these:
>
> * should not ship /usr/bin/vtkWrapPythonInit


I vote for this one.
what is your experience with this script ?


>
> * rename it to something else
> * conflict with python-vtk
>
> Otherwise it seems to work fine both on i386 and amd64 except the
> python interface. The i386 package is in my repository:
>
> http://debian.certik.cz/
>
> Yes, let's create dummy man pages, that's ok.

Agreed


> The python-vtk issue is
> also minor - just removing the python-vtk package fixes this. So now
> we need to concentrate on the python issue.

Ok


>
>
> One thing that needs  to work is
>
> 1) starting paraview
> 2) Tools -> Python Shell
> 3) it should print
>


> <snip>
>
> But we are getting close, even now the package is usable. BTW, did you
> try the MPI support?

yup and it works :)
I have some PDE results solved in // on my dual cpu laptop and visualized
them using the SOS/CASE ensight format


> Do you have some easy usecase?

yes

File -> Connect -> add SErver ->
type the command
mpirun -np 2 pvserver

then you re ready to load data sets spread distributed over 2 processes
 the thing is that the pvserver is typically run on the machine where you
did the computation using as many processes you used for the computation
while paraview is run on your "loca" visualisation host

Let's write it to
> README.Debian?

good idea


> I would like to try it (I mean we should try that it
> works), but I didn't yet find time to figure this out how to use it.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.alioth.debian.org/pipermail/pkg-scicomp-devel/attachments/20080218/dd86b199/attachment.htm