[Paraview] pvbatch vs. pvpython

Burlen Loring bloring at lbl.gov
Thu Aug 9 12:20:31 EDT 2012


this sounds more and more external to pv. For example if you compiled 
your program with one version of mpi but tried to run it with another, 
you'd very likely see exactly what you are seeing here. A few of things 
you could look at to boost your confidence: "module list", "which 
mpirun", "echo $LD_LIBRARY_PATH", "ldd pvbatch" all of which will let 
you confirm that your environment is using the mpi and library versions 
that you expect, and that PV itself can find it's dependencies.

Sometimes sys admins in attempt to make things easy load modules for 
you. So when you are put into a new shell you environment is  changed. 
And of course the default modules will change as packages are updated. 
This could lead to unexepected weird problems over time. So, when 
running something like paraview that has a lot of dependencies, it can 
be helpful if you "freeze" your PATH and LD_LIBRARY_PATH. copy these 
into a bash script (or module file) which you source when using 
paraview, and then pass these into your qsub script with "-v" option.

On 08/09/2012 07:47 AM, Ganesh Vijayakumar wrote:
>
>
> On Wed, Aug 8, 2012 at 5:25 PM, Burlen Loring <bloring at lbl.gov 
> <mailto:bloring at lbl.gov>> wrote:
>
>     assuming that you've already examined all the output, and there's
>     no other info, you could do a couple of things: put some print
>     statements in your python script to see how far it makes it. Also
>     you could use MPT specific environment variables that tell mpt to
>     print a stack trace, see man mpi. A debug build may be helpful for
>     this.
>
>     before all this debugging effort you may want try a very simple
>     script to gain confidence that things are indeed working correctly
>     with your build. I'll send a simple test script that might be of
>     use for that.
>
>
> It doesn't even get past the first line i.e the module import line. 
> I'm unable to get it to force a core dump as "ulimit -c unlimited" is 
> not working. I've asked the system administrator about the issue.
>
> But in the meanwhile can I force a debug compilation? How can I do 
> that? Just add "-g" to CMAKE_C_FLAGS and CMAKE_CXX_FLAGS?
>
> ganesh

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20120809/098d0b86/attachment.htm>


More information about the ParaView mailing list