Hi,

I thought I would include an update. I was unable to get the path to the target in spite of the useful suggestions. However, I found that MPICH2 (which uses a python version of mpiexec) does not have a problem; LAM, however, does with its bourne shell script.

Best Regards,
Rob

On Thu, Jun 12, 2008 at 11:56 AM, Robert Langlois <ezralanglois@gmail.com> wrote:
Hi Volodya,

I am creating a shared python library that handles MPI functionality.  The library compiles fine and seems to run but I want to add unit testing to my Jamfile.

So I do the following:

python-extension foo : bar.cpp ;
 # Since the compiling works fine, I left out the mpi specific requirements.

make-test run-pyd : foo foobar.py
                         : <testing.arg>-lib=?? <python.interpreter>mpiexec
                      : test-foo ;

For the above test, I added "<python.interpreter>mpiexec", which prefixes mpiexec on the command line like so:
mpiexec python foobar.py

This all works properly (at least on linux). The problem is that before "mpiexec python foobar.py" is executed, PYTHONPATH is exported to the system but for some reason, this exporting does not work when I add <python.interpreter>mpiexec. If I remove <python.interpreter>mpiexec, it works fine but I need mpiexec in the test.

So to work around this, I added <testing.arg>-lib=?? where ?? will be the path of the shared library created by "python-extension foo", but I cant seem to get the path in a straightforward manner.

Best Regards,
Rob


On Thu, Jun 12, 2008 at 11:41 AM, Vladimir Prus <ghost@cs.msu.su> wrote:
On Thursday 12 June 2008 19:51:41 Robert Langlois wrote:
> Hi Everyone,
>
> How can I get the build path of a target? For example, say I create a
> target:
>
> python-extension test-lib ;
>
> This file ends up in the following path:
> bin/programs/learn/python/gcc-4.0.2/debug/threading-multi
>
> I wanted to know if there is a rule that already exists to get this path in
> a platform independent way.
>
> The reason I need this path is that I am trying to prefix the python
> interpretor with another command (mpiexec) and the PYTHONPATH environment
> variable is nullified in some way when I do this. So, I propose to pass the
> path of the shared file to a python script to get around this problem.

Your email is a little bit obscure as to what you're trying to do. To answer
your question -- during Jamfile parsing, it's not possible to get the path
of the target, because you cna build your project with 10 different property
sets. Inside a generator, you have a specific property set, and can get a
path for a target (see virtual-target.path method). Inside an rule/action
pair you also can get a path to the target.

I suggest you describe your use-case more concretely.

- Volodya
_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost-build



--
Robert Langlois



--
Robert Langlois, Ph.D.
Instructor/PostDoc
University of Illinois at Chicago
Department of Bioengineering
Bioinformatics Program
835 S. Wolcott,
Suite W103
Room 164B, M/C 563
Chicago, IL 60612-7340
(312) 996-6514
ezra@uic.edu