dolfin team mailing list archive

[Andre Massing <massing@xxxxxxxxx>]

 Hi!

Den 12.10.10 17.09, skrev Johan Hake:
> Andre!
>
> Are you able to compile a c++ program using mpich2?
Yes, that went fine, and the program ran through without any problems  
up to that point
where an indirect call to the pydolfin occurred (via the plot function), 
and of course I could not
load the dolfin module in python.
> I am not sure it is related but we figured that we needed to load the
> (open)mpi library before we loaded the compiled module. This is done in:
>
>    dolfin/common/dynamicloadmpi.py
>
> We might need to alter this, but I have no clue of how, as it should
> presumable load the mpich2 library if libmpi.so points to mpich2.
Yes, that might be the reason why the dolfin python module crashed while 
being loaded. Will look at this but first I have to fix my by now
somehow messed installation... :'(
--
Andre


> Johan
>
> On Monday October 11 2010 06:04:07 Andre Massing wrote:
> >    Hi!
> >
> > I tried to use DOLFIN mpich2 on my Mac OS X, but the swig generated
> > python bindings are
> > broken with mpich2: Loading _cpp.so during the import results into the
> > following error message:
> >
> > Attempting to use an MPI routine before initializing MPICH2
> >
> > I got a similar file when compiling dolfin using MPICH2 on Ubuntu Lucid:
> >   >>>  import dolfin
> >
> > Fatal error in MPI_Comm_test_inter: Invalid communicator, error stack:
> > MPI_Comm_test_inter(110): MPI_Comm_test_inter(comm=0x1cf92200,
> > flag=0x7fffd831c57c) failed
> > MPI_Comm_test_inter(67).: Invalid communicator
> >
> > Is this a known issue? I could not find a related bug report...
> >
> > Andre
> >
> > _______________________________________________
> > Mailing list: https://launchpad.net/~dolfin
> > Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
> > Unsubscribe : https://launchpad.net/~dolfin
> > More help   : https://help.launchpad.net/ListHelp