Ok. It was actually not that that caused the problem. I can remove it and I will start another thread.
Johan On Tue, Nov 25, 2014 at 12:03 PM, Garth N. Wells <[email protected]> wrote: > > On Tue, 25 Nov, 2014 at 10:59 AM, Johan Hake <[email protected]> wrote: > >> Hello, >> >> there is a barrier in the constructor of Function.cpp (line 221) that >> uses MPI_WORLD. Any good reasons it is not using the one provided by the >> FunctionSpace? >> > > > Looks like it's left over from some debugging. It can be removed. > > Garth > > > This line create a deadlock for code constructing different dolfin >> problems on different MPI_GROUPS. >> >> Johan >> > >
_______________________________________________ fenics mailing list [email protected] http://fenicsproject.org/mailman/listinfo/fenics
