Re: [Libmesh-devel] The "must be called in parallel" problem

[John P. <pet...@cf...>]

Roy Stogner writes:
 > On Mon, 5 Nov 2007, John Peterson wrote:
 > 
 > > Do you just mean place a parallel_only() macro similar to what we
 > > have now for stop() in any function you are about to make parallel only?
 > 
 > I don't think there's any sort of macro that can give us a useful
 > error message here, unless it's solve_the_halting_problem().  How do
 > you tell whether a function has been called in parallel or not?

A necessary condition would be libMesh::n_processors() > 1.
If you flag a function as parallel_only() and the above
condition doesn't hold ... is that enough to be an error?

But I guess you're worried about the case where you're in parallel,
and *every* CPU must call the same function?  Isn't there an
MPI_Wait or something that does that?

-John