Re: [netcdf-hdf] bug in MPI cleanup

On Sat, Sep 15, 2007 at 10:39:48AM -0500, Quincey Koziol wrote:
>       That's a clever way to help the problem.  It's a bit "weird" in the  
> sense that nc_open_par/nc_create_par are per-file and the H5close is  
> per-library/per-application (more like MPI_Init/MPI_Finalize).  It  
> should work fine if the library is opened and closed repeatedly, but  
> it's not something we stress a lot in our tests.

I definitely get what you're saying:  in MPI-IO, the communicator
passed to the library is valid for a given file, but we're trying to
do init/cleanup once.

The global state variable helps make sure we don't initialize
ourselves repeatedly.  I think nc_open_par and nc_create_par could do
something similar.  There is already some global per-library state
maintainig identifiers when opening/closing files, after all.

==rob

-- 
Rob Latham
Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
Argonne National Lab, IL USA                 B29D F333 664A 4280 315B