Re: datatype decoding & GR return codes

Steve Huss-Lederman (lederman@cs.wisc.edu)
Mon, 21 Apr 1997 20:04:20 -0500 (CDT)

I think Rajeev is right about the MPI_TYPE_RESIZE. UB & LB are
treated as fundamental datatypes. They are different since they are
only inputs to a constructor. RESIZE creates a new datatype. I want
to change the chapter to reflect Rajeev's suggest. This means adding
a new combiner for the datatype creation function RESIZE. Any
thoughts or objections?

Another open issue is the meaning of the return code in user callbacks
for GRs. They appear to be a hold over from when MPI controlled the
state of GR. I am not sure what they are needed for in the new GR
proposal. I have not had enough time to get it completely thought
through so anyone who has a reason to have return codes should let us
know why. We also need to digest the rest of Linda's note.

Thanks,
Steve