[Mpi-forum] C++ issues

[Jeff Squyres]

1. For those of you who missed it, please note that Rolf has made a very important MPI-2.2 errata proposal regarding supporting C++ datatypes:

    http://lists.mpi-forum.org/mpi-forum/2012/06/1848.php

Note that I agree that this is was actually an error: we just previously goofed by not listing some of the MPI:: datatypes with MPI_ language-neutral equivalents.  Nevertheless, this is important and everyone should realize that this is being proposed as a *single-vote MPI-2.2 errata*.  It has implications on most (all?) existing MPI implementations.

*** SUPPORTING RATIONALE: Some users have pointed out (rightfully, IMNSHO) that if we're deleting the C++ bindings in MPI-3 #281, we need to ensure to still support C++ datatypes with the C bindings.  Hence, Rolf's MPI-2.2 errata proposal is pretty important.

*** Doug/Rolf: there *may* be some side-effects of Rolf's proposal on #281 (e.g., leaving some language about C++-specific datatypes in, even if they get changed from MPI::<foo> to MPI_CXX_<foo>).  #281 and Rolf's proposal should be examined together to ensure that they don't conflict in areas like this.

2. I have a dim recollection of someone saying that you can't always replace MPI::ERRORS_THROW_EXCEPTIONS with a user-defined error handler that throws an exception.  Aside from the fact that MPI implementations need to ensure that they can transparently pass exceptions (e.g., compile a C-based MPI implementation with --allow-passing-cxx-exceptions-through-c-functions types of compiler flags), is my faulty memory remembering a real issue?  I have a dim recollection that it may have been Darius or Dave G. who brought up this issue...?

If this is a real issue, do we need to provide MPI_ERRORS_THROW_EXCEPTIONS?

-- 
Jeff Squyres
jsquyres at cisco.com
For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/