[mpich-devel] Two issues with MPI_File error handling

[Rob Latham]

On 05/07/2015 06:58 AM, Lisandro Dalcin wrote:
> This is using the preview release 3.2b2, I discovered this issue
> months ago and it seems I forgot to report it.

We really should get mpi4py into the nightly tests.  You seem to find a 
lot of interesting cases.

> 1) MPI_File_call_errhandler(file, value) should return MPI_SUCCESS
> instead of value under normal operation (as the MPI spec says), and
> particularly if the error handler is set to MPI_ERRORS_RETURN. This
> bug is trivial to fix.

Trivial to fix...if we only had to worry about  C and Fortran!

(and maybe we can ignore C++ regressions now that the bindings are 
deprecated? )

It looks like we have hijacked the error code (years and years ago, in 
to send information to the C++ binidngs:

http://git.mpich.org/mpich.git/blob/HEAD:/src/mpi/errhan/file_call_errhandler.c#l81

The change dates back to 2008:
http://git.mpich.org/mpich.git/commit/abbe596a

and it looks like the Comm and Window error handlers stomp on the error 
code, too.

This is ticket #2259 if you want to follow along

https://trac.mpich.org/projects/mpich/ticket/2259

I'll respond to #2 in another email

==rob

-- 
Rob Latham
Mathematics and Computer Science Division
Argonne National Lab, IL USA