[svn-r5368] Use the correct type in the segment calls - it must be an MPI_Aint, not an MPIDI_msg_sz_t, particularly when the size of MPI_Aint is changed to match MPI_Offset (the segment calls specify an MPI_Aint at the last argument, for example.  This is part of the changes needed to make attributes work properly when --with-aint-size=8 is selected

[svn-r5367] Added pragma to tell gcc (if a recent enough version is available) to ignore the -Wfloat-equal warning diagnostic

[svn-r5365] Fix the make targets for configure files so that they are correctly built in the source directory

Buffers weren't being freed and the MPI_IN_PLACE argument to
MPI_Alltoall{v,w} required an MPI-2.2 #ifdef guard.

No reviewer.

This fixes the cases in MPI_Alltoallv/MPI_Alltoallw where the send
counts are 0 with a non-zero-sized send type and the recv counts are
non-zero with a zero-sized recv type.  Prior to this change MPICH2 would
hang when this situation was encountered.  This commit also includes a
regression test for this behavior.

Thanks to Michael Hofmann for reporting the bug and providing a succinct
test case that illustrates the problem.

No reviewer.

configure. This also adds support in the release tarball creation
script to check for potential ABI mismatch (i.e., changes in mpi.h.in
and language bindings).

Reviewed by goodell.

[svn-r5357] Explicitly declaring sendcount and recvcount that are logged by the rlog macros - without this fix build with rlog fails after r5276

[svn-r5356] Adding Scalable graph topology, MPI_Reduce_scatter_block, MPI_Reduce_local and MPI_Op_commutative - implemented in r5307, r5253, r5277 - to windows. Also look at svn forum tickets 24, 27 and 33 for details

[svn-r5355] Added (or moved) MPIU_THREADPRIV_DECL so that the thread-private area is acquired only once within a routine, since the data (pointer to the thread-private area) is fixed during the execution.  This reduces the overhead of the nest increment/decrement, for example

[svn-r5354] Ensure that a build failure in the mpi-io directory causes the make to fail (without this change, the make succeeded on Mac OSX, even though the make failed in the mpi-io directory

preparation for the move to web-only bug report submission.

Nemesis internode communication still has some sort of problem, this
just makes it so that the test will play nicely with the rest of the
MPICH2 test suite.

No reviewer.

The MPIR_Status_set_empty macro doesn't set the MPI_ERROR field, so we
were making comparisons against uninitialized data that only showed up
on some machines.  Valgrind catches this reliably on all platforms.

No reviewer.

[svn-r5344] Adding support for C99 fixed-width types, C99 complex types, MPI_AINT and MPI_OFFSET types on windows - these were added in r5142 . Also see mpi forum ticket 18. As of this commit these types are disabled on windows

its own "correct behavior" definition. Added one for Hydra as well.

Reviewed by buntinas.

Thanks to Bill for catching this.  It would be nice if we could figure
out a way to implement the macros that requires that the commit/reap
macros are in place if the decl/alloc macros are ever used.

No reviewer.

No reviewer.

This simplifies the handling of NULL weight arrays in the dist graph
case and just makes sense in general.

No reviewer.

The MPIU_MEM_CHECK_MEMCPY function shouldn't have checked pointer values
when len==0.

No reviewer.

Realloc can return either NULL or a unique value that is safe to pass to
free().  The NULL interpretation is in some ways more conservative, so
we now implement it in MPIU_trrealloc and handle the case in
the MPIU_REALLOC_OR* macros.

No reviewer.

arguments to get corrupted while printing.