Oops, needed to up the version to rc2 since we released rc1 and
have fixed a few more bugs since then.

Update the NEWS file entry for 1.2.9

Closes #1715: Fix for "Wrongly detect overlapping regions"

Submitted by bosilca, Reviewed by brbarret, RM-approved by tdd

r19757: add comment on what ompi_convertor_need_buffers() does.

r20120:
A fix for http://www.open-mpi.org/community/lists/users/2008/12/7502.php.
The solution is not to compute the OVERLAP flag, as the best we can do
is an approximative answer. Without this flag the unpack can leads to
unexpected answers if the data-type contain any overlapping regions.
As such datatypes are illegal in MPI, this became a user responsability.

r20123:
Correct the disp array when creating a data based on the
MPI_COMBINER_INDEXED_BLOCK combiner.

Closes #1738: Datatype misalignment fix

Submitted by bosilca, Reviewed by tdd, RM-approved by tdd

Patch applied corresponding to:

r20133:
Fix an issue where the datatype description introduce a buffer
misalignment. Because some architectures (read SPARC64) require
aligned accesses, we increase the storage space when we pack a
datatype description to keep the fields aligned. This has to
be done on both sided in order to be consistent.

r20134:
Some 64 bits architectures support pointer aligned on 4 bytes
(where the sizeof(long) is 8). On such architectures dont
assert if the datatype representation is not aligned on 64 bits.

r20140:
TYPE shouldn't be surrounded by parens because it causes
issues for some versions of gcc when the construct
{{{ a = ((int)) sizeof(b) }}} comes up...

Closes #1736: LOCK_SHARED fix

Submitted by rolfv, Reviewed by tdd and jsquyres, RM-approved by tdd

Applied r19533 with a few minor manual edits:
Add a check for LOCK_SHARED in the sys/synch.h file. If it exists
then smash it to avoid problems with preprocessor and C++.

Resync the v1.2 NEWS file with the trunk.

Closes #1722: Take into account active MTU vs Max MTU for HCA

Submitted by apant, Reviewed by jsquyres, RM-approved by tdd

Patch applied corresponding to r20131:
Check the active port MTU against the MTU statically configured
for the HCA. QLogic HCA's capable of 4K MTU had an issue when
connected to switches running at 2K.

Closes #1677: Fix lower bound datatype problem

Submitted by bosilca, Reviewed by jsquyres, transitively RM-approved by bosilca

r20033: Correctly use the snprintf. Don't zap the last char

r20034:
When copying data using the same datatype don't ignore the gap in
the begining. Thanks to Peter Kjellstrom for the bugreport.

Get ready for v1.2.9rc1: update NEWS and VERSION files

Closes #1654: Fix v1.2 warnings

Submited by jsquyres, Reviewed by pasha & rhc, RM-approved by tdd

Patch applied which fixes warnings that came up from the use of
RedHat's gcc extension D_FORTIFY_SOURCE=2 which is used by OFED.