GitLab

directory. This will reduce on dependency of Hydra on the MPICH2
directory structure. A million more to go.

string against NULL. Use empty strings instead to keep it happier.

No reviewer.

See also r3982 and ticket #429.

Reviewed by buntinas@.

elements are there on the system and wrap around once we reach the
maximum core. This requires us to make sure that we can read the
system topology information (also through PLPA); so for cases where we
cannot read this information, even if there is no over-provisioning,
we do not perform any binding. Though we are being more conservative
on the systems that we can function correctly, this should be more
robust.

binding part not to build even when explicitly configured to.

[svn-r4123] Added an optional test for handling of faults in MPI programs - MPICH2 fails this, but now there is something to test

than one process to fail. This commit should fix ticket #442.

sockets code can also use it in the future.

passing now to break.

Reviewed by balaji@.

[svn-r4104] Explicitly specifying path to executables in the fortran project template file. This makes it easier for the windows test scripts

No reviewer.

[svn-r4102] Updating the product code & ver numbers in windows installers for the upcoming 1.1b1 release

could hold. So, we were "losing" some data. This should fix ticket
#468.

[svn-r4100] Handle unhandled defns - don't use _ALL_SOURCE for struct ifconf, no lustre config files on windows - in windows configure script

[svn-r4099] Unpack to the *right* iov buffers. The original code ignored iov offsets. - refer ticket #462 for details - No reviewer

[svn-r4098] Prevent sending the same data twice - Since we are sending 2 MPI_INTS put/get 'SIZE/2 -2' elements from/to array A. The last two elements are used for Accumulate - No reviewer

Akingunola. Thanks!

try to interpret it. Single quotes just passes it along to the remote
side. This should fix the nightly test failures with Hydra.