Meaningful Dynamic Analysis of MPI jobs and Staging Transports
Currently we are doing no valgrind analysis of SST (and Dataman?) because CMake isn't running valgrind on the right thing. For the ADIOSSst* family of tests, they are run with valgrind doing analysis of the perl script that forks the MPI jobs that make up the actual test. For the TestStagingMPMD* family, valgrind is only analyzing mpiexec. It would be better to sort out how to run dynamic analysis on all tests, but if we can get valgrind on the MPI jobs "by doing "mpiexec valgrind" rather than "valgrind mpiexec", we'd at least be examining the staging transports in some modes.