the H5Fed file data structure can't be created by >1 MPI task, so it is disabled in parallel mode (for now)
the H5Block3dSetHalo function was inverting the halo radii
the h5b_test wasn't writing the correct data in the halo case
Finished integrating H5Block and the Fortran interface (untested).
Started adding an H5Block regression test with a few simple tests (all pass).
Added automatic detection of stripe information on lustre, and config option to compile against lustre API.
Moved buffers for H5Block ghost zone disolving out of the file handle and into the h5b_3d_set_view function.
Fixed bug with pointers in the H5Fed file data not being initialized to NULL.