further nc_test progress

[Robert Latham]

I've taken care of the read cases (test_get.m4 ) so that they either
call the _all version of the function or (in the case of the var1
functions) go into indep data mode before making the function call.

So now the tests go pretty far, and we do get a lot of "good
comparisons", but we do get a few discrepancies.  We also fail to
complete the test.  ncmpi_get_vara_short kills the test with this
error: 

MPI_Type_create_subarray: Invalid value in array_of_subsizes
ABORT - process 0: Invalid buffer pointer
Return code = -1, signaled with Aborted

If anyone wants to take a closer look at nc_test, i've committed all
my collective and indep-mode changes to cvs.

==rob

-- 
Rob Latham
Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
Argonne National Labs, IL USA                B29D F333 664A 4280 315B