parallel-netcdf 0.9.4 released

[Robert Latham]

It has been quite a few months since our last release, but we are
pleased to announce the release of parallel-netcdf version 0.9.4.

http://www.mcs.anl.gov/parallel-netcdf/

This version makes some changes to the API: all routines now take
MPI_Offset types, instead of size_t and ptrdiff_t.  This change brings
the implementation in line with the API document, and also helps us
solve some precision issues on 64 bit platforms.  Code using size_t or
ptrdiff_t types should instead use the MPI_Offset type.  

More detailed changes:

. fix some build issues on NEC
. removed some unused files
. better cleanup of resources in error cases
. modify some limits to stay in sync with netcdf-3.5.1
. compiler warning cleanups
. added support for uchar/schar types (NC_BYTE)
. implement ncmpi_delete
. use well-known constants instead of magic numbers 
. Changed all size_t for dimension size, array index[], start[],
  count[], stride[]in the API to MPI_Offset.  changed size_t to
  MPI_Offset to match the new API
. fix the comments in pnetcdf.inc and nfconfig.inc to be more
  Fortran-friendly


As usual, please report any problems to the mailing list

==rob
(on behalf of the parallel-netcdf developers)

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