parallel-netcdf 1.0.2 released

[Patrick Jöckel]

Dear developpers,

the compilation on AIX5.3 (p690 system) with

   setenv OBJECT_MODE 64
   setenv MPICC  mpcc_r
   setenv MPIF77 mpxlf_r
   setenv F77    xlf_r
   setenv F90    mpxlf90_r
   setenv F90FLAGS -qsuffix=f=f90
   setenv FC     xlf_r
   setenv CC     xlc_r
   setenv CXX    xlC_r
   setenv CFLAGS -q64
   setenv FFLAGS -q64
   setenv F90FLAGS -q64
   setenv CXXFLAGS -q64

was successful.
However, I detected three issues:

1) The tool ncvalid terminates either with
     Error @ [0x00000000]:
        Unknow magic number, while (C D F \001) is expected!
    or
      Error @ [0x0003a4f4]:
        Unexpected EOF, while fetching the name string of variable[1] in     
        VARIABLE list!
    for valid netCDF files.

2) The "-v" option of ncmpigen produces
     - CDF 002 files for "-v 1" or "-v classic" or without "-v"
     - CDF 001 files for "-v 1" or "-v 64_bit_offset",
     which is somewhat counterintuitive.
     (Checked with "od -An -c -N4 file.nc").

3) "gmake testing" does not work (gmake complains about the "-w" option),
    however "make testing" is OK

Yours,

 Patrick


On Friday 09 November 2007 23:38, Robert Latham wrote:
> Well, I didn't hear too many complaints about the pnetcdf
> pre-releases.  In that case....
>
>
> The Parallel-NetCDF team is please to release 1.0.2, just in time for
> the SC07 rush.
>
> It's been almost two years since our last "official" release.  No
> groundbreaking new features since 1.0.1, but we've accumulated bug
> fixes and various improvements in the meantime.  Plus, we've got our
> code under SVN now (complete with public read-only access) and a
> trac-based defect-tracker.
>
> Come and get it: http://www.mcs.anl.gov/parallel-netcdf
>
> - bug fixes in nonblocking I/O functions
> - code cleanups
> - better error handling
> - CDF-2 format bugfixes
> - even better cross-compliation support
> - speedups from serial NetCDF attribute access code
> - use proper MPI fortran-to-c conversion routines in Fortran bindings
> - configure knows about 'gfortran' compiler now
> - assorted configure checks for sicortex
> - fixed reference counting bug
> - handle null stride to vars routines
> - Clean up Itanium warnings
> - Better interaction with HP-MPI
>
> I should note for our AIX-using friends that AIX is still a problem
> area for some:  do give it a shot and let us know what problems
> remain.
>
> ==rob (and the rest of the pnetcdf gang)

-- 
---------------------------------------------------------------
Dr. Patrick Joeckel
Max Planck Institute for Chemistry
Joh.-J. Becher Weg 27         
55128 Mainz                   phone:++49-6131-305452      
Germany                       fax  :++49-6131-305436     
                             e-mail:joeckel at mpch-mainz.mpg.de
                         http://www.mpch-mainz.mpg.de/~joeckel
---------------------------------------------------------------