[MITgcm-devel] the "stubs" idea for MNC

[Ed Hill]

Hi Alistair & Chris,

Executive summary:

  The stubs idea doesn't work.
  We must compile a NetCDF library for EACH COMPILER.


Details:

  Get my implementation of the stubs idea for MNC at:

    http://mitgcm.org/eh3/MNC/mnc_stubs.tar.gz

  and use it as a drop-in replacement for pkg/mnc.

  Try out the code and notice that:

    1) it compiles cleanly (so I got all the name mangling bits 
       right), but

    2) it does not run.

  The reason is that the NetCDF calls make extensive use of 
  strings on both sides of the Fortran/C interface.  The 
  NetCDF implementors solved the problem of passing strings 
  by using (including a copy of) the "cfortran.h":

    http://www-zeus.desy.de/~burow/cfortran/

  header as part of their C library.

  The upshot of all this seems to be that, unless we're using 
  two compilers that are binary compatible with each other, 
  we must compile the NetCDF library for each compiler.  
  Alternatively, we're going to have to write a compatibility 
  layer thats at least as clever as "cfortran.h" because it 
  will have to know how the different compilers encode strings 
  (with *ONLY* the knowledge we can glean from a previously-
  compiled library--remember that "cfortran.h" has the 
  advantage of being able to generate C code to work with the 
  Fortran interface, not the other way around) and then 
  translate between them.

Ed

-- 
Edward H. Hill III, PhD
office:  MIT Dept. of EAPS;  Rm 54-1424;  77 Massachusetts Ave.
             Cambridge, MA 02139-4307
emails:  eh3 at mit.edu                ed at eh3.com
URLs:    http://web.mit.edu/eh3/    http://eh3.com/
phone:   617-253-0098
fax:     617-253-4464