Re: g95 netCDF Fortran status

On Tue, 11 Nov 2003, Charlie Zender wrote:

> Hi Andy,
> 
> The status of g95-compiled netCDF Fortran code is as follows:
> 
> g95 does not yet compile the netCDF Fortran90 test suite.
> 
> [email protected]:/data/zender/tmp/netcdf-3.5.1-beta10/src/f90$ g95 -o netcdf_test.o netcdf_test.f90
> netcdf_test.f90: In function `MAIN__':
> netcdf_test.f90:112: error: aggregate value used where a float was expected
> 
> This program netcdf_test.f90 is rather short, 112 lines,
> so I include it as an attachment below.
> You may recognize why it generates an error on the last line, or not.
> 
> g95 now compiles, links, and runs the netCDF Fortran77 test suite.
> There are huge numbers of run-time errors.
> The test suite is rather extensive.
> I attach messages below in case you find they give useful clues. 
> I will try re-compiling the self-test with a g95 snapshot every few
> weeks and let you know when it looks close to running successfully.
> 
> At this point, you might find it helpful to try the netCDF fortran
> installation yourself:
> 
> Get netCDF tarball from http://www.unidata.ucar.edu/packages/netcdf
> tar xvzf netcdf-3.5.1-beta10.tar.Z
> cd /data/zender/tmp/netcdf-3.5.1-beta10/src
> make distclean
> # NB: Use PGI Fortran definition because cfortran.h does not yet support Intel ifc 
> export CC='/usr/bin/gcc'
> export CXX='g++'
> export CPPFLAGS='-DNDEBUG -DpgiFortran'
> export CFLAGS='-O'
> export FC='g95'
> export FFLAGS='-O'
> ./configure --prefix=/usr/local
> make
> make test

  Hey, this great news!  Although from past experience, I can tell you
that the fun has only just started.  If there are a vast number of errors,
then there is likely some idiom that the netcdf author(s) use a lot that
is messing things up.  Find it and many things are fixed at once.  Joost
and I had this experience with linpack.  If there were scattered errors
over the test suite, then that tends to point to lots of different things.  
I couldn't glean anything from the example you sent, since it uses a
module that wasn't included.

  If you could track one of the errors down to a small program that still
fails, that is more helpful than most people realize.  It's likely that
you'll pick one of the major issues very quickly.  While I'll certainly
get around to investigating the problems in netcdf, mails in the mailbox
prompt me better than anything else.

  Great news,
      Andy



-------------------------------------------------------
This SF.Net email sponsored by: ApacheCon 2003,
16-19 November in Las Vegas. Learn firsthand the latest
developments in Apache, PHP, Perl, XML, Java, MySQL,
WebDAV, and more! http://www.apachecon.com/