[MITgcm-devel] gfortran and namelist handling

[Ed Hill]

Hi folks,

Someone un-subscribed me from MITgcm-devel so this email may or may not
get through.  *sigh*

In any case, I spent some time this weekend building netCDF and related
packages for the upcoming Fedora 7 release.  In the past, I've tried to
fully support both g77 and gfortran with netCDF due to some gfortran
problems--mostly, the namelist extension as documented at:

  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24459

The good news is that newer gfortran releases have improved.  Attached
is a sample MITgcm output run with the latest gfortran as shipped in
the Fedora Core 6 updates:

  gcc version 4.1.1 20070105 (Red Hat 4.1.1-51)

and it now works on almost all of the MITgcm tests.  At least one of
the failures is probably due to a sloppy input file that can probably
be easily fixed:

  Experiment:  deep_anelastic
  At line 1926 of file ini_parms.f
  Fortran runtime error: namelist not terminated with / or &end

And the gfortran to be shipped with Fedora 7 will be even newer
(4.1.2) so it will have all the above fixes.


As a result, I'm tempted to drop netCDF support for g77 within Fedora 7
when it ships in a few months.

Do you folks have any opinions on the matter?  That is, how badly do
you want or need netCDF for g77 support in Fedora?

Ed

-- 
Edward H. Hill III, PhD  |  ed at eh3.com  |  http://eh3.com/
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: mitgcm_summary.txt
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20070319/3fe7f978/attachment.txt>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20070319/3fe7f978/attachment.sig>