(j3.2006) [ukfortran] (SC22WG5.4347) Good news for the Fortran 2003 standard

[N.M. Maclaren]

On Nov 6 2010, Toon Moene wrote:
>
>ECMWF tries to move to a programming paradigm where at least the outer 
>structure of the model(s) use object oriented design (the reason for 
>this is complicated, but it stems from the fact that they noticed they 
>copied -and then slightly modified- far too much code).

That seems (semi-)rational.

>Obviously, it would be easiest if they could use Fortran for this - 
>their "toy model" code at the moment has both a C++ variant and a 
>Fortran variant.

Agreed.

>Although I haven't looked into it yet, my impression is that the Fortran 
>variant is rather crippled by the fact that they want to (also) compile 
>it with Gfortran (See http://gcc.gnu.org/wiki/Fortran2003Status) for 
>where Gfortran is, 2003-wise.

I rather wonder what they want that isn't there, and suspect that they
may be using several extremely tricky and non-portable C++ features
without realising just how problematic they are.  As usual, the gotcha
is that the problems don't show up in toy programs :-(


Regards,
Nick Maclaren.