[petsc-dev] Fortran source format

[Satish Balay]

On Sat, 23 Jan 2010, Jed Brown wrote:

> The nominally fixed-format fortran source contains continuations written like
> 
>       call PetscSetFortranBasePointers(PETSC_NULL_CHARACTER,            &
>      &     PETSC_NULL_INTEGER,PETSC_NULL,PETSC_NULL_SCALAR,             &
>      &     PETSC_NULL_DOUBLE,PETSC_NULL_REAL,PETSC_NULL_OBJECT,         &
>      &     PETSC_NULL_TRUTH,PETSC_NULL_FUNCTION)
> 
> The trailing ampersand is in column 73, so new gfortran is generating
> loud warnings about line truncation.  Was this done because of some
> compiler quirks, or is it feasible to just make sure that *.F files are
> treated as fixed-form and *.F90 are treated as free-form (default on the
> compilers I've seen)?

The above is the recommended portable fixed/free form code. [i.e it
complies with both fixed form and free form standards.

We wanted PETSc to compile [and example work] even if someone uses
--with-fc='compiler --free-form' of with FFLAGS - hence this usage.

I guess - Gfortran should be smart to detect that - and not warn that
the line is truncated.. [It happens with -Wall only]. The alternative
is to make it illegal to use --free-form or --fixed-form compiler
flags. But thats probably more drastic than the current state. [well
currenlty --with-fixed=132 is illegal anyway]

Satish