Barry Smith <[email protected]> writes:
>     Add a configure option that causes PETSc to compile Fortran code
>     with the free-form option. 

Isn't that just FFLAGS=-ffree-form?  We currently write everything in
polyglot so it should work, right?

This doesn't do any good for readability of examples or easy of
contributing.  I'd be in favor of having all new examples use .F90, and
perhaps more that use --with-fortran-datatypes.

Is there any way we can defer --with-fortran-datatypes to the user's
compile time?  Like let them include different headers to decide which
to use?  It sucks to require every component in a simulation to make the
same choice here.  For example, if you wanted to couple Blaise's code to
someone else's Fortran that didn't use datatypes, you'd be dead in the
water right now.

Attachment: pgpScSR2QKxMS.pgp
Description: PGP signature

Reply via email to