Re: possible gfortran.mk changes

[Greg Troxel <gdt%lexort.com@localhost>]

"Dr. Thomas Orgis" <thomas.orgis%uni-hamburg.de@localhost> writes:

> I'm only reacting as I'm somewhat involved in packges that use Fortran,
> but my platform is Linux with my own provided toolchain (GNU or Intel,
> maybe AMD in future).
>
> So I avoid all this machinery. But I wonder about this:
>
> Am Tue, 31 Oct 2023 09:54:46 -0400
> schrieb Greg Troxel <gdt%lexort.com@localhost>:
>
>>   - Change the logic that says "only if gcc" to not say that.  Clearly
>>     we don't mean it.
>
> The comment above that rather clearly formulates that it is meant. When

It does, but several other comments were wrong.

> using clang, one probably wants to use flang (or one of the differing
> flangs that exist) instead of gfortran. Similar for OpenMP. You can
> have varying combinations. Or maybe you don't

That's a completely valid point, but right now there is zero support for
using flang.  gfortran.mk gets included if fortran is in USE_LANGUAGES.
So if you do build anything fortran on a system with fortran, you get
gfortran.

And, because of the if gcc, that means that on any system with clang,
you get gfortran 7.  Except on macOS, in which case you get 12.

> Before touching that logic, it would be good to have someone around who
> uses a clang-based toolchain and wants Fortran stuff (math/blas, for
> example).

Yes, that would indeed be good.   I was hoping to get such a report.
Is anybody on FreeBSD here and could they test?