Re: static FORTRAN compilation set-up in libtool.m4

[Ralf Wildenhues]

Hello Wolfgang, all,

please don't top-post, thank you.

* Wolfgang Kilian wrote on Thu, Sep 02, 2010 at 08:58:54AM CEST:
> On Wed, 1 Sep 2010, Ralf Wildenhues wrote:
> >* Juergen Reuter wrote on Wed, Sep 01, 2010 at 09:31:12PM CEST:
> >>     nagfor*)
> >>        # NAG Fortran compiler
> >>        _LT_TAGVAR(lt_prog_compiler_wl, $1)='-Wl,-Wl,,'
> >>        _LT_TAGVAR(lt_prog_compiler_pic, $1)='-pic'
> >>        _LT_TAGVAR(lt_prog_compiler_static, $1)='-Bstatic'
> >>        ;;

> >Thank you for the suggestion.  I found the compiler manual at
> >http://www.nag.co.uk/nagware/np/r52_doc/index.asp
> >and I kind of wonder whether the -Wl,-Wl,, in your change is a typo.

> I may step in here, since I wrote these lines:
> 
> The '-Wl,-Wl,,' is NOT a typo.  The current nagfor version 5.2
> invokes gcc for linking, not ld directly.  Therefore the linker
> flags have to be wrapped twice.  I discussed this with the NAG
> developers and they agree that this was a poor choice which they may
> change in a future version, but we have to live with this for now.

OK, thanks for the valuable information, that's definitely something I
wouldn't have ever guessed from looking at the documentation alone.

> The double wrapper works for me so far.  I could not test this on OS
> other than Linux, however.

That's fine.  Any support for nagfor is an improvement over no support.

> I'll try to run the test suite later.

Cool.

> We have implemented quite a few Fortran autoconf macros, in
> particular for testing various Fortran 2003 (and 2008) features,
> which might be useful for the autoconf distribution.  We have tried
> to stay in line with the autoconf conventions, but it might be
> necessary that an autoconf expert looks after this before they could
> enter the distribution.

Just send them to autoconf-patches, I will be glad to review them.

Thanks,
Ralf