Re: Fortran 9x support Case and underscoring

[Steven G. Johnson]

Phillipe,

All of the issues you raise are completly orthogonal to Fortran 9x
support, since they apply equally well to the existing F77 macros.  Please
try not to confuse issues, and I would suggest a separate email per bug
report or feature request.

On Tue, 5 Aug 2003, Philippe Trottier wrote:
>   In the __AC_FC_NAME_MANGLING it would be easy to add the preserve
> case in the test but this might break some compiler ???

I fail to understand the reason for invoking a nonstandard flag to
preserve case (not to mention that these would break the existing F77_FUNC
wrapper macros).  The whole point of the name-mangling macro is to allow
you to follow the Fortran conventions from C, not vice versa.  (For
example, to link to a pre-existing Fortran library like LAPACK whose
compilation flags you have no control over.)

If you want a new macro to find a case-preserving flag (if one exists at
all for a given compiler), feel free to submit one to the macro archive.

> The verbose output of ifc if not recognised :

I know about this bug, it gets confused by an extra quote (") character in
the output; the macro needs to be fixed to strip out this quote like it
does for the cray compiler; this is on my TODO list.  For now, the
workaround is to set the FLIBS environment variable manually.

> Version options ...
> --------------------------
> 
> I knew I saw it somewhere in the fortran.m4:355
> 
> # Provide some information about the compiler.
> [...] 
> This is not working with broken compiler like Absoft f77 and f95, they 
> need a dummy source file.

This is just some optional extra info for the config.log, it doesn't break
anything if the compiler doesn't support it.  Making it support Absoft is
therefore very low on my list of priorities.

> Other compatibility questions, are all Fortran compiler using CHARACTER 
> like this : REFMAKERF95(platform,strlen(platform));

Some Fortran compilers pass strings via an extra length argument at the
end of the argument list, some create a data structure that holds the
string pointer and a length, and there are probably other variations.

A macro to detect Fortran string-passing conventions, for C
interoperability, would be a nice addition, probably to the macro archive
initially.  (My current policy is just to avoid passing strings between
languages.)

Cordially,
Steven G. Johnson