960530: UNICOS netcdf fortran/ build fails: m4:-:581 can't open file

Dear Carlie,

> To: address@hidden
> From: Carlie Coats <address@hidden>
> Subject: netcdf "make all" and m4
> Organization: mcnc
> Keywords: 199605302059.AA15404

In the above message you wrote:

> I recently had the following sequence happen during a 
> "make all" of netCDF 2.4.2 on a Unicos8 Cray C-90.  It
> was quite distressing, and I never did figure out what
> the problem was (I wound up copying netcdf.inc and jackets.c
> from a Unicos9 T-90 where the build _had_ succeeded).
> 
> I'd like to suggest that this situation be considered 
> fatal, rather than a warning followed by an attempt to
> continue.
> 
> fwiw.  Thanks!
> 
> - ----------------------  cut here  -----------------------------
> 
> making `all' in directory /home/xcc/netcdf/src/fortran
> 
>         ./fortc -L . -O unicos common.inc > netcdf.inc
>         ./fortc -L . -O unicos jackets.src > jackets.c
> 
> m4:-:581 can't open file

For what it's worth, this is a well-known problem.  It's due to a bug
in the m4(1) utility on *some* UNICOS systems.  Unfortunately, there's
not much we can do about it (the m4(1) utility doesn't even report an
error).  The workarounds are:

    1.  Get CRI to fix their m4(1) utility.

    2.  Install and use another m4(1) utility (e.g. GNU's).

    3.  Cross-build the relevant files in the fortran subdirectory on a
        platform with a working m4(1) utility and then copy them to the
        system with the non-working m4(1) utility, e.g.

            $ cd fortran
            $ rm -f jackets.c netcdf.inc ftest.f
            $ make jackets.c netcdf.inc ftest.f OS=unicos
            ...
            $ rcp jackets.c netcdf.inc ftest.f \
                user@bad_m4_system.domain/netcdf/src/fortran

--------
Steve Emmerson   <address@hidden>