Re: netCDF 2.3.2 on Solaris

> Organization: Sandia Labs
> Keywords: 199408311755.AA09807 netCDF 2.3.2 Solaris configure

Hi Darin,

> I am trying to install netcdf 2.3.2 pl2 on a Sun SS10 as netcdf is the basis
> for the Sandia Nat'l Labs exodus II format for FEA codes which SNL shares
> with us. Unfortunately I have not had much luck. Chris Webster of ucar has
> attempted to help me for the last couple of days and has recommended that I
> turn to support@unidata as a best/last resort.

I just tried installing netcdf 2.3.2 pl2 on a SPARCstation 10 running
Solaris 2.3, and it seemed to work fine.  After uncompressing and un-tarring
the source distribution from
ftp://ftp.unidata.ucar.edu/pub/netcdf/netcdf.tar.Z (which is just a link to
ftp://ftp.unidata.ucar.edu/pub/netcdf/netcdf-232pl2.tar.Z), I just typed

    ./configure --prefix=/home/russ/build

and then

    make

There were a few compiler warnings of the form

    warning: semantics of ">" change in ANSI C; use explicit cast

for the files fortran/jackets.c, ncdump/ncdump.c, ncgen/generate.c, but no
compile errors.  When I then ran 

    make test

all tests were passed with no more compile errors or warnings and no
problems reported.

> I can send a copy of a plain vanilla install log. However to summarize my
> problem seems to be during the make that error.c, genlib.c, and dumplib.c
> all complain of an argument mismatch and fail to compile. Looking at the
> fortan file I do not find a solaris install file so it defaults to sunos and
> in CUSTOMIZE Chris suggested changing to OS=solaris but this did not help. I
> do not have a solaris.m4 install file.

When you run `configure', it should determine for itself that OS='sunos',
and that should be enough information for the subsequent make to run
correctly.  It should not be necessary to set OS explicitly.  The
definition OS='sunos' is correct for Solaris 2.3, since the operating
system is just SunOS 5.3.  In building netCDF it is not necessary to
distinguish between SunOS 4 and SunOS 5, since they both use the same
conventions for Fortran programs calling C libraries (which happens to be
the only use made of the OS macro).

If you find, after running

    make distclean
    configure --prefix=<whatever>
    make
    make test

that you get a compile error or an error in the output of "make test",
please email me the resulting `config.status' file that `configure' produces
and the output of `make'.  It might also be possible that some environment
variable you have set is inadvertently being interpreted by the `configure'
script, so I'd also like to see the output of `env'.  Otherwise, if things
work OK after doing the above, then you should just be able to `make
install' and be finished.

> P.S. my email address is probably reported wrong by my sun. It should be
> address@hidden However, I am sure that address@hidden
> works and is checked daily.

It came out address@hidden when I replied, so I assume it's
being reported correctly.

--
Russ Rew                                                   UCAR Unidata Program
address@hidden                                             P.O. Box 3000
http://www.unidata.ucar.edu/                             Boulder, CO 80307-3000