960207: netCDF on BSD

Masato,

>Date: Wed, 7 Feb 96 14:52:41 JST 
>From: address@hidden (Masato Shiotani)
>Organization: Div. of Ocean and Atmospheric Sci., Hokkaido Univ.
>Subject: Re: 960206: netCDF on BSD 
>Keywords: 199602010918.AA01928

In the above message you wrote:

> >> The above error messages from the m4(1) utility are not good.  They
> >> indicate that the utility is broken (I've seen this error before from
> >> the UNICOS m4(1) utility).  I will send you the files `netcdf.inc' and
> >> `jackets.c' in separate messages.  Put them in the fortran/ subdirectory
> >> and execute the command `make'.
> 
> Do you mean that I should put your 'netcdf.inc' as 'common.inc' and
> 'jackets.c' as 'jacket.src'?  This didn't work, so I just put your
> files as you indicated; is this correct?

You did the right thing.  The files 'netcdf.inc' and 'jackets.c' are
created from the files 'common.inc' and 'jacket.src' by the fortc(1)
script in the fortran/ subdirectory.  This script uses the m4(1) utility,
which appears to be broken on your system.  This is why I sent you the
files 'netcdf.inc' and 'jackets.c' that I created here.

> The following is the output:
> 
> # make
> gcc -c -O -I../libsrc jackets.c
> ar rcuv ../libsrc/libnetcdf.a jackets.o
> a - jackets.o
> ranlib ../libsrc/libnetcdf.a

The above looks good.
 
> >> Did you execute the command `make clean'?  Judging from the debugger 
> >> output, it appears that the files in the xdr/ subdirectory were not
> >> compiled with the `-g' option (the argument lists do not have variable
> >> name information).  Would you please do the following:
> 
> Yes, I tried to 'make clean' then 'make'.  But it happened nothing.  I
> suppose during the configuration it is assumed that the system xdr is
> appropriate, and the resulting Makefile in xdr acts to do nothing.
> (This can be seen in the output of the configuration and 'make' as I
> sent you before.)  Probably, I should know how to compile the xdr
> library based on resources coming with the NetCDF.

It appears that the XDR library on your system is broken.  We can test
this by having the program `xdrtest' in the xdr/ subdirectory use the
XDR implementation that comes with the netCDF package.  Try the
following:

    1.  Go into the xdr/ subdirectory.

    2.  Edit the file `Makefile': change the value of the macro LIBOBJS
        from

            LIBOBJS     =

        to

            LIBOBJS     = xdr.o xdrfloat.o xdrstdio.o xdrarray.o

    3.  Execute the command `make clean'.

    4.  Execute the command `make all'.  Trap the output and send it to
        me.

    5.  Stop here if the above fails (it should not).

    6.  Execute the command `make test'.  Trap the output and send it to
        me.

    7.  Stop here if the above fails.

    8.  Go up one directory -- to the top-level source directory.

    9.  Execute the command `make test'.  Trap the output and send it to
        me.

    10.  Stop here if the above fails.

    11.  Execute the command `make install'.

    12.  Stop here if the above fails.

    13.  Execute the command `make clean'.

> Masato Shiotani (Div. of Ocean and Atmospheric Sci., Hokkaido Univ.)

Are you in Oceanographic or Atmospheric Science?  We have several
meteorologists here; I'm the only oceanographer.

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