Compiling netCDF under Linux (answer)

• To: netcdfgroup@xxxxxxxxxxxxxxxx
• Subject: Compiling netCDF under Linux (answer)
• From: mmccarri@xxxxxxxxxx (Mike McCarrick)
• Date: Tue, 12 Apr 1994 07:58:44 -0700 (PDT)

I have had no trouble compiling the stock netCDF source under Linux
with a couple of minor changes. 

1) The lex available with Linux is gnu flex -- it defines yywrap() as
   a macro with a constant value of 1, where ncgen/ncgen.y is trying
   to define a function called yywrap() that returns 1. Either 
   #undef yywrap or #ifdef out the yywrap() function.

   Note to developers: you might want to include this in your 
   distribution for those folks using gnu utilities. That is,
   in ncgen.y:
        #ifndef yywrap
        int yywrap() 
        {
            return 1;
        }
        #endif

2) The "fortran compiler" distributed with linux is really a shell
   script that calls f2c, the fortran to C converter. It doesn't work
   on the fortran source in the netCDF distribution. I don't remember
   enough fortran to tell why. The problem is, configure will find your
   f77 script and set FC=f77 in the Makefiles despite the fact that you
   set FC="" in CUSTOMIZE. Just edit the master Makefile and set FC="" 
   after running configure or hide your f77 before running configure.

3) The -traditional flag should *not* be used with gcc. The distributed
   CUSTOMIZE file shows "CC = gcc -traditional" as a possible value
   (commented out) but this will not work. Just use CC=gcc.

4) There is no number 4 - that's all we need!  "make all" proceeds with
   no errors. "make test" runs with no errors. 

The library works under Linux without a hitch. I have a number of 
data analysis routines that I can compile under DOS or Linux. They
all run twice as fast under Linux because of the 32 bit code and the
faster filesystem. I keep datafiles on an NFS server so they can be 
accessed by a number of users. If I need to do some processing, it
is much faster to reboot my machine to Linux, process all of the files,
then reboot to DOS to look at them (my nice display routines are 
Windows programs -- I haven't figured out Motif yet).


Good Luck!

Mike

-- 
Dr. Michael McCarrick
ARCO Power Technologies, Washington DC

>From owner-netcdfgroup@xxxxxxxxxxxxxxxx    23 Wed, Oct
Date:    Wed, 23 Oct 1991 11:38:46 EDT
From: GUMLEY@xxxxxxxxxxxxxxxxx (Liam E. Gumley)
To: netcdfgroup@xxxxxxxxxxxxxxxx
Subject: netCDF on IBM 3090
Received: by unidata.ucar.edu id AA26676
  (5.65c/IDA-1.4.4 for netcdfgroup-send); Wed, 23 Oct 1991 10:09:51 -0600
Received: from ltp.gsfc.nasa.gov by unidata.ucar.edu with SMTP id AA26672
  (5.65c/IDA-1.4.4 for <netcdfgroup@xxxxxxxxxxxxxxxx>); Wed, 23 Oct 1991 
10:09:49 -0600
Message-Id: <911023113846.2280012c@xxxxxxxxxxxxxxxxx>
X-Vmsmail-To: SMTP%"netcdfgroup@xxxxxxxxxxxxxxxx"

We are looking into installing the netCDF software on an IBM 3090 system.

If anyone has previously attempted such an installation, I would appreciate
very much any information regarding such details as

netCDF source code version
transporting the netCDF source code to the IBM 3090
IBM 3090 operating system version
IBM compiler versions (C, FORTRAN)
problems in compiling the netCDF source
problems in executing the netCDF test programs
patches necessary to the netCDF code

Thanks very much for you time.

Cheers,
Liam.

--
Liam E. Gumley                                  
MODIS Science Data Support Team
Research and Data Systems Corporation              Disclaimer:
Greenbelt MD, USA                                  Opinions I express here are
gumley@xxxxxxxxxxxxxxxxx                           my own, not the company's.