20030509: McIDAS on Red Hat 9

>From:  Andrew Tupper <address@hidden>
>Organization:  ABoM
>Keywords:  200305090727.h497Rp7U015039 McIDAS-X v2002d RedHat 9 Linux f2c

Andrew,

>I'm having a go at installing McIDAS 2002d on a Red Hat Linux 9 machine,

You are the first McIDAS users I have heard of that is trying to build
McIDAS under RedHat 9.0 Linux.  Even though we don't have a RH 9
machine in-house, I will chime in with some thoughts.

>and have stuck a problem the same as that described in #11961 (for Red
>Hat 7.3) and #12098 (Red Hat 9) in the MUG inquiry database.

The problem that Deb Molinar reported in INQ 11961 was a bit different
from yours.  Her build under RH 7.3 would hang at the 'convdlm mdxgra'
step.  A quick look at your makelog tells me that the convdlm executable
is most likely not being build on your machine:

./mccomp -O -gcc -I. -I../netcdf/libsrc -I../hdf/hdf/src -I../hdf/mfhdf/libsrc 
-I../jpeg -I../zlib -c convdlmf.f
mccomp: Using default optimizations
rm -f convdlmf.o
mcfc -c -O -I. -I../netcdf/libsrc -I../hdf/hdf/src -I../hdf/mfhdf/libsrc 
-I../jpeg -I../zlib convdlmf.f
convdlmf.f:
   convdlm:
   nchars:
   len_trim:
compile convdlmf.f:             done
./mccomp -O -Wl,-s -gcc -o convdlm convdlmc.o convdlmf.o
mccomp: Using default optimizations
mcfc -Wl,-s -O -o convdlm convdlmc.o convdlmf.o -L/usr/X11R6/lib -lm
/usr/local/lib/libf2c.a(lread.o)(.text+0x1ae): In function `l_R':
: undefined reference to `__ctype_b'
/usr/local/lib/libf2c.a(lread.o)(.text+0x2d1): In function `l_R':
: undefined reference to `__ctype_b'
/usr/local/lib/libf2c.a(lread.o)(.text+0x399): In function `l_R':
: undefined reference to `__ctype_b'
/usr/local/lib/libf2c.a(lread.o)(.text+0x952): In function `l_L':
: undefined reference to `__ctype_b'
/usr/local/lib/libf2c.a(lread.o)(.text+0xba4): In function `l_CHAR':
: undefined reference to `__ctype_b'
/usr/local/lib/libf2c.a(lread.o)(.text+0xd12): more undefined references to 
`__ctype_b' follow
collect2: ld returned 1 exit status
link    convdlm:                done

One of the unfortunate side effects of using the mcfc script wrapper for
gcc/f2c is that the make process does not die when errors are encountered.
This is one of the smaller reasons that led me to adapt McIDAS code to
use the GNU g77 compiler instead.

>The
>problem is that the compilation fails at the 'convdlm mdxgra' stage (last
>few lines reproduced below).&nbsp; I've tried it with two different
>versions of f2c, and I've verified that f2c and gcc are in the mcidas
>path and looked at the other material on the Unidata page at:
>
>http://www.unidata.ucar.edu/packages/mcidas/780/mcx/warnings_mcx.html#gcc

The error talked about in this Unidata inquiry was related to the linker
using the shared f2c library, libf2c.so, instead of the non-shared
version, libf2cd.a.  The shared library is missing the MAIN__ entry
point, and that was the problem referenced above.

>I was wondering if anybody has successfully installed McIDAS on a Red Hat
>7.3, 8 or 9 machine, or can offer any more suggestions for the
>problem?

Several Unidata sites routinely build under 7.3 and 8.0.  You are the first
brave lad that has vendured into the RH 9 arena :-)

>The last few lines of the log are:
>
>./convdlm mdxgra.dlm
>./mclog: line 251: ./convdlm: No such file or directory
>convdlm mdxgra.dlm: FAILED

Yes, but the real problem is actually in the creation of the convdlm
executable.  What I normally do in cases like this is try building
the executable by hand so I can get a better idea of exactly what
is missing.  In your case, the __ctype_b entry point referenced in
libf2c.a is the missing quantity.  So, the first thing to do is find
out what library this entry point is in, and add it to your link line
for convdlm.  If you can get past this step, the next hurdle will likely
be seeing if your convdlm executable can convert the various .dlm
files.  For this you may have to run the make by hand from the
mcidas2002d/src directory (e.g., 'make all').

>The full log is at:
>http://arist.edu.kagoshima-u.ac.jp/andrewt/makelog.txt
>
>Thanks for any help!

I don't know if the above helps, but is should be a starting point.

>Andrew Tupper
>Bureau of Meteorology, Northern Territory Regional Office,
>Australia.
>Currently at: Faculty of Education, Kagoshima University, Kagoshima,
>Japan
>mobile phone: (+81) 090 448 34958
>E-mail: address@hidden

Cheers,

Tom
--
+-----------------------------------------------------------------------------+
* Tom Yoksas                                             UCAR Unidata Program *
* (303) 497-8642 (last resort)                                  P.O. Box 3000 *
* address@hidden                                   Boulder, CO 80307 *
* Unidata WWW Service                             http://www.unidata.ucar.edu/*
+-----------------------------------------------------------------------------+