5.9.0rc3 build failure in numerical/slatec/zbesy.lisp, gcl 2.4.3, gcc 3.2

[Sorry if anybody gets this twice, i had a problem with Valerij's
address, maybe it was set up to avoid spam?  Thanks again for the replies.]

Hi Valerij,

Thanks for the advice.  Also thanks to Jim (amundson@fnal.gov) and
RJF (fateman@cs.berkeley.edu) for their helpful remarks.

It looks like the compilation process generates c code which is then
compiled (and deleted).  But since there's a failure to compile the
zbesy.c, it sticks around.  And at the top of zbesy.c is
#include <cmpinclude.h>, so some file named cmpinclude.h must be
visible.  And by setting C_INCLUDE_PATH appropriately
(to /home/lisp/src/gcl-2.4.3/h in my case), that obstacle can
be overcome.  (I suppose that the later-versioned gcl version must make
its own include directory as part of the build process, and perhaps
configure could detect this and set up the makefiles suitably.)

Unfortunately, zbesy.c is truncated in the middle.  And that may
be the origin of the complaint about not printing a non-number: whatever
software is emitting the c code wants to emit a number but cannot
for whatever reason, perhaps some variable is undefined.  (I guess
the rust in my recollection of lisp is showing.)

Anyhow thanks to all three of you for the advice, and i'm open
to any other suggestions, especially if there's some way to avoid
the cvs route: i have limited bandwidth, and it looks like the cvs
tree for gcl is at least an order of magnitude larger than the tarball.

Maybe there's a tarball for gcl 2.5.0 somewhere?  (Or maybe there's
some cvs command that will generate one?)

Thanks again.

dan
 



   From: Valerij Pipin <"vpipin at irk, pip"@iszf.irk.ru>
   Cc: hitt@panix.com, maxima@www.ma.utexas.edu
   Organization: ISZF
   Sender: maxima-admin@www.ma.utexas.edu
   Date: Thu, 7 Nov 2002 09:41:29 +0800

   Hello!

   I'm not sure but at least one thing should be done here. 
   For  building maxima with gcl-2.4.3 you have to put the file "cmpinclude.h" (that is made during gcl building)
   somewhere in include path. 
   Note, if you will try the cvs gcl you need to upgrade the gcc-3.2 up to the latest version.

    best regards
   Valerij

   On 06 Nov 2002 14:03:47 -0600
   James Amundson <amundson@fnal.gov> wrote:

   > On Wed, 2002-11-06 at 13:22, Dan Hitt wrote:
   > > Hi,
   > > 
   > > I'm trying to build maxima 5.9.0rc3.
   > > 
   > > I'm using a red hat 7.3 (GNU/Linux) system, and i managed to build
   > > gcl 2.4.3 (whose prompt says 2.5.0, but the tar ball is labelled 2.4.3)
   > > using gcc 3.2, and i'm trying to use this gcl to build maxima.
   > 
   > I'm afraid you need to use GCL 2.5.0, i.e., the current CVS version of
   > GCL. The file README.lisps says so, but perhaps not forcefully enough.
   > 
   > It doesn't make me happy to rely on a CVS version, but I'm afraid that's
   > the way it is at this time.
   > 
   > If anyone has had *success* with GCL 2.4.3, I would like to hear about
   > it.
   > 
   > --Jim
   > _______________________________________________
   > Maxima mailing list
   > Maxima@www.math.utexas.edu
   > http://www.math.utexas.edu/mailman/listinfo/maxima
   > 
   _______________________________________________
   Maxima mailing list
   Maxima@www.math.utexas.edu
   http://www.math.utexas.edu/mailman/listinfo/maxima