Path: utzoo!attcan!uunet!mcsun!unido!altger!york
From: york@altger.UUCP (york)
Newsgroups: comp.sys.atari.st
Subject: Re: Absoft Fortran
Summary: Experiences while implementing NAG graphical library
Message-ID: <2066@altger.UUCP>
Date: 3 Dec 89 14:32:30 GMT
References: <8912010814.AA04387@ucbvax.Berkeley.EDU>
Reply-To: york@altger.UUCP (york)
Organization: Altos Computer Systems Munich
Lines: 28

I also understand your problems. Some of them may be solved by
using the 'script' commando from the library manager. May be, the
linker also has cuch a command. There is still no way to do it via
command line options.

But it is even worse. Recently I tried to impement the NAG graphics
library (*only* about 350 routines). For I had impemented it before
on my UN*X system, I created two one pass libaries (using the output
of lorder and tsort) on my ST. The linker map told me, that no routine
was missing. So I started the test program. There was neither a label,
nor a loop construct in the main program, but the code managed to 
build up a never ending loop.

Ok, I took the linker map from the ST and concatenated all modules
which should be neccessary to build the program to one file. I compiled
and started the program. I was somewhat surprised, that the run-time
link-loader now told me, that one routine was missing.

To make the long story shorter: The linker had found about 22 modules
in the libraries, but additional 25 ones had to be linked to build a
working program.

Conclusion: The compiler of Absort Fortran may be nice, but beware of
the linker. May be I 'll be using the Prospero product despite its
longer compilation and linking times.

Ulrich Liesenfeld     s=uli; ou=analyt; ou=chemie; p=uni-bochum;
					  a=dbp; c=de;