Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!cs.utexas.edu!uunet!mcsun!cernvax!julian
From: julian@cernvax.cern.ch (julian bunn)
Newsgroups: comp.lang.fortran
Subject: comp.lang.fortran
Keywords: fortran90
Message-ID: <5821@cernvax.cern.ch>
Date: 27 Jun 91 09:49:08 GMT
Organization: CERN, Geneva, Switzerland
Lines: 23



Perhaps I can add a measured data point to the compiler versus
translator discussion. For the purposes of tool testing, I have
constructed a 200-line set of garbage Fortran code. Using a mainframe
compiler, this code generates 2 warning errors, and is thus
presumably considered ready for execution. Converting the code
to f90 source form and running it through NAG's f90chk, I get
47 error messages and 33 warning messages, all of them correct.
This opens up new vistas in code testing, and is clear evidence
that the product is certainly more than a dumb translator.

For information, the compiler is invoked in three steps, just like
f77:
     f90c name.f90
     f90link name.c
     a.out
so the C step is invisible. However, the C code contains the necessary
links back to the Fortran code.

                           Mike Metcalf

                           metcalf@cernvm.cern.ch