Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!uunet!philmtl!philabs!trotter!dean1!aj4640
From: aj4640@dean1.usma.edu (Robertson John MAJ)
Newsgroups: comp.lang.fortran
Subject: Re: Microsoft Fortran
Keywords: optimizer
Message-ID: <215@dean1.usma.edu>
Date: 9 Apr 90 14:30:20 GMT
References: <1990Apr6.230725.17580@cs.dal.ca>
Reply-To: aj4640@dean1.acd.usma.edu
Organization: US Military Academy, West Point, NY
Lines: 21
Summary:
Expires:
Sender:
Followup-To:
Distribution:

I ran into a strange problem with the new Microsoft 5.0
Fortran compiler.

I recompiled a well-used piece of production code with
5.0 and immediately began generating runtime erros.
Curiously, when I re-compiled with the debug switch
thrown, the errors went away--thus debugging with Code
View was impossible.  After painstakingly tracking the
bug, I discovered that the compiler (with default
settings) was generating code that spuriously changed
the value of an integer*4 variable named NOLD to an
apparently random value.  Changing the variable name
solved the problem.  Also, running the compiler with
the optimizer disabled (/Od, I think) also stopped the
problem.

Compilers at relase 5.0 should not have such odd
behavior.  Any other horror stories out there?

Jack Robertson
aj4640@dean1.acd.usma.edu