Path: utzoo!mnetor!uunet!husc6!uwvax!dogie!uwmcsd1!ig!agate!ucbvax!decwrl!purdue!i.cc.purdue.edu!j.cc.purdue.edu!pur-ee!uiucdcs!uxc.cso.uiuc.edu!uxe.cso.uiuc.edu!hirchert
From: hirchert@uxe.cso.uiuc.edu
Newsgroups: comp.lang.fortran
Subject: Re: question about constants in FORTRAN
Message-ID: <50500048@uxe.cso.uiuc.edu>
Date: 24 Apr 88 20:17:00 GMT
References: <1801@ut-emx.UUCP>
Lines: 19
Nf-ID: #R:ut-emx.UUCP:1801:uxe.cso.uiuc.edu:50500048:000:976
Nf-From: uxe.cso.uiuc.edu!hirchert    Apr 24 15:17:00 1988


>The next most helpful thing would be for the compiler to choose an
>implementation strategy that avoided truly bizarre resukts.  In this
>case, what I would do is create at each call of "sub", a temporary
>variable initialised with the value zero, and pass that address as
>the parameter.  Then there is no way an illegal redefinition within
>sub can change the value of a subsequent occurrence of the literal
>zero.  I submit that a compiler writer who cared for the users of
>his work would at least go this far.

If a compiler writer cares about me as a user, he won't use this technique
to slow down my correct programs, just so your incorrect programs don't get
bizarre results.  (Perhaps, this should be a compiler option.  Then I could
turn it on for programs that I suspect of doing such illegal things, but I
could still avoid the performance hit on programs I know to be correct.)

>Robert Firth

Kurt W. Hirchert    National Center for Supercomputing Applications