Path: utzoo!utgpu!water!watmath!clyde!att!rutgers!rochester!uhura.cc.rochester.edu!ur-valhalla!galaxy.ee.rochester.edu!davis
From: davis@galaxy.ee.rochester.edu (Al Davis)
Newsgroups: comp.lang.fortran
Subject: Re: Fortran vs C for computations
Message-ID: <1443@valhalla.ee.rochester.edu>
Date: 6 Sep 88 00:52:29 GMT
References: <535@nikhefh.hep.nl>
Sender: usenet@valhalla.ee.rochester.edu
Reply-To: davis@ee.rochester.edu (Al Davis)
Organization: UR Dept. of Electrical Engg, Rochester NY 14627
Lines: 26


J.Vermaseren says....
>        float x = -2.0;
>	        z = sqrt(x);
>		gives back some value.

>The better thing would be to have a parameter that can be set by the user
>that determines whether an overflow or sqrt of a negative number should be
>fatal (with intelligent errormeassage please). Most fortran compilers have
>this option.

I have never seen a fortran compiler that had this option.  They all abort
the program with a non-friendly message.  Any option that any fortran would
have would necessarily be non-portable.

C has a function "matherr" that defines how these errors are handled.  When
an error occurs, it calls matherr.  The vendor supplies a default handler,
that returns a value, and sets the global variable "errno", etc.  If you
don't like it, replace it with your own.  You have full control.  It is
portable, standard, and well documented.

On another subject, I suspect our postnews doesn't work correctly.  Will
someone far away please send me mail if you read this?

al davis.
davis@ee.rochester.edu