From: "A.Appleyard" <A DOT APPLEYARD AT fs2 DOT mt DOT umist DOT ac DOT uk>
To: djgpp AT sun DOT soe DOT clarkson DOT edu
Date: Fri, 24 Feb 1995 16:10:27 GMT
Subject: dodging the INF-amous NaN-nygoat

On Thu, 23 Feb 1995, Erwin Kok wrote:
  > ... I made up a program compiled it and it worked (under DJGPP
FPU-emulation), but when I bought a co-processor, I compiled it and a strange
floating-point exception appeared... After a few debugging hours I found that
the result from an arithmetic statement was NAN (Not A Number) ... My advise
is check if any variables are NAN's.

Bill Davidson <bdavidson AT ra DOT isisnet DOT com> replied:-
  > A couple of weeks ago (?) someone posted a lovely little CTRL87.C program
that allows one to diddle the '87 control word. It works like a charm and
allows one to do just this sort of thing without resorting to asm.

  Again, please, either by re-posting this CTRL87.C program or in some other
way, let the djgpp user be able to enter a state where floats and doubles that
become INF or NaN are passed through arithmetic normally and don't make the
program abort. I used to run on a CDC mainframe where I could do that, and it
was very handy: 1/0=INF; 1/INF=0; 0/0=NaN; log(0)=-INF; exp(-INF)=0; etc.