From owner-freebsd-current  Wed Aug 26 05:43:56 1998
Return-Path: <owner-freebsd-current@FreeBSD.ORG>
Received: (from majordom@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id FAA23335
          for freebsd-current-outgoing; Wed, 26 Aug 1998 05:43:56 -0700 (PDT)
          (envelope-from owner-freebsd-current@FreeBSD.ORG)
Received: from gilgamesch.bik-gmbh.de (gilgamesch.bik-gmbh.de [194.233.237.91])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id FAA23330
          for <current@FreeBSD.ORG>; Wed, 26 Aug 1998 05:43:54 -0700 (PDT)
          (envelope-from cracauer@gilgamesch.bik-gmbh.de)
Received: (from cracauer@localhost) by gilgamesch.bik-gmbh.de (8.8.8/8.7.3) id OAA21218; Wed, 26 Aug 1998 14:43:03 +0200 (MET DST)
Message-ID: <19980826144302.C20511@cons.org>
Date: Wed, 26 Aug 1998 14:43:02 +0200
From: Martin Cracauer <cracauer@cons.org>
To: Stephen Hocking-Senior Programmer PGS Tensor Perth <shocking@prth.pgs.com>,
        current@FreeBSD.ORG
Subject: Re: Floating Point Exceptions, signal handlers & subsequent ops
References: <199808250753.PAA29567@ariadne.tensor.pgs.com> <19980825180559.A9890@cons.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
X-Mailer: Mutt 0.93.1i
In-Reply-To: <19980825180559.A9890@cons.org>; from Martin Cracauer on Tue, Aug 25, 1998 at 06:05:59PM +0200
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

In <19980825180559.A9890@cons.org>, Martin Cracauer wrote: 
> In <199808250753.PAA29567@ariadne.tensor.pgs.com>, Stephen Hocking-Senior Programmer PGS Tensor Perth wrote: 
> > 
> > I've noticed in one of my applications that the first FP operation after 
> > return from a caught SIGFPE is invalid. I have a signal handler installed that 
> > just prints out some basic info (like "SIGFPE caught"). The first FP op after 
> > this (in my case, converting a long to a double) just gives garbage. Repeat 
> > the same statement and it gives a sensible result. Has anyone else seen this 
> > before I file a PR with code to reproduce the problem?
> 
> I just did a short test and for me the next FP operation after a
> caught SIGFPE (division by zero) is still sane (without
> setjump/jumpjump).

I'm afraid I was talking nonsense here :-( This test program shows the
problem. Now the question, why did my other test work well...?

#include <stdio.h>
#include <stdlib.h>
#include <signal.h>

void
handler(int s)
{
  write(1, "sig\n", 4);
}

int
main(void)
{
  double f0;
  double f2;
  double f4;
  double f8;
  double res;

  /* Prevent optimization */
  f0 = atof("0.0");
  f2 = atof("2.0");
  f4 = atof("4.0");
  f8 = atof("8.0");

  signal (SIGFPE, handler);

  res = f2 / f0;

  printf("After division by zero: %g\n", res);

  res = f4 + f8;

  if (res != 12.0) {
    printf("Next operation didn't work: %g\n", res);
  }
  else {
    printf("Next operation worked: %g\n", res);
  }

  return 0;
}


-- 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Martin Cracauer <cracauer@cons.org> http://www.cons.org/cracauer
BSD User Group Hamburg, Germany     http://www.bsdhh.org/

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message