From owner-freebsd-hackers  Sun Jul 23 12:42:24 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.11/8.6.6) id MAA10717
          for hackers-outgoing; Sun, 23 Jul 1995 12:42:24 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34])
          by freefall.cdrom.com (8.6.11/8.6.6) with ESMTP id MAA10709
          for <hackers@freebsd.org>; Sun, 23 Jul 1995 12:42:16 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id FAA20837; Mon, 24 Jul 1995 05:38:50 +1000
Date: Mon, 24 Jul 1995 05:38:50 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199507231938.FAA20837@godzilla.zeta.org.au>
To: graichen@sirius.physik.fu-berlin.de, hackers@freebsd.org
Subject: Re: man-pages ...
Sender: hackers-owner@freebsd.org
Precedence: bulk

>i think we need some manpages for:

>* matherr(3) - can somebody who knows how this function works write a page for
>it (or maybe adapt a sun page)

I'm surprised to see that we have it.  It is currently broken (not even
reachable).

It allows you to replace the default error handling for the standard
math functions by your own error handling.  Unfortunately the default
error handling forgets the correct (IEEE) return value and there is
no reasonable way to recover.  Forgetting the return value also breaks
normal (IEEEish) error handling, e.g., acos(2.0) is 0 instead of NaN.

The default error handling is actually unreachable because lib/msun
is compiled with the wrong CFLAGS (should be CFLAGS += -D_POSIX_MODE).

Math functions normally never return an error because in FreeBSD math
errors normally generate a fatal SIGFPE.

Bruce