From owner-freebsd-hackers Mon Nov 6 02:16:57 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id CAA19268 for hackers-outgoing; Mon, 6 Nov 1995 02:16:57 -0800 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id CAA19254 for ; Mon, 6 Nov 1995 02:16:48 -0800 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id VAA11226; Mon, 6 Nov 1995 21:11:28 +1100 Date: Mon, 6 Nov 1995 21:11:28 +1100 From: Bruce Evans Message-Id: <199511061011.VAA11226@godzilla.zeta.org.au> To: bde@zeta.org.au, jkh@time.cdrom.com Subject: Re: NPX still broken in 2.1.0-951104-SNAP... Cc: freebsd-hackers@freebsd.org, julian@TFS.COM, julian@ref.tfs.com, markd@grizzly.com Sender: owner-hackers@freebsd.org Precedence: bulk >FWIW, I think the argument for "greater compatibility" should win out >over all others anyway. If SCO, Linux, Unixware, Solaris and >god-knows-what else all do it this way, then we're just being >pig-headed not to follow precedent and in any comparison people might >make, it won't be FreeBSD that's lauded for taking the idealist's >stand that thrust it in a different direction. Er, we are following (i386) precedent (trapping on errors). The math libraries are just inconsistent with that precedent, and no one has has time or care enough to fix them. The precedent is somewhat braindamaged although it has its advantages, so the correct fix is not obvious. The idealist's stand is that everything uses IEEE arithmetic and that programs know enough about floating point arithmetic to check for errors in the few cases where the IEEE defaults aren't good enough. This has been the standard for a long time in non-i386 Unixes (especially Sun's). Bruce