From owner-freebsd-ports Thu Apr 20 17:00:14 1995 Return-Path: ports-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id RAA14368 for ports-outgoing; Thu, 20 Apr 1995 17:00:14 -0700 Received: from gold.tc.umn.edu (root@gold.tc.umn.edu [128.101.115.11]) by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id RAA14362 for ; Thu, 20 Apr 1995 17:00:11 -0700 Received: by gold.tc.umn.edu; Thu, 20 Apr 95 18:59:57 -0500 Date: Thu, 20 Apr 1995 18:59:57 -0500 (CDT) From: Jan E Backlund Subject: Re: Floating point exceptions in Tcl port. To: Satoshi Asami | =?ISO-2022-JP?B?GyRCQHUbKEI=?= =?ISO-2022-JP?B?GyRCOCsbKEIgGyRCOC0bKEI=?= cc: ports@FreeBSD.org In-Reply-To: <199504200429.VAA11816@silvia.HIP.Berkeley.EDU> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: ports-owner@FreeBSD.org Precedence: bulk On Wed, 19 Apr 1995, Satoshi Asami | =?ISO-2022-JP?B?GyRCQHUbKEI=?= =?ISO-2022-JP?B?GyRCOCsbKEIgGyRCOC0bKEI=?= wrote: > * When I run 'make test' after making the Tcl port, I get a number of > * messages from the kernel about processes exiting with masked floating > * point exceptions. Could someone please explain the meaning and > * significance of these messages. > > I don't know how you can do a "make test", as there is no "test" > target in the Makefile. If you mean you went into work/tcl7.3 (or > something like that) and did a "make test" and got FP errors, that > sounds like a serious problem. Please send a detailed report (output > of "make test", "dmesg" or anything we can use to identify the > problem). > > To ports people: by the way, someone want to maintain tcl? It's way > too important to leave it MAINTAINER-less, with me answering questions > with shots in the dark like this. ;) > > Satoshi > You are right, I went into work/tcl7.3 and did 'make test'. The messages, I've discovered, are system messages (Perhaps a slogan: 'You learn something new everyday -- when you use Unix!'). The format is: April 20 18:30:01 sacc /kernel: pid 1175 (tcltest) exited with masked floating point exceptions 0x29 My vague glimmerings of an understanding suggest that in 4.4BSD, unlike some systems, the kernel will flag floating point exceptions and exit with an error code unless the process explicitly requests that the offending exception should be ignored. In any case the event is logged to the system log. When jkh (Jordan Hubbard?) ported tcl, the following lines were added to tclBasic.c: fpsetround(FP_RN); fpsetmask(0L); such that tcl wouldn't die when a floating point exception occurs. (What does the L in 0L do, by the way?) What I still don't understand is whether the exceptions are par for the course and that most operating systems are too dumb to notice that something has gone wrong, or if they are specific to the port. Jan Erik back0003@gold.tc.umn.edu