From owner-freebsd-current Fri May 2 23:19:44 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.5/8.8.5) id XAA15206 for current-outgoing; Fri, 2 May 1997 23:19:44 -0700 (PDT) Received: from zibbi.mikom.csir.co.za (zibbi.mikom.csir.co.za [146.64.24.58]) by hub.freebsd.org (8.8.5/8.8.5) with ESMTP id XAA15201 for ; Fri, 2 May 1997 23:19:41 -0700 (PDT) Received: (from jhay@localhost) by zibbi.mikom.csir.co.za (8.8.5/8.8.5) id IAA06658; Sat, 3 May 1997 08:19:26 +0200 (SAT) From: John Hay Message-Id: <199705030619.IAA06658@zibbi.mikom.csir.co.za> Subject: Re: -current build is now broken.. In-Reply-To: <199705022339.QAA24331@austin.polstra.com> from John Polstra at "May 2, 97 04:39:28 pm" To: jdp@polstra.com (John Polstra) Date: Sat, 3 May 1997 08:19:26 +0200 (SAT) Cc: current@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL31 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-current@FreeBSD.ORG X-Loop: FreeBSD.org Precedence: bulk > > I think with a little care it will be possible to keep building > > 3.0 snaps and releases on a 2.2 machine. > > Have you forgotten how this thread got started? The thing that > you say will be possible "with a little care" got broken by Bruce. > Nobody on this project is more careful than Bruce. (And his > competence is hardly in doubt, either.) > > A little care doesn't cut it in software -- you need testing. It's > unreasonable to expect all the developers to have what's needed > for testing 3.0 release builds on 2.2 machines. It's much more > reasonable to ask that the people who build releases -- a dramatically > smaller group -- do it on up-to-date systems. > Maybe first a little background. I normally build the releases that go onto ftp.internat.freebsd.org - the international crypto releases. So in the last few months I had to build various 2.1.x and 2.2.x releases and a 3.0 snap or two, but I only have one machine (our server) that have the capability (mostly disk space) to build releases. I do have a machine at home that is running current, but while I have the cvs tree there and can do a make world, it doesn't have enough diskspace to do a make release. O as an added bonus, none of these two machines actually run the crypto stuff. :-) When a make release breaks, I make sure that make world still works on my -current box at home and if it works, I know it is a make release problem and if I can figure out what is wrong I send in patches or have a chat with Jordan or Mark Murray. Up to now this has worked for me and while it did not guarantee a buildable make release everytime, we did get there again. :-) By far the most problems that I had was dependency related or assuming that something will be there (which is also kind of dependency I guess). I don't doubt Bruce's capabilties. Maybe what we need to do, is when we change the build behaviour (adding and using a new option in install for instance) that we also add it to the appropriate bootstrap targets and if we can't test a make world and make release, we give a "heads up" warning so that people building releases know about it. Remember that even if you decide not to worry about people building 3.0 releases on 2.2 machines, we still need to support source upgrades from 2.2 to 3.0 and that will generally take us 90% of the way. John -- John Hay -- John.Hay@mikom.csir.co.za