Date: Thu, 23 Jul 1998 04:21:43 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: jb@cimlogic.com.au (John Birrell) Cc: nik@iii.co.uk, dwhite@resnet.uoregon.edu, jb@cimlogic.com.au, current@FreeBSD.ORG Subject: Re: Does building current on 2.2.x still work? Message-ID: <199807230421.VAA15337@usr05.primenet.com> In-Reply-To: <199807221002.UAA04499@cimlogic.com.au> from "John Birrell" at Jul 22, 98 08:02:57 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > Every time I see this suggestion posted, someone else jumps in with "But > > it didn't work properly for me.", which is why I haven't put it in the > > tutorial yet. > > I sent a message (the one with the *grumble*) to this list asking people > having trouble to send me private mail with the details of the problem. > So far I've received a total of ZERO private mails on this subject. > I can only conclude that when people follow the procedure, there is no > problem. I can resend this as private email if you don't read it here... 8-). The failure I see is in the if_de.c driver, which uses the version number of the build environment instead of the version number of the target environment. This is because there was code with conditional compilation checked into the 3.x source tree. Specifically, the version number is a gcc and preprocessor builtin that is statically defined, and is not dynamically determined, so even running a FreeeBSD 2.x compiler on a FreeBSD 3.x system will trigger the bug. I realize that this was imported directly to avoid maintenance issues, since the main developement platform is apparently NetBSD and the author wants to maintain a single set of sources. However (and it's a big "however"), this type of this is precisely why CVS supports vendor branches. > I don't think this is the way to word it. Anyone building -current needs > to allow for the possibility that changes might have been made to the > .mk files since the last installation on their machine. We all expect that > sources in the tree are consistent so that /usr/src/Makefile should only > be interpreted by make in conjunction with the .mk include files that are > also in the tree. To ensure that this consistency is maintained and the > world has the best chance of building, novice world builders should > _always_ use the -m argument "just in case" IMHO. Or the buildworld target should add the argument and recall make in a subshell on the user's behalf. > The fact that it is often possible to build without the -m argument > should be the thing that is left undocumented. We shouldn't need to > make special mention about building -current on 2.2.6 when there is > a "command for all seasons" that is backed by logic. Hm. 8-|. The default command (ie: no arguments) should always "do the right thing", in my book... the subshell hack is pretty trivial. > FWIW, when we've got an ELF only build, people building on an aout > system will get even worse errors if they have out of date .mk files. 8-(. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199807230421.VAA15337>