From owner-freebsd-current Sun Sep 7 07:40:04 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id HAA20711 for current-outgoing; Sun, 7 Sep 1997 07:40:04 -0700 (PDT) Received: from usr09.primenet.com (tlambert@usr09.primenet.com [206.165.6.209]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id HAA20693 for ; Sun, 7 Sep 1997 07:40:02 -0700 (PDT) Received: (from tlambert@localhost) by usr09.primenet.com (8.8.5/8.8.5) id HAA08754 for current@freebsd.org; Sun, 7 Sep 1997 07:40:01 -0700 (MST) From: Terry Lambert Message-Id: <199709071440.HAA08754@usr09.primenet.com> Subject: Bug in committed pci/if_de.c To: current@freebsd.org Date: Sun, 7 Sep 1997 14:40:00 +0000 (GMT) X-Mailer: ELM [version 2.4 PL23] Content-Type: text Sender: owner-freebsd-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk The committed if_de.c incorrectly references FreeBSD version information. This is incorrect, because version information is already encoded in the act of commision to the source tree. One does not check a driver out of -current, and then compile it in an old tree... one checks out the updated driver from the old tag. That's what "stable" is for. It's also incorrect because it's using __FreeBSD__, which is a compiler compile-time constant, and has nothing to do with the version of FreeBSD that is being target by the compile, only (really) which version of FreeBSD the compiler was compiled on. In point of fact, without two "build world"'s, the first build of the tools will take place on the older version of OS (say 2), and have the older number, even when the build world is (supposedly) a newer version (say 3). This is less of a problem for people who track -current, and do so with full world builds, but for people who upgrade via "build world" only at release time, it's an issue. Anyway, since I like to cross-compile and keep a -stable older machine, it's an issue for me. Arguably, there should be no non-contrib code (and that, in user space) that even references __FreeBSD__ at all, and the gcc/cpp values should come from somewhere else. Somewhere configurable, prefferably. Maybe a sysctl that defaults to the real version and can be lowered manually? Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.