From owner-freebsd-current Thu Jun 27 16:09:05 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id QAA05277 for current-outgoing; Thu, 27 Jun 1996 16:09:05 -0700 (PDT) Received: from rocky.mt.sri.com (rocky.sri.MT.net [204.182.243.10]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id QAA05272 for ; Thu, 27 Jun 1996 16:09:02 -0700 (PDT) Received: (from nate@localhost) by rocky.mt.sri.com (8.7.5/8.7.3) id RAA12141; Thu, 27 Jun 1996 17:08:50 -0600 (MDT) Date: Thu, 27 Jun 1996 17:08:50 -0600 (MDT) Message-Id: <199606272308.RAA12141@rocky.mt.sri.com> From: Nate Williams To: "Jordan K. Hubbard" Cc: scott@statsci.com, current@freebsd.org Subject: Re: Building inside of /usr/src? In-Reply-To: <7890.835914838@time.cdrom.com> References: <7890.835914838@time.cdrom.com> Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk > > I don't know enough about the internals and all of the places that this is > > used to be able to say if it'd be safe to revert the behavior in 'make' > > itself, but maybe there could be a function added that just returns the > > "real" getcwd() (or just grab the output from a /bin/pwd run). > > It did do this, the problem is that it then overlayed a perfectly good > "trusted" value for the current directory and then spammed over it with > the value of $PWD. Which according to the code we saw should also be a 'trusted' value. I think changing the stat call to an lstat (as Poul suggested) would give the best of both worlds. In it's current incarnation (w/out the $PWD stuff), I suspect that if you used symlinks in your source tree you'd break any chance of getting a working build because your relative paths would be screwed up. # ls -l /usr/src/bin lrwxr-xr-x 1 root wheel 11 Jun 25 23:27 /gnu@ -> gnu-stable/ So anything that used files in gnu and hoped that ../../gnu/lib/libgnumalloc.a would work will now break. Anything that uses .PATH is subject to break now. Nate