From owner-freebsd-ports@FreeBSD.ORG Sat Apr 23 22:18:44 2005 Return-Path: Delivered-To: freebsd-ports@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id 37C3C16A4D0; Sat, 23 Apr 2005 22:18:44 +0000 (GMT) Received: from smtpq2.home.nl (smtpq2.home.nl [213.51.128.197]) by mx1.FreeBSD.org (Postfix) with ESMTP id E1B5F43D46; Sat, 23 Apr 2005 22:18:42 +0000 (GMT) (envelope-from danny@ricin.com) Received: from [213.51.128.136] (port=43347 helo=smtp5.home.nl) by smtpq2.home.nl with esmtp (Exim 4.30) id 1DPSxV-0004Xw-JB; Sun, 24 Apr 2005 00:18:41 +0200 Received: from cp464173-a.dbsch1.nb.home.nl ([84.27.215.228]:58469 helo=desktop.homenet) by smtp5.home.nl with esmtp (Exim 4.30) id 1DPSxU-0005Vz-GA; Sun, 24 Apr 2005 00:18:40 +0200 From: Danny Pansters To: freebsd-questions@freebsd.org Date: Sun, 24 Apr 2005 00:18:07 +0200 User-Agent: KMail/1.8 References: <20050420155742.GA83965@thought.org> <20050423212902.29985d3c.flynn@energyhq.es.eu.org> In-Reply-To: <20050423212902.29985d3c.flynn@energyhq.es.eu.org> X-Face: "0Qv=,p:+]LvuqrtS4U\z3k"qN=.1]@=?utf-8?q?=258=3F=3BPoab=23v=27F=7E=0A=09!Wm=5Fe-=24=7EL=5D=3B?=>[c*L^Qoladj)x@mH}Bqz"vLO?Zdl}[@V@=?utf-8?q?U=3Fx3=23lI=3A=0A=09=24DN=7E!Hr?=@K`-mNv"zXm MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-6" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200504240018.08315.danny@ricin.com> X-AtHome-MailScanner-Information: Please contact support@home.nl for more information X-AtHome-MailScanner: Found to be clean cc: freebsd-ports@freebsd.org Subject: Re: mplayer/mplayer-plugin question X-BeenThere: freebsd-ports@freebsd.org X-Mailman-Version: 2.1.1 Precedence: list Reply-To: freebsd-questions@freebsd.org List-Id: Porting software to FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 23 Apr 2005 22:18:44 -0000 On Saturday 23 April 2005 21:29, Miguel Mendez wrote: > On Sat, 23 Apr 2005 14:14:56 -0500 > > "Jeremy Messenger" wrote: > > >> Personally, I'd love to see /usr/X11R6 folded into /usr/local, but > > >> until then, I think it's nothing short of retarded for apps to install > > >> into unusual locations to prove a point. > > > > > > It might be interesting looking at the work the pkgsrc people have done > > > wrt $PREFIX enforcement. On my NetBSD boxen xorg lives under > > > /usr/pkg/xorg and all packages are installed under /usr/pkg, not > > > /usr/X11R6 or /usr/local. > > > > I disagree with NetBSD's PREFIX. I would go with the global prefix, > > /usr/local. 85% of configure has the /usr/local by default and FreeBSD > > already has /usr/local (to kill the colour discussions), so all we have > > to do is remove /usr/X11R6. Before you ask how we can test if the port is > > respect the prefix, we should be able to find out very easy when you work > > with pkg-plist by follow the porter handbook. > > I wasn't advocating the use of /usr/pkg, but rather the way they > enforce it for every package. I personally don't mind what the prefix > is called, but for the sake of POLA /usr/local would certainly serve > FreeBSD better. FWIW, well behaved software should always be $PREFIX > clean. Good discussion. Wanted to add some thoughts... [ I stopped CC'ing everyone, for those who read the lists anyway and don't want to get 3 or 4 copies, but I don't mind if anyone CCs me again in a response ] At netbsd they have an "xwedge" package that basically maps any /usr/X11R6 to /usr/pkg come install time. We could easily have something likewise (or borrow it from net) but it still has the problem: what are you going to do with users who already have the /usr/X11R6 "bonus" tree. Also (minor?) the xorg distribution should install into PREFIX also then, of course. xwedge seems to be great if it's the first thing you install/setup, I don't know how/if it can cope if installed after one already has 200 packages installed. If it copes with that, borrowing it as a starting point would make sense. Any port that's PREFIX clean should be no problem if a similar "xwedge" scheme is used. Then eventually it could be dropped after everything caught up with there being only one prefix. Still, I wonder if just ruthlessly making the X target a hard link to the local target (and maybe later fase out X11BASE in ports) wouldn't be the best way to go about this. Would be completely POLA agnostic at first _and_ at last, for those cases that will/can not be stomped into conforming to LOCALBASE you could always retain a simple hard link. That's one inode pointing to one other. It won't saturate our disks ;-) The only problem I can think of is maybe there will be name clashes somewhere. But it may very well be the case that things go a lot deeper, and there's no easy solutions. In that case, well, we can already live with it now... Of course that is if people _WANT_ to do away with the X prefix, up until now my impression was that most folks didn't mind it or even thought it was better. I never really cared very much about it apart from aestetic [sp?] reasons, but I can certainly imagine it being confusing to some (new) users. One other thing, which doesn't concern end users is that it can make things easier for porters. If you have a port that needs to put something into LOCALBASE and something into X11BASE you're always going to have an interesting plist and more error prone littering in your Makefile. Of course you can always cope, but simpler is better. OK, enough babble :) Regards, Dan