From owner-freebsd-ports@FreeBSD.ORG Tue Apr 12 09:41:27 2011 Return-Path: Delivered-To: ports@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id DE9C4106566C for ; Tue, 12 Apr 2011 09:41:27 +0000 (UTC) (envelope-from aehlig@linta.de) Received: from linta.de (isilmar-3.linta.de [188.40.101.200]) by mx1.freebsd.org (Postfix) with ESMTP id 541CD8FC1B for ; Tue, 12 Apr 2011 09:41:26 +0000 (UTC) Received: (qmail 32765 invoked by uid 10); 12 Apr 2011 09:41:25 -0000 Received: from curry.linta.de by isilmar.linta.de with BSMTP; 12 Apr 2011 09:41:25 -0000 Received: by curry.linta.de (Postfix, from userid 1001) id 341771CC1D; Tue, 12 Apr 2011 10:41:09 +0100 (BST) Date: Tue, 12 Apr 2011 10:41:09 +0100 From: "Klaus T. Aehlig" To: Eitan Adler Message-ID: <20110412094109.GC92254@curry.linta.de> References: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.21 (2010-09-15) Cc: FreeBSD Ports Subject: Re: Removing Cruft from the ports tree X-BeenThere: freebsd-ports@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting software to FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 12 Apr 2011 09:41:27 -0000 Hi, I'm disappointed that there are plans to remove the ports framework. I always considered this to be one of the very strong points of the ports infra structure, precisely because I believe... > Not only that but because maintainers would be > able to choose the best possible configuration for the their port > users would no longer have to mess around. ...there is no "best configuration" for every port. It depends a lot depends on what you use your machine for. On my desktop I care for full features, on my laptop for little dependencies, servers WITHOUT_X11, ... Or, what I've been recently doing in my day job. There probably is no "best" implementation of the message passing interface; that's why there are different MPI Implementations in the ports tree (e.g., openmpi, mpich2, ...). Now there are a lot of scientific libraries (mainly in the math and science category) that also support MPI. To benefit from this, I need them to build against the same mpi implementation (because I use several of them in the same executeable). There I found it very convenient, that the WITH_MPI and WITH_OPENMPI options from /etc/make.conf were honored, and that I could easily switch between MPI implementations just by changing a few configurations and rebuilding. Maybe I misunderstood your sugesstion. I'm not saying we need the blue dialog boxes. I'm pefectly happy with configuring ports by adding things like .if !empty(.CURDIR:M*/ports/foocategory/barport*) WITH_BAZ=YES BAZ_OTIONS=... .endif in /etc/make.conf or any other means of chosing what I need. What I'm saying is, that we're losing a lot, if we give up the possibility of ports to install different sets of files and register different sets of dependencies depending user configuration; just diffent configure arguments are not enough. I know, it is a lot of work to support the different builds for a single port, but I think it's worth is; especially, as it is probably not less work (but more confusing for the user) to multiply each port for the different options (e.g., the mpi implementation used -- or maybe without mpi for serial computations). Best regards, Klaus