Date: Sun, 26 Sep 2010 23:47:14 -0700 From: perryh@pluto.rain.com To: talon@lpthe.jussieu.fr, m.seaman@infracaninophile.co.uk Cc: questions@freebsd.org Subject: Re: Free BSD 8.1 Message-ID: <4ca03df2.lQjjNnRah4BJhw4Y%perryh@pluto.rain.com> In-Reply-To: <4C9F3BBA.2060809@infracaninophile.co.uk> References: <20100926123019.GA41450@lpthe.jussieu.fr> <4C9F3BBA.2060809@infracaninophile.co.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
Matthew Seaman <m.seaman@infracaninophile.co.uk> wrote: > On 26/09/2010 13:30:19, Michel Talon wrote: > > Matthew Seaman said > >> Be aware that installing the ports tree from the DVD images > >> is not the ideal way to do it ... it is better to ... grab > >> an up-to-date copy of the ports directly from the net. > > > > I disagree with that ... Another option is to install > > the ports tree from the DVD,and install corresponding > > precompiled packages ... and *not* updating the ports > > tree ... I suspect the best results can be had from an approach in between these; details below. > ... being up-to-date with the ports tree generally *does* > give you better results than not. > Ports are a moving target, dependent entirely on upstream changes. This last is an oversimplification. Not all ports even _have_ an upstream, and those that do (granted, the great majority) depend not only on upstream changes but also on the maintainer's and committers' ability to keep up with those changes. > Expecting that a snapshot taken months or weeks ago will work > just as well as one updated in the last hour is plain daft ... > ported software generally does improve over time. Updates that > fix problems are way more common that updates that introduce them > ... Couldn't this as well be said of FreeBSD itself? If it were universally accepted, there would be no need for the stable or security branches and the considerable effort that goes into maintaining them: everyone would just run -CURRENT. One _huge_ advantage of starting with a release _and its corresponding set of ports & packages_ is that everything is self-consistent. This tends not to be true of snapshots taken between releases, if only because no one has time to do that much release engineering for every update of every port. I tried to follow the OP's approach a few years ago, and got burned rather badly. By the time I had the system working well enough to start on the project I had intended to work on, the time budgeted for the setup _and_ the work had been almost entirely consumed in setup! I get the impression that M. Talon may have had similar experiences. I've recently started on a new system, and am planning to install 8.1-RELEASE, including the corresponding ports tree; then install what ports I can from packages and also fetch the corresponding distfiles; and finally build -- from release-corresponding ports -- any that aren't available as packages or where I want non-default OPTION settings. That approach should avoid most nasty surprises while getting things set up and working. _After_ everything is installed and configured properly will be plenty soon enough to consider whether any ports need to be updated -- and the already- installed-and-working package collection will provide a fallback in case of trouble trying to build any updated versions.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4ca03df2.lQjjNnRah4BJhw4Y%perryh>