Date: Tue, 9 Aug 2005 20:03:59 +0200 From: Thierry Herbelot <thierry@herbelot.com> To: freebsd-current@freebsd.org Cc: freebsd-ports@freebsd.org, Colin Percival <cperciva@freebsd.org> Subject: Re: Portsnap is now in the base system Message-ID: <200508092004.09405.thierry@herbelot.com> In-Reply-To: <42F8E5B0.9000709@freebsd.org> References: <42F871B4.6000703@freebsd.org> <200508091902.57044.thierry@herbelot.com> <42F8E5B0.9000709@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Le Tuesday 9 August 2005 19:19, Colin Percival a écrit : > > > > I've had a look at the man page(s), and at the web page on > > http://www.daemonology.net/portsnap/ I'm still missing one piece of > > information : how are the ports snapshots initially built ? > > Magic. :-) I *knew* it. > > Seriously, I checkout a copy of the ports tree, run `make describes` three > times (for 4.x, 5.x, and 6.x), package up the resulting files, build some > patches, and then throw everything onto my web server. From there it gets > mirrored by another server (and more mirrors will follow). > > Once I've ironed out all the bugs in the building and mirroring, I'll make > that code available via the projects repository. fine > > > One misfeature of cvs is the possibility to fetch incomplete updates to > > the repository (no atomic commits in cvs, and this must be carried over > > to cvsup). Do "your" snapshots behave better in this domain ? > > No. I take a snapshot of the files on cvsup-master (via cvsup over an ssh > tunnel -- being a committer hath its privileges :-) ), and that's what I > package up, with the one exception that if the INDEX build is broken then > users will get the most recent unbroken INDEX instead. > > > As one last question : I assume the same process of building "coherent" > > snapshots could be also applied to the core cvs repository of the full > > FreeBSD project, and a cvs-snap utility could be imagined ? > > This process doesn't build coherent snapshots; and the same process > wouldn't work very well for anything other than ports due to a number of > reasons concerning the structure of the ports tree and the lack of > structure of the non-ports trees. too bad : as you demonstrated, the project makes good use of "assured" methods of disseminating new software - these are two very important stepping stones and it's just a pity not to have the same kind of tool for the core of the OS = the source (and it's too late for a SoC project ...) > > Colin Percival TfH
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200508092004.09405.thierry>