On 9/27/07, Manny Calavera <[EMAIL PROTECTED]> wrote: > On Mon, Sep 24, 2007 at 03:52:21PM -0500, Aaron Griffin wrote: > > On 9/24/07, RedShift <[EMAIL PROTECTED]> wrote: > > > > > > > > > Aaron Griffin wrote: > > > > Hey all, > > > > * 'Continuous Integration' setup / machine > > > > > > > > This is a pipedream of mine. I'd like to setup a CI server somewhere. > > > > For those unfamiliar with the term, what this means is that the server > > > > will, periodically, check out our PKGBUILDs and try to build ever > > > > single package. Clean chroots don't matter, what we're checking is > > > > whether it will build as the instructions say. > > > > I guess a pacbuild instance COULD be used here, but this should be a > > > > simple bash script. > > > > If anyone wants to do something fun, feel free to jump on this one, > > > > but otherwise it's not important. > > > > > > > > > > I have an x86_64 server at home that's regularly doing nothing, I can > > > supply such a service if the necessary software already exists. > > > > Yeah I have a machine that I wanted to install x86_64 on too, so I > > could have a build machine for that arch. > > > > I guess "the software" could be pretty simple due to the fact that we > > don't care about real output all that much (right now, we could always > > combine this with pacbuild at a later date), but for right now we > > would simply be making sure packages build with relation to other > > packages. > > > > So it'd be a cvs update, list dirs changed in the update, build > > packages, bamf. Technically we'd want to use local repos for packages > > and have the PKGDEST point there, with some db rebuilding after each > > success. > > > > All we're looking at, really, is success/fail and the output (makepkg -L). > > > > I don't have time at the moment, but if someone wants to write a > > 100-200 line bash script for this, that'd be awesome. > > > > Thanks, > > Aaron > > What will you do when there's some sort of conflicting package? whats > the point of installing something but not see if it actually works? Will > you run each program specifically or just have it install correctly? > There's a difference. Many things will install errorless but then refuse > to run after a ton of tweaking.
This is _not_, I repeat, _NOT_ a substitute for testing. You are talking about testing. I am, as the title of the email suggests, talking about integration. http://martinfowler.com/articles/continuousIntegration.html Continuous Integration is a real term, not something I made up _______________________________________________ arch mailing list [email protected] http://archlinux.org/mailman/listinfo/arch
