On Sat, Apr 28, 2012 at 12:12 PM, Hilco Wijbenga <hilco.wijbe...@gmail.com>wrote:
> On 27 April 2012 18:20, Chris Graham <chrisgw...@gmail.com> wrote: > > On Sat, Apr 28, 2012 at 11:06 AM, Hilco Wijbenga > > <hilco.wijbe...@gmail.com>wrote: > > > >> On 27 April 2012 17:51, Chris Graham <chrisgw...@gmail.com> wrote: > >> > On Sat, Apr 28, 2012 at 7:04 AM, Hilco Wijbenga < > >> hilco.wijbe...@gmail.com>wrote: > >> > > >> >> > >> >> Actually, I've been working on a Maven extension that uses checksums > >> >> to determine whether a particular project needs to be rebuilt (taking > >> >> all its dependencies into account). We are currently using a Bash > >> >> script for that purpose. It simply invokes Maven for each project > that > >> >> needs to be (re)built but, obviously, doing this from Maven directly > >> >> is much easier, faster, and more reliable. > >> >> > >> >> > >> > Use Hudson/Jenkins for this, that's what I use. > >> > >> This is for local development. The build server isn't in the picture > >> yet. It would not be smart to have the build server skip parts of the > >> build anyway. > > > > No, that's not true. If the project build resolution is deterministic, ie > > it will always result in the same build result. This is good or standard > > SCM practice. > > What isn't true? > > You are not seriously suggesting we now do all local development via > the build server, are you? That would be insanely inefficient. Not to > mention the total chaos that would ensue. > The opposite. If you trust what you're doing locally, it should be used on the CI server. The point is simple: If you don't trust it, don't use it. > Local development is *local*, on your local machine, it hasn't reached > anything or anyone else yet. I don't think we are talking about the > same thing. :-) Or at least, I hope so. ;-) > > > Jenkins fingerprinting (not that I've used that one much) or M2 job type > > (that I have almost always used) dependency analysis does exactly this. > > In my experience, Jenkins' dependency analysis only works if nothing > (POM wise) has changed. As soon as you add/delete/move dependencies, > it gets confused and builds things in the wrong order resulting in > failed builds. Given that it doesn't know about all changes at the > same time, this is not entirely unexpected, I suppose. But now we're > veering off topic. :-) > I've never seen this behaviour. > > However, by definition, a release build, would (and should) result in a > > full build. > > > > -Chris > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@maven.apache.org > For additional commands, e-mail: dev-h...@maven.apache.org > >
