I've just finished correcting the pom inconsistencies. The conclusion of this work is that ALL samples had wrong parent poms specified... Modules were correctly specified so when building from a top-level folder (i.e. samples/) there was no problem. However, when trying to build a single sample with a local maven repo that didn't have the required poms (i.e. apache-7.pom), the paths and artifactIds to the parent poms were incorrect causing MissingArtifactException.
I've seen in various places tweaks like <relativePath>../../pom.xml</relativePath> even tough it had a direct parent only one level up. I think these point out that the fine granularity of the poms makes it too hard for us to maintain. If we run into the same problem in the future I suggest we reconsider and move to either a single pom in samples/ or poms in each direct child of samples (getting-started, learning-more, ...) and samples. On Fri, Dec 10, 2010 at 2:03 AM, ant elder <antel...@apache.org> wrote: > On Thu, Dec 9, 2010 at 4:08 PM, Florian MOGA <moga....@gmail.com> wrote: > > On Thu, Dec 9, 2010 at 5:22 PM, ant elder <antel...@apache.org> wrote: > >> > >> A regular hudson build of the samples with an empty repo sounds > >> useful, though i worry it might take quite a long time to run. I'll > >> set up a test one to see how long it does take. > >> > > > > I agree, it mustn't be the Hudson build, but any mechanism that prevents > > from getting tricked by the local repo containing the artifacts already. > I'm > > thinking we can setup this kind of build and run it just before doing a > > release, it's not necessary on a daily basis. > > > >> > >> I have just tried building all the samples from trunk with an empty > >> repo and it worked ok, how are you building them, from an svn checkout > >> or from one of the src/bin distribution snapshots? > >> > > > > svn checkout... take a look at the comet binding sample, it specifies the > > parent as tuscany-samples with relativePath ../pom.xml. The artifact > found > > one directory level upper is tuscany-sample-binding-comet. Similar things > > happened with the introduction of the learning-more, running-tuscany etc. > > directories. So, it has to fail. > > Does it worth having poms in each and every directory? Wouldn't it be > much > > easier to maintain a single pom at the samples/ directory level to which > all > > the samples point? > >> > > It does sound simpler to have the single parent. I vaguely remember we > talked about doing that before when talking about the samples, guess > we just haven't gotten to it yet. > > ...ant >