I switched to using a testing version of the plugin (which BTW is only resolvable ever in the repo created during the integration testing phase) so the ITs themselves wouldn't have to:
a. adjust to new plugin versions in order to run b. rely on metadata resolution for the plugin version, which can lead to using an incorrect version of the plugin This way, it is absolutely invariant which plugin version they will test. The "testing" version doesn't exist anywhere else, and the modded POM that is used to build the testing version is marked as deleteOnExit(). I guess I don't understand why everyone's nervous about this? -john On 12/20/06, Jason van Zyl <[EMAIL PROTECTED]> wrote:
On 20 Dec 06, at 6:32 AM 20 Dec 06, Fabrizio Giustina wrote: > The vote has ended, with 5 positive votes from PMC members. The maven > eclipse plugin will be released with version 2.3. > > Jason, I will leave the actual release to you as discussed, you can > now proceed. > Great I can take care of this. > > (about the way the plugin is tested: I totally agree that installing a > test version into the local repository is something that should be > absolutely avoided. Till recently plugin tests used the embedder, but > if I'm not wrong such tests have been refactored by John using this > "new way", probably to overcome the limitation that the embedder was > not able to run multiproject builds. The embedder runs multiple projects. At any rate, I've briefly talked with John and we will align everything to one set of tools and improve them. > This is not a problem for users > but I'll try to refactor again tests after this release.) > > fabrizio > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
