I always build my features by hand. It's just easier, IMHO. A few errors on the console telling you what's missing, and you're off to the races. You can also try using PAX exam to test your features, but I've found that to be more trouble than it's worth (it always chooses to fail when I'm trying to do something important, like cut a release).
On Sun, Feb 12, 2017 at 8:31 AM Thomas Vandahl <[email protected]> wrote: > Hi Jean-Baptiste, > > On 05.02.17 16:36, Thomas Vandahl wrote: > > Reading this makes me think that the list of Maven dependencies > > (including transitive ones) should be sufficient to resolve everything. > > Is this a valid assumption to make? The karaf-maven-plugin gives me for > > example > > > > <bundle>mvn:org.apache.camel/camel-blueprint/2.18.1</bundle> > > > > Is this identical to > > > > <feature version="2.18.1">camel-blueprint</feature> > > > > and if not, how do I get the plugin to generate the latter from a > > dependency entry in the POM? Is this possible at all or do I need to > > manage the feature dependencies manually? > > Ok, I got my feature artifact to verify successfully. Upgrading to Karaf > 4.0.8 helped a lot. The error messages are much more helpful. > > However, it was *not* possible to get the feature to verify without > adding feature dependencies to the feature.xml-template such as > aries-blueprint, aries-proxy, eventadmin and spring - even though most > of them were listed as wrapped bundles in the dependency tree. Moreover, > I had to add the standard-feature-descriptor and the > spring-feature-descriptor as "runtime"-dependencies to the POM. > > One last thing I learned was that a bundle dependency of > camel-ftp/2.18.1 is obviously not identical to a feature dependency of > camel-ftp. In the first case, the Karaf Maven plugin complains about a > missing dependency to com.jcraft.jsch.*, in the second case, it doesn't. > > I now have a verified feature file but when I try to install it in a > clean Karaf 4.0.8 instance, all cores go up to 100% and nothing else > happens. No entries in the log file, no network activity, just CPU. I > gave up after a little bit over an hour. > > Could you please explain how this is supposed to work and what debug > approach would be the best? I can provide my code examples if you want, > I'm just a bit reluctant to post them here. > > Bye, Thomas. > >
