It was the lein-immutant plugin killing off my peer as a workaround for C-c'ing on windows.
With 107, hornetq load balancing is strange; certain messages are only consumed on one node, all the time. Worse, killing a master node kills its peers.
… [IMMUTANT-147] We won't know for sure if this fixes the issue until we have quite a few successful CI runs.
We have to do this to make the release build happy, since the ns jar isn't (yet) a real maven artifact, causing the release build to complain about a snapshot dependency.
That way we don't have to keep the docs up-to-date in two places with differing release schedules.
Also clean up redundant tests.
If created, the calls return true. If they already existed and need not be created, false is returned.
The pojo subsystem apparently tries to deploy beans it finds in any jboss-beans.xml file it finds on the classpath. The netty jar has one, of course, but it references a bean that's already installed in the AS, because it depends on netty, too. A side benefit of removing the pojo subsystem is the disappearance of this annoying log message: "Found legacy bean/pojo namespace: urn:jboss:bean-deployer:2.0 - might be missing some xml features (potential exceptions)." It remains to be seen whether Immutant users will rely on this now-missing feature. I kinda doubt it. :)
This fix has the happy side effect of allowing the dev ns to be used in clojure 1.3. It also includes some test re-org in core.