If plugin A declares <depends plugin="B" useClasses="false"> and plugin B does not exist then plugin A's deployment fails. This should be made more lenient as the deployment of A should still be valid and useful. It may still be able to run independently and may still inject successfully into other, existing plugins.
So, from greg, one use case for this feature would be (4:37:50 PM) ghinkle: i want to be able to inject my plugin into other people's plugins even if i'm not sure the user will have all of them (4:41:03 PM) ccrouch: but will you really get inject just through the <depends> clause (4:41:04 PM) ccrouch: ? (4:41:47 PM) ghinkle: not just through it (4:41:55 PM) ghinkle: but its needed if you want to inject (4:42:14 PM) ccrouch: ok, necessary but not sufficient (4:43:02 PM) ghinkle: e.g. the hibernate plugin could written to work in both jbossas and tomcat (4:43:08 PM) ghinkle: (and probably should at some point)
This has been at least partially implemented. A final solution may be more robust but rev3810 implemented leniency towards missing plugin dependencies and rev3816 prevented invalid deploymentService execution.
No more work on this defined at this time, so we'll resolve and create new jiras if and when we have more improvement work for plugin dependency
A hot-plug test with the Hibernate and JBossAS plugin. The Hibernate plugin registration failed first, but was successful after a while when the JBossAS plugin was registered. rev3819
This bug was previously known as http://jira.rhq-project.org/browse/RHQ-1935 This bug is related to RHQ-571 This bug is related to RHQ-1205