[jboss-dev-forums] [JBoss Microcontainer Development] - Re: Testing Deployers with new Reflect + ClassPool

It will always be a mistery why the tests passed before (but I swear they did ;) Yes, the classpools are never cleaned up, and this is generating the failures. This is a basic thing that I completely missed when doing the ClassPools. When I tested those against Branch5_x and JBoss AOP, they worked, but the point is that JBoss AOP has the mechanisms to clean the ClassPools. So, this is how JBoss AOP regiesters and unregisters ClassLoaders: 1. The AspectDeployer registers the ClassLoader when an aspect is deployed. 2. The AspectManager translator is is attached to the ClassLoaderSystem, which invokes the unregisterClassLoader everytime a unit is undeployed. On our side of the story, the ClassLoader is registered on an as-needed basis, by RepositoryClassPoolFactory.getPoolForLoader(ClassLoader). I dunno it would be a good idea to create a translator just for the sake of unregistering classloaders. What would be a better approach? Maybe a LifecycleCallback? View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4260714#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...