Minor Bugs Fixes

[prosgarz35]

Have a look

[chibenwa]

registerMBean(): hides a real error instead of fixing it

If the MBean name (org.apache.james:type=server,name=) is already registered, the PR unregisters the existing MBean and registers its own in its place. The problems:

• It hides misconfiguration. In production, a collision means two servers share the same jmxName, for example two smtpserver entries in smtpserver.xml. Today startup fails with a clear error. With this PR, the second server silently replaces the first in JMX, and the first one can no longer be managed.
• It breaks shutdown. The first server's destroy() calls unregisterMBean(), which removes the second server's MBean. The second server's destroy() then hits InstanceNotFoundException, which gets wrapped in a RuntimeException inside @PreDestroy.
• synchronized (mbeanServer) protects nothing. It's the JVM-wide PlatformMBeanServer. Nothing else synchronizes on it (not the jmx/*Stats classes, not other components), so the isRegistered → unregister → register race is still there.
• The catch (InstanceAlreadyExistsException) just repeats the same steps. It runs the same unregister/register sequence again, and it uses a fully qualified class name instead of an import.

Could you describe the issue you hit ?We'd rather fail fast with a clearer error message, or fix the lifecycle that leaks the registration.

[chibenwa]

It does nothing new.

super.configureBootstrap(bootstrap) already sets it in protocols/netty/src/main/java/org/apache/james/protocols/netty/AbstractAsyncServer.java:182

The added line sets it a second time, and the new comment wrongly suggests it wasn't set before.

[prosgarz35]

Yeah looks like i made mistake