vertx.eventBus.send() always registers a reply handler

[Narigo]

IIRC this is a leak, if you don't send a reply back from the server side. It would register a handler but that would never be executed.

I guess the default value should just be null and it be disregarded by Vert.x in the Java core. Not sure whether this is the best approach, though. Probably better use an Option and then call the according Java API function.

[galderz]

Yeah, Option with a default value of None sounds good to me.

The Java API uses null as a way to figure out whether the reply handler needs to be registered or not. An Option would allow us to transform the optional reply handler from something that's Scala friendly to Java API.

On the Java side, EventBus.send() javadoc should be clearer on what values are allowed for replyHandler :|

[ddossot]

While running load tests, throwing 500K string messages on the bus, I run into a memory leaks related to using vertx.eventBus.send() with vert.x 2.0.1-final and mod lang Scala 0.1.1-CR2. Note I was running in a 3 nodes cluster and the single message handler was registered locally in each node.

Basically, with either a no-op reply handler { ignoredReply => {} } or (null), I'm facing a memory leak while if I use vertx.internal.eventBus.send instead, no such leak shows up.

[Narigo]

Can you test this again with #63 applied, please?

[Narigo]

Please reopen / open a new issue if you still find leaks with registering (or unregistering) handlers.