JBMESSAGING-432 - Connection Leaking on Factory (1 connectio

[clebert.suconic]

In regard to http://jira.jboss.com/jira/browse/JBMESSAGING-432, is this in regard to the Client created inside ClientConnectionFactoryDelegate::invoke?

Also, what was the intent of having a separate client here. I mean, why we didn't use a single client for this here?


Clebert

[clebert.suconic]

I'm removing the comment from ClientConnectionFactoryDelegate to fix this.

 if (remotingConnection == null)
 {
 //Not a call to createConnectionDelegate - disconnect the client

 client.disconnect();
 }

I"m not sure why we are using a separate client here thought. (maybe because it's hard to access properties from the client in a different layer). It would take a lot of work to change this. (to not have to open a new client here).

I will just disconnect the client, since it's created every time. (So, the leak it's a little bigger. (It's still a minor issue though).

[ovidiu.feodorov]

Not disconnecting the client there was done on purpose, read the initial comment on http://jira.jboss.com/jira/browse/JBMESSAGING-432

[clebert.suconic]

Sorry, I forgot that Remoting/Locator would have a cache for this socket.

Client will be GCED byt the underlying socket will sit somewhere in a static HashMap inside remoting. (the Locator property will be used for that).

A proper fix would be to have client created before, and always reuse the same client.

But that would change internal API. I don't think we should do it now.

I guess we should mark this as Won't fix.