<Swing Dev> Is the following code supposed to leak?

[Pavel Porvatov]

Hi Roman,
> Hi all,
>
>>> The vm doesn't terminate.
>>>
>>> One VM to rule them all: the instances are separated via the
>>> appcontext, but there is just one server vm running.
>> I'm not sure that's correct behavior. According to javadoc of
>> java.lang.System#exit: "Terminates the currently running Java Virtual
>> Machine.".
> We are running the Notepad in a slightly different context: we run the
> app on a server machine and send all the rendering (Java2D, etc) to the
> browser which renders on an HTML5 canvas. Example:
>
> http://icedrobot.de:9091/SessionInitializer?cls=SwingSet2
>
> In such a (very unusual) setting, we don't terminate the VM. User
> sessions are separate by AppContext, and we run many instances of the
> app inside one VM. We never let the user terminate the VM (since that
> would kill the VM for all users). We know that this is not the kind of
> usage that AWT was designed for, but we like to push Java to the borders
> and boldly go where no AWT has gone before ;-)
So if your SecurityManager denies app closing, then the application 
should be alive. It looks that your implementation contradicts javadoc 
contracts.
> Besides, I find it a bit weird that addXZYListener() secretly registers
> listener to some sort of internal global object. That sounds very leaky.
> Is it documented that such listeners must always be carefully removed?
> Or am I missing something?
As I said normally listeners shouldn't be removed manually, until this 
listeners are added to static instances (or delegates). The 
JTextComponent#getActions method invoked on non-static object, but it 
uses inside static editor kit. So listener removing depends on 
implementation and programmers should take such nuances into account.

Regards, Pavel