When WispService.stop is called, its internal ActorPools are stopped. This causes a problem for restarting the service:

Fantom Shell v1.0.54 ('?' for help)
fansh> wisp::WispService { port = 8080 }.start
fan.wisp.WispService@1309e87
fansh> [18:33:07 05-Oct-10] [info] [web] WispService started on port 8080

fansh> Service.find(wisp::WispService#).stop
fan.wisp.WispService@1309e87
fansh> Service.find(wisp::WispService#).start
sys::Err: ActorPool is stopped
	at fan.sys.Err.make(Err.java:78)
	at fan.sys.Err.make(Err.java:75)
	at fan.concurrent.Actor._send(Actor.java:140)
	at fan.concurrent.Actor.send(Actor.java:91)
	at fan.wisp.WispService.onStart(WispService.fan:39)
	at fan.sys.Service$.start(Service$.java:205)
	at fan.wisp.WispService.start(WispService.fan)
	at fan.sh2.FanshEval._eval(fansh:5)
      ...
fan.wisp.WispService@1309e87

Uninstalling WispService has no effect, as the old service instance does not release the port:

Fantom Shell v1.0.54 ('?' for help)
fansh> wisp::WispService { port = 8080 }.start
fan.wisp.WispService@19ce060
fansh> [18:38:58 05-Oct-10] [info] [web] WispService started on port 8080

fansh> Service.find(wisp::WispService#).uninstall
fan.wisp.WispService@19ce060
fansh> wisp::WispService { port = 8080 }.start
fan.wisp.WispService@5a67c9
fansh> [18:39:17 05-Oct-10] [err] [web] WispService cannot bind to port 8080
  sys::IOErr: java.net.BindException: Address already in use
    java.net.PlainSocketImpl.socketBind (PlainSocketImpl.java)
    java.net.PlainSocketImpl.bind (PlainSocketImpl.java:365)
    java.net.ServerSocket.bind (ServerSocket.java:319)
    inet::TcpListenerPeer.bind (TcpListenerPeer.java:83)
    inet::TcpListener.bind (TcpListener.fan)
    inet::TcpListener.bind (TcpListener.fan)
    wisp::WispService.listen (WispService.fan:59)
    wisp::WispService.onStart (WispService.fan:39)
    fan.sys.Func$Indirect0.call (Func.java:121)
    concurrent::Actor.receive (Actor.java:99)
    concurrent::Actor._dispatch (Actor.java:225)
    concurrent::Actor._work (Actor.java:196)
    concurrent::ThreadPool$Worker.run (ThreadPool.java:255)

Ideally WispService should be able to resume after a stop / start cycle. If this is not possible (which seems likely due to the const requirement on sys::Service and the fact that an concurrent::ActorPool cannot be restarted once stopped) then the documentation should state this limitation.

In any case, the fact that WispService does not unbind from the port when stop is invoked is a bug and should be fixed. Looks like wisp::WispService.listen is missing a call to TcpListener.close.

Not too good.

I pushed a changeset which does a couple of things:

• releases the port on shutdown and logs a messages
• makes shutdown more robust in case root WebMod raises an exception
• cleared up exception/documentation about reuse after stop

Should this be promoted to a ticket for tracking purposes?

Should this be promoted to a ticket for tracking purposes?

Yeah good question. I think once we have a real release we probably want to track everything formally.

The loose process I am using now is basically this: if I am not going to fix immediately, I turn it into a ticket just to track for myself. If I fix it immediately then I don't promote it into a ticket, although I do note the topic number in the changeset and in the changelog.