On 03/12/2010 05:45, Mladen Turk wrote: > We would need that back or even better axe running and paused > and creating an "state" enum. > > For example the connector will report as running on start() > even if missconfigured cause it is set before Acceptor or Pollers > are created. > > So internally both for paused and running we actually have > pauseStarted -> paused, runningStarted -> running=true > For stop() we have > pausePending -> paused, runningPending -> running=false > > Currently it is possible to start() while stop() is still running > which could only happen if stop() actually ended. > > Think that single eg. EndpointState enum will cover all the > states during start(),pause(),resume(),stop() replacing multiple > booleans (at least running and paused) with a single enum. > > > WDTYT?
What we really need is the Lifecycle stuff that is in Catalina. We can't use that directly since that would create a dependency from the connectors to Catalina. We could refactor the Lifecycle stuff out of Catalina. Put the interfaces in o.a.tomcat and the implementation (i.e. the base classes) in o.a.tomcat.lifecycle. I'm not completely comfortable doing that but I like duplicating the code even less. I'm leaning towards making that change. Regardless of the above, I think the answer to bug 50360 is a bindOnInit property. If true, socket is bound on init() and released on destroy() If false, socket is bound on start() and released on stop() Question is, should the default be true (as current, works for jsvc, issues when embedding) or false (change from current, breaks jsvc, works when embedding). I'm leaning towards a default of true. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
