This has brought to light some interesting effects (aka bugs) both in
the general implementation as well as in previous fixes.
SupervisorHierarchySpec is without TODOs now and GREEN.
- always suspend/resume for Suspend/Resume/Recreate, no matter which
state the actor is in, to keep the counter balanced
- preRestart failures are logged but otherwise ignored; there’s nothing
else (apart from terminating the actor) which we could do at that
point
- preRestart/postRestart exceptions have their own distinguishable
subtype of ActorKilledException now
- fix some race conditions in tests to make them produce fewer false
failures
- remove cruft from SupervisorStrategy and add methods which can
actually be used to implement your own (with proper warning signs)
- introducing RepointableActorRef, which starts out with an
UnstartedActorCell which can cheaply be created; the Supervise()
message will trigger child.activate() in the supervisor, which means
that the actual creation (now with normal ActorCell) happens exactly
in the right place and with the right semantics. Messages which were
enqueued to the dummy cell are transferred atomically into the
ActorCell (using normal .tell()), so message sends keep working
exactly as they used to
- this enables getting rid of the brittle synchronization around
RoutedActorRef by replacing that one with a RepointableActorRef
subclass which creates RoutedActorCells upon activate(), with the nice
benefit that there is no hurry then to get it right because the new
cell is constructed “on the side”
misc fixes:
- InvalidMessageException is now actually enforced when trying to send
“null”
- Mailboxes may be created without having an ActorCell, which can come
in handy later, because the cell is only needed when this mailbox is
going to be scheduled on some executor
- remove occurrences of Props(), which is equivalent to Props[Nothing],
which is equivalent to «bug»
- add test case which verifies that context.actorOf is still synchronous
- plus all the stuff I have forgotten.
- this enables using any MessageQueue in BalancingDispatcher,
CallingThreadDispatcher and in general leads to less conflation of
concepts
- add MessageQueue.cleanUp(owner, deadLetterQueue) for the benefit of
durable mailboxes
- change MailboxType.create to take an optional owner and generate only
a MessageQueue, not a Mailbox
split systemDispatch(Create()) into systemEnqueue(Create()) directly
after createMailbox and registerForExecution from within
Dispatcher.attach() (resp. CallingThreadDispatcher.register() does its
own thing)
- add provider, guardian, systemGuardian and deathWatch to it
- make ActorSystemImpl extend ExtendedActorSystem
- use ExtendedActorSystem for creating extensions, thereby limiting the
access extensions get to just those four published methods.
* Changed signatures and constructor of MessageDispatcherConfigurator
* Changed Dispatchers.lookup, keep configurators instead of dispatchers
* Removed most of the Dispatchers.newX methods, newDispatcher is still there because of priority mailbox
* How should we make it easy to configure priority mailbox?
* Changed tons tests
* Documentation and ScalaDoc is not updated yet
* Some tests in ActorModelSpec are temporary ignored due to failure
- it was telling all children to stop(), then waited for the
ChildTerminated messages and finally terminated itself
- this worked fine, except when the stop came from the supervisor, i.e.
the recipient was suspended and did not process the ChildTerminated
- so, as the mirror of Supervise() that it is, I changed
ChildTerminated() to be a system message and instead of stopping
processing normal messages by checking the stopping flag, just suspend
the actor while awaiting the ChildTerminated's to flow in.
* Dispatchers need Scheduler to be able to shutdown themselves. Stop Scheduler after dispatchers.
* Changed CallingThreadDispatcher global object to Extension, since it holds map of references to mailboxes. Will be GC:ed when system is GC:ed.
* Made testActor lazy, since it is not used in all tests, and it creates CallingThreadDispatcher.
* Activated some java tests that were not running
* Many tests were not stopping created ActorSystems. VERY IMPORTANT TO STOP ActorSystem in tests. Use AkkaSpec as much as possible.
* Used profiler to verify (and find) dangling ActorSystemImpl and threads from dispatchers.
* FutureSpec creates ForkJoinPool threads that are not cleared, but number of threads don't grow so it's not a problem.
* Fixed obvious
* Created tickets for several, #1408, #1409, #1410, #1412, #1415, 1416, #1418
* Moved LoggingReceive from akka.actor to akka.event
* Touched several of the FIXME to make them visible in code review
- ensure that no “complex” things are attached to a LogEvent (think
serialization)
- ensure no escaping the “this” reference via LoggingBus during
constructors (e.g. ActorSystem)
- change it so that
+ Actor/ActorRef are represented by their address
+ Class[_] by simpleName
+ String by itself
- this means that people need to think a little more while deciding how
“this” should look like in logging (which I think is a good thing)
(cascading into all dispatchers, mailboxes, other stuff; had to move
deadLetter stuff to ActorSystem again and split its initialization due
to cyclic dependency)
- the same information is transmitted as sender, hence enabling
ChildTerminated to become a singleton
- make lastSender accessible in TestKit (needed now for DeathWatchSpec)
- fix nasty infinite loop when logging at the wrong moment during
shutdown
(since I know now what’s causing these Jenkins failures ;-) )
- include recipient in DeadLetter
- include recipient in calls to enqueue/systemEnqueue
- move DeadLetterMailbox to ActorSystem (saves some space, too)
- hook up DeadLetterMailbox so it sends DeadLetters to app.deadLetters,
which publishes them on the eventStream
- subscribe TestEventListener to DeadLetter and turn it into Warning
The generated warnings about ChildTerminated are very much correct, they
remind us that we still need to fix supervisor.stop() to await all
children’s death before actually committing suicide.
