* scala-library.jar had changed name to scala-library-2.10.4.jar
* use wildcard to avoid same problem when scala-library version is changed
* the wilcard will only include jar files in the lib directory, and not the
lib/akka directory, see 'Understanding class path wildcards'
http://docs.oracle.com/javase/6/docs/technotes/tools/windows/classpath.html
(cherry picked from commit 996de05b44531bda517ef8ee274a34ad76e9716f)
AtLeastOnceDelivery can delivery out-of-order, and that's OK.
Although, in the case of message replay followed by taking user land
commands which may trigger `deliver` calls, it is nicer to at least once
try to send the replayed but not confirmed deliveries *first*, before
sending the completely new deliveries.
This change acomplishes this by triggering redelivery explicitly when
recovery has finished, and setting the timestamps on these messages a
bit in the past, so they hit their redelivery deadline right away during
this recovery induced redelivery.
Resolves#15590
* The filter is used by the LoggingAdapter before publishing
to the event bus
* Slf4jLoggingFilter uses backend log level configuration
(e.g. logback.xml)
Since the updated github issues views, it shows all "removed tags",
events in the issue even if the removed tag was never set on the issue.
So this script would always show up as "removed validating,needs-attention"
Even if the issue never had the label needs-attention set to it,
which is confusing visually.
This change introduces that we only remove tags that actually are set on
the issue.
* the lower bound was rather racy, depends on "where in it's Tick"
time the throtteler currently was. In general the upper bound is also
not exact, but "good enough" because the `.5` is an estimation of "the
throtteler must finish it's previous tick, and then it sends the data"
* lower redelivery interval does not change semantics here (no one
answers anyway), and lessens the changes to be "unlucky" with gc pauses
* was unable to reproduce error with snapshotting in 80 builds on
jenkins. Brainstormed possible errors but not found yet..
For example in Java 7 you can now join a multicast group:
case class JoinGroup(group: InetAddress, networkInterface: NetworkInterface) extends SocketOption {
override def afterConnect(c: DatagramChannel): Unit = {
c.join(group, networkInterface)
}
}
IO(Udp) ! Udp.Bind(self, new InetSocketAddress(MulticastListener.port),
options=List(ReuseAddress(true),
JoinGroup(MulticastListener.group, MulticastListener.interf)))
Other minor changes:
- changed all methods in SocketOption to take a Channel instead of a Socket. The socket can be gotten from the Channel but not the reverse.
- all methods that are called before the bind are now called beforeBind for consistency.
- All network connections now call the beforeBind and afterConnect.
When in `A`:
* `goto(A)` will trigger `onTransition(A -> A)`
* `stay()` will NOT trigger `onTransition`
Includes:
* migration guide
* docs updates
* test
