the combination with concurrencyLevel and LinkedBlockingQueue never
worked as intented. The idea was that the cachedExecutorService would
spawn new threads until maximumPoolSize (=concurrencyLevel) is reached,
and then start queing the Runnables.
But this was not the case, since ThreadPoolExecutor does not take into
consideration if the worker threads is busy, i.e. executing a Runnable,
or idle, i.e. waiting for a Runnable.
This means that if a busy Worker would execute a Runnable, which would
block, because it's waiting for an event (e.g. an incoming IQ
request), then the handling of those incoming IQ request would be
queued by ThreadPoolExecutor, because no fewer threads then corePoolSize
are running and the task can be queued (since the LinkedBlockingQueue is
unbounded).
Since e6045c6593 the cached executor
services maximum pool size is limited to concurrencyLevel (was
previously Integer.MAX_VALUE). In order to prevent
RejectExecutionException we need to use an queue which max size is
greater than 1 (i.e. nont an SynchronousQueue).
Connection closed with error java.util.concurrent.RejectedExecutionException: Task org.jivesoftware.smack.tcp.XMPPTCPConnection$2@41dce200 rejected from java.util.concurrent.ThreadPoolExecutor@41d59150[Running, pool size = 3, active threads = 3, queued tasks = 0, completed tasks = 4]
at java.util.concurrent.ThreadPoolExecutor$AbortPolicy.rejectedExecution(ThreadPoolExecutor.java:2011)
at java.util.concurrent.ThreadPoolExecutor.reject(ThreadPoolExecutor.java:793)
at java.util.concurrent.ThreadPoolExecutor.execute(ThreadPoolExecutor.java:1339)
at org.jivesoftware.smack.AbstractXMPPConnection.asyncGo(AbstractXMPPConnection.java:1583)
at org.jivesoftware.smack.tcp.XMPPTCPConnection.processHandledCount(XMPPTCPConnection.java:1655)
at org.jivesoftware.smack.tcp.XMPPTCPConnection.access$2300(XMPPTCPConnection.java:137)
at org.jivesoftware.smack.tcp.XMPPTCPConnection$PacketReader.parsePackets(XMPPTCPConnection.java:1083)
at org.jivesoftware.smack.tcp.XMPPTCPConnection$PacketReader.access$200(XMPPTCPConnection.java:896)
at org.jivesoftware.smack.tcp.XMPPTCPConnection$PacketReader$1.run(XMPPTCPConnection.java:911)
at java.lang.Thread.run(Thread.java:841)
and remove getConnectionID().
Also make streamId a field of AbstractXMPPConnection. Most XMPP
connection types have a streamId, it appears to be optional when BOSH
is used though.
instead of just logging a warning if the XMPP domain has no DNS SRV
lookups, create the failedAddresses list now within DNSUtil and add the
information that the SRV lookup failed.
Also limit the max pool size of the remove callbacks service and use the
same keep alive time for idle threads as the cached executor service
uses.
Note that it would be possible to merge those two. But this could lead
to tasks from the cached executor service blocking the removal of
callbacks, which we don't want.
if connect() was not previously called. Previously calling login() with
arguments would not check for the preconditions.
The check to throw needs to be performed in AbstractXMPPConnection
before every 'abstract login(Non)Anonymously()' call. That's the two
lines that check the preconditions are duplicated.
Also fix NPE in
XMPPTCPConnection.throwNotConnectedExceptionIfAppropriate() when
packetWrite is null (i.e. if the connection was never connected before).
This is useful for cases where a result set is requested, as it's the
case in XEP-13 and XEP-313.
Also adds
XMPPConnection.createPacketCollector(PacketCollector.Configuration).
after at most 12 hours.
Also set a keep alive time for the removeCallbacksService to 30 seconds
and add AbstractXMPPConnection.schedule(Runnable, long, TimeUnit).
isAnonymous() is now a bit weakened since it also considers
allowNullorEmptyUsername.
SaslExternalMechanism of smack-sasl-provided now also doesn't throw an
UnsupportedOperationException.
Follow up on 7e4e3699a1
to ConnectionConfiguration.Builder().
And prepare SASL EXTERNAL for empty or null usernames.
Also clarify some parts regarding the user field.
Fixes SMACK-627
This also moves the logic to send error IQ replies from "when there is
no IQ provider registerd" to "when there is no IQ request handler
registered". Which has for example the advantage that IQ parsing no
longer asks for a connection instance.
It's important to know if the stream was resumed. authenticated() is the
ideal callback for Managers to reset their state (e.g. cached values of
the connection state). But if the stream was resumed, the cached values
don't have to be reset.
notably add a cache for the active and default privacy list to avoid
IQ get/response round-trips.
Also add a few methods to PrivacyListManager to get the privacy list
names. The already existing methods always returned the whole list
together with the name, which caused two round-trips.
Simplified some code.
Properly escape Privacy XML.
Differentiate between asynchronous and synchronous ones. Asynchronous
are the ones where the invocation order may not be the same as the order
in which the stanzas arrived.
Since it's no longer guaranteed that when a unit test calls
processPacket(stanza)
the stanza will be completely processed when the call returns, it was
necessary to extend the unit tests (mostly Roster and ChatManager) with
a packet listener that waits for his invocation. Since we now also use
LinkedHashMaps as Map for the packet listeners (SMACK-531, SMACK-424),
adding a packet listeners as last also means that it will be called as
last. We exploit this behavior change now in the unit tests.
Rename 'recvListeners' to 'syncRecvListeners' in AbstractXMPPConnection.
Rename 'rosterInitialized' to 'loaded' in Roster.
Add Roster.isLoaded().
Reset 'loaded' to false in
Roster.setOfflinePresencesAndResetLoaded() (was setOfflinePresences()).
Fixes SMACK-583, SMACK-532, SMACK-424