The NPE is caused by an inbound presence stanza without the 'from'
attribute set. The stacktrace of the NPE is:
FATAL EXCEPTION: Smack Cached Executor
Process: de.fhg.ivi.senetz.mobile.android.mbk.debug, PID: 13365
java.lang.NullPointerException: Attempt to invoke virtual method 'int java.lang.Object.hashCode()' on a null object reference
at java.util.concurrent.ConcurrentHashMap.get(ConcurrentHashMap.java:944)
at org.jivesoftware.smack.roster.Roster.getPresencesInternal(Roster.java:374)
at org.jivesoftware.smack.roster.Roster.getOrCreatePresencesInternal(Roster.java:388)
at org.jivesoftware.smack.roster.Roster.access$1100(Roster.java:94)
at org.jivesoftware.smack.roster.Roster$PresencePacketListener$1.run(Roster.java:1519)
at org.jivesoftware.smack.AsyncButOrdered$Handler.run(AsyncButOrdered.java:121)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1162)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:636)
at java.lang.Thread.run(Thread.java:764)
Thanks to Marcel Heckel for reporting this.
Fixes SMACK-861.
Fix for SMACK-841
Since Smack 4.2.4, the getPresencesInternal method in the Roster class
can return a LruCache object, which is a LinkedHashMap with access
order. This means that any access using get or getOrDefault will be a
modification of the Map. If you loop over the keySet of the Map and
there are more than one, the second call to get will throw a
ConcurrentModificationException!
Since the keys are only used here to obtain the corresponding
values, the simplest solution is to just loop over the values instead.
Daniele Ricci reporting the following NPE using Smack 4.1.9
java.lang.NullPointerException: Attempt to invoke virtual method 'int java.lang.String.indexOf(int)' on a null object reference
at org.jxmpp.util.XmppStringUtils.parseBareJid(XmppStringUtils.java:124)
at org.jivesoftware.smack.roster.Roster$RosterPushListener.handleIQRequest(Roster.java:1416)
at org.jivesoftware.smack.AbstractXMPPConnection$2.run(AbstractXMPPConnection.java:1061)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1113)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:588)
at java.lang.Thread.run(Thread.java:818)
This is possibly caused by a service sending roster pushes for unbound
connections, i.e. where getUsers() returns 'null'. We now log such
situations instead throwing an NPE.
This adds supports for an experimental protocol flow where a pending
friend request's decission is later on deliverd to the requestor after
the owner made its decission.
allow multiple of them to be installed, instead of at most one. Fixes
deadlock in LowLevelRosterIntegration test because
IoTProvisioningManager's SubscribeListener would not come up with a
decission.
