by implementing a new logic how the messages are retrieved.
Previously in case the node list has exactly one item, the method
would perform an unnecessary call to nextResult() causing a delay.
Fixes SMACK-785.
Remove the "if (!joined) return" guard in leave() this allows to
resync the instances state with the real world state in case they ever
get out of sync.
Also call userHasLeft() in even if leave() throws and in certain
situations if destroy() throws.
Thanks to Дамян Минков and Ingo Bauersachs for pointing this out.
and some further minor jingle fixes:
- deprecate getJingleTransport() in favor of getTransport()
- Jingle.Builder now checks if the session ID is not empty
The test failed because the ArrayList - in contrast
to the underlying Set - did not check for duplicates
on insert. Under certain circumstances this lead to
an index out of bounds exception because the list in
the test contained duplicated entries which were not
present in the set of the Socks5Proxy.
I fixed the issue by only inserting the address when
it was not in the list before.
- Change visibility of some Socks5Bytestreams code.
- Add central ThreadPool
- Move FullJidAndSessionId in own class
- More complete JingleSession class
- More complete JingleUtil class
- Improved tests
A start for the new Jingle API. Since Jingle is a single IQ with many
plugable extensions, there are some particularities we need to deal
with, e.g. jingle users have to register with JingleManager.
This is untested code. There may be drangons.
Previously publish() was "asynchronously" in that sense that the response
of the IQ as *not* awaited, and send() would wait for the
responses. It makes no sense to have that functionality this way.
Instead we now make publish() to the right thing, i.e., wait for an IQ
result, make send() a proxy for publish(), and mark send() deprecated.
In the future, there may be a publishAsync() method which returns a
Future instance.
by extending Manager.
Because FileTransferNegotiator will hold a strong reference to the
StreamNegotiators, which will eventually prevent XMPPConnection from
being GC'ed if no weak references in StreamNegotiator are used.
Thanks to Werner Glanzer for pointing this out.
To prevent
java.util.ConcurrentModificationException
at java.util.HashMap$HashIterator.nextEntry(HashMap.java:788)
at java.util.HashMap$KeyIterator.next(HashMap.java:815)
at org.jivesoftware.smackx.ping.PingManager.pingMyServer(PingManager.java:252)
at org.kontalk.service.msgcenter.MessageCenterService$3.run(MessageCenterService.java:1114)
at java.lang.Thread.run(Thread.java:818)
Thanks to Daniele Ricci for reporting this.
instead of an EntityFullJid, because according to XEP-0045 § 7.8.1.:
"The <room@service> itself MUST then add a 'from' address to the
<invite/> element whose value is the bare JID, full JID, or occupant
JID of the inviter …"
The jids list doesn't have to be lazy initialized, because every IQ of
that type is guaranteed to contain at least one JID.
Also use ParserUtils.getJidAttribute().
for the last message. We now count the number of messages we want to
retrieve, and don't wait for another message if we have already
received all.
Thanks to King Jeong Hun for reporting this.
capturing all outoing presences of type 'available' would also capture
presences not used for presence broadcast, e.g., MUC presences.
This caused the EntityCaps integration test (localEntityCaps) to fail
if the MUC integration test was run before.
the manager must use a copy of the BlockListIQ's JID list, since it may
be the empty list which is not modifiable.
Also rename the getter methods from getJids() to getBlockedJids().
and not by '!error' as this there are presence stanzas containing the
muc#user namespace also send to the client, which results in a
ClassCastException in the StanzaListener.
Returning a boolean is not really expressive in such cases. For example,
if it returns 'false' then the callee has no information *what* went
wrong. Instead throw an SmackException with some more information.
A unauthorized user sents a update room subject message which is
rejected by the server. It responds with an error message which
is not picked up by the ChangeSubjectListener.
Mostly remove the helper utils. The server is required to present the
client with a consisent state of the block list and corresponding
modifications, so we should not end up with duplicate entires if we
don't check for them.
SMACK-731
Smack's previous entity caps implementation assumed that an entity lost
its entity caps feature as soon as a presence without caps from that
entity was received. But according to XEP-0115 § 8.4, this is a
perfectly normal optimization technique. We now reset the caps state
after an available presence becomes unavailable.
Also introduce PresenceEventListener, which is required for this
feature.
Also make Roster.preApprove() take a BareJid as argument.
Fixes SMACK-723.
Some servers respond to disco#info requests with
<iq id='npGtO-21' type='result' to='phone@xxx/MAXS' from='xxx'>
<query xmlns='http://jabber.org/protocol/disco#info'>
<identity type='pep' name='Prosody' category='pubsub'/>
<identity type='im' name='Prosody' category='server'/>
<feature var='urn:xmpp:blocking'/>
<feature var='urn:xmpp:carbons:2'/>
<feature var='urn:xmpp:carbons:1'/>
<feature var='vcard-temp'/>
<feature var='http://jabber.org/protocol/commands'/>
<feature var='urn:xmpp:mam:0'/>
<feature var='jabber:iq:private'/>
<feature var='http://jabber.org/protocol/pubsub#publish'/>
<feature var='http://jabber.org/protocol/disco#info'/>
<feature var='http://jabber.org/protocol/disco#items'/>
<feature var='urn:xmpp:ping'/>
<feature var='msgoffline'/>
<feature var='jabber:iq:roster'/>
<feature var='urn:xmpp:archive:auto'/>
<feature var='urn:xmpp:archive:manage'/>
<feature var='urn:xmpp:archive:pref'/>
<feature var='http://jabber.org/protocol/rsm'/>
<feature xmlns='urn:xmpp:archive'><optional><default/></optional></feature>
</query></iq>
Note the
<feature
xmlns='urn:xmpp:archive'><optional><default/></optional></feature>
which will cause the current parser implementation to parse it as
DiscoInfo Feature resulting in an Exception because some attributes are
missing.
This commit prevents this.
The previously used approach of
project(':smack-core').sourceSets.test.runtimeClasspath
caused the 'eclipse' target to produce duplicate classpath entries in
.classpath when run with Gradle >= 2.6. It also relied on Gradle
internals.
Instead we now use
project(path: ":smack-core", configuration: "testRuntime")
project(path: ":smack-core", configuration: "archives")
to be able to use test classes from other subprojects (usually
smack-core) in e.g. smack-extensions. The 'archives' configuration
includes the test jar.
See also https://discuss.gradle.org/t/11784
Thanks to Lari Hotari for helping with this issue.
So that we can wrap the original exception within the
SmackException. It's not possible to wrap it into the IOException
because of the used min Android API level.
