[
https://issues.apache.org/activemq/browse/CAMEL-1633?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=52209#action_52209
]
Edward Campbell commented on CAMEL-1633:
----------------------------------------
Found out why the order mattered in this case...
According to the smack documentation if you register the same listener multiple
times, only the most recent filter is used.
Since the XmppConsumer is being used for all listeners in this case only the
last add<Message|Packet>Listener call is honored.
It would be wiser to have the listeners as separate classes, then the handling
of each packet/message type can be addressed separately.
A ConnectionListener can be added as well to handle connections that go down
and can be retried later.
> XMPPConsumer.processPacket does not correctly handle received non-message
> packets.
> ----------------------------------------------------------------------------------
>
> Key: CAMEL-1633
> URL: https://issues.apache.org/activemq/browse/CAMEL-1633
> Project: Apache Camel
> Issue Type: Bug
> Components: camel-xmpp
> Affects Versions: 1.6.1
> Reporter: Edward Campbell
> Assignee: Claus Ibsen
> Fix For: 1.6.2, 2.0.0
>
> Attachments: XmppConsumer.java
>
>
> I am currently porting an application using an in house Smack integration to
> Apache Camel.
> Specifically, I am getting failures with multiuser chat where messages stop
> being received in the middle of the message stream I am sending.
> I have yet to verify the issue exists with private chat as well, but the
> XMPPConsumer source looks like there will be a similar issue.
> The XMPPConsumer class registers itself for all packet types in the doStart
> method, but in the processPacket method immediately casts the received Packet
> to Message.
> I have found with the in house integration that Smack sends several types of
> Packets, and I could not find assurance that it would not call the packet
> listener with a null message.
> A simple if((null != packet) && (packet instanceof Message)) should be used
> to prevent improper packets from being utilized.
> FYI: the above if statement should also prevent packets from building up in
> the Smack message queue, since all messages will be processed without
> throwing an exception.
> So a call to muc.nextMessage() is unnecessary, and actually detrimental
> (since if the next packet is a message, it will be dropped without
> processing).
> It may be wise to actually use a try/catch block to prevent exceptions from
> being thrown by the processPacket method, since messages that throw
> exceptions seem to stay in the Smack message queue.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
