[
https://issues.apache.org/jira/browse/DIRMINA-675?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12682610#action_12682610
]
Steve Ulrich commented on DIRMINA-675:
--------------------------------------
IMHO it's the encoder, not the decoder.
Currently the IoSession.write does call the FilterChain asynchronously instead
of using the MessageQueue (or am I missing something?).
So there are asynchronous are calls to
ProtocolCodecFilter.ProtocolEncoderOutputImpl (they definitely exists) which
does several offers and polls at the CircularQueue at same time and resulting
in loss of messages or duplicate poll results (which means that a single
WriteRequest is proposed more than once to the next filter).
The interesting part is, that I can't see anything that changes the logic
between M3 and M4, so it could be pure luck, that it worked with M3.
> Duplicate data from decoder
> ---------------------------
>
> Key: DIRMINA-675
> URL: https://issues.apache.org/jira/browse/DIRMINA-675
> Project: MINA
> Issue Type: Bug
> Affects Versions: 2.0.0-M4, 2.0.0-RC1
> Environment: Ubuntu Linux 8.10 x86, Java(TM) SE Runtime Environment
> (build 1.6.0_10-b33)
> Reporter: Gris
> Priority: Blocker
> Fix For: 2.0.0-RC1
>
> Attachments: MinaIssue.tar, test.tar
>
>
> I got duplicate messages in messageReceived(...) in my io handler adapter,
> but there are no such problem while decoding message (each message decoded
> only once). Rolling back to M3 seems solved problem.
> Note. SslFilter is used.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
