You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by "Martin Ritchie (JIRA)" <qp...@incubator.apache.org> on 2008/10/14 13:35:46 UTC
[jira] Commented: (QPID-1262) [AMQP-0-8/9] Client can incorrectly
report CommitOK if previous commit timed out.
[ https://issues.apache.org/jira/browse/QPID-1262?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12639391#action_12639391 ]
Martin Ritchie commented on QPID-1262:
--------------------------------------
Whilst the CommitOk from the previous loop can cause this failure when we loop the tests. I don't see how it can cause a problem when we have closed the connection between test run as happens in our test suite.
> [AMQP-0-8/9] Client can incorrectly report CommitOK if previous commit timed out.
> ---------------------------------------------------------------------------------
>
> Key: QPID-1262
> URL: https://issues.apache.org/jira/browse/QPID-1262
> Project: Qpid
> Issue Type: Bug
> Components: Java Broker
> Affects Versions: M2.1, M3
> Reporter: Martin Ritchie
> Fix For: M4
>
>
> Summary:
> As shown by the new SyncWaitTimeoutDelayTest there is a bug in AMQP. The TxCommitOk bodies do not have a correlation id so it is impossible to tell if the TxCommitOk received is for the last TxCommit sent if you have timed out a previous commit.
> Attached is a log for this behaviour. As this is an AMQP bug it is unlikely we can resolve it in Qpid.
> Question is does 0-10 also exhibit this behaviour.
> Work around is to close the session after a commit timeout this will ensure that the TxCommitOk does not spuriously arrive on the client. It does however leave you unsure of the commit state, but presumably there was a reason you gave up waiting for the TxCommitOk and are prepared for the operation actually succeeding but never being notified about it.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.