[
https://issues.apache.org/jira/browse/ARTEMIS-1276?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16093473#comment-16093473
]
ASF GitHub Bot commented on ARTEMIS-1276:
-----------------------------------------
Github user jdanekrh commented on the issue:
https://github.com/apache/activemq-artemis/pull/1407
> messages only expire on client on activemq5 while it's on server & client
on artemis
Did you mean expire on _server_ on activemq5? If it expired on client, how
would it then get to DLQ queue on server?
---
I think get it now. When the test receives for the first time (first `for`
loop), the client gets first 10 messages sent, which are those with short
expiration time. One of them is delivered ok, the 9 remaining expire on the
client while it is in sleep(). That means these 9 never get into the DLQ,
because for that they would have to expire on the server. Which explains the
`actual: 90, expected: 99` result when it is then receiving from DLQ.
I did not know about this difference between 5 and Artemis.
From user point of view, the fact that prefetch changes behavior (and is
not just performance optimization) and that it can cause what can be thought of
as a message loss (these 9 messages have irrevocably disappeared, right?) could
be considered a bug, I think.
I guess there is nothing to be done about it here, so removing [wip].
> [openwire-tests] Enable more OpenWire tests (from ActiveMQ 5)
> -------------------------------------------------------------
>
> Key: ARTEMIS-1276
> URL: https://issues.apache.org/jira/browse/ARTEMIS-1276
> Project: ActiveMQ Artemis
> Issue Type: Test
> Components: OpenWire
> Affects Versions: 2.2.0
> Reporter: Jiri Danek
> Priority: Minor
>
> I've been looking into the currently disabled OpenWire tests, if there is a
> way to implement support for more of them and enable them.
> I believe I've made some progress on this front, so I'd wish to discuss
> 1) what is the endgame about these tests; do we want to eventually run them
> all, or is it expected to to ignore some?
> 2) what tests would be useful to enable now? I've been looking at them from
> the point of how easy it is to enable them, not how useful it is.
> 3) would it be ok to add mockito as a test dependency? it seems useful,
> because sometimes it is necessary to wrap something in ActiveMQ5 class, and
> doing a proxy avoids having to invoke superclass constructors (when the thing
> that needs to be implemented is a class and not an interface). Mockito is IMO
> nicer to use than using java.lang.reflect.Proxy.newProxyInstance directly.
> and propose PR based on that discussion.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
