[
https://issues.apache.org/jira/browse/ROCKETMQ-134?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15901121#comment-15901121
]
ASF GitHub Bot commented on ROCKETMQ-134:
-----------------------------------------
Github user StyleTang commented on the issue:
https://github.com/apache/incubator-rocketmq/pull/72
@Jaskey
I found pullStatus is final, and I don't want to redefine it, so I judge
the msgList size.
may be change the value of pullStatus will be better.
As you say, the problem occurred when your consumer client get messages
every pull request, and the message can be filter by tag.
The probability is very low.If everyone think it doesn't matter,I will
close this PR.
But I think have a check is better.
> the offset of message filter by tags may be not commit to broker
> ----------------------------------------------------------------
>
> Key: ROCKETMQ-134
> URL: https://issues.apache.org/jira/browse/ROCKETMQ-134
> Project: Apache RocketMQ
> Issue Type: Bug
> Components: rocketmq-client
> Affects Versions: 4.0.0-incubating
> Reporter: Jie.Tang
> Assignee: Xiaorui Wang
> Priority: Trivial
>
> when different string has a same hash code.the message commit offset of
> filtered message may be not commit to broker.
> for example:
> 1.consumer pull message from broker, broker return status FOUND and messages
> filter by tags hash code
> 2.consumer client get the messages and than processPullResult will filter
> message by tags.
> 3.PullCallback may get a pullResult which status is FOUND but messageList is
> empty.(filter by tags)
> but only NO_MATCHED_MSG and NO_NEW_MSG will correctTagsOffset
> we can't commit the right with status of FOUND(for messageList is empty).
> Is that so?
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
