[
https://issues.apache.org/jira/browse/FLINK-31710?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17707969#comment-17707969
]
Ran Tao commented on FLINK-31710:
---------------------------------
In fact, we have encountered other similar double references in related
projects of the flink ecology, such as guava and flink-shaded-guava and hadoop
and flink-shaded-hadoop. It is a good practice to constrain the use of the
shaded scheme. I'd be happy to work on this issue. Can u assign this ticket to
me? (If we decide to use shaded-curator)
> Remove <curator.version/> and rely on the curator dependency that's provided
> by flink
> -------------------------------------------------------------------------------------
>
> Key: FLINK-31710
> URL: https://issues.apache.org/jira/browse/FLINK-31710
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Coordination
> Affects Versions: 1.17.0, 1.16.1, 1.18.0
> Reporter: Matthias Pohl
> Priority: Major
>
> Currently, we're relying on a dedicated curator dependency in tests to use
> the {{TestingZooKeeperServer}} (see [Flink's parent
> pom|https://github.com/apache/flink/blob/97cff0768d05e4a7d0217ddc92fd9ea3c7fae2c2/pom.xml#L143]).
> Besides that, we're using {{flink-shaded}} to provide the zookeeper and
> curator dependency that is used in Flink's production code.
> The flaw of that approach is that we have to maintain two curator versions.
> This Jira issue is about investigating whether we could just remove the
> curator test dependency and rely on the {{flink-shaded}} curator sources.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
