[
https://issues.apache.org/jira/browse/FLINK-5855?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15981456#comment-15981456
]
ASF GitHub Bot commented on FLINK-5855:
---------------------------------------
Github user kl0u commented on the issue:
https://github.com/apache/flink/pull/3758
Hi @zhangminglei, thanks for looking into this! I tend to agree with
@tillrohrmann that there is not need for locking there, as there is no sharing
of the data-structure between different threads. The same holds for the lock at
line 726. So these 2 `synchronized()` clauses could go away.
I will have a more thorough look on the whole locking scheme in the class
in the following days. But for now I think that this PR could be closed and the
JIRA could be updated.
What do you think?
> Unprotected access to pendingFilesPerCheckpoint in BucketingSink
> ----------------------------------------------------------------
>
> Key: FLINK-5855
> URL: https://issues.apache.org/jira/browse/FLINK-5855
> Project: Flink
> Issue Type: Bug
> Components: Streaming Connectors
> Reporter: Ted Yu
> Assignee: mingleizhang
> Priority: Minor
>
> {code}
> handlePendingFilesForPreviousCheckpoints(restoredState.pendingFilesPerCheckpoint);
> synchronized (restoredState.pendingFilesPerCheckpoint) {
> restoredState.pendingFilesPerCheckpoint.clear();
> {code}
> Lock on pendingFilesPerCheckpoint should be obtained prior to the call to
> handlePendingFilesForPreviousCheckpoints().
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
