Github user GJL commented on the issue:
https://github.com/apache/flink/pull/4356
Thanks for your contribution. I have a few questions because the submitted
code does not match with what the ticket currently says:
From https://issues.apache.org/jira/browse/FLINK-5486:
>The handlePendingFilesForPreviousCheckpoints() call should be enclosed
inside the synchronization block. Otherwise during the processing of
handlePendingFilesForPreviousCheckpoints(), some entries of the map may be
cleared.
I am confused because the submitted code encloses
`bucketState.pendingFilesPerCheckpoint.clear();` in a synchronized block.
@tedyu Did you run into actual production issues, or did you find the
problem while reviewing the code? Can you elaborate why the additional
synchronized block is needed?
---
