[
https://issues.apache.org/jira/browse/FLINK-10827?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16682631#comment-16682631
]
ASF GitHub Bot commented on FLINK-10827:
----------------------------------------
tillrohrmann commented on a change in pull request #7061: [FLINK-10827][tests]
Add test for duplicate() to SerializerTestBase
URL: https://github.com/apache/flink/pull/7061#discussion_r232467908
##########
File path:
flink-core/src/main/java/org/apache/flink/api/java/typeutils/runtime/PojoSerializer.java
##########
@@ -181,11 +181,13 @@ public boolean isImmutableType() {
}
}
- if (stateful) {
- return new PojoSerializer<T>(clazz,
duplicateFieldSerializers, fields, executionConfig);
- } else {
- return this;
+ if (!stateful) {
+ // as a small memory optimization, we can share the
same object between instances
+ duplicateFieldSerializers = fieldSerializers;
}
+
+ // we must create a new instance, otherwise the
subclassSerializerCache can create concurrency problems
Review comment:
I'm wondering whether the statefulness check is simply wrong. Since the
`PojoSerializer` could also use a subclass serializer we should also check
`subclassSerializerCache` and `registeredSerializers` for statefulness.
However, the way this is done currently by duplicating and checking for
referential equality might be quite inefficient...
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
> Add test for duplicate() to SerializerTestBase
> ----------------------------------------------
>
> Key: FLINK-10827
> URL: https://issues.apache.org/jira/browse/FLINK-10827
> Project: Flink
> Issue Type: Test
> Components: Tests
> Reporter: Stefan Richter
> Assignee: Stefan Richter
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.7.0
>
>
> In the past, we had many bugs from type serializers that have not properly
> implemented the {{duplicate()}} method in a proper way. A very common error
> is to forget about creating a deep copy of some fields that can lead to
> concurrency problems in the backend.
> We should add a test case for that tests duplicated serializer from different
> threads to expose concurrency problems.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
