You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Alan Sheinberg (Jira)" <ji...@apache.org> on 2025/04/24 16:22:00 UTC
[jira] [Created] (FLINK-37721) Fix Janino Bug which resulted in
Alan Sheinberg created FLINK-37721:
--------------------------------------
Summary: Fix Janino Bug which resulted in
Key: FLINK-37721
URL: https://issues.apache.org/jira/browse/FLINK-37721
Project: Flink
Issue Type: Sub-task
Reporter: Alan Sheinberg
The bug comes from the fact that previously, we were creating an anonymous function callback within the janino generated code. This function was then referencing the local {{{}DelegatingAsyncResultFuture{}}}, effectively creating a closure. When the function was called back from different threads, they seemed to both reference the same {{DelegatingAsyncResultFuture}} rather than their own local ones. The result was returning incorrect "synchronous" results, so queries like the following might have issues:
```
SELECT f1, func(f1, f2, f3) FROM Table1;
```
Here, the {{f1}} column returned was retrieved by asking for it from the local {{{}DelegatingAsyncResultFuture{}}}, and so getting the wrong instance meant getting the wrong result.
To fix it, we should remove the inner anonymous function and just added the same logic on {{{}DelegatingAsyncResultFuture{}}}. It requires setting it up with all of the metadata (indexes, sync results), and when the result is complete, rather than calling the generated function, create the result {{RowData}} itself.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)