[
https://issues.apache.org/jira/browse/SPARK-59242?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tengfei Huang updated SPARK-59242:
----------------------------------
Affects Version/s: 4.3.0
(was: 4.2.0)
> NPE in SQLExecution.withNewExecutionId masks the real query failure when the
> SparkContext is stopped concurrently
> -----------------------------------------------------------------------------------------------------------------
>
> Key: SPARK-59242
> URL: https://issues.apache.org/jira/browse/SPARK-59242
> Project: Spark
> Issue Type: Task
> Components: SQL
> Affects Versions: 4.3.0
> Reporter: Tengfei Huang
> Priority: Major
> Labels: pull-request-available
>
> `SQLExecution.withNewExecutionId` dereferences sc.dagScheduler inside its
> finally block. SparkContext.stop() sets _dagScheduler to null before it stops
> the listener bus, so a query that finishes while the context is being torn
> down reaches that line with a null receiver and throws a
> `NullPointerException`.
> Because the NPE is thrown from a finally, it replaces the query's original
> exception rather than being suppressed, and it skips every remaining cleanup
> step in the block.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
