[
https://issues.apache.org/jira/browse/KAFKA-12699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17326223#comment-17326223
]
Walker Carlson commented on KAFKA-12699:
----------------------------------------
I would not say its a bug per-say but I just wanted anyone surprised by the
behavior to know that it was known. Also I am not sure what else to categorize
it as
We might be able to handle the EOS corner cases by changing around where the
catch block is. I am not sure a no-op would be better, but I do think that
since the thread died exceptionally we should have the exception raised again
to make sure it it handled properly. Also we should do our best to make sure
the exception and thread are paired. I could see a case where the stream user
and the owner of the runtime are not entirely equal so might have different
requirements. I do agree calling System.exit might be a bit extreme but they
could be tracking threads deaths and exceptions for some reason.
> Streams no longer overrides the java default uncaught exception handler
> -------------------------------------------------------------------------
>
> Key: KAFKA-12699
> URL: https://issues.apache.org/jira/browse/KAFKA-12699
> Project: Kafka
> Issue Type: Bug
> Components: streams
> Affects Versions: 2.8.0
> Reporter: Walker Carlson
> Priority: Minor
>
> If a user used `Thread.setUncaughtExceptionHanlder()` to set the handler for
> all threads in the runtime streams would override that with its own handler.
> However since streams does not use the `Thread` handler anymore it will no
> longer do so. This can cause problems if the user does something like
> `System.exit(1)` in the handler.
>
> If using the old handler in streams it will still work as it used to
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
