[jira] [Commented] (HIVE-3084) Hive CI failing due to script_broken_pipe1.q

["Gunther Hagleitner (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HIVE-3084?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13526278#comment-13526278 ] 

Gunther Hagleitner commented on HIVE-3084:
------------------------------------------

I ran into the same issue. It affects both script_broken_pipe1.q as well as minimr_broken_pipe.q. script_broken_pipe3.q has the same problem, but it won't cause the test to fail - just to not test what it was intended to do.

The test seems to assume that the user process (bash's "true") exits and closes its stdin before the ScriptOperator closes the associated output (in the close method). There is however no guarantee that things happen in this sequence and if they don't there will be no exception thrown.

I have rewritten the tests using two ScriptOperators in the pipeline. The first one will exit properly, the second one will throw the exception. Closing the first ScriptOperator (flushing buffers, waitFor...) should ensure that the second user process had a chance to exit, although strictly speaking there is still no guarantee. I've run this a number of time and wasn't able to reproduce the issue with the fix. 
                
> Hive CI failing due to script_broken_pipe1.q
> --------------------------------------------
>
>                 Key: HIVE-3084
>                 URL: https://issues.apache.org/jira/browse/HIVE-3084
>             Project: Hive
>          Issue Type: Bug
>            Reporter: Travis Crawford
>            Assignee: Travis Crawford
>         Attachments: HIVE-3084_disable_failing_test.1.patch, HIVE-3084_two_script_operators.patch
>
>
> Hive's [CI job|https://builds.apache.org/job/Hive-trunk-h0.21/] regularly fails due to this test:
> {code}
> org.apache.hadoop.hive.cli.TestNegativeCliDriver.testNegativeCliDriver_script_broken_pipe1
> {code}
> Running just that test in a loop I'm not able to reproduce the issue locally. This may be related to state from a previous test causing this one to fail?
> {code}
> ant clean package
> for x in $(seq 1 10); 
>   do ant test -Dtestcase=TestNegativeCliDriver -Dqfile=script_broken_pipe1.q;
> done
> {code}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira