[
https://issues.apache.org/jira/browse/UIMA-4168?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14266604#comment-14266604
]
Jim Challenger commented on UIMA-4168:
--------------------------------------
It turns out to be even simpler. Just be sure to flush stdout in viaducc
before calling other python scripting. This way there is only one process with
i/o pending and the python libs don't get confused.
> DUCC scripting uncaught exceptions
> ----------------------------------
>
> Key: UIMA-4168
> URL: https://issues.apache.org/jira/browse/UIMA-4168
> Project: UIMA
> Issue Type: Bug
> Components: DUCC
> Affects Versions: 1.1.0-Ducc
> Reporter: Jim Challenger
> Assignee: Jim Challenger
> Priority: Minor
> Fix For: 2.0.0-Ducc
>
>
> If ducc scripts that use ducc_util.spawn() participate in a unix pipeline,
> and that pipeline is broken, either accidentally or deliberately, the
> underlying subsystem.popen() might throw an exception when components
> elsewhere in the pipeline close(). Couldn't find any way to catch this
> exception in a reliable manner. Solution is to set the sys.excepthook method
> to a custom method that deals with the problem. In this case, it's ok to
> simply ignore the exception so it doesn't spew to the console, disconcerting
> the user. The test for this is to pipe the output of ducc_process_submit
> into 'head', which reads a few lines and then closes its pipe even if
> ducc_process_submit isn't completed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
