[ 
https://issues.apache.org/jira/browse/FLINK-8067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16508057#comment-16508057
 ] 

ASF GitHub Bot commented on FLINK-8067:
---------------------------------------

Github user pnowojski commented on the issue:

    https://github.com/apache/flink/pull/6081
  
    @yanghua why is it difficult? Can not you create more or less similar test 
to @GJL's `testSetsUserCodeClassLoader` from here: 
https://github.com/apache/flink/pull/4980/files. With couple side notes:
    
    1. build `MockEnvironment` using `MockEnvironmentBuilder` with a custom 
user class loader - do not use mockito for that
    2. create a stream task (reusing one of the implementation in 
`StreamTaskTest.java` or create a new one)
    3. register a timer service, wait for it's triggering and assert the user 
class loader in `ProcessingTimeCallback#onProcessingTime`


> User code ClassLoader not set before calling ProcessingTimeCallback
> -------------------------------------------------------------------
>
>                 Key: FLINK-8067
>                 URL: https://issues.apache.org/jira/browse/FLINK-8067
>             Project: Flink
>          Issue Type: Bug
>          Components: Streaming
>    Affects Versions: 1.4.0
>            Reporter: Gary Yao
>            Assignee: vinoyang
>            Priority: Minor
>             Fix For: 1.6.0, 1.5.1
>
>
> The user code ClassLoader is not set as the context ClassLoader for the 
> thread invoking {{ProcessingTimeCallback#onProcessingTime(long timestamp)}}:
> https://github.com/apache/flink/blob/84a07a34ac22af14f2dd0319447ca5f45de6d0bb/flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/StreamTask.java#L222
> This is problematic, for example, if user code dynamically loads classes in 
> {{ProcessFunction#onTimer(long timestamp, OnTimerContext ctx, Collector<O> 
> out)}} using the current thread's context ClassLoader (also see FLINK-8005).



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to