[jira] [Commented] (REEF-1506) Check behaviour of IMRU tests that run in process

["Mariia Mykhailova (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/REEF-1506?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15752734#comment-15752734 ] 

Mariia Mykhailova commented on REEF-1506:
-----------------------------------------

These tests use {{InProcessIMRUClient}} instead of {{REEFIMRUClient}}, which effectively uses a separate IMRU implementation in {{IMRURunner.cs}}. {{IMRURunner.Run}} calls map and update tasks directly, not through {{UpdateTaskHost}}, so naturally changes to {{UpdateTaskHost}} don't affect these tests.

{{InProcessIMRUClient}} seems to be an old testing implementation which probably predates the one we're working on now. [~juliaw] [~markus.weimer] [~dkm2110] Is this implementation used in practice? Could it cause any problems, other then just confusing people?

> Check behaviour of IMRU tests that run in process
> -------------------------------------------------
>
>                 Key: REEF-1506
>                 URL: https://issues.apache.org/jira/browse/REEF-1506
>             Project: REEF
>          Issue Type: Bug
>            Reporter: Sergey Dudoladov
>            Priority: Minor
>
> We have several test such as TestMapperCountInProcess or TestNaturalSumInProcess that are not always affected by the changes in the IMRU code base. To reproduce, make the UpdateTaskHost sleep after  broadcasting to mappers (in the line 121). The tests will still complete successfully unaffected by the Update() sleep timeout despite the fact that evaluator processes will still be running.
> Tests that use local runtime work fine.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)