[ 
https://issues.apache.org/jira/browse/BEAM-2734?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16115114#comment-16115114
 ] 

Eugene Kirpichov commented on BEAM-2734:
----------------------------------------

Another issue is also something I broke with 
https://github.com/apache/beam/pull/3360 because I forgot to rebuild and 
release the Dataflow worker after the changes to runners-core in that PR.

All failing tests are :

 org.apache.beam.sdk.transforms.SplittableDoFnTest.testSideInput
 org.apache.beam.sdk.transforms.SplittableDoFnTest.testAdditionalOutput
 org.apache.beam.sdk.transforms.SplittableDoFnTest.testLifecycleMethods
 
org.apache.beam.sdk.transforms.SplittableDoFnTest.testPairWithIndexWindowedTimestamped
 org.apache.beam.sdk.transforms.SplittableDoFnTest.testOutputAfterCheckpoint
 org.apache.beam.sdk.transforms.SplittableDoFnTest.testPairWithIndexBasic

 org.apache.beam.sdk.metrics.MetricsTest.testUnboundedSourceMetrics
 org.apache.beam.sdk.metrics.MetricsTest.testBoundedSourceMetrics
 org.apache.beam.sdk.metrics.MetricsTest.testAttemptedCounterMetrics
 org.apache.beam.sdk.transforms.windowing.WindowTest.testMergingCustomWindows
 
org.apache.beam.sdk.transforms.windowing.WindowTest.testMergingCustomWindowsKeyedCollection


> Dataflow ValidatesRunner broken at HEAD
> ---------------------------------------
>
>                 Key: BEAM-2734
>                 URL: https://issues.apache.org/jira/browse/BEAM-2734
>             Project: Beam
>          Issue Type: Bug
>          Components: runner-dataflow
>            Reporter: Eugene Kirpichov
>            Assignee: Thomas Groh
>            Priority: Critical
>
> https://builds.apache.org/view/Beam/job/beam_PostCommit_Java_ValidatesRunner_Dataflow/
> Last stable build (#3567), 23 days ago
> We need to fix this.
> One of the issues is something I broke recently myself: SplittableDoFnTest is 
> broken due to serialization incompatibility by 
> https://github.com/apache/beam/pull/3649 which slightly changed that 
> BoundedToUnboundedSourceAdapter class, which changed its serialVersionUID. I 
> don't know why this is so fragile - we should also fix the fact that it is so 
> fragile.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Reply via email to