@singlis Thanks Scott for pointing out the test failures. I have address them by either revising the test or removing it due to the test no longer making sense. More specifically, REEF.NET assumed that if a task failed then the Evaluator should crash. This, however, does not happen on the Java side. In earlier conversations, I pointed out this semantic mismatch, and decided to correct it in the new bridge. As such, tests on the old bridge that assumed the incorrect semantics have been either removed or revised to match the correct semantics.
At this point, I believe tests that fail either do so in main or pass on my location machine: one instance of this is: ``` Org.Apache.REEF.Tests.Functional.IMRU.TestFailMapperEvaluatorsOnInit.TestFailedMapperOnLocalRuntime [FAIL] ``` Which passes on my local machine but seems to fail on AppVeyor because in can't read the log: ``` Cannot read from log file C:\projects\reef\lang\cs\bin\.netcore\Release\Org.Apache.REEF.Tests\net461\REEF_LOCAL_RUNTIME7253ce0e\reef-BroadcastReduceDriver-20180907140304298\driver\driver.stdout ``` Moving forward, it is my understanding that all of these tests will go away when the new bridge is in place, which is why I ported @motus tests over to the new bridge. That will resolve the issue with the tests that fail in main. Thank you again for your extremely helpful feedback! [ Full content available at: https://github.com/apache/reef/pull/1477 ] This message was relayed via gitbox.apache.org for [email protected]
