Stephan Ewen created FLINK-1203: ----------------------------------- Summary: ClassCast exceptions in parallel tests (Surefire Bug?) Key: FLINK-1203 URL: https://issues.apache.org/jira/browse/FLINK-1203 Project: Flink Issue Type: Bug Components: Build System Affects Versions: 0.8-incubating Reporter: Stephan Ewen Fix For: 0.8-incubating
I am frequently seeing weird (non-deterministic) class cast exception in the tests, where apparently casts fail in serial parts of the program. The casting attempts have nothing to do with the program context. {code} java.lang.ClassCastException: org.apache.flink.api.common.operators.base.DeltaIterationBase cannot be cast to org.apache.flink.api.common.operators.base.GroupReduceOperatorBase at org.apache.flink.api.scala.operators.translation.DistinctTranslationTest.testCombinable(DistinctTranslationTest.scala:39) {code} I am wondering whether that might be a strange bug in the forked execution, a possible bug in maven surefire. I propose to deactivate the "reuseFork" option in surefire, to create a clean JVM (and class loaders) for each test. -- This message was sent by Atlassian JIRA (v6.3.4#6332)