Good call. Adding a @NotThreadSafe annotation to the test class seems to solve the problem. The setup and teardown of CassandraUnit is still shared between all the tests and each test (after setup) runs rather quickly so I don't think running the tests serially is a problem. I've never seen the error you're seeing before. A full stack trace might help.
-- Michael Mior mm...@apache.org Le mer. 25 juil. 2018 à 18:16, Andrei Sereda <and...@sereda.cc> a écrit : > Did you run the tests from IDE or command line (maven) ? Do you get some > behaviour in both ? > > Integration tests (-Pit) run by maven (surefire) are executed in parallel > (see pom.xml <https://github.com/apache/calcite/blob/master/pom.xml#L876 > >). > This may contribute to non-determinism (eg. sharing some static variable > like Hook). > > I will try to reproduce it myself to better understand what's happening > with cassandra adapter. Still getting NPE in maven for some reason. > > On Wed, Jul 25, 2018 at 4:53 PM Michael Mior <mm...@apache.org> wrote: > > > This is unrelated to that branch (except that I don't want to merge the > > branch with CassandraUnit until tests pass). > > > > -- > > Michael Mior > > mm...@apache.org > > > > > > > > Le mer. 25 juil. 2018 à 16:36, Andrei Sereda <and...@sereda.cc> a écrit > : > > > > > Mike, > > > > > > Is your branch michaelmior:2419-cassandra-unit up to date ? I'm getting > > > NPE: > > > > > > > > > > Caused by: java.lang.NullPointerException > > > > at > > > > > > > > > > org.apache.cassandra.config.DatabaseDescriptor.getDiskFailurePolicy(DatabaseDescriptor.java:1881) > > > > at > > > > > > > > > > org.apache.cassandra.utils.JVMStabilityInspector.inspectThrowable(JVMStabilityInspector.java:82) > > > > at org.apache.cassandra.io.util.FileUtils.<clinit>(FileUtils.java:79) > > > > ... 13 more > > > > > > > > > What error are you getting ? > > > > > > Regards, > > > Andrei. > > > > > > On Wed, Jul 25, 2018 at 1:29 PM Michael Mior <mm...@apache.org> wrote: > > > > > > > I'd appreciate some help with the JIRA below. I'm using Hook.TRIMMED > > > > to expose materialized views to the adapter and a test which checks > > > > that the view can be used for query rewriting is failing about 50% of > > > > the time. I don't see why there should be any non-determinism here so > > > > any ideas appreciated :) > > > > > > > > https://issues.apache.org/jira/browse/CALCITE-2423 > > > > > > > > -- > > > > Michael Mior > > > > mm...@apache.org > > > > > > > > > >
