[ https://issues.jboss.org/browse/JBWS-4146?page=com.atlassian.jira.plugin.... ] r searls commented on JBWS-4146: -------------------------------- There is some randomness to this failure. There is no consistent iteration of the testing loop in which a test fails with this error. However the location of the failure is consistent. It will always fail in test testSetConfigProperties line 140, {code:java} 139 cfg.setConfigProperties(port, clientConfig); 140 141 assertEquals(null, client.getEndpoint().get("A")); {code} test testAddInterceptorsThroughSetMapOfProperties line 339, {code:java} 338 InterceptorUtils.addInterceptors(client, properties); 339 340 assertEquals("1", client.getEndpoint().get("A")); {code} test testAddInterceptorsThroughSetConfigProperties line 469. {code:java} 468 cfg.setConfigProperties(port, clientConfig); 469 470 assertEquals(null, client.getEndpoint().get("A")); {code} In each of these tests there are previous calls to client.getEndpoint() that succeed. The initial client created in each test is used in all calls to getEndpoint within a test. Method getEndpoint is in class org.apache.cxf.endpoint.AbstractConduitSelector. Variable, endpoint is a protected field in this class. It is only accessed through its getter and setter methods and it is not otherwise altered within the class. I have verified method setEndpoint never receives a NULL input value and it is always called with a valid value before method getEndpoint is called. I have confirmed that the instance of AbstractConduitSelector is not replaced during the run of a test. The class is running in the same thread for calls to setEndpoint and getEndpoint. This appears to be an IBM jdk-1.8.0_201 issue. -- This message was sent by Atlassian Jira (v7.12.1#712002)