The unit test IdleTimeoutTest always fails when run with jcov-instrumented jars in our nightly test runs. The failure occurs with the RMI/JRMP connector but not with RMI/IIOP or JMXMP. This *might* be a flaw in the test, but I am suspicious of the idle timeout logic because we have had problems there before. So we should investigate this problem.

Here is part of the test output, showing the rmi protocol, which fails, and the iiop protocol, which passes.

[exec] Idle timeout test for protocol rmi
     [exec] Waiting for notifs
     [exec] Got notification: d:type=mlet,instance=0
     [exec] Got notification: d:type=mlet,instance=1
     [exec] Got notification: d:type=mlet,instance=2
     [exec] Got notification: d:type=mlet,instance=3
     [exec] Waited 571ms for notifs to arrive
     [exec] Requesting conn id
     [exec] Client id changed: <rmi: 1> -> <rmi: 2>
     [exec] Connection id list on server after client close: []
     [exec] Test for protocol rmi FAILED
     [exec] Idle timeout test for protocol iiop
     [exec] Waiting for notifs
     [exec] Got notification: d:type=mlet,instance=0
     [exec] Got notification: d:type=mlet,instance=1
     [exec] Got notification: d:type=mlet,instance=2
     [exec] Got notification: d:type=mlet,instance=3
     [exec] Waited 989ms for notifs to arrive
     [exec] Requesting conn id
     [exec] Removing notification listener
     [exec] Waiting for id list to drop ours
     [exec] Client id changed as expected: <iiop: 3> -> <iiop: 4>
     [exec] Connection id list on server after client close: []
     [exec] Test passed for iiop