On Fri, 3 Sep 2021 20:32:20 GMT, Mark Sheppard <mshep...@openjdk.org> wrote:
> A number of nio DatagramChannel tests are intermittently failing on > macosx-aarch64. > In some instances this is a receive call blocking indefinitely waiting on > data which has > already been sent, and should be available immediately to the receive method > call. > Other test failure scenarios are problems during the test compilation phase > with a SocketException being thrown and the message: > "test result: Error. Agent communication error: java.net.SocketException: No > buffer space available; check console log for any additional details" > > The ManySourcesAndTargets and other tests execute in agentvm mode. This > results in certain test diagnostic > Output being lost during the test failure handling capture process. To > mitigate this lost diagnostics, the > ManySourcesAndTargets test has been amended to execute in othervm mode. > > Additionally, to assist in the buffer allocation issue, the netstat command > executed by the test > failure_handler has an extra argument added to obtain additional details on > mbuf usage. > The failure handler will now execute with netstat -mm Marked as reviewed by alanb (Reviewer). test/jdk/java/nio/channels/DatagramChannel/ManySourcesAndTargets.java line 28: > 26: * @summary Test DatagramChannel send/receive and that receive returns > the expected > 27: * sender address > 28: * @run main/othervm ManySourcesAndTargets This change looks okay. Do you mind change L143 to use sender.send rather than reader.send while you are there? That will avoid at least some questions while trying to track down the underlying issue and will ensure that the test is printing out the actual sender address. ------------- PR: https://git.openjdk.java.net/jdk/pull/5366
