This is a review request for the following bug fixes in the SCTP area. The changes for each fix are quite small, but affect similar files and tests, so I have built a webrev containing all four MQ patches together.
8232097: (sctp) SctpNet.init() results in java.lang.UnsatisfiedLinkError [1] This is a regression in JDK 13 resulting from some refactoring. It was not noticed since the SCTP test are not reliably runnable on all systems. The fix simply ensures that the classes referring to native methods now take charge of ensuring that the appropriate native libraries are loaded. 8232101: (sctp) Add minimal sanity tests for SCTP [2] This issues adds a trivial test to the jdk_net test group that would have caught the issue above, by simply trying to create a new SCTP channel. 8233845: (sctp) Relax assertion in sun.nio.ch.sctp.AssociationChange::association While testing, I noticed an incorrect firing assertion. The fix is to relax the assertion a little, in line with the spec. ( more details in JIRA ) 8233847: (sctp) Flx link-local IPv6 scope handling and test cleanup. [4] Again, while testing I noticed intermittent failures because of link-local addresses being used without the appropriate scope id. As well as some resource exhaustion issues because of unclosed sockets. Finally, the socket option tests were failing, rarely, because they incorrectly used the peer local addresses rather than the client’s view of the peer’s remote addresses. Webrev: https://cr.openjdk.java.net/~chegar/sctp_14/webrev.00/ With these changes, I’ve run the SCTP tests several thousands of times without failure. -Chris. [1] https://bugs.openjdk.java.net/browse/JDK-8232097 [2] https://bugs.openjdk.java.net/browse/JDK-8232101 [3] https://bugs.openjdk.java.net/browse/JDK-8233845 [4] https://bugs.openjdk.java.net/browse/JDK-8233847
