On Fri, 26 Aug 2022 19:09:45 GMT, Roger Riggs <rri...@openjdk.org> wrote:
> CleanerTest is failing intermittently on Aarch64, with -Xcomp, and using a > VirtualThread for the Cleaner in the test. > The extensively relies on references processing and invokes GC on a very > short cycle. > The existing 10ms delay between GC requests is too short for the changes to > reference states to be reliably visible. > > The delay between gc requests is extended to 200ms and the tests pass > reliably. > There are small improvements in diagnostic messages and cleanup of unused > imports. Hello Roger, The change related to increasing the time to 200ms looks fine to me and even the typo fix. The prefix to the error message isn't too clear to me. Is the "Should (not) have been run" prefix implying that the "Cleaner should (not) have been run"? While in this test, do you think we should change it to use `@enablePreview` instead of the `--enable-preview` options to the test definition section? ------------- PR: https://git.openjdk.org/jdk/pull/10048
