On Tue, 30 Sep 2025 11:37:37 GMT, Jaikiran Pai <[email protected]> wrote:
> Can I please get a review of this change to a jtreg failure handler > configured in the JDK? > > The change proposes to generate a thread dump much sooner than previously > whenever a test times out. This should thus capture a much more accurate > state of the test process when the test is considered timed out. > > Due to the recent changes in the default timeout factor, we have noticed some > tests which timeout and the jtreg failure handler actions start execution. > While those are being executed the test sometimes completes. So by the time > the "jstack" failure handler action is executed (can be several seconds > later), the test's state will no longer be accurate. > > The change here generates a thread dump using jstack as the first action in > the set of failure handler actions. It does it only once and then moves to > the rest of the actions, one of which subsequent "jstack" which generates > thread dumps more than once. > > I have verified that this change works as expected when a test times out. The > action is named "thread_dump" instead of just reusing the "jstack" name > because the current HTML rendering of the processes.html runs into trouble if > there are more than one action with the same name. > > I wanted to reorder some of the other commands in that set, but it causes > some trouble in the rendering of the HTML and would require some changes to > that part. So I decided to keep this simple and have it done sooner to help > investigating timeout failures in our CI. Marked as reviewed by erikj (Reviewer). ------------- PR Review: https://git.openjdk.org/jdk/pull/27574#pullrequestreview-3284673134
