On Mon, 8 Jun 2026 16:05:51 GMT, Alexander Zvegintsev <[email protected]> wrote:
>> I must admit that I don't quite understand why this would be safer. My >> assumption is that if the latch is never counted down, something has already >> gone seriously wrong, and we would end up hitting the jtreg timeout >> regardless. But I may be missing some aspect of the reasoning here. > > Yes, if the latch is never counted down, then something is already wrong, but > my concern is with the failure mode: > > An unbounded `await` causes the test to hang, relying on the global jtreg > timeout for diagnosis. When run outside of jtreg directly(`java > test/jdk/javax/swing/SwingWorker/TestDoneBeforeDoInBackground.java`), the > test simply hangs indefinitely. > > A bounded `await` would fail at the specific point of failure with a clear > message (e.g., "worker did not start in time"). This also ensures consistency > with the existing bounded `await` for `doneLatch`. > > Overall, this approach is easier to triage, faster in CI, and more robust > when running tests under different timeout configurations. Thanks a lot for the explanation - I understand now and fully agree. I added a timeout with a failure message using a timeout value that is slightly higher than it was before and adjusted by the timeout factor of jtreg. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/31348#discussion_r3380477546
