On Tue, 23 Feb 2021 10:16:26 GMT, Daniel Fuchs <dfu...@openjdk.org> wrote:
>> Hi, >> >> I've removed the problematic "process" handling logic and have stripped the >> test back to only use threads. >> >> The problem was reproducible (intermittently on my Windows machine) using >> this smaller test and should make the test more stable. > > Does the new version of the test still occasionally catches the issue and > fails if the fix is not present? Hi, locally this test reproduces the issue on my Windows machine. I believe our internal testing Windows boxes are too powerful and handle the threading too efficiently to reproduce this error. I can reproduce locally approx 10% of the time, but after over 100 runs I cannot reproduce as part of our internal testing. As this is not verifiable in a completely deterministic way, I believe having the test as a stable, passing test is a more appropriate approach instead of selecting this as no-reg hard. It adds more test coverage to the code also. ------------- PR: https://git.openjdk.java.net/jdk/pull/2572
