On Fri, 1 Sep 2023 06:57:36 GMT, Serguei Spitsyn <sspit...@openjdk.org> wrote:
> The fix looks okay to me but it'd be nice to find out why the > `exclusiveAccess.dirs` were set for these tests initially. I'm not sure, > anyone remembers it though. Yeah. I cannot find a good discussion on these tests. I suspect it is mostly historical, from the times where machines were small and resource-constrained, so that a stress test made life hard for any other concurrently running test. Today, that is not the case. We have been doing e.g. GC stress tests in tier3, which can hog CPUs for minutes, without ill effects. Anyway, this is similar to other parallelism improvements we did over the previous years, e.g. https://bugs.openjdk.org/browse/JDK-8273433, https://bugs.openjdk.org/browse/JDK-8273438. Would you like to run this PR in your test systems to see if this particular relaxation is problematic? ------------- PR Comment: https://git.openjdk.org/jdk/pull/15506#issuecomment-1704958195
