On Tue, 5 Dec 2023 09:08:59 GMT, Andrew Haley <a...@openjdk.org> wrote:
>> We've seen some rare failures of the CLQ Whitebox test on "less-strong" >> architectures, and the only thing which -- given my research -- could be the >> culprit is spuriously failing weakCAS (which is correct in terms of the >> implementation of CLQ). >> >> After discussion with @DougLea, it was decided as the CLQ implementation >> does not guarantee what the failing test tests, and modifying the test would >> mean that it would generally not be able to enforce anything, the test is >> invalid and should be removed -- hence this PR. > > Few AArch64 HotSpot systems implement weak CAS as anything other than plain > CAS. In order to get to the root cause of this problem, it would help to know > on which AArch64 hardware this test failed. > @theRealAph I think the problem in this case was that the whitebox test in > this case relied on a presumption that was only true for stronger consistency > architectures, and rewriting the test would essentially be asserting that "a > lot of permutations are valid, and only internally observable" which is a > low-value test. Oh right, so nothing to do with weak CAS, then. Fair enough. ------------- PR Comment: https://git.openjdk.org/jdk/pull/16786#issuecomment-1842584686
