RFR: 8271195: Use largest available large page size smaller than LargePageSizeInBytes when available [v11]

[Albert Mingkun Yang]

On Mon, 7 Mar 2022 12:08:48 GMT, Thomas Schatzl <tschatzl at openjdk.org> wrote:

> The test is not able to distinguish failures from problems with the VM logic and other processes spuriously allocating enough large pages to make this test fail afaict.

I don't see any easy way fixing this and including this test in CI can cause undesirable false-positives.

> also created a minimal test case as per @albertnetymk suggestion to test a point scenario.

I personally like the minimal test case you posted above, as it makes debugging easier, but it still suffers from the same false-positive problem. It's probably best to study/discuss that in its own ticket/PR.

In summary, I believe it's very unlikely that someone accidentally breaks the large-page fall-back logic this test tries to verify. Even if the large-page fall-back logic were broken somehow, it's trivially reproducible and should be simple to fix as well. Therefore, the benefit of this test is questionable. OTOH, the cost of maintaining this test (e.g. dealing with false positives) is very real.

Given such low benefit-cost ratio, I suggest dropping the test in this PR.

-------------

PR: https://git.openjdk.java.net/jdk/pull/7326