On Tue, 17 Nov 2026 16:07:00 GMT, Ashay Rane <[email protected]> wrote:
>> Just to be clear this is only necessary because you are forcing >> `UseBranchProtection` on for the GHA tests - right? The normal expectation >> is that such a flag is only set when the person launching the VM knows the >> hardware supports it - hence the warning when you ask for it but the machine >> doesn't have it. Changing the test as part of this PR seems reasonable to me. >> >>> In the process, I discovered a couple of tests (DataDumpDcmdTest.java and >>> LoadAgentDcmdTest.java) that can be improved to use OutputAnalyzer's >>> stderrShouldBeEmptyIgnoreVMWarnings(). >> >> Those tests don't run in the GHA tier1 testing though do they? We are >> cautious with accepting warnings in more sophisticated tests because it >> could hide actual test configuration issues. > >> Just to be clear this is only necessary because you are forcing >> UseBranchProtection on for the GHA tests - right? > > Yes, IO.java needs changes because we could be asking for branch protection > on a machine that doesn't support it and because this patch adds that option > in our GitHub Actions. > >> Those tests don't run in the GHA tier1 testing though do they? > > Actually, they do. IO.java is run as part of the jdk/tier1 part 1 tests (see > [log](https://github.com/raneashay/openjdk-jdk/actions/runs/44025777656/job/133879065428#step:10:2931) > for an example) and both DataDumpDcmdTest.java and LoadAgentDcmdTest.java > tests are run as part of HotSpot tier 1 serviceability tests (see > [log](https://github.com/raneashay/openjdk-jdk/actions/runs/44025777656/job/133879065436#step:10:879)). Okay. I thought the Dcmd tests ran in tier3 and higher. I'm reluctant to ignore all warnings in these tests. But I'm also reluctant to add tolerance of this specific warning to those tests as well. I don't see any neat solution here. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/33595#discussion_r4608950191
