Integrated: 8306446: java/lang/management/ThreadMXBean/Locks.java transient failures
Kevin Walls
kevinw at openjdk.org
Fri Jul 28 09:49:28 UTC 2023
On Thu, 15 Jun 2023 20:34:53 GMT, Kevin Walls <kevinw at openjdk.org> wrote:
> This test iterates an array of ThreadInfos in a few places (e.g. in the method doCheck()), and needs to tolerate and ignore nulls, in case a thread finishes and the test hits an NPE.
>
> There are other calls like "TM.getThreadInfo(tid).getLockName()" which might often be risky, but if the threads are blocked as they are here, they can't be terminating, so this usage is safe.
>
>
> The test has additional problems when started in a virtual thread. ThreadMXBean.getThreadInfo() methods only return a ThreadInfo for platform threads. The test needs to avoid some checks if mainThread is virtual.
>
> In assertNoLock, it needs to not object to a thread holding a lock on a VirtualThread object is not relevant.
> Also the loop in doChecks which follows a chain of locks... This needs to recognise that ForkJoinPool thead is not worth pursuing. It's not one of the very narrow set of threads this test cares about.
>
> Despite these exclusions, the test does some reasonable verification work when MainThread is virtual. This test historically cam in with a general "JVM monitoring and management API" change, it is not testing a particular fix.
>
>
> There's a failure condition in doCheck() which will not make the test fail: if it logs "TEST FAILED" in its final for loop, there is no failure. Make the loop count the failures, and throw if there are any.
>
> Also, while looking into this... The variable names in some methods are confusing. In checkBlockedObject(), let's use "threadName" rather than "result" if we are finding a thread name, and let's not reuse the same result variable for a lockName later in the method.
>
> The logs from this test are hard to read and verify, I find it better if the lock objects OBJB and OBJC are of classes other than Object, so you get to read, e.g.:
> LockAThread blocked on Locks$ObjectB at 4691fdfd
> (ObjectB, not just Object).
This pull request has now been integrated.
Changeset: 4ae5a3e3
Author: Kevin Walls <kevinw at openjdk.org>
URL: https://git.openjdk.org/jdk/commit/4ae5a3e39b681bfd001df1483d8a6d1fce0bc7f8
Stats: 54 lines in 1 file changed: 31 ins; 1 del; 22 mod
8306446: java/lang/management/ThreadMXBean/Locks.java transient failures
Reviewed-by: cjplummer, sspitsyn
-------------
PR: https://git.openjdk.org/jdk/pull/14501
More information about the serviceability-dev
mailing list