RFR: 8298081: DiagnoseSyncOnValueBasedClasses doesn't report useful information for virtual threads [v2]

[Patricio Chilano Mateo]

On Thu, 15 Dec 2022 05:15:27 GMT, David Holmes <dholmes at openjdk.org> wrote:

>> The mechanism used by DiagnoseSyncOnValueBasedClasses to show where the sync occurs only reports platform/carrier thread stack details. That is no use if the code is being executed by a virtual thread. We can use the code being introduced by [JDK-8292674](https://bugs.openjdk.org/browse/JDK-8292674) for JNI error/warning reporting to show the virtual thread stack instead.
>> 
>> Before the fix the logging shows:
>> 
>> [0.173s][info][valuebasedclasses] Synchronizing on object 0x00000007ff880e68 of klass java.lang.Character
>> [0.173s][info][valuebasedclasses] at jdk.internal.vm.Continuation.run([java.base at 21-internal](mailto:java.base at 21-internal)/Continuation.java:257)
>> [0.173s][info][valuebasedclasses] at java.lang.VirtualThread.runContinuation([java.base at 21-internal](mailto:java.base at 21-internal)/VirtualThread.java:216)
>> [0.173s][info][valuebasedclasses] at java.lang.VirtualThread$$Lambda$8/0x0000000801046f70.run([java.base at 21-internal](mailto:java.base at 21-internal)/Unknown Source)
>> [0.173s][info][valuebasedclasses] at java.util.concurrent.ForkJoinTask$RunnableExecuteAction.exec([java.base at 21-internal](mailto:java.base at 21-internal)/ForkJoinTask.java:1423)
>> [0.173s][info][valuebasedclasses] at java.util.concurrent.ForkJoinTask.doExec([java.base at 21-internal](mailto:java.base at 21-internal)/ForkJoinTask.java:387)
>> [0.173s][info][valuebasedclasses] at java.util.concurrent.ForkJoinPool$WorkQueue.topLevelExec([java.base at 21-internal](mailto:java.base at 21-internal)/ForkJoinPool.java:1312)
>> [0.173s][info][valuebasedclasses] at java.util.concurrent.ForkJoinPool.scan([java.base at 21-internal](mailto:java.base at 21-internal)/ForkJoinPool.java:1843)
>> [0.173s][info][valuebasedclasses] at java.util.concurrent.ForkJoinPool.runWorker([java.base at 21-internal](mailto:java.base at 21-internal)/ForkJoinPool.java:1808)
>> [0.173s][info][valuebasedclasses] at java.util.concurrent.ForkJoinWorkerThread.run([java.base at 21-internal](mailto:java.base at 21-internal)/ForkJoinWorkerThread.java:188)
>> 
>> after the fix:
>> 
>> [0.177s][info][valuebasedclasses] Synchronizing on object 0x00000007ff880e68 of klass java.lang.Character
>> [0.177s][info][valuebasedclasses] at SyncOnValueBasedClassTest$VTTest.lambda$main$0(SyncOnValueBasedClassTest.java:195)
>> [0.177s][info][valuebasedclasses] - locked <0x00000007ff880e68> (a java.lang.Character)
>> [0.177s][info][valuebasedclasses] at SyncOnValueBasedClassTest$VTTest$$Lambda$1/0x0000000801000c10.run(Unknown Source)
>> [0.177s][info][valuebasedclasses] at java.lang.VirtualThread.runWith([java.base at 21-internal](mailto:java.base at 21-internal)/VirtualThread.java:335)
>> [0.177s][info][valuebasedclasses] at java.lang.VirtualThread.run([java.base at 21-internal](mailto:java.base at 21-internal)/VirtualThread.java:305)
>> [0.177s][info][valuebasedclasses] at java.lang.VirtualThread$VThreadContinuation.lambda$new$0([java.base at 21-internal](mailto:java.base at 21-internal)/VirtualThread.java:177)
>> [0.177s][info][valuebasedclasses] at java.lang.VirtualThread$VThreadContinuation$$Lambda$7/0x0000000801046d48.run([java.base at 21-internal](mailto:java.base at 21-internal)/Unknown Source)
>> [0.177s][info][valuebasedclasses] at jdk.internal.vm.Continuation.enter0([java.base at 21-internal](mailto:java.base at 21-internal)/Continuation.java:327)
>> [0.177s][info][valuebasedclasses] at jdk.internal.vm.Continuation.enter([java.base at 21-internal](mailto:java.base at 21-internal)/Continuation.java:320)
>> 
>> 
>> The existing test for `DiagnoseSyncOnValueBasedClasses` is adapted to provide a basic sanity test using virtual threads, which suffices to test the stack information. Unfortunately the `--enable-preview` has a viral impact.
>> 
>> Other testing: tiers 1-3 sanity check
>> 
>> Thanks.
>
> David Holmes has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Move dispatch logic into new JavaThread method

Looks good.

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

Marked as reviewed by pchilanomate (Reviewer).

PR: https://git.openjdk.org/jdk/pull/11641