RFR: 8291429: java/lang/Thread/virtual/ThreadAPI.java timed out on single core system

Alan Bateman alanb at openjdk.org
Wed Oct 5 19:36:26 UTC 2022


On Wed, 5 Oct 2022 19:02:04 GMT, Chris Plummer <cjplummer at openjdk.org> wrote:

>> This is a test only change for two tests for virtual threads that hang/timeout on single core systems. The two tests involve pinning and require at least two carrier threads. The test lib used by these tests is updated to define a new method that ensures parallelism is at least a given value and both tests are updated to use this. There are a number of tests in the debugger area that may potentially use this in the future.
>
> test/jdk/java/lang/management/ThreadMXBean/VirtualThreadDeadlocks.java line 30:
> 
>> 28:  *   platform and virtual threads in deadlock
>> 29:  * @enablePreview
>> 30:  * @modules java.base/java.lang:+open java.management
> 
> Can you explain the need for this change?

That is jtreg foo to open java.lang, needed to access a private field.

> test/jdk/java/lang/management/ThreadMXBean/VirtualThreadDeadlocks.java line 68:
> 
>> 66:     public static void main(String[] args) throws Exception {
>> 67:         // need at least two carrier threads due to pinning
>> 68:         VThreadRunner.ensureParallelism(2);
> 
> In this test case why is there no need to maintain a reference to the returned AutoCloseable? Isn't there a chance it can be collected and the old parallelism value restored while the test is running.

This test doesn't need to restore it so it doesn't keep a reference and not an issue if it is GC'ed.

> test/lib/jdk/test/lib/thread/VThreadRunner.java line 176:
> 
>> 174:                     if (!closed) {
>> 175:                         closed = true;
>> 176:                         pool.setParallelism(parallelism);
> 
> What is the rationale for restoring the parallelism? It's just a test. Is this really necessary? Are we reusing the JVM to run other tests, and even if we are does it matter?

Look at the ThreadAPI test as an example. It's a TestNG with dozens of test methods. All but one can run on single core systems.

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

PR: https://git.openjdk.org/jdk/pull/10562


More information about the core-libs-dev mailing list