RFR: 8274597: [TESTBUG] Two of the dnd tests times out and fails

[Sergey Bylokhov]

On Fri, 1 Oct 2021 19:19:47 GMT, Sergey Bylokhov <serb at openjdk.org> wrote:

>> @mrserb I can see three non-daemon threads in 'WAITING' state(given below). So looks like 'AWT-Shutdown' thread created by AWTAutoShutdown.java is the hook which is getting blocked. But I think this is an expected behaviour only as it will block this thread until all AWT tasks are complete. 
>> 
>> 
>> "main" #1 prio=5 os_prio=0 cpu=78.13ms elapsed=240.13s tid=0x0000017129532000 nid=0x1518 in Object.wait()  [0x0000001fb8ffe000]
>>    java.lang.Thread.State: WAITING (on object monitor)
>> 	at java.lang.Object.wait(java.base at 11.0.14-ea/Native Method)
>> 	- waiting on <0x00000000ff504d78> (a java.lang.Thread)
>> 	at java.lang.Thread.join(java.base at 11.0.14-ea/Thread.java:1305)
>> 	- waiting to re-lock in wait() <0x00000000ff504d78> (a java.lang.Thread)
>> 	at java.lang.Thread.join(java.base at 11.0.14-ea/Thread.java:1379)
>> 	at java.lang.ApplicationShutdownHooks.runHooks(java.base at 11.0.14-ea/ApplicationShutdownHooks.java:107)
>> 	at java.lang.ApplicationShutdownHooks$1.run(java.base at 11.0.14-ea/ApplicationShutdownHooks.java:46)
>> 	at java.lang.Shutdown.runHooks(java.base at 11.0.14-ea/Shutdown.java:130)
>> 	at java.lang.Shutdown.exit(java.base at 11.0.14-ea/Shutdown.java:174)
>> 	- locked <0x00000000ff50a4e0> (a java.lang.Class for java.lang.Shutdown)
>> 	at java.lang.Runtime.exit(java.base at 11.0.14-ea/Runtime.java:113)
>> 	at java.lang.System.exit(java.base at 11.0.14-ea/System.java:1750)
>> 	at com.sun.javatest.regtest.agent.AStatus.exit(AStatus.java:199)
>> 	at com.sun.javatest.regtest.agent.MainWrapper.main(MainWrapper.java:84)
>> 
>> 
>> "AWT-Shutdown" #24 prio=5 os_prio=0 cpu=0.00ms elapsed=239.97s tid=0x000001714cd43800 nid=0x29c4 in Object.wait()  [0x0000001fba6fe000]
>>    java.lang.Thread.State: WAITING (on object monitor)
>> 	at java.lang.Object.wait(java.base at 11.0.14-ea/Native Method)
>> 	- waiting on <0x00000000ff69b6c0> (a java.lang.Object)
>> 	at java.lang.Object.wait(java.base at 11.0.14-ea/Object.java:328)
>> 	at sun.awt.AWTAutoShutdown.run(java.desktop at 11.0.14-ea/AWTAutoShutdown.java:291)
>> 	- waiting to re-lock in wait() <0x00000000ff69b6c0> (a java.lang.Object)
>> 	at java.lang.Thread.run(java.base at 11.0.14-ea/Thread.java:834)
>> 
>> 
>> 
>> "AWT-EventQueue-0" #27 prio=6 os_prio=0 cpu=109.38ms elapsed=239.96s tid=0x000001714d006800 nid=0x1f70 waiting on condition  [0x0000001fba8fe000]
>>    java.lang.Thread.State: WAITING (parking)
>> 	at jdk.internal.misc.Unsafe.park(java.base at 11.0.14-ea/Native Method)
>> 	- parking to wait for  <0x00000000ff6eda50> (a java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject)
>> 	at java.util.concurrent.locks.LockSupport.park(java.base at 11.0.14-ea/LockSupport.java:194)
>> 	at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(java.base at 11.0.14-ea/AbstractQueuedSynchronizer.java:2081)
>> 	at java.awt.EventQueue.getNextEvent(java.desktop at 11.0.14-ea/EventQueue.java:566)
>> 	at java.awt.EventDispatchThread.pumpOneEventForFilters(java.desktop at 11.0.14-ea/EventDispatchThread.java:190)
>> 	at java.awt.EventDispatchThread.pumpEventsForFilter(java.desktop at 11.0.14-ea/EventDispatchThread.java:124)
>> 	at java.awt.EventDispatchThread.pumpEventsForHierarchy(java.desktop at 11.0.14-ea/EventDispatchThread.java:113)
>> 	at java.awt.EventDispatchThread.pumpEvents(java.desktop at 11.0.14-ea/EventDispatchThread.java:109)
>> 	at java.awt.EventDispatchThread.pumpEvents(java.desktop at 11.0.14-ea/EventDispatchThread.java:101)
>> 	at java.awt.EventDispatchThread.run(java.desktop at 11.0.14-ea/EventDispatchThread.java:90)
>
>>  I can see three non-daemon threads in 'WAITING' state(given below). So looks like 'AWT-Shutdown' thread created by AWTAutoShutdown.java is the hook which is getting blocked. But I think this is an expected behaviour only as it will block this thread until all AWT tasks are complete.
> 
> This issue is unrelated to the non-daemon threads since the System.exit() is called, it should exit the whole JVM. But before exit it executes some number of registered ShutdownHook, we need to check what ShutdownHook is executed and why it hangs. The AWT-Shutdown is not a hook, it is a thread which prevents vm exit if there are some UI peers, but it does ont prevent the System.exit().

> @mrserb I can see only two shutdown hooks registered, Thread[ToolkitShutdown,6,system] and Thread[ScreenUpdater,6,system]. But I don't know which one of this is getting hanged, any idea how can I check it?

You can add logging to each of them and check, or you can debug the native code.

>Also if we put a Thread.sleep(100) or just a System.out.println() at the end of the main(), its not hanging and the tests are passing.

maybe the test found a deadlock in the product and you just workaround it?

>So for the time being in order to stabilise the test run on mach5 systems and also considering the fact that this issue is currently reproducible only with Windows 11 systems, can I go ahead and fix the tests?.

The algorithm to use in such cases is something like this:
 - If the bug is in the test, then fix the test
 - If the bug is in the product then fix the product
 - If the root cause is unknown then add test to the problem list
 - Continue to work on the bug
 - Find the root cause
 - Fix the test/product and remove the test from the problem list

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

PR: https://git.openjdk.java.net/jdk/pull/5777