I think I may have figured this out. I reopened and added a comment to
POOL-411.
Phil
On Thu, Jul 6, 2023 at 2:41 PM Phil Steitz <phil.ste...@gmail.com> wrote:
> <snip>
>
> I guess it's good news that CI hit the error below when reviewing the PR
> that I had prepared for the POOL-391 fixes. I only saw it once in many
> test runs and only on OpenJDK 20.0.1. Looks like CI is running 17 on
> azure-linux. I am pretty sure it has nothing to do with the changes in the
> PR, partly because I saw it once running the first RC code.
>
> This is very strange and troubling. More eyeballs most welcome. Here is
> the problem:
>
> The NPE below happens inside GKOP addIdleObject, which is called by
> addObject. The code for addObject follows the standard pattern:
> register(key);
> try {
> addIdleObject(key, create(key));
> } finally {
> deregister(key);
> }
>
> So addIdleObject is called while the owning thread has the key
> "registered." To register is key is basically sayinig you are interested
> in / about to do something to its associated pool. There is a
> numInterested counter (attached to the keyed pool) that gets incremented
> when a thread registers a key and decrremented when the key is
> deregistered. Registration creates a pool and iniitializes its counter if
> there is no pool under the given key. When pools are deregistered, the
> counter is checked and if the pool has no instances under management and
> the counter is zero, the pool is removed. In this case, the registration
> in addObject should prevent the pool being removed before or during
> execution of addIdleObject, but the NPE means it has been removed. Somehow
> numInterested is getting corrupted.
>
> I will keep trying to get this to happen and see if I can find scenarios
> where deregister is somehow called twice for one register. Any suggestions
> or success getting this to happen with the code now in master most
> appreciated.
>
> Phil
>
>> 2. On MacOS 13.4.1, OpenJDK 20.0.1, I got the following test failure just
>> one time and can't reproduce:
>>
>> java.util.concurrent.ExecutionException: java.lang.NullPointerException:
>> Cannot invoke
>> "org.apache.commons.pool2.impl.GenericKeyedObjectPool$ObjectDeque.getIdleObjects()"
>> because the return value of "java.util.Map.get(Object)" is null
>>
>> at java.base/java.util.concurrent.FutureTask.report(FutureTask.java:122)
>> at java.base/java.util.concurrent.FutureTask.get(FutureTask.java:191)
>> at
>> org.apache.commons.pool2.impl.TestGenericKeyedObjectPool.lambda$testConcurrentBorrowAndClear$2(TestGenericKeyedObjectPool.java:1056)
>> ... 71 more
>> Caused by: java.lang.NullPointerException: Cannot invoke
>> "org.apache.commons.pool2.impl.GenericKeyedObjectPool$ObjectDeque.getIdleObjects()"
>> because the return value of "java.util.Map.get(Object)" is null
>> at
>> org.apache.commons.pool2.impl.GenericKeyedObjectPool.addIdleObject(GenericKeyedObjectPool.java:307)
>> at
>> org.apache.commons.pool2.impl.GenericKeyedObjectPool.addObject(GenericKeyedObjectPool.java:332)
>> at
>> org.apache.commons.pool2.KeyedObjectPool.addObjects(KeyedObjectPool.java:136)
>> at
>> org.apache.commons.pool2.KeyedObjectPool.addObjects(KeyedObjectPool.java:113)
>> at
>> org.apache.commons.pool2.impl.TestGenericKeyedObjectPool.lambda$testConcurrentBorrowAndClear$0(TestGenericKeyedObjectPool.java:1036)
>> at
>> java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:577)
>> at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:317)
>> at
>> java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1144)
>> at
>> java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:642)
>> at java.base/java.lang.Thread.run(Thread.java:1623)
>>
>>
>>
>>
