On Tue, 27 Oct 2026 11:06:20 GMT, Viktor Klang <[email protected]> wrote:
>> test/jdk/java/util/concurrent/forkjoin/AsyncShutdownNowInvokeAnyRace.java
>> line 114:
>>
>>> 112: Future<Void> shutdownResult =
>>> CompletableFuture.runAsync(shutdown);
>>> 113:
>>> 114: try (pool) { // Ensure that the pool is terminated after
>>> each iteration
>>
>> Replacing
>>
>> try {
>> try {
>> invokeAny();
>> } catch (...) {
>> ...
>> }
>> } finally {
>> pool.shutdown();
>> }
>>
>> with
>>
>> try (pool) {
>> invokeAny();
>> } catch (...) {
>> ...
>> }
>>
>> has a semantic difference: earlier, `catch` clauses were kicking in right
>> after an `invokeAny()` failure; now, `catch` clauses kick in after
>> `pool.close()`, which might block due to stale workers, and consequently
>> short-circuit catch clauses, e.g., `catch (TimeoutException ...)` might get
>> skipped.
>>
>> @viktorklang-ora, is this change intentional or harmless?
>
>>and consequently short-circuit catch clauses, e.g., catch (TimeoutException
>>...) might get skipped.
>
> I might be missing something, but the contents of the try-body needs to
> execute before the finally, so the TimeoutException will be thrown *before*
> pool.close() is called (and then subsequently be handled by the catch). What
> might be affected though, is the dumping of threads, which is indeed an
> oversight. Let me open another issue to do the double-try (which also means
> that I can update the copyright year.
Opened https://bugs.openjdk.org/browse/JDK-8395898
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/33453#discussion_r4387950522
