On Sun, 3 Dec 2023 09:50:53 GMT, Alan Bateman <al...@openjdk.org> wrote:
>> test/jdk/java/util/stream/BuiltInGatherersTest.java line 331:
>>
>>> 329: case Integer n when n ==
>>> config.streamSize - 1 -> {
>>> 330: awaitSensibly(firstReady);
>>> 331:
>>> while(tasksWaiting.getQueueLength() < tasksToCancel) {
>>
>> @AlanBateman This is the only part of this I really don't like. Any better
>> suggestion as to "wait for N things waiting" in a test? 🤔
>
>> @AlanBateman This is the only part of this I really don't like. Any better
>> suggestion as to "wait for N things waiting" in a test? 🤔
>
> Semaphore::getQueueLength is more for monitoring purposes so the usage does
> seem a bit unusual here. There are a number of ways this could be done, maybe
> the simplest is for tasksWaiting to be an AtomicInteger and have the default
> just increment it and sleep-for-a-day.
> @AlanBateman Unfortunately that doesn't really work as there's then a window
> between incrementing the counter and going to sleep. So while it might
> execute correctly in 99.9999% of cases I don't want to have a jbs bug filed
> which will be one of those things where you can't recall it 6-12 months down
> the line 😓
Thread.sleep throws if called with the interrupt status set so a cancel in the
window between before the sleep should be okay. I think in this case, it's
whatever is simplest is reliable and easy to read/maintain.
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/16928#discussion_r1413551130
