BulkheadAsynchRetryTest.testRetriesJoinBackOfQueue seems wrong

My reading of the test BulkheadAsynchRetryTest.testRetriesJoinBackOfQueue suggests that it’s implementation-dependent. That is, I believe it’s possible to write an compliant implementation that fails this test. I’m specifically thinking:

1. The beginning of the test reads:

AsyncBulkheadTask taskA = newTask();
Future resultA = retryQueueAsyncBean.test(taskA);

AsyncBulkheadTask taskB = newTask();
Future resultB = retryQueueAsyncBean.test(taskB);

AsyncBulkheadTask taskC = newTask();
Future resultC = retryQueueAsyncBean.test(taskC);

taskA.assertStarting(resultA);
taskB.assertNotStarting();
taskC.assertNotStarting();

This test expects the first call will be the first to make it into the bulkhead, while there’s nothing in the test that would affect thread ordering. Clearly this presumes one implementation strategy, while other implementation strategies are also valid.

1. The test javadoc also says

Test that when an execution is retried, it goes to the back of the bulkhead queue.

This itself again presumes an implementation strategy, while there are other valid implementation strategies.

WDYT?

[Ladicek]

I’ll submit a PR.

[Ladicek]

I forgot to mention the issue number in the PR, but this has been resolved in #410.