RFR: 8368625: com/sun/net/httpserver/ServerStopTerminationTest.java fails intermittently in tier7 [v2]

[Mark Sheppard]

On Fri, 10 Oct 2025 17:33:53 GMT, Daniel Fuchs <dfuchs at openjdk.org> wrote:

>> Mikhail Yankelevich has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains four additional commits since the last revision:
>> 
>>  - race condition fix
>>  - Merge branch 'master' into JDK-8368625
>>  - race condition fix
>>  - JDK-8368625: com/sun/net/httpserver/ServerStopTerminationTest.java fails intermittently in tier7
>
> test/jdk/com/sun/net/httpserver/ServerStopTerminationTest.java line 164:
> 
>> 162:         // Complete the exchange 10 second into the future.
>> 163:         // Runs in parallel, so won't block the server stop
>> 164:         final Duration exchangeDuration = Duration.ofSeconds(Utils.adjustTimeout(10));
> 
> Looks good, but if this test fail we could envisage bumping that delay though. A better implementation could be to complete the exchange *after* exiting from server.stop() - without using any virtual thread, and just verifying that the server.stop() waited at least for 1s.

The second temporal constraint is something that could possibly fail. If it fails why is it considered a test failure?

Consider a slightly convoluted scenario:

The Exchange will activate after 1 second. That thread is scheduled to execute but remains RTR due to excessive load on the system
The server.stop is invoked with a 20 second delay, indicating this is the max allowed time for graceful shutdown i.e. the time in which any extant
exchanges should complete. If that timeout expires in the finishLatch.await returns — no check is made to determine if this was a timeout expiry or not —
then the server proceeds to shutdown.

Thus, in this scenario, the test will fail. As such, how is this a failure? Your test has placed a temporal condition which has a possibility of not being met.
Your expectation is that because the shutdown delay is 20 seconds the test should always pass, with the Exchange having completed, as such deterministic. This
is not unreasonable, but it is not a condition that can be 100% met, and is not deterministic.

The semantics of server.stop doesn’t consider a delay timeout as an error condition, so why is this the case in the test.
So rather than it being a test failure, should a delay expiry just be observational ?

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

PR Review Comment: https://git.openjdk.org/jdk/pull/27670#discussion_r2444705202