[PATCH v4 1/4] windows.media.speech/tests: Add test to check if IAsyncInfo_Close non is blocking.

[Alexandre Julliard]

Rémi Bernon <rbernon at codeweavers.com> writes:

> On 4/19/22 11:57, Bernhard Kölbl wrote:
>>>> -        WaitForSingleObject(compilation_handler.event_finished,
>>> INFINITE);
>>>> +        todo_wine
>>> ok(!WaitForSingleObject(compilation_handler.event_finished, 1000), "Wait
>>> for event_finished failed.\n");
>>>>            CloseHandle(compilation_handler.event_finished);
>>>>
>>>>            hr =
>>> IAsyncOperation_SpeechRecognitionCompilationResult_put_Completed(operation,
>>> NULL);
>>>
>>>
>>> I don't understand why you need to change this now? If it is going to
>>> fail and timeout after you stubbed some method, maybe it'd be better to
>>> make the change with the stub.
>>>
>> The reason for this change is so the tests can't be blocked forever
>> and you
>> get a message if it failed. 1000ms is long enough for the finished event to
>> fire. Makes debugging the tests easier imho.
>> 
>
>
> I think having a non-INFINITE timeout makes the intent less clear. If
> the wait is never supposed to timeout, INFINITE makes it obvious
> imho. It also makes sure you will never get spurious failures even if
> for some reason a run gets particularly slow.
>
> If for some reason the test breaks and the wait starts blocking, the
> testbot timeout is there to catch it, although yes, it's a bit longer.

When I'm running make test before pushing commits, I don't get the
testbot timeout, so it's nice if it doesn't get stuck forever.

-- 
Alexandre Julliard
julliard at winehq.org