> On July 5, 2016, 3:08 p.m., Alexander Rukletsov wrote: > > src/tests/hierarchical_allocator_tests.cpp, line 3100 > > <https://reviews.apache.org/r/49612/diff/2/?file=1436764#file1436764line3100> > > > > I would keep these 3 instances in this test. It's not immediately clear > > that `AWAIT` happens in `handleAllocationsAndRecoverResources()`. > > Neil Conway wrote: > I don't agree with keeping `Clock::settle` in this case. For one thing, > it implies the wrong thing: i.e., keeping `Clock::settle` implies that > `handleAllocationsAndRecoverResources()` does not do sufficient > synchronization on its own, which is incorrect. > > Second, `Clock::settle` is a kludge; it should only be used when there is > _no other way_ to achieve the synchronization required by the test. So I'd > prefer to discourage its use whenever possible.
In general, `Clock::settle` is not a kludge: it allows to make sure a certain event is processed before another event is scheduled. It is not the case here, though. We usually write tests in a way that people can easily follow main steps without looking into functions (hence a lot od copy paste code instead of fixture methods). `handleAllocationsAndRecoverResources()` already breaks this convention by modifying the caller's context; removing `Clock::settle` makes things even worse IMO. - Alexander ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/49612/#review140790 ----------------------------------------------------------- On July 5, 2016, 8:51 a.m., Neil Conway wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/49612/ > ----------------------------------------------------------- > > (Updated July 5, 2016, 8:51 a.m.) > > > Review request for mesos and Alexander Rukletsov. > > > Repository: mesos > > > Description > ------- > > If a test case calls `Clock::settle` and then immediately waits for a > future to be completed, settling the clock is usually unnecessary. > > > Diffs > ----- > > src/tests/hierarchical_allocator_tests.cpp > 0498cd5e54b0e4b87a767585a77699653aa52179 > > Diff: https://reviews.apache.org/r/49612/diff/ > > > Testing > ------- > > `mesos-tests --gtest_filter=HierarchicalAllocatorTest.* --gtest_repeat=500 > --gtest_break_on_failure` > > > Thanks, > > Neil Conway > >
