> On May 25, 2017, 3:58 a.m., Anand Mazumdar wrote: > > src/tests/master_allocator_tests.cpp > > Lines 719-720 (patched) > > <https://reviews.apache.org/r/59541/diff/1/?file=1732021#file1732021line736> > > > > hmm, not immediately clear to me what guarrantees that this method is > > invoked once before the end of the function? > > > > - `driver.stop()` would send the teardown message to the master and > > immediately trigger the latch making `driver.join()` return. > > - If the function finishes before the message is processed by the > > master, `removeFramework()` might not be invoked.
Good catch -- I'm not sure offhand if that sequence is possible, but the intent here is just to reset the expectation to the default value, so `DoRepeatedly()` is probably better anyway. - Neil ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/59541/#review176038 ----------------------------------------------------------- On May 24, 2017, 9:32 p.m., Neil Conway wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/59541/ > ----------------------------------------------------------- > > (Updated May 24, 2017, 9:32 p.m.) > > > Review request for mesos and Anand Mazumdar. > > > Bugs: MESOS-7552 > https://issues.apache.org/jira/browse/MESOS-7552 > > > Repository: mesos > > > Description > ------- > > This test did not pause the clock; hence, batch allocations could occur > at unpredictable times and cause the test to fail. > > Fix this by pausing the clock and explicitly advancing it as needed to > trigger batch allocations or other events. > > > Diffs > ----- > > src/tests/master_allocator_tests.cpp > 750e030d388d5469399bc79fa723095e62b8f2c5 > > > Diff: https://reviews.apache.org/r/59541/diff/1/ > > > Testing > ------- > > `./src/mesos-tests --gtest_filter="MasterAllocatorTest/0.FrameworkExited" > --gtest_break_on_failure --gtest_repeat=5000` # Linux > > Without this RR, the test fails within a few hundred iterations. > > > Thanks, > > Neil Conway > >
