----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/66799/#review202015 -----------------------------------------------------------
Ship it! Ship It! - Greg Mann On April 25, 2018, 4:46 p.m., Benno Evers wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/66799/ > ----------------------------------------------------------- > > (Updated April 25, 2018, 4:46 p.m.) > > > Review request for mesos and Greg Mann. > > > Bugs: MESOS-8687 > https://issues.apache.org/jira/browse/MESOS-8687 > > > Repository: mesos > > > Description > ------- > > This test used to be sporadically segfault as described in MESOS-8687. > The suspected cause is that a in a master actor, the `httpSequence` > field was lazily initialized in `ProcessBase::consume()` and afterwards > a call to `ProcessBase::_consume()` was dispatched, where it was > assumed that `httpSequence` is already initialized. > > However, during this test the master actor would be destroyed and a > new actor would be spawned with the same PID. The dispatched method > would be called on this new actor and find `httpSequence` to be not > initialized, leading to a crash. > > This patch introduces a call to `Clock::settle()` after the master > is shut down to ensure the outstanding `_consume()` gets discarded > before starting the new master actor. > > > Diffs > ----- > > src/tests/api_tests.cpp dd8e221d8fd1b2a241505345337897e4ee4a6347 > > > Diff: https://reviews.apache.org/r/66799/diff/1/ > > > Testing > ------- > > `./src/mesos-tests --gtest_filter="*MasterAPITest*MasterFailover*" > --gtest_repeat=100 --gtest_break_on_failure` > > > Thanks, > > Benno Evers > >
