-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/41986/
-----------------------------------------------------------
Review request for mesos, Alexander Rukletsov, Jie Yu, and Michael Park.
Bugs: MESOS-4208
https://issues.apache.org/jira/browse/MESOS-4208
Repository: mesos
Description
-------
Fixed race in persistent volume tests. This race occurred because
`scheduelerDriver.reviveOffers` will trigger an offer immediately, while the
tests originally assumed that settling and advancing the clock after reviving
was necessary to produce an offer.
Diffs
-----
src/tests/persistent_volume_tests.cpp
2fb57814b2805bc76981d1877603a1a033f29289
Diff: https://reviews.apache.org/r/41986/diff/
Testing
-------
In order to reproduce this bug, insert a short sleep onto line 1204 or line
1402 of `src/tests/persistent_volume_tests.cpp` - note that these line numbers
refer to the file *before* applying this patch.
The race is due to the fact that calling `reviveOffers` triggers a new offer,
while the code is assuming that the clock must be settled and advanced before
an offer will be sent. Thus, the offer may arrive before the `EXPECT_CALL` is
executed, causing a previous `EXPECT_CALL` to be triggered more times than
expected.
By calling `EXPECT_CALL` before `reviveOffers`, this race is avoided. The
unnecessary `Clock:settle` and `Clock::advance` calls have been removed as well.
To test,
`GTEST_FILTER="PersistentVolumeTest.BadACLDropCreateAndDestroy:PersistentVolumeTest.BadACLNoPrincipal"
bin/mesos-tests.sh` was run both with and without `sleep(1);` inserted before
the relevant `EXPECT_CALL`s.
Thanks,
Greg Mann
