> On Mar 29, 2017, at 11:59 AM, Kirk Lund <kl...@apache.org> wrote: > > Yes, the test failure is intermittent. It's up to the group if we want to > release with intermittently failing tests. These are new tests, so we do > seem to be creating new flaky tests. I would tend to take a more hardline > approach which says if the build is RED we don't ship. >
> Jared and I have been digging into the cause, and it seems to be these new > DUnit Locator/Server starter rules which appear to be more complicated than > they need to be and are promoting pollution of later tests. They are being > used in several dozen tests. We'll have to put other work on hold in order > to refactor these rules sufficiently to make the tests pass consistently. > I agree that we need consistency in our test suite. I think the best place to do that refactoring is on develop. Retrofixing a maintenance branch could be a pretty open-ended activity. How far do we go? My suggestion: vote on the merits of the changes in this patch release. And start a thread on dev@ to gather consensus on release criteria going forward. $0.02, Anthony
