First of all, I believe all gating tests should be run in precheckin. If we make jdk11 tests gating, we should make it part of the precheckin. If we don't put them in precheckin, they should not be gating.
Secondly, If we don't make jdk11 tests gating, soon they will become like windows tests, people only look at them after it's been failing for days, which is not good. Thirdly, for non-gating tests, we probably should run them after all the gating tests are done. On Wed, Nov 7, 2018 at 11:39 AM Owen Nichols <onich...@pivotal.io> wrote: > Now that tests are passing under Java 11, it was recommended last week to > make Java 11 tests gating for the develop pipeline. [Fyi, Windows tests > are not yet gating, meaning the pipeline will success and publish artifacts > even if a Windows tests fails.] > > Three topics merit discussion: > > 1) For the Geode 1.8 release, should the release notes advertise > “experimental support for Java 11” or no support? If the latter, should > Java 11 tests still be gating on the 1.8 release branch, or only on develop? > > 2) As of now, pre-checkin runs tests only against Java8. Now that Java11 > is gating in develop, should we now be testing against both Java8 and > Java11 as part of validating pull requests? > > 3) In the develop pipeline, should non-gating jobs continue to be run in > parallel with gating jobs? Or would it be better to change the develop > pipeline to only run the non-gating tests after all gating jobs have passed? > > Thanks, > -Owen > > P.S. after a brief trial of combining Java8 and Java11 tests into a single > pipeline job, that was reverted and it looks like separate jobs are here to > stay. Combined jobs made it difficult to look at dual scrolling outputs > and dual archive-results steps. If you feel there’s a better solution, > please speak up. > > -- Cheers Jinmei
