Apologies to Neil & the Dali team - when last I looked at the 3.9.3 page it hadn't been updated - I guess it was updated a couple hours later according to the wiki history. This process of having to hunt for checkmarks seems needlessly difficult. Why can't we just assume the smoke passes unless someone announces on this list that there's a problem before a predefined cutoff time? If we all know there's a smoke test coming, surely we can plan to allocate an hour in a 24hr period to do the smoke test?
What if we announced the smoke test further in advance, so people can better plan?
In JBoss Tools, we announce a code freeze a few days before the Wed at the end of our 3-week sprints, build the final build on the Thursday, and it's ready for testing/smoke/staging/integration by noon on Thursday, giving people a few days to test it (and call for a respin if needed) before it's released the following Tuesday. This would align well with the Tuesday +2 dates for Oxygen and Photon milestones/RCs.
Where we're different from WTP is that we assume all is green unless someone (usually dev or QE) calls for a respin between the Wed code freeze and the Tuesday release day.
----
Regarding the build churn, yes, I agree that was unfortunate. That was partly caused by changes in M5 / Tycho 1.1, which I'm still solving. I hope there won't be any more breaking changes for M6, but I'll work more closely with mistria and akurtakov to find out if things are in the pipeline which could impact us. Being forced to move to Tycho 1.1 (because of Dali, platform M5, and Webservices) introduced new issues since Tycho 1.1 is more strict about how it validates source feature/plugin generation. So expect more gerrits to fix that, coming soon.
Some of the M5 build churn was also caused by stalled gerrits. For example, there's a broken test in servertools [1] which I attempted to disable since it's testing an obsolete/deprecated update site format for wtp server adapter discovery, and which once merged, will result in more passing gerrits for servertools. This gerrit has been sitting waiting for review and merge since Jan 15. Two weeks should be unacceptable at any time, but is particularly painful during a milestone smoketest & release period.
What can we do to prevent these stalls in future? I have some ideas, but I'm wondering what you all think we could do.
Thanks,
Nick