[xwiki-devs] [Retrospective] Big slippage on all aspects for 5.0M2 - Chance to learn

Hi devs, I'd just like to point out that 5.0M2 seems to be a very bad release WRT the roadmap we had defined, although a lot of hard work has been done. It's been a long time since we've diverged that much actually. Let's take this as a chance to learn and do a quick retrospective. Roadmap: http://www.xwiki.org/xwiki/bin/view/Roadmaps/WebHome Misses from Roadmap: * SOLR implementation missing the scheduler and thus cannot replace the default search. Main reason: Edy is sick currently. * AppWithinMinutes work * Scalable import/export * Home page improvements (XE) and usability improvements in general * Horizontal Menu management (add, remove pages) * XEM Homepage / Portal investigation * New Activity Stream Investigation Hits: * EM upgrade of subwikis. Mostly done but still missing a few things * virtual=1 by default * WYSIWYG Editor improvements * 60% of listed important bugs * General Flavours Investigation And WRT dates we're slipping by 10 to 15 days…. Other miss: the build stability. The whole build has been down for over 10 days and there are still failing modules. There has been a lot of stuff done of course but they were not planned (i.e not on the roadmap). This is our biggest miss and what I want to point out: we've not followed the defined roadmap and instead we've done other stuff like putting the new security module on by default which took us very long. We need to learn from what happened to handle it better next time and we now need to get back on track and go back to following our defined roadmap. IMO the following should have been done better (and that's what we should learn for next time): * Discuss replacing some items from the roadmap by others from the beginning. For ex don't take the task of setting the new security as the default without putting it in the roadmap * Setting the new security module should have been done on a branch and the tests/code fixed there before merging it on master. As a general rule nobody should commit anything on master knowing that it would break the build. Because that impacts everyone and prevents everyone from working on what they have to work on. * Similarly, if a commit is done and we realize afterwards that it breaks a lot of stuff, we should revert and put on a branch (I have the JUnit 4.11 upgrade in mind, which changed test order) * Removing xwiki/1.0 syntax by default with the breakage of several tests should also not have been committed directly without fixing the tests first, either locally or on a branch if we want to share the workload WDYT? Do we agree about handling it this way for next time? Since I wasn't there during most of this timeframe (holiday for 1 week and then conference for another week) I may have missed some other stuff an I may even be wrong about what happened. Please correct me or add what I missed. Thanks -Vincent