Mozmill 1.5 Post Mortem Notes - Sep 2, 2010
- Attendees: Bob, Clint, Geo, Aaron, Anthony, Henrik, Al, Matt
What were the objectives and were they achieved?
- (Achieved) Another small point release to fix issues for QA folks
- (Achieved) Release targeted for mid quarter so QA would be able to get their stuff done
- (Mostly achieved) Set a base line for new features which involved refactoring, so that smaller dot releases are more doable
- (Achieved) Bring in correct features to allow mozmill-tests team to meet goals
- (Mostly achieved) Community outreach
- Working module loading chain still needed for add-ons tests
What went well and should be continued?
- Auto-tools team constrained scope of release and deferred features as necessary
What didn't go well and needs revision or elimination?
- A more formal planning process is needed w/ some level of planning documentation generated
- A clear roadmap is needed, not only for 2.0 but also for upcoming dot releases
- Define a clear set of features for a dot release, push out invasive changes to the at minimum next release
- There is no fixed schedule for dot releases yet
- A 6 week duration should be possible (Clint has to talk with the team)
- Testing patches is quite intensive without existence of a complete set of tests for Mozmill
- Clint is working on it in the next weeks. Can hopefully make it for 1.5.x
- Team members should escalate to managers if/when necessary to work out priority issues
- Or tiebreak within the teams by bringing in a third opinion as needed
- Was raised that a feature inclusion process needed to be there. There is a process involving bugzilla whiteboards and periodic meetings, but perhaps may not be as explicit/written-down as it could be for reference between meetings.
- Releases of the add-on have to be faster available (review process on AMO takes too long)
- Use QMO as a temporary and trusted place to host the add-on
- A number of issues surfaced as a result of bug 583604 Synthesizing events doesn't return any information if event has been successfully processed. One of those issues were around the fact that tests were failing on Clint's environment and not on Henirk's. Two issues here:
- The patch introduced instability, but it is not know for sure why. Root cause should be determined.
- For confidence of future releases, tests for mozmill need to be very reliable and used as part of the patch landing process. (added by Matt)
Note: we didn't really get to these, although points 1 & 2 were touched on in-meeting.
- Are scheduled maintenance releases with clear set of fixes and features possible?
- Fixed releases eventually 1 per month?
- Those shouldn't be any big new features but bug fixes we need to continue our work.