This page contains notes from a retrospective discussion held during the Eclipse project weekly architecture call.
- Patch review days were good. Increase in contributions, some community feedback on improved patch acceptable from platform
- Patch reviews depends on team, some have low patch volume and can review as they come in others have high patch volume and patch review days work well.
- Platform UI team did 2 days at start of each milestone, with additional follow-up as required to iterate patches
- Gerrit is great for managing patches, rebasing easily to keep them up to date rather than getting stale in bugzilla
- Gerrit tempts you to merge directly from web UI, but this is a dangerous practice because you're not pushing what you tested (or worse, you didn't test)
- CBI was a net loss for us so far. It was a very large volume of work just to get our build back to its old state. There is benefit for others in the community to make it easier for them to run builds independently. For committers it hasn't delivered value yet.
- Still lots of rough edges in our build that we need to address. For example we can't currently diff JARs reliably between builds.
- Performance testing was non-existent. We need to get those back in place.
- Get them running, and then get useful ongoing data out of them.
- Performance pass during M6 - dedicated time built into plan to work on performance.
- Consider doing performance test pass in earlier milestone such as M3
- Do further automated analysis on performance data to get more useful signals out of data
- Planning happened very late. We need to get the plan in place much earlier.
- Add topics to call in advance so we have time to think about it
- Far too many platforms to test, not enough people/time to test them.
- Consider dropping 32-bit for Mac, AIX, Solaris, HP-UX
- JDK matrix - consider only having Java 7 reference platforms for next release (and possibly Java 8)