Attending: Oliver, Paul, Chris, Joanna, Eugene
- Paul noted that Harm will be late
Oliver asks if there are any objections to last weeks summary
Joanna has hard stop at 10:30a
- anything else needs to cover before leaves?
Oliver asks if we are well and truely done with 4.5.1
- Joanna notes that it is done
- Joanna noted that there are still some issues around profiling/probes to handle for future
Paul brought up the subject of branching out to start 4.5.2 checkins
- Joanna noted that the team is having branch issues.
- Empty plugins do not seem to want to branch
- Joanna has a question into the webmaster.
- Jerome has some stuff he'd like to start testing when it is done (currently sitting in patch in bugzilla)
Defects requested by consumer that had targeted 4.5.1 but were not done now have their target set to 4.5.2
4.5.1 Post mortem
Oliver asked if we can talk a bit about the lessons learned from 4.5.1. Joanna made a number of excellent comments
- We need to put a bit of diligence into achieving better test coverage
- We didn't have enough variety in testing (e.g., parts of integrated AC was missed)
- Some other areas found by consuming product (failure under hundreds of repeated executions)
- Joanna noted that some of this repeated testing would probably be outside of scope of "expected" testing within TPTP
- We should not overbook our resources such that there is not time for late arriving but critical defects.
- We should plan some percent of time for issues that will arise
- And just promote some defects if we don't need that reserved time
- We need to examine the build infrastructures to see if it can be made a bit more robust
- We need backup expertise for code areas that are complex (e.g., profiler and AC)
Oliver asks for a bit more discussion around coordinating with Beijing team and achieving this backup expertise.
- Joanna is working to identify a few folks on her team to be potential backups
- Chris needs to identify right names for Intel
- Will want to do some code walkthrus for Martini (w/ Asaf) and AC/profiler walkthrus (w/ Beijing team)
We ran into calendaring issues with 4.5.1 because consumer was not aligned to Eclipse simultaneous releases.
- Would be ideal to ensure consumers align to Eclipse better
- Oliver notes that this is good goal but is not always possible (there are other constraints in product schedule :-)
- Best we can probably do is try to talk to consumers about our schedules well in advance
- And see if can hash out a coordinated plan
- In worst case, may want to just do our own schedule and worry about consumer patches after the fact.
- Joanna did talk a bit about when consumers start testing with TPTP builds and hoping that this happens earlier in releases.
Eugene would like to raise possibility of bringing in Peer review more consistently
- Question to Joanna: Are there enough skills to do this?
- Has a few folks with native code skills who can review each other
- For java skillsets are not a problem.
- Oliver is behind this idea.
- Review may be a combination of going over the lines of code themselves (which may be really complex construct) to identifying areas that should definately be tested.
- Paul mentions that some of items he finds relate to internal API usage and API contract
- The review flag can be used as it was back in the old days.
Paul would like to see locking down codebase thruout release to enforce review policies
There was discussion around ensuring that test case is in place for feature
- Bugzilla does not have test case field
- Could be a check during review that reviewer can verify.
XML Roadmap Collateral
Joanna pretty much done with draft
- Joanna decided not to create a full 4.5.2 plan
- Noted that historically we did not create full plans for point releases like this
- She took the 4.5 plan and tweaked it so that it renders correctly in firefox
- Oliver asks if there is anything controversial in the content or approach
- Joanna notes that the content should not be controversial since it was already approved for TPTP 4.5
- Joanna notes that the approach should not be controversial because the posted instructions suggests said that some may just want to convert existing html plans to xml
University project has approached TPTP interested in looking at bug reopen rates, patch size rates, etc.
- Might provide some interesting statistics
Paul wanted to particularly recognize Eugene for being a driving force on the newsgroup recently.
- Kudos to Eugene!
Before the main meeting started Oliver and Paul were talking about the build time test infrastructure and its status.
- Paul notes that Joel and Josh are still working on it
- they are running Junit tests on J2SE 1.4, 1.5, 1.6 today!
- There are some Test project dependences to make it even more productive
- In particular ome issues running AGR between different systems
Oliver asked a question about whether other test harnesses (e.g., Rational) were considered for TPTP and others.
- Paul mentioned that EMO insists that the tests for projects must be checkout-able and runnable (and possibly updatable) via cvs.
- This tends to limit the ability to use some 3rd party tooling/harnesses.
Soliciting Eclipse Plugin Profiler Users
Eugene sent out link to introduction wiki page (responded to Oliver's mail)
- Goal to minimize reading before users can start going
- AI to Oliver's team and Chris to provide feedback
Eugene notes that planning is starting now
- Open for submissions in October
- Oliver asks who might be going
- Oliver will go
- Joanna hoping to get travel budget for 1-2
- Chris hoping to get travel budget for 1-2
- AI Chris and Joanna to follow up with their companies to try to get more firm detail
- Oliver notes that EclipseCon is really good opportunity to get a feel for the overall lay of the land for Eclipse.