Platform Team Meeting 14 Nov 2006

Agenda

• Q of the fortnight: What is your most interesting or challenging problem for α5?
• α5, α6
  • build support expected in α5? Bugs you are relying on getting fixed?
  • performance

Notes

• Q: What is your most interesting or challenging problem for α5?
  • Andi: 6 is afraid because 789! Still in α4 mode.
  • Grant: Recurrence is interesting, writing documentation is challenging
  • Heikki: Performance improvements -- python 2.5 support could be interesting. Have been holding discussion until we ship α4. (Andi wants the "with" statement).
  • Morgen: Coordinating w/cosmo team closely -- never had to work this closely. Will be relatively straightforward to implement. How to support both new and old world in transition.
    • Test plan for sharing -- work with Aparna and Dan?
  • Bear: Interesting is Python 2.5, Ubuntu, switching stuff to python from shell scripts, keeping up wiht build related bugs. Build changes that don't break at least one developer every time a change is made.
    • Readline? breaks pdb, which breaks a developer, back to where we started. Noise on what the best way to do it, windows console integration required to get it done. (3 year old discussions). Grant: readline in cygwin? There is, but you can't depend on it, not the same Python. (Wing debuggers don't use readline anyway).
    • PJE: check and see what he is using for readline -- don't know how that is working. His Chandler has readline somehow. Hmm that is weird. Can arrow up and go back and edit lines, not GNU readline, using windows keys for editing. He'll research and investigate.
    • Bear: Jeffrey is running into a pdb problem when he sets a breakpoint. Bear and PJE will follow up.
  • PJE: sharing related work not challenging yet, unless we run into perf issues. Main area, getting setuptools changeover to direct installation (more doc changes and packaging changes, hesitating to pull the trigger just yet).
• Build bugs expecting to get fixed?
  • do_tests stop at first failure by default -- want "unit tests" to keep running by default? Andi: seeing all errors at once is a timesaver. Grant: I like to see just the first error, that is what he has to go fix. PJE: anything that is a unit test, want it to stop immediately. (His take is the opposite of Andi). Functional tests, want to know the % of tests that pass. Bear: include a switch so that Andi can run them. Andi is not stuck today.
  • readline
  • converting do_tests to python, will allow to simplify tinderbox scripts
• Heikki: Grant had proof of concept fix for sharing performance? (Lisa did some analysis, Grant took a look). What is the status?
  • Grant: hasn't thought about it lately. We make new connections, which is the problem. Grant tried to cache the connections. Bug for that?
  • Morgen: when we move to the new protocol, subscribe will be one big GET.
  • Grant: we still have to detect that it is Cosmo. Or are we only supporting Cosmo?
  • Morgen: osaf.us is fast enough, not too worried about sharing performance. How do sharing performance tests do on osaf.us?
  • Heikki: all tests were running inside internal network (qa-cosmo). Small calendar, large repository, large times for publishing.
  • Andi: 9 connections where there should only be one -- a problem! Seems like there is something to be looked at.
  • Grant to pick it up and go look at it again.
  • Morgen: haven't nailed down schedule with Cosmo. Chandler will be ready in a few weeks, don't know Cosmo's schedule yet.
• Andi: increasingly uncomfortable with search -- we haven't built up enough data to search on. Once you have real data, become glaring that we don't have search.

Status

Katie

• Progress
  • α4 dogfooding, bug councils, trying to move bugs along
    • moved to osaf.us
  • reviewed bug lists with Morgen, Grant for α5, α6
  • α5 dashboard design meetings
  • Dev list summaries:
    • Sept 18 - 24
    • Sept 25 - Oct 1
    • Oct 2 - 8
    • Oct 9 - 15
    • Oct 16 - 22
    • Oct 23 - 29
  • Business strategy position interviews
  • Branding exercises
  • Review of Chandler α5, α6, Preview plan with Mitch and PPD (will show up on lists and at staff meeting)
  • Cosmo/Chandler sharing meeting
• Plan
  • Review of Cosmo Preview plan with Mitch and PPD (will show up on lists and at staff meeting)
  • α4
    • clarify steps to reproduce bug that screws up indexes (Bug:7348)
    • write up dogfood feedback
  • α5
    • Draft of Item/ContentItem/Stamping section of PreviewDomainModelOverview
    • Catch up on specs
    • Performance targets for Preview, whatever else we need to get unblocked
    • Get started on test egg, menu fixes
    • Summarize list of loose ends for chandler-dev
  • Branding exercises
  • Business strategy position interviews

Andi

• http://kumu.osafoundation.org/~vajda/logs.cgi

Morgen

• Progress
  • Fixed bug 7239 (items not synchronizing to the server; plus importing an ical file which had icalUIDs the same as existing non-event items was causing problems)
  • Fixed bug 6555 (Update Cosmo Sharing Service info in Chandler accounts)
  • Fixed bug 7199 (Incomplete events in office calendar on cosmo-demo)
  • Fixed bug 7224 (Add platform to HTTP user-agent)
  • Fixed bug 7230 (Incomplete events when subscribing to a calendar that a non-Chandler client like CosmoUI^? has created events in)
  • Fixed bug 7330 (Found a corner case in incomplete event detection)
  • Fixed bug 7354 (Incomplete events being published into office calendar)
  • Fixed bug 7356 (Chandler started to sync shares after I tried to quit the app)
  • Met with Cosmo team to go over new protocol and record types
  • Defined EIM record types
  • Added UUIDType and DecimalType^?
• Plan
  • Continue to define EIM record types
  • Implement importer/exporter callbacks for record types
  • Reorganize sharing layer around EIM and morsecode

Grant

• Progress
• 6329: Triage: Purge button commits pending triage-status changes
• 7296: 'just this event' not an option when deleting 1st recurring event
• 7314: Traceback window: Stamp.mail error on Cosmonauts2 share
• 7318: Traceback when changing occurrence
• Made a patch to help out with 7196 (deferred deletion during repository merge).
• 7299 (PPD collection 412 error message): Added log to diagnose problem further
• Helped ErnestoRivera with his presentation/demo for all-hands meeting.>
• Analyzed tinderbox crash logs and collected Mac crash logs from various users to determine whether 6732 and 7269 are impacting α4 users or not.
• Put together patches to deal with Mac OS X 10.4 compiler build issues for 7269).
• Started to work on Domain Model Overview and SVK notes documents.
• Various meetings: Chandler engineering, deferred delete discussion, dashboard/recurrence discussions, α5 domain model next steps, Cosmo sharing, Cosmo domain model.
• Wrote some (slow) code to support per-attribute modifications to recurring events (6700).

• Plan
  • Work on per-attribute recurring event modifications.
  • Domain Model Documentation.

Bear

• Progress
• 7273: setuptools requires dos-style path for pyegg installs
• Started working on 6159 (Create an Ubuntu .deb installer package).
• Research for 7321 (No warning appears when starting PPC Chandler on Intel Box).
• Many Checkpoints.
• Reviewed bug list with Heikki
• α4 dogfooding (migrated collections to osaf.us).

• Plan
  • More α4 dogfooding
  • Finish .deb installer
  • Work on do_tests.sh rewrite with Dan
  • Movement on new tinderbox scripts