Desktop Release Meeting, Thursday May 31st 2007

Present: Bear, Grant, Philippe, Aparna, Sheila, Heikki

Agenda

• Performance
  • Collaborative session feedback (Philippe)
  • Next steps
• Bug count and schedule
• Dogfoodable checkpoint
  • What's left to do? Blocking issues?
• QA/Build

Notes

Performance

• Collaborative session feedback
  • Got quite a bit of data (100 or so measures). That was good.
  • Not much of a variety of machines. Too many Intel macs. Only 1 Windows, no Linux so we're still in the dark for those platforms. That was not good.
  • Interesting find: loading of a dump is way slow. Need to add that as a test case certainly. Some issues addressed since then (svn 14481 / Bug 9360 created a progress bar).
  • Tests description short but this didn't confuse people much. Need to describe some of them more precisely though
  • Note: very serious perf problem if machine has less than 512 MB RAM... 1 Gig recommended. This is really annoying. We should have someone investigate even if we won't hold off Preview for this.
  • Note: Big difference between starting after a reboot or on a warmed up machine (i.e. a machine on which Chandler has been launched, quit then relaunched). Need to specify in perf eval wiki what to measure.
  • Practically everything could benefit from some perf boost (nothing is ideally snappy...) but only few scenarios are really slow
  • Scenarios where the app feels sluggish:
    • 1st time overlay (building indexes)
    • drawing calendar canvas (when switching from dashboard view, when scrolling, when moving between weeks)
    • triaging
  • Idea: add a progress bar (in the status bar?) that for anything with time > 5s that we don't think we're going to improve much before Preview.
• Next steps
  • Philippe to email the list with the results
  • Philippe to encourage more people to run the tests by themselves and send data
  • Need to do a session to perf test sharing
  • Need to log bugs and discuss with devs approaches for triaging and drawing calendar

Bug count and schedule

• Check the curves: going down but still high count. Stale bugs accumulating (we fix newfound bugs fast but we let old bugs accumulate).
• P4 and P5 to punt if possible, Philippe doing this but slowly (not high prio in his list of things to do and some of those bugs turned out to be more serious than their P4 status make them look like)
• Need to triage more aggressively the P3/Medium and above. We'll do that with the Bug Council Tuesday @ 3pm (will leave time for some of us to prep, reviewing those bugs before hand)

Dogfoodable checkpoint

• Philippe to tweak the query on the dev page so that we get only the Chandler's bugs (done)
• We triaged the leftover blockers
• Only 4 bugs left. Looks like we'll hit Monday.

QA/Build

• Bear to integrate the Windmill tests in TBoxes, shouldn't have too much interference with current Chandler schedule though

-- PhilippeBossut - 31 May 2007