Apps Team Meeting May 3rd, 2007

This is our first joined meeting, or our second if you count the the perf meeting we had 2 weeks ago. In any case, the format is not wildly different from what we had before.

Agenda

• Question du Jour : (5 min)
• Performance work : (10 min) We did a lot of progress in those last 3 weeks though we haven't reached all the acceptable target values. Let's review who has still perf work on his plate and what's the plan and likelyhood to improve those numbers.
• Debug : (10 min) Most of us are now entering the debug phase. Some points:
  • The bug load is very high: PPD has been triaging and pushing to P4 lots of bugs, I and Katie need to do that too. We certainly won't fix all of them. We need to focus on the really problematic ones (that's the meaning of the prioritization).
  • Load not equally distributed: the workload is very different between devs. I'll work with those with a high load to try to rebalance that.
• Plan : (10 min) So, what's the plan? (and dates...). We haven't changed them yet but it's certainly a good moment to review this. What needs to be done to get there:
  • SWAG: we need to get some estimate of those bugs, especially the P1 to P3 ones
  • Rebalance bug load: once the SWAG is done, we'll rebalance the load and see where we land
• Issues and Announcements : (5 min)
  • Reviews: tomorrow is last day...
  • Have something to announce? Issues to bring to the group? Please take the stage.

Minutes

Performance work

• We need to add a sync scenario: what's the target? unclear. Morgen says that the sync is different from Subscribe and Publish.
• Who's working on profiling Publish and Subscribe: Morgen says "there's a wiki page with script tests for it". No one seems to be willing to own this and Morgen is overbooked. Philippe to get this decided off line.
• Who's working on perf? Most devs done or not sure what to dig through next. Philippe to review the open perf bugs and assess with devs.
• Other things:
  • Startup time (John). We're getting to the "acceptable" target but that target is pretty low anyway. Static linking of wx and Python would help (Andi). Build nightmare but feasible. Heikki and Bear to check that.
  • BKirsch, Andi: Great perf improvements on mail.

Debug & Plan

• SWAG: devs to update SWAGs for all P1, P2 and P3. See the doc for how to use and update SWAG on Bugzilla.
• For devs with more than 20 bugs, review and prioritize them with your manager before SWAGging

Status

Andi

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

Brian (bkirsch)

• Checked in additions to EggTranslations project including a hasTranslation API that returns True if a localization gettext .mo file is loaded for the given locale.
• Augmented the Chandler I18nManager to default to en_US if 1. no locale was specified via preferences or the command line 2. Chandler does not have a translation for the locale returned from the Operating System. Adding this code prevents the case where wxWidgets and PyICU localize to say German but Chandler strings are in English.
  
• Huge improvements to mail download performance related to Bug: 7177 Improve mail download performance and repository storage. Refactored repository centric code to a MailWorker thread which shares a view for all downloads. There is still work to do to complete the MailWorker logic including migrating POP to use it. There is also some Repository issues that need to be resolved once more than 3500 mail messages are downloaded. And finally since the code in mail has been significantly refactored I need to do extensive testing of all Preview functionality (Chandler Folders, Chandler Headers, Edit / Update) before checking in.
• On completion of mail performance work will augment the Chandler Localization scripts (createEggTranslation.py) to work with Chandlers new plugin framework. With the plugin framework introduction putting eggs in develop mode no longer results in the localization being picked up by Chandler.
  

Bryan (stearns)

• Finished performance work
• Fixed a flurry of byline bugs (and closed a few others as worksforme)
• Working on reviews

Grant

• Progress
  • Fixed 8348 Feedback window popped - wasn't doing anything
  • Improved various Calendar-related performance tests by having a single event query for the all-day and timed canvases.
  • Diagnosed a problem that was causing the performance tinderbox failures, and suggested a fix.
  • Reduced the number of time-related indexes we create to search for events in the calendar.
  • Fixed 8844 (right-click to get context menu on a new collection causes Internal Program Error).
  • Fixed a raise reported by a user on IRC.
  • Fixed 8815 Title changes made in calendar do not get reflected in detail view
  • Fixed 8732 changing timezone messes up indexes
• Plan
  • Reviews
  • Recurrence fixes

Jeffrey

• Fixed Bug#5868, fix icons for read-only
• Fixed a timezone turn on issue with recurring events
• Fixed Bug#8904, calendar was getting stale items with new notification handling
• Tweaked vobject to handle UTC equivalent timezones when serializing, Bug#8967
• Performance work with notifications, Bug#6091
• Fixed Bug#8977, wonky calendar geometry (most of the work was done by John tracking down the issue)
• Paint the calendar less often performance work

John

26 April 2007

• I worked on notifications and concluded that we don't really have enough information to use them to keep the table "incrementally" up to date. An example of this problem is that when an Item is deleted we don't know what location it was deleted from, which is required to tell the wxGrid which row to delete. I spend a bunch of time looking into how we could fix this and got bogged down in lots of Andi's repository code. I think it's going to require some help from him for me to succeed.
• I made a pass through the recent recorded scripts bugs and the following appear to have been fixed by my recent changes:
  • bug#8448: Recorded scripts don't properly handle setting focus to sidebar
  • bug#8678: Recorded script assertion error on right click in sidebar
  • bug#8449: Recorded scripts don't properly record shifted non-alphabetic characters
  • bug#8601: Assert during script playback when switching collections
  • bug#8389: Script recording assert when clicking in location field
  • bug#8861: Script recording crashes Chandler when user attempts to add recurrence
• I think there are only two outstanding script recording bugs in bugzilla.
• I tracked down a difficult Windows performance bug which causes line scrolling in the table to redraw the entire table on each scroll. It appears that the Win32 ::ScrollWindow API invalidates the entire screen when the window is a child of one with the WS_EX_COMPOSITE flag (used for double buffering). This happens even when you tell :ScrollWindow to not invalidate anything. If I manually calculate the correct invalidation region the scrolling problem is fixed. I still need to do a bit more investigation to make sure that this "bug" isn't really a "feature" for some reason I don't understand.

3 May 2007

• Fixed bug#4453: summary table scrolling is too slow.
• Tracked down but#8977:Calendar view has wonky geometry after restart on Windows for Jeffrey
• Researched rare failures of recorded tests, but still don't understand the source of the problem.
• I embarked on getting rid of DynamicContainers, which we think will yield a big performance improvement. It also gets rid of boat loads of code and the new code will be much easier to understand. It also involves serious refactoring of toolbars and menus, which will with some luck yield a simple solution to the "calclulated menu" feature. It's a risky change, so I may or may not be able to pull it off, but hopefully I'll know whether or not it will work in a few days.

Morgen

• Progress
  • Added auto-conflict-resolution hook, applied it to triage/createdOn
  • Sharing dialog fixes
  • Read/unread status fixes
  • Conflict dialog formatting improvements
  • Syncs are retried when server has database deadlocks now
  • Glossing over the fact there is a default sharing account yet no way to assign one
  • Functional/Perf tests now using morsecode+eim
  • Improved EIM layer support for read-only shares
• Plan
  • Continue bug fixing
  • Write reviews

Philippe (pbossut)

• Perf: checked with the team, organizing the "perf collaborative session" with QA
• 0.7Release: lots and lots of Bugzilla time...
• Interns: Interviews, reviews, discussions: done with filling the last spot!
• Reviews: starting to gather the reviews and writing mine

Phillip (PJE)

• Your status here...

Reid

• Spending a lot of time trying to use wx.GraphicsContext to draw items in the calendar
• results are mixed -- see the excel spreadsheet attached to Bug#8758
• not sure how to proceed :-/ -- any feedback would be great, especially if any performance-savvy people could try my patch and compare
  • to disable my changes, set testing_graphicsContext to False on line 893 of parcels/osaf/framework/blocks/calendar/CollectionCanvas.py
  • note that my patch has hotshot profiling in it, so it sholdn't be committed as-is
• MultiStateButton already had the optimization I was thinking of (must have done it a while ago)
  • further optimizations would be more work than the simple fix I was thinking of

Robin

• Spent a day and a half dealing with a failing hard drive in one of my development machines. I managed to recover everything except for a current Windows registry, but I had one from a few weeks ago so it wasn't too bad of a loss.
• Assisted various OSAF developers with perf related wx questions, alternative implementation ideas, etc.
• Closed Bug#7076 (Infinite redraws when editing in sidebar)
• Merged and tested several patches from various wxPython contributors for improvements to FloatCanvas, XRCed, MaskedEdit controls, etc.
• Backported some fixes that have been made on the wx 2.9 branch to the 2.8 branch.
• Preparing for a 2.8.4.0 release. Published a preview build for my test group.
• Various community support activities via email and IRC.

-- PhilippeBossut - 03 May 2007