Electrolysis/Meetings/2014-12-11

From MozillaWiki
Jump to: navigation, search


« previous week | index | next week »

Triage Proposal

  • break needinfo triage bugs out of the main triage list
  • we can better track these as a stand alone list
  • once a user responds, the bug falls back into the main list
  • mixed into the triage list, we have to re-triage these every week
  • we can afford to skip this list when triage meetings run late
  • Updated Triage List: http://is.gd/geixlc
    • Updated to strip out bugs that hit on summary filters but have needinfos
  • Needinfo List: http://is.gd/vsPqdk
  • ACTION: cpeterson to investigate permanent named queries
  • Suggestion: people should pre-triage crap bugs before Tuesday/Thursday meetings
    • Dupes
    • Needinfo bugs until they become actionable, or close them if they never get there.

jimm

  • bug 1095754 (m4) - plugin geometry, wip. working on moving geometry information into the layer tree.
  • bug 1059775 (m4) / bug 1071751 (m6) - trying to sort out plugin-container shutdown issues that still exist. While I've never reproduced the wedged 100% cpu plugin-container bug, I have reproduced the slow shutdown in 1071751. In one case this happened with html5 video and content, plugins weren't involved.
  • e10s use by platform on nightly
    • windows: 70%
    • linux: 60%
    • mac:75%

ally

  • PDX workweek happened.
    • good job mconley & mrbkap!
      • spent lots of time with the amo folks
        • they will start requiring e10s compliance in new addons when we uplift to aurora
  • closed bug 1080159 - [e10s] investigate JS errors thrown by sync with e10s
  • closed bug 1080212 - [e10s] Add Telemetry probe for js exceptions that occur in the addon js compartment
    • telemetry team has verified that data is arriving
      • keyed histograms are a very new type of histogram
        • cpp support for them had to be added to complete
        • ui to view it will not be in place for a long time
        • may need to update probe at the telemetry group's request as they figure stuff out
  • bug 1071562 - [e10s] Support non-text content types for content process clipboard
    • copy/paste image will be back!
    • wip is posted
    • review count: 1.5 (3 waiting)

mconley

  • Talk slides: http://mikeconley.github.io/e10s-MM-CPOW-talk/
  • M4 - bug 1102410 - AboutProtocolChild attempts to register the same factory twice for multiple nsIAboutModule's
    • Fixed and landed!
    • This fixes the BugzillaJS bug with Interest Dashboard.
    • This appears to have fixed browser-child.js init bustage that caused a number of other bugs (history not working, autocomplete popups not working, etc)
  • M4 - bug 1108761 - Add class to let menulist > menupopup stylings apply to other popups
  • M4 - bug 1068349 - IPC message handlers that return false cause child process to crash without a crash report
    • Got review, but then bsmedberg showed ted and I that we aren't actually annotating the crash reports with the child process minidumps properly. I'm digging into why that is.
    • crash-stats, except for hangs, is not giong to show us the minidumps for both processes in the fancy web UI just yet. bsmedberg said there are bugs to make Socorro display minidumps from multiple processes, and I've needinfo'd him for the bug numbers.
  • M4 - bug 1090602 - <option> events do not bubble up to the parent <select>
    • ally, smaug and I came up with a tentative solution at the work week - I'm going to try to convince nsListControlFrame to just deal with the fact that it has no access to the combobox frame when running in the content process.
  • Question from ttaubert during the work week: are we at all worried about add-ons taking advantage of shared state in JSMs in the content process? If so, this advantage is, obviously, obliterated when we switch to multi-content process... is this worth worrying / messaging about?
  • Actions:
    • File a bug to try/catch around AddonShim initialization to prevent bug 1102410-like fallout (done -- bug 1110495)
    • File a bug to get Socorro to display minidumps from multiple processes and just let bsmedberg dupe (done -- bug 1110498)
    • Communicate with WIll Bamberg about documenting dangerous storage patterns in JSM's in the content process.

billm

  • almost done with CPOW timeout patch; more complicated than expected. Tests pass.
  • helping jchen on telemetry
  • slow script dialog?
    • make it harder for people to disable slow script dialog?
    • allow script script dialog to break into debugger?
  • investigating IPC hang on grooveshark

gw280

  • talking to UX about changing feed subscribe UI; land temporary workaround
  • debugging layers corruption bug

juanb

  • retest old plugin bugs new OS X plugin changes