Labs/Jetpack/Weekly Meeting/2011-02-01

From MozillaWiki
Jump to: navigation, search



FlightDeck 1.0a8

  • All deliverables are on track for completion
  • Docs issue was found in the SDK, update to branch-1.0b2 latest for fix
  • Minor Skywriter issue found, Skywriter bug filed in Github

SDK 1.0b2 Review

  • the bad
    • the 1.0b2 tag was pushed to the canonical repository a couple days after the release, and pushing it isn't on the release checklist;
      • pushing it should be added to the checklist, and it should be pushed in conjunction with the release
    • the branch-1.0b2 branch didn't get deleted, and deleting it isn't on the release checklist;
      • deleting it should be added to the checklist, and it should be deleted in conjunction with the release
    • the 1.0b2 cycle started at the beginning of December, and some work was done the first week of December (before a MoCo work week and the holidays disrupted further development), but it was difficult to pick that work back up after the work week/holidays;
      • after the last release of the year, it might be better to give folks time to work on side projects for the rest of December and start the development cycle for the next release at the beginning of January
    • documentation code changes caused a regression we didn't catch before the release, which made API reference pages fail to load;
      • we should have some sort of automated testing of those pages, perhaps a cfx test that calls "cfx sdocs" to generate the page set, a python server to serve the pages, and a python spider to crawl them and check their links
    • the release driver finished writing the release blog post a few hours before the release, which didn't give anyone (including the release driver himself but also the PMM) a chance to review it;
      • for Wednesday releases, the release driver should finish the blog post by the Monday before the release to give folks review time
    • the release driver put together the development plan after the cycle had already begun, and it mostly reflected what folks were already working on rather than advance planning;
      • the release driver should put together the development plan before the cycle begins in order to plan the priorities and deliverables of the release in advance
    • deliverables that miss a release sometimes get lost instead of being reassigned to the next release as appropriate;
      • deliverables that miss a release should be reassigned to the next release as appropriate (i.e. not by default, but if those deliverables remain a priority)
  • the good
    • we released almost on time
    • we waved off large patches at the end of the cycle instead of landing them and scrambling to fix regressions
    • we didn't have to do a second RC (perhaps because we waved off large patches)

automated testing update

  • Brian, Myk, RelEng, Tools, and others continue to work on fixing infrastructure
  • currently waiting to see if Clint Talbert's patch fixes win7 buildslave hangs
  • after that must get all SDK tests working
  • will disable tests exhibiting intermittent failures if necessary to get green so Firefox/Gecko devs can trust SDK tests to indicate whether or not their changes cause regressions

/js/jetpack/ situation

  • bugs in /js/jetpack/ were noticed by platform developers last week, causing jetscramble to determine extent of problems and what to do about them
  • investigation revealed that /js/jetpack/ is not ready to ship in Firefox 4 and can't be made ready in time
  • thus /js/jetpack/ pulled from Firefox 4, which means SDK can't support OOP addons on that release
  • options moving forward include:
    • dropping OOP addons goal for SDK 1.0
    • retargeting SDK 1.0 to Firefox 5
    • staying in beta until OOP addons lands
  • myk to post to discussion group about the situation

SDK 1.0b3 status

  • most deliverables on track, except for Myk's e10s ones, which are all at risk
  • plan (particularly e10s goals) may change given /js/jetpack/ fallout


  • myk to post to discussion group about /js/jetpack/ situation
  • Create a bug for the future testing of SDK docs, Brian will be looking for a python script to aid in crawling the generated docs to verify things are operating as expected.