Releases/Thunderbird 3.0b2/Post Mortem

From MozillaWiki
Jump to: navigation, search

Post Mortem for the Thunderbird release of 3.0b2

What went well


  • Process went musch smoother than with beta1
  • Benefitted from beta1 learning curve
  • Added 5 locales compared with beta1
  • In general the release schedule worked well for us
  • mail/locales/shipped-locales could/should be updated by someone from l10n

What could have gone better, and lessons learned

Leading up to the release

  • In reference to the Development Page page.
    • Weekly milestones are necessary
      • Create a milestone for every week in release cycle
      • Name milestones: ${REL}week${NUM} ex: b2week1, b2week2, etc
      • Plan landing of items (features/bugs) in each week
      • Set status whiteboard of bugs with milestone names (makes them searchable)
      • Work with Q/A to plan sessions testing milestone one week after it lands
      • Create search links for each week milestone
    • Release dates should have indicator of strength
      • Something like keywords strong, weak, or likely to slip to communicate how solid the dates are at this point
  • should release process planning meeting be removed from 3.0b2? (hasn't been done for last 2 releases) Should it simply be replaced with references to prior releases' postmortems?

String Freeze

Code Freeze

Release Notes


  • Overall, best build so far.
  • Branding changes could be better automated: bug 481374
  • --official-branding needs to be watched for more carefully
  • Signing is still blocking on MoCo, need to get that done ourselves
  • Need a tracking bug for landing on the release branch


  • Need to avoid mixing Litmus runs and Test Days - as it confuses people. This should be fixed with better communication (see 2.1 above).
  • Need to run upgrade from previous version tests to avoid things like bug 480556.


  • Lost 3 locales that were in beta1, hopefully can get them back for beta3 or the final
  • Still need to work harder on the communication between the release driver and the l10n coordinator (on both sides). Need to work harder to ensure that l10n is in everyone's thought process when it comes to releasing
  • release schedule finalization came very late. Would be good to get this fixed early on in the beta3 cycle
  • Lessons learned: Compare-locales (l10n dashboard) does not see everything, especially with regards to the windows installer. Need to watch the tinderboxen as well early on to more effectively communicate necessary fixes to localizers
  • Need to clarify the difference between slushy and firm string freeze. We did string-only landings again. We need to avoid this in the upcoming releases
  • Reported 2009-03-21 on IRC "KaiRo wsmwk/sid0: from the error message, it looks like addressBookCmd.commandkey is missing on the L10n" for de_DE, and the build process was slightly off (see below). A second issue that let this get out is most l10n RC builds were probably not at least smoketested.
<KaiRo>	yes, as he said it's after the revision he posted in .l10n so we should be safe
of course, only if that revision was also tagged and used for b2 and not the "current" state of the tree at release time
<sid0>	so anyway, the de release of b2 has a missing entity
<KaiRo>	and I suspect that's the problem here, probably the manual process gozer still used this time doesn't care about using specific revisions
<sid0>	sigh
<KaiRo>	the automated process Firefox is already using and which we will use in the future does actually go with a list of specific signed-off revisions

Anything else

  • need to firm up crash-stats.m.c update step, seems to be insufficient to ask for adding the release and next development release. ref bug 480728 MTBF missing Thunderbird 3.0b2 milestone, and development 3.0b3pre & 3.1a1pre // 3.0b2 topcrash no results