Build 1

Hit this in release-runner/release-sanity:

2016-05-19 16:29:39,222 : INFO : Comparing l10n changesets on dashboard https://l10n.mozilla.org/shipping/l10n-changesets?ms=tb45_beta_b1 to on-disk buildbot-configs/mozilla/l10n-changesets_thunderbird-beta ...
2016-05-19 16:29:39,273 : ERROR :       locale ja-JP-mac revisions not matching: 2f89a61f1eed (config) vs. cdfbfaf4e90a (dashboard)
2016-05-19 16:29:39,273 : ERROR :       locale it revisions not matching: dcbcdc8d2cee (config) vs. b8f605dce379 (dashboard)
...
2016-05-19 16:29:39,275 : ERROR :       locale uk revisions not matching: 5cef470a2a46 (config) vs. 34582a9552a0 (dashboard)
2016-05-19 16:29:39,275 : ERROR : Error verifying l10n dashboard changesets

The code doesn't know how to cope with this versioning scheme so it makes a bad request to l10n. We hit this for 45.1b1 too and ended up disabling the l10n dashboard check. nthomas went ahead and did that, but at the same time wswmk - the latter submitting build2 and starting it. Tagging and source for build2 were cancelled while still in the mock install step, so no damage done.

Build 2

• (TB Drivers) Ship the l10n milestone (build1 only)
• (TB Drivers) Submit to Ship It
• Publish in Balrog
• Post-release tasks

Issues

• repacks started retrying a lot, hundreds of times a lot; filed as bug 1274454. This turned out to be an issue with the hg bundles for build/tools expiring on S3, bug 1274456. Once that was resolved almost the repack jobs were rebuilt by forcing jobs with script_repo_revision and release_config set as properties. Only 4/10 for linux and {2,7,10}/10 for linux64 had succeeded earlier, by virtue of running in us-east-1 where a bundle was still available.
• linux and linux64 update verify tests are failing due to missing gtk3 libs as expected
• win32 update verify tests have failures but the tb team is progressing with user testing
• anti-virus failures too
• some repacks were rerun and finished after the push to mirrors, recovery on the gisthub