RelMan set this going speculatively, prior to TBPL test results coming in. There were some failures which resulted in two changesets being backed out, and RelMan asked for build1 to be stopped. The only jobs that ran to complete were Firefox tagging and bouncer submission. Everything else was early enough in the job that it could be cancelled.
nthomas considered redoing build1 to avoid any steps that test for buildNumber > 1, but
- tagging should cope fine for Fennec even though we didn't tag build1, and it's like a normal build2 for Firefox
- Firefox bouncer_submitter runs for build1 only, but I deliberately let it run
- bumping of the patcher config doesn't test for buildNumber, checks if the 'to' version is stale instead, so this will happen in build2
- we'd have to reach into the ship-it db to change the release state, in order to modify the build1 already marked as completed there. And muck around with release configs to skip a tag for Firefox
So better to let build2 run and deal with any unforeseen bustage.
The android build failed in make buildsymbols, after running out of disk space. bug 1032637 for us to require more at the start, bug 1032632 for the build system to stop immediately on failure. Build was auto-retried.