This week in Mozilla RelEng – February 14th, 2014

Ben Hearsum

Highlights:

Aki spent most of his week braindumping Mozharness and VCS Sync knowledge before he goes on PTO, which will help to ensure we don't drop the ball on anything while he's gone.
Pete migrated repositories from the legacy vcs2vcs system to the much more stable VCS Sync system. Moving these reduces the chance of a tree closure.
Geoff turned on full ADB logcat logs for many Android tests. This should help significantly when debugging test failures on RelEng machines.
Glandium enabbled sccache for some try builds, which is expected to improve build times by having a much better cache hit rate than ccache.
Many of us started early work on jacuzzis, which will also improve build times when we're ready to deploy it more widely.
Rail increased our usage of AWS spot instances, which lowers our AWS bill with no impact to turnaround times.
Mike thinks he found the reason why "hg purge" fails in some circumstances. Once addressed, we can spend less time recreating full source directories, especially during try builds.
Armen enabled testing of b2g reftests on EC2 machines. Once problems with these are sorted out, we can move all of these tests to EC2, which allows us to achieve better reliability and scale for them.