Harvesting releases

This month was a very interesting one.

I had a chance to be involved into 6 (!) release processes: 3.7a1, 4.0b1 (2 builds), 3.6.6, 3.6.7 and 3.5.11. All of these builds were unique (at least for me).

3.7a5

Last alpha with a different naming (MozillaDeveloperPreview). We introduced linux64 and macosx64 platforms in this release. Lucky me, the build environment for these platforms was carefully prepared and tested by Armen and Bear beforehand. During the preparation for this release, RelEng resolved some annoying bugs, which reduced manual intervention into the release process.

4.0b1

Not released yet. First branded version of Firefox 4 built for 5 platforms. Due to some discovered bugs we had to wait a day or two and produce build2.

3.6.6

Stable release with some fixes. Nothing unusual except the previous product version, 3.6.4 (not 3.6.5), and some fun with forcing L10N repacks. Despite of the fact that the time when we started the build wasn't ideal (Friday night, my Saturday morning) we released it in less than 24 hours.

It is the fastest release in RelEng history. It's pleasure being a part of history. :)

3.6.7

Not released yet. Available for the beta users. We had to run this release in parallel with 3.5.11. Needed some sed magic for snippets (thanks to Nick Thomas) to reduce server load and use to the mirrors for the beta channel updates. A lot of fun with producing Major Updates (MU) for Firefox 3.0.19 manually.

3.5.11

Not released yet old stable version. Available for the beta users. The build was done in parallel with 3.6.7. As a part of this build we also produced MUs for 3.5.x -> 3.6.7. MUs were done by release automation.

As a result, now I have much more clearer understanding of the release process, the release work flow and the release infrastructure.

Special thanks go to Ben Hearsum, Chris AtLee and Nick Thomas for being great supervisors!

Comments

Comments powered by Disqus