I’d like to welcome John Hopkins to Release Engineering. I apologize that this welcome is a very belated welcome; John started in RelEng on 19th Sept and coop already posted a welcome here.
From one perspective, its not really any change here, John had already been coming to the RelEng group meetings since earlier this year when Mozilla Messaging folded back into Mozilla Corporation. Even before that, John was a familiar face to us, working on lots of the same buildbot and release automation code as the rest of RelEng.
From another perspective, there is potential for significant change here. Are there economies of scale if we merge the existing Thunderbird release engineering systems and the existing Firefox release engineering systems together? Now that we are in rapid-release cycles for both products, could we sim-ship Thunderbird with Firefox? How would we handle long term support issues for both products? Some things seem obvious, like how the tool chains for compiling-and-linking are identical for both products, so sharing machines and setup would help, and how we can improve Mozilla’s busfactor – jhopkins is the only person in RelEng who has shipped Thunderbird release since the old TB2 automation. Some things are less obvious, like what to do with the different spec hardware which was being used by MoMo. All this will play out over the coming months. Suggestions, comments, ideas all welcome!
As a good sign of how well everyone already works well together, the same week jhopkins officially joined RelEng, we discovered a serious problem: the production MoMo machines for Thunderbird were sitting on a shelf in the old Vancouver office, and these production machines were a few days away from being powered off as part of the move to a new Vancouver office. As the new office wasnt ready, so Vancouver people had to move to a temp-workspace for some weeks, and the Thunderbird production machines couldnt come to the temp-workspace offices. This meant the Thunderbird production machines would be offline, and the trees closed for the duration. No-one liked the idea of closing the Thunderbird trees for weeks, so we all piled in to help.
An immense volume of behind-the-scenes work took place, the Thunderbird continuous integration processes were migrated to our existing RelEng colos, toolchains setup on slightly different spec machines, and then production builds+tests enabled in the RelEng colo – all before the old Vancouver office was vacated, and all without closing the Thunderbird trees at all.
Bug#688230 has all the details. As I said in the “Friends of the Tree” section of the Mozilla Foundation weekly call at the time, this was an impressive volume of behind the scenes work, and developers using production systems never hit any problems!
Welcome John!