Node / io.js Build Convergence Meeting #98

There's a meeting today between some @iojs/build folk and some joyent/node folk including Alexis Campala / @orangemocha and Julien Gilli / @misterdjules, approximately 3 hours from now.

YouTube feed: http://www.youtube.com/watch?v=8dxkM9vHmrY

Anyone wanting to participate should contact me directly for a hangout link

Here's an initial agenda that @orangemocha put together:

High level convergence plan (work in progress):

• Take io.js as starting point, bring over few things from Node.js
  • What to bring over from node.js?
  • Hardware assets? The list of Jenkins agents is available here: http://jenkins.nodejs.org/computer/.
    • Most agents run on containers or VMs hosted in Joyent's public cloud. osx-build is a mac machine that is also owned by Joyent.
    • The Windows agents run on an Azure subscription sponsored by Microsoft. Currently 4 machines running for a total of 28 cores.
  • Repo credentials to ensure we can push to Node LTS repo
  • Users. Grant accounts on io.js Jenkins
  • Distribution site? Shall we unify?
  • Jenkins Domain name?
  • Automatic merge job (node-accept-pull-request)?
  • Other Node.js Jenkins jobs?
• Extend this infrastructure to Node LTS release branches, and convergence repo?
• Move the build repository (and its issues) to the Node.js foundation org. Follow the same strategy as per repo convergence plan.
• Release methodologies might differ for a while (Node.js just started doing pre-releases) and are beyond the scope of the CI convergence, even though it might put requirements on CI.

Build working group membership:

• Does it make sense for some folks from Node.js (eg Julien and/or Alexis) to join, at least for the sake of bringing stuff back together?
• How do you guys collaborate? Do you have meetings? On Australian time? :)

Info on node-accept-pull-request:

• A bit of documentation is here: https://nodejs.org/documentation/workflow/#index_md_using_jenkins_to_build_test_and_merge_every_pr
• Pretty much 0 test regressions since we adopted it
• Flaky tests removed in io. PR: test: remove flaky test functionality node#812
• Need to do a better job at documenting how this works. But let’s first agree on the high level plan
• High priority improvements:
  • Clarify/document process for handling flaky tests
  • Add metadata for issues fixed
  • Make it easier to enter reviewer names. Parse LGTM? Ideally just pre-populate a form
  • Enable simple edits of commit message(s). Maybe just for single-commit PRs.
  • Extend to more branches
  • Comment on PR with link to results

Discovery (mostly for the benefit of Alexis & Julien, can be done asynchronously after the call):

• Did you start from the Node.js Jenkins jobs or something else?
• All jobs triggered by commits? On all branches?
• “Commits to the repository are tested on the full set while pull requests to the Node.js and libuv projects from non-core contributors are tested on a smaller, more secure subset.” How is this accomplished?
• Where is the process about committing changes (running CI) documented?
• test-simple vs test-all. Did you define different test configurations, with overlap? How?
• https://github.com/iojs/build#configurations-all-code So you don’t test commits from non-team members on all core platforms?
• Do regressions happen? Some failure on Windows: https://jenkins-iojs.nodesource.com/job/iojs+any-pr+multi/lastCompletedBuild/nodes=win2012r2/tapTestReport/
• State of the build (State of the build (io.js) April 2015 #77) mentions automated PR testing. Status?
• Still using VC 2012? I thought io.js’ version of v8, required 2013.
• Job configuration backups?
• Handling passwords. Should they all be in a centralized spreadsheet?