Rolling, rolling, rolling back nothing I love more than communicating why we had to roll back again (⁠ノ⁠ಠ⁠益⁠ಠ⁠)⁠ノ

  • kkj@lemmy.dbzer0.comEnglish
    2·
    14 hours ago

    What do you do if you have code that isn’t complete enough to work? Do you have to just leave it untracked?

    • chellomere@lemmy.world
      1·
      12 hours ago

      I don’t know what others do, but I personally whip out git commit -n and bypass the hooks in this situation.

    • neonred@lemmy.world
      21·
      13 hours ago

      If you have code that is not complete it is not qualified to be deployed. Cut work items into smaller chunks but never deploy not fully, 100% working and tested stuff. Not even on dev.

      • kkj@lemmy.dbzer0.comEnglish
        2·
        12 hours ago

        Every branch you have deploys on commit? You have to fully QA all of your code before it goes into any sort of source control?

        • neonred@lemmy.world
          1·
          12 hours ago

          Not quite.

          • Every commit is just a local commit
          • Every push runs pre-hooks which execute bunch of checks, for example linters, style checkers, etc. and prevent a push if something is not perfect
          • After every push the CI/CD pipeline runs on origin
          • Every run of the pipeline executes again checks with linters but also securoty checks for CVEs on dependencies and runtime
          • Every pipeline run also executes all tests such as unit tests, scenario tests, integration tests
          • If any of the above fails, the pipeline fails and stops
          • Only if everything is okay, one can deploy on dev, the first stage
          • Only if this is okay, the artifact gets pushed to the central artifact store
          • Only if this suceesa a prod deployment can run, which pulls the artifact from the store
          • Runners for dev and prod are distinct and don’t have rights the other has, the only common contact point is the artifact store

          That’s an extremely very basic overview with many steps and concepts omitted but you get the idea.

          • kkj@lemmy.dbzer0.comEnglish
            2·
            11 hours ago

            That seems reasonable to perform on protected branches, but I’m not a fan of protecting all branches. That could leave valuable code with a single copy on a dev machine. I’d rather have it pushed to an unprotected branch and then be checked on merge instead of push.

          • chellomere@lemmy.world
            1·
            12 hours ago

            So, what if I want to push some debug or preliminary code to a topic branch, would this system prevent this if all tests don’t pass?

            • neonred@lemmy.world
              1·
              10 hours ago

              No, it does not prevent pushing (as long as the pre-hooks work) but you cannot deploy from a failed pipeline/branch because you have defective software, as proven by failed tests.