What's next release?



  • Started using https://danielkummer.github.io/git-flow-cheatsheet/index.ru_RU.html ♪ I'm supposed to pick a branch. production release:

    Which branch should be used for bringing forth production releases?
       - DVR-1570
       - DVR-1573
       - develop
       - master
    

    Branch name for production releases: [master]

    It's clear here. It's not clear following:

    Which branch should be used for integration of the "next release"?

    • DVR-1570
    • DVR-1573
    • develop
      Branch name for "next release" development: [develop]

    What? next release? We've always had a bad release. They were all tested and left for the master, and the container broke.

    Tell me what it is. next release?



  • next release The branch is needed on a large team to dissolve the design and testing.

    Imagine, you have a team of ten developers. Each of them starts working on F1, F2, F10. Each of them is generated by some R1.

    Let's say it's a simple process when the branch of the figs goes straight into the master and turns on the food.

    When F1 is ready, it's fine, it's all right. But there's still problems. In F2, we have no change from F1 (the net is created from R1). There's a dilemma here: hope that changes in F1 and F2 do not affect each other (then F2 can be tested in the branch and slide into development) or slide down master F2 and test what F1 and F2 are working as necessary.

    Since the work is parallel, and we have a lot of flow of F2, there are problems with the fact that the same function (in this case F1) potentially needs to be tested in F2-F10 several times. And if you count what after the merger, say, F2 in master It will be necessary to do the same for F2, the number of such problems increases.

    There are a few ways to solve (or mitigate) the problem and schema. next release It's one of them. In this approach, each fitch is tested separately in a branch with a fig, then rolls in. next release One round of testing is carried out, and when all F1-F10 slits are carried out (this place is open). If it's okay, a release is made.

    Alternative ways to solve this problem (and wiping nets with fictits directly):

    1. Make changes in F1 - F10 unable to influence each other
    2. Good coverage of automatic tests (in fact, the need for re-testing is not a problem, because everything is automated and we can afford it).
    3. Keep small teams. At that time, the overhead costs of cross-testing are not as large (especially if paragraphs 1-2 are still at some level) as the costs of maintaining a more complex and slow process.

Log in to reply
 


Suggested Topics

  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2
  • 2