There are a number of levels you have to go through before your continuous integration pipeline becomes perfect. I found eight of them and presented my findings at DevOpsDays in Salt Lake City a few weeks ago (watch the video). Now it's time to write them down and ask you — Which level are you at? Post your answer below.
Twins (1988) by Ivan Reitman
- Source Code. Here you just write source code on your computers and maybe somewhere on the server. The best you can do here is to build it manually, say, every day. Is it continuous integration? Well, to some extent, provided you don't forget to compile and package everything regularly.
- Automated Build. At this level, your build is automated, which means you can compile, test, and package the entire product with just one line at the command line. Pay attention; one line. You must be able to hit one button and either get an error or a successful build.
- Git. At this level, you keep your source code in Git. You can keep it in some other source control system, but that would be strange — Git is the status quo at the moment. You should be able to get a new computer, with nothing in it, check out the source code from a Git repository, and run a full build.
- Pull Requests. Each and every change to your source code must be submitted through a pull request, which means that you host your repository on GitHub. You may host it somewhere else, but again, that would be strange because GitHub is the status quo at the moment. Again, nobody should be able to commit anything directly to the master branch except through forks and pull requests.
- Code Reviews. Every pull request must pass a mandatory code review before it gets merged. You must have some code review policy that explains who does reviews, what happens if the author doesn't agree with the reviewer, etc. But no pull request may be merged unless it has been reviewed.
- Tests. At this level, your code is covered by unit tests (and integration tests), and every change comes with a new test. Your automated build runs all tests together and fails if any of them fail.
- Static Analysis. Checking the quality of your code without running it is what static analysis is about. At this level, the quality of your code is checked by the automated build. If the quality is lower than the threshold, the build fails.
By the way, at the presentation, I also mentioned what problems you may encounter at each maturity level.