The discussions regarding the Julia release process started last year when it hit Julia 1.0. Yesterday, Stefan Karpinski, one of Julia's core developers shared its finalized release process giving details on the kind of releases, the stages of the release process, the phases of a release, and more.
“This information is collected from a small set of posts on discourse and conversations on Slack, so the information exists “out there”, but this blog post brings it all together in a single place. We may turn this post into an official document if it’s well-received,” Stefan wrote.
As with most programming languages that follow Semantic Versioning (SemVer), Julia has three types of releases: Patch, Minor, and Major.
A patch release will be represented by the last digit of Julia’s version number. It will include things like bug fixes, low-risk performance improvements, and documentation updates. The team plans to release a patch every month for the current active release branches, however, this will depend on the number of bug fixes. The team also plans to run PackageEvaluator (PkgEval) on the backports five days prior to the patch release. PkgEval is used to run tests for every registered package, update the web pages of Julia packages, and create status badges.
A minor release will be represented by the middle digit of Julia’s version number. Along with some bug fixes and new features, it will include changes that are unlikely to break your code and the package ecosystem. Any significant refactoring of the internals will also be included in the minor release. Since minor releases are branched every four months, developers can expect three minor releases every year.
A major release will be represented by the first digit of Julia’s version number. Typically, major releases consist of breaking changes, but the team assures to introduce them only when there is an absolute need, for instance, fixing API design mistakes. It will also include low-level changes that can end up breaking some libraries but are essential for fundamental improvements to the language.
There are three phases in the Julia release process. The development phase takes up 1-4 months where new features are introduced, bugs are fixed, and more. Before the feature freeze, alpha (early preview) and beta (later preview) versions are released for developers to test them and to share their feedback. After the feature freeze, a new unstable release branch is created. In the development phase, the new features will be merged onto the master branch, while the bug fixes will go on the release branch.
The second phase, stabilization, also takes up 1-4 months where all known release-blocking bugs are fixed and release candidates are built. Then they are checked for any more release-blocking bugs for one week and if there are none a final release is announced. After this, starts the maintenance phase where bug fixes are backported to the release branch. This continues till a particular release branch is declared to be unmaintained.
To ensure the quality of releases and maintaining a predictable release rate the Julia team overlaps the development and stabilization phases. “The development phase of each release is time-boxed at four months and the development phase of x.(y+1) starts as soon as the development phase for x.y is over. Come rain or shine we have a new feature freeze every four months: we pick a day and you’ve got to get your features merged by that day. If new features aren’t merged, they’re not going in the release. But that’s ok, they’ll go in the next one,” explains Karpinski.
Talking about long term support, Karpinski wrote that there will be four active branches. The master branch is where all the new features, bug fixes, and breaking changes will go. The unstable release branch will include all the active bug fixing and performance work that happens prior to the next minor release. The stable release branch is where the most recently released minor or major version exists. The fourth one is the long term support (LTS) branch, which is currently Julia 1.0. This branch continues to get applicable bug fixes until it is announced to be unmaintained.
Karpinski also shared the different fault tolerance personas in Julia. Check out his post on the Julia blog to get a better understanding of the Julia release process.
Julia announces the preview of multi-threaded task parallelism in alpha release v1.3.0
Julia Angwin fired as Editor-in-Chief of The Markup prompting mass resignations in protest
Creating a basic Julia project for loading and saving data [Tutorial]