The development of Miralis is done through pull requests against the main
branch.
We strive to maintain a clean linear history for the main
branch, we rebase all PRs before merging and expect PRs to be rebased against the latest main
branch.
An explicit goal is to ensure that all commits in the main
branch pass the test suite of that commit, in other words just test
must always succeed.
To enforce this the CI run the tests against each new commit when submitting a PR, if you see a failure in the CI check for the details to find out which commit caused the issue.
Of course writing code requires iteration, a good rule of thumb is to write a first version while committing along the way, and to rework those commits in a second time using tools such as git rebase --interactive
or jj.
This section describes the style we strive to enforce across the code base, and serves as a reference when arbitrary choices need to be made.
Comments:
Comment starts with a leading white space and a capital letter, like this:
// Our comment style
But not like this:
//We avoid this
// or this
//and this
Rust code:
We always insert a blank line between two functions:
fn foo() {
bar();
}
fn bar() { // See how there is a blank line before
baz();
}
The only exception is within trait definitions for function signatures:
trait MyTrait {
// This is OK
fn foo();
fn bar();
// But this is **not** OK
fn baz() {
foo();
}
fn fuzz() { // See how there is no blank line above? This is **not** OK
baz();
}
}
Markdown:
We write markdown documentation with exactly one sentence per line, like this:
We always have only one sentence per line.
This makes it easy to review changes as the diff shows exactly which sentences changed.
In particular we do not wrap sentences across multiple line in .md
files (but we do in Rust comments!):
We **never** wrap a sentence
across multiple lines.