Hello fellow Rustacean! Great to see your interest in compiler internals and lints!
First: if you're unsure or afraid of anything, just ask or submit the issue or pull request anyway. You won't be yelled at for giving it your best effort. The worst that can happen is that you'll be politely asked to change something. We appreciate any sort of contributions, and don't want a wall of rules to get in the way of that.
Clippy welcomes contributions from everyone. There are many ways to contribute to Clippy and the following document
explains how you can contribute and how to get started. If you have any questions about contributing or need help with
anything, feel free to ask questions on issues or visit the #clippy
on Zulip.
All contributors are expected to follow the Rust Code of Conduct.
If you're new to Clippy and don't know where to start, the Clippy book includes a developer guide and is a good place to start your journey.
- Find something to fix/improve
- Change code (likely some file in
clippy_lints/src/
) - Follow the instructions in the Basics docs to get set up
- Run
cargo test
in the root directory and wiggle code until it passes - Open a PR (also can be done after 2. if you run into problems)
All issues on Clippy are mentored, if you want help simply ask someone from the Clippy team directly by mentioning them in the issue or over on Zulip. All currently active team members can be found here
Some issues are easier than others. The good-first-issue
label can be used to find the easy
issues. You can use @rustbot claim
to assign the issue to yourself.
There are also some abandoned PRs, marked with S-inactive-closed
.
Pretty often these PRs are nearly completed and just need some extra steps
(formatting, addressing review comments, ...) to be merged. If you want to
complete such a PR, please leave a comment in the PR and open a new one based
on it.
Issues marked T-AST
involve simple matching of the syntax tree structure,
and are generally easier than T-middle
issues, which involve types
and resolved paths.
T-AST
issues will generally need you to match against a predefined syntax structure.
To figure out how this syntax structure is encoded in the AST, it is recommended to run
rustc -Z unpretty=ast-tree
on an example of the structure and compare with the nodes in the AST docs.
Usually the lint will end up to be a nested series of matches and ifs, like so.
But we can make it nest-less by using let chains, like this.
E-medium
issues are generally pretty easy too, though it's recommended you work on an good-first-issue
first. Sometimes they are only somewhat involved code wise, but not difficult per-se.
Note that E-medium
issues may require some knowledge of Clippy internals or some
debugging to find the actual problem behind the issue.
T-middle
issues can be more involved and require verifying types. The ty
module contains a
lot of methods that are useful, though one of the most useful would be expr_ty
(gives the type of
an AST expression). match_def_path()
in Clippy's utils
module can also be useful.
Unfortunately, IntelliJ Rust
does not (yet?) understand how Clippy uses compiler-internals
using extern crate
and it also needs to be able to read the source files of the rustc-compiler which are not
available via a rustup
component at the time of writing.
To work around this, you need to have a copy of the rustc-repo available which can be obtained via
git clone https://github.com/rust-lang/rust/
.
Then you can run a cargo dev
command to automatically make Clippy use the rustc-repo via path-dependencies
which IntelliJ Rust
will be able to understand.
Run cargo dev setup intellij --repo-path <repo-path>
where <repo-path>
is a path to the rustc repo
you just cloned.
The command will add path-dependencies pointing towards rustc-crates inside the rustc repo to
Clippy's Cargo.toml
s and should allow IntelliJ Rust
to understand most of the types that Clippy uses.
Just make sure to remove the dependencies again before finally making a pull request!
For rust-analyzer
to work correctly make sure that in the rust-analyzer
configuration you set
{ "rust-analyzer.rustc.source": "discover" }
You should be able to see information on things like Expr
or EarlyContext
now if you hover them, also
a lot more type hints.
To have rust-analyzer
also work in the clippy_dev
and lintcheck
crates, add the following configuration
{
"rust-analyzer.linkedProjects": [
"./Cargo.toml",
"clippy_dev/Cargo.toml",
"lintcheck/Cargo.toml",
]
}
clippy_lints/src/lib.rs
imports all the different lint modules and registers in the LintStore
.
For example, the else_if_without_else
lint is registered like this:
// ./clippy_lints/src/lib.rs
// ...
pub mod else_if_without_else;
// ...
pub fn register_plugins(store: &mut rustc_lint::LintStore, sess: &Session, conf: &Conf) {
// ...
store.register_early_pass(|| box else_if_without_else::ElseIfWithoutElse);
// ...
store.register_group(true, "clippy::restriction", Some("clippy_restriction"), vec![
// ...
LintId::of(&else_if_without_else::ELSE_IF_WITHOUT_ELSE),
// ...
]);
}
The rustc_lint::LintStore
provides two methods to register lints:
register_early_pass and register_late_pass. Both take an object
that implements an EarlyLintPass
or LateLintPass
respectively. This is done in
every single lint. It's worth noting that the majority of clippy_lints/src/lib.rs
is autogenerated by cargo dev update_lints
. When you are writing your own lint, you can use that script to save you some time.
// ./clippy_lints/src/else_if_without_else.rs
use rustc_lint::{EarlyLintPass, EarlyContext};
// ...
pub struct ElseIfWithoutElse;
// ...
impl EarlyLintPass for ElseIfWithoutElse {
// ... the functions needed, to make the lint work
}
The difference between EarlyLintPass
and LateLintPass
is that the methods of the EarlyLintPass
trait only provide
AST information. The methods of the LateLintPass
trait are executed after type checking and contain type information
via the LateContext
parameter.
That's why the else_if_without_else
example uses the register_early_pass
function. Because the
actual lint logic does not depend on any type information.
Clippy is following the Rust triage procedure for issues and pull requests.
However, we are a smaller project with all contributors being volunteers currently. Between writing new lints, fixing issues, reviewing pull requests and responding to issues there may not always be enough time to stay on top of it all.
Our highest priority is fixing crashes and bugs, for example an ICE in a popular crate that many other crates depend on. We don't want Clippy to crash on your code and we want it to be as reliable as the suggestions from Rust compiler errors.
We have prioritization labels and a sync-blocker label, which are described below.
- P-low: Requires attention (fix/response/evaluation) by a team member but isn't urgent.
- P-medium: Should be addressed by a team member until the next sync.
- P-high: Should be immediately addressed and will require an out-of-cycle sync or a backport.
- L-sync-blocker: An issue that "blocks" a sync. Or rather: before the sync this should be addressed, e.g. by removing a lint again, so it doesn't hit beta/stable.
We use a bot powered by Homu to help automate testing and landing of pull requests in Clippy. The bot's username is @bors.
You can find the Clippy bors queue here.
If you have @bors permissions, you can find an overview of the available commands here.
Contributions to Clippy should be made in the form of GitHub pull requests. Each pull request will be reviewed by a core contributor (someone with permission to land patches) and either landed in the main tree or given feedback for changes that would be required.
All PRs should include a changelog
entry with a short comment explaining the change. The rule of thumb is basically,
"what do you believe is important from an outsider's perspective?" Often, PRs are only related to a single property of a
lint, and then it's good to mention that one. Otherwise, it's better to include too much detail than too little.
Clippy's changelog is created from these comments. Every release, someone gets all commits from bors with a
changelog: XYZ
entry and combines them into the changelog. This is a manual process.
Examples:
- New lint
changelog: new lint: [`missing_trait_methods`]
- False positive fix
changelog: Fix [`unused_peekable`] false positive when peeked in a closure or called as `f(&mut peekable)`
- Purely internal change
changelog: none
Note this it is fine for a PR to include multiple changelog
entries, e.g.:
changelog: Something 1
changelog: Something 2
changelog: Something 3
All code in this repository is under the Apache-2.0 or the MIT license.