Clippy
A collection of lints to catch common mistakes and improve your Rust code.
There are over 550 lints included in this crate!
Lints are divided into categories, each with a default lint
level. You can choose how
much Clippy is supposed to annoy help you by changing the lint level by
category.
Category | Description | Default level |
---|---|---|
clippy::all | all lints that are on by default (correctness, suspicious, style, complexity, perf) | warn/deny |
clippy::correctness | code that is outright wrong or useless | deny |
clippy::suspicious | code that is most likely wrong or useless | warn |
clippy::complexity | code that does something simple but in a complex way | warn |
clippy::perf | code that can be written to run faster | warn |
clippy::style | code that should be written in a more idiomatic way | warn |
clippy::pedantic | lints which are rather strict or might have false positives | allow |
clippy::nursery | new lints that are still under development | allow |
clippy::cargo | lints for the cargo manifest | allow |
More to come, please file an issue if you have ideas!
The lint list also contains "restriction lints", which are for things which are usually not considered "bad", but may be useful to turn on in specific cases. These should be used very selectively, if at all.
Installation
If you're using rustup
to install and manage you're Rust toolchains, Clippy is
usually already installed. In that case you can skip this chapter and go to
the Usage chapter.
Note: If you used the
minimal
profile when installing a Rust toolchain, Clippy is not automatically installed.
Using Rustup
If Clippy was not installed for a toolchain, it can be installed with
$ rustup component add clippy [--toolchain=<name>]
From Source
Take a look at the Basics chapter in the Clippy developer guide to find step by step instructions on how to build and install Clippy from source.
Usage
This chapter describes how to use Clippy to get the most out of it. Clippy can
be used as a cargo
subcommand or, like rustc
, directly with the
clippy-driver
binary.
Note: This chapter assumes that you have Clippy installed already. If you're not sure, take a look at the Installation chapter.
Cargo subcommand
The easiest and most common way to run Clippy is through cargo
. To do that,
just run
cargo clippy
Lint configuration
The above command will run the default set of lints, which are included in the
lint group clippy::all
. You might want to use even more lints or you might not
agree with every Clippy lint, and for that there are ways to configure lint
levels.
Note: Clippy is meant to be used with a generous sprinkling of
#[allow(..)]
s through your code. So if you disagree with a lint, don't feel bad disabling them for parts of your code or the whole project.
Command line
You can configure lint levels on the command line by adding
-A/W/D clippy::lint_name
like this:
cargo clippy -- -Aclippy::style -Wclippy::double_neg -Dclippy::perf
For CI all warnings can be elevated to errors which will inturn fail
the build and cause Clippy to exit with a code other than 0
.
cargo clippy -- -Dwarnings
Note: Adding
-D warnings
will cause your build to fail if any warnings are found in your code. That includes warnings found by rustc (e.g.dead_code
, etc.).
For more information on configuring lint levels, see the rustc documentation.
Even more lints
Clippy has lint groups which are allow-by-default. This means, that you will have to enable the lints in those groups manually.
For a full list of all lints with their description and examples, please refer to Clippy's lint list. The two most important allow-by-default groups are described below:
clippy::pedantic
The first group is the pedantic
group. This group contains really opinionated
lints, that may have some intentional false positives in order to prevent false
negatives. So while this group is ready to be used in production, you can expect
to sprinkle multiple #[allow(..)]
s in your code. If you find any false
positives, you're still welcome to report them to us for future improvements.
FYI: Clippy uses the whole group to lint itself.
clippy::restriction
The second group is the restriction
group. This group contains lints that
"restrict" the language in some way. For example the clippy::unwrap
lint from
this group won't allow you to use .unwrap()
in your code. You may want to look
through the lints in this group and enable the ones that fit your need.
Note: You shouldn't enable the whole lint group, but cherry-pick lints from this group. Some lints in this group will even contradict other Clippy lints!
Too many lints
The most opinionated warn-by-default group of Clippy is the clippy::style
group. Some people prefer to disable this group completely and then cherry-pick
some lints they like from this group. The same is of course possible with every
other of Clippy's lint groups.
Note: We try to keep the warn-by-default groups free from false positives (FP). If you find that a lint wrongly triggers, please report it in an issue (if there isn't an issue for that FP already)
Source Code
You can configure lint levels in source code the same way you can configure
rustc
lints:
#![allow(clippy::style)] #[warn(clippy::double_neg)] fn main() { let x = 1; let y = --x; // ^^ warning: double negation }
Automatically applying Clippy suggestions
Clippy can automatically apply some lint suggestions, just like the compiler.
cargo clippy --fix
Workspaces
All the usual workspace options should work with Clippy. For example the
following command will run Clippy on the example
crate in your workspace:
cargo clippy -p example
As with cargo check
, this includes dependencies that are members of the
workspace, like path dependencies. If you want to run Clippy only on the
given crate, use the --no-deps
option like this:
cargo clippy -p example -- --no-deps
Using Clippy without cargo
: clippy-driver
Clippy can also be used in projects that do not use cargo. To do so, run
clippy-driver
with the same arguments you use for rustc
. For example:
clippy-driver --edition 2018 -Cpanic=abort foo.rs
Note:
clippy-driver
is designed for running Clippy and should not be used as a general replacement forrustc
.clippy-driver
may produce artifacts that are not optimized as expected, for example.
Configuring Clippy
Note: The configuration file is unstable and may be deprecated in the future.
Some lints can be configured in a TOML file named clippy.toml
or .clippy.toml
. It contains a
basic variable = value
mapping eg.
avoid-breaking-exported-api = false
disallowed-names = ["toto", "tata", "titi"]
cognitive-complexity-threshold = 30
See the list of lints for more information about which lints can be configured and the meaning of the variables.
To deactivate the "for further information visit lint-link" message you can define the CLIPPY_DISABLE_DOCS_LINKS
environment variable.
Allowing/denying lints
You can add options to your code to allow
/warn
/deny
Clippy lints:
-
the whole set of
Warn
lints using theclippy
lint group (#![deny(clippy::all)]
) -
all lints using both the
clippy
andclippy::pedantic
lint groups (#![deny(clippy::all)]
,#![deny(clippy::pedantic)]
). Note thatclippy::pedantic
contains some very aggressive lints prone to false positives. -
only some lints (
#![deny(clippy::single_match, clippy::box_vec)]
, etc.) -
allow
/warn
/deny
can be limited to a single function or module using#[allow(...)]
, etc.
Note: allow
means to suppress the lint for your code. With warn
the lint will only emit a warning, while with deny
the lint will emit an error, when triggering for your code. An error causes clippy to exit with an error code, so is
useful in scripts like CI/CD.
If you do not want to include your lint levels in your code, you can globally enable/disable lints by passing extra flags to Clippy during the run:
To allow lint_name
, run
cargo clippy -- -A clippy::lint_name
And to warn on lint_name
, run
cargo clippy -- -W clippy::lint_name
This also works with lint groups. For example you can run Clippy with warnings for all lints enabled:
cargo clippy -- -W clippy::pedantic
If you care only about a single lint, you can allow all others and then explicitly warn on the lint(s) you are interested in:
cargo clippy -- -A clippy::all -W clippy::useless_format -W clippy::...
Specifying the minimum supported Rust version
Projects that intend to support old versions of Rust can disable lints pertaining to newer features by specifying the minimum supported Rust version (MSRV) in the clippy configuration file.
msrv = "1.30.0"
The MSRV can also be specified as an inner attribute, like below.
#![feature(custom_inner_attributes)] #![clippy::msrv = "1.30.0"] fn main() { ... }
You can also omit the patch version when specifying the MSRV, so msrv = 1.30
is equivalent to msrv = 1.30.0
.
Note: custom_inner_attributes
is an unstable feature so it has to be enabled explicitly.
Lints that recognize this configuration option can be found here
Clippy's Lints
Clippy offers a bunch of additional lints, to help its users write more correct and idiomatic Rust code. A full list of all lints, that can be filtered by category, lint level or keywords, can be found in the Clippy lint documentation.
This chapter will give an overview of the different lint categories, which kind of lints they offer and recommended actions when you should see a lint out of that category. For examples, see the Clippy lint documentation and filter by category.
The different lint groups were defined in the Clippy 1.0 RFC.
Correctness
The clippy::correctness
group is the only lint group in Clippy which lints are
deny-by-default and abort the compilation when triggered. This is for good
reason: If you see a correctness
lint, it means that your code is outright
wrong or useless and you should try to fix it.
Lints in this category are carefully picked and should be free of false
positives. So just #[allow]
ing those lints is not recommended.
Suspicious
The clippy::suspicious
group is similar to the correctness lints in that it
contains lints that trigger on code that is really sus and should be fixed. As
opposed to correctness lints, it might be possible that the linted code is
intentionally written like it is.
It is still recommended to fix code that is linted by lints out of this group
instead of #[allow]
ing the lint. In case you intentionally have written code
that offends the lint you should specifically and locally #[allow]
the lint
and add give a reason why the code is correct as written.
Complexity
The clippy::complexity
group offers lints that give you suggestions on how to
simplify your code. It mostly focuses on code that can be written in a shorter
and more readable way, while preserving the semantics.
If you should see a complexity lint, it usually means that you can remove or replace some code and it is recommended to do so. However, if you need the more complex code for some expressiveness reason, it is recommended to allow complexity lints on a case-by-case basis.
Perf
The clippy::perf
group gives you suggestions on how you can increase the
performance of your code. Those lints are mostly about code that the compiler
can't trivially optimize, but has to be written in a slightly different way to
make the optimizer's job easier.
Perf lints are usually easy to apply and it is recommended to do so.
Style
The clippy::style
group is mostly about writing idiomatic code. Because style
is subjective, this lint group is the most opinionated warn-by-default group in
Clippy.
If you see a style lint, applying the suggestion usually makes your code more
readable and idiomatic. But because we know that this is opinionated, feel free
to sprinkle #[allow]
s for style lints in your code or #![allow]
a style lint
on your whole crate if you disagree with the suggested style completely.
Pedantic
The clippy::pedantic
group makes Clippy even more pedantic. You can enable
the whole group with #![warn(clippy::pedantic)]
in the lib.rs
/main.rs
of
your crate. This lint group is for Clippy power users that want an in depth
check of their code.
Note: Instead of enabling the whole group (like Clippy itself does), you may want to cherry-pick lints out of the pedantic group.
If you enable this group, expect to also use #[allow]
attributes generously
throughout your code. Lints in this group are designed to be pedantic and false
positives sometimes are intentional in order to prevent false negatives.
Restriction
The clippy::restriction
group contains lints that will restrict you from
using certain parts of the Rust language. It is not recommended to enable
the whole group, but rather cherry-pick lints that are useful for your code base
and your use case.
Note: Clippy will produce a warning if it finds a
#![warn(clippy::restriction)]
attribute in your code!
Lints from this group will restrict you in some way. If you enable a restriction
lint for your crate it is recommended to also fix code that this lint triggers
on. However, those lints are really strict by design and you might want to
#[allow]
them in some special cases, with a comment justifying that.
Cargo
The clippy::cargo
group gives you suggestions on how to improve your
Cargo.toml
file. This might be especially interesting if you want to publish
your crate and are not sure if you have all useful information in your
Cargo.toml
.
Continuous Integration
It is recommended to run Clippy on CI with -Dwarnings
, so that Clippy lints
prevent CI from passing. To enforce errors on warnings on all cargo
commands
not just cargo clippy
, you can set the env var RUSTFLAGS="-Dwarnings"
.
We recommend to use Clippy from the same toolchain, that you use for compiling
your crate for maximum compatibility. E.g. if your crate is compiled with the
stable
toolchain, you should also use stable
Clippy.
Note: New Clippy lints are first added to the
nightly
toolchain. If you want to help with improving Clippy and have CI resources left, please consider adding anightly
Clippy check to your CI and report problems like false positives back to us. With that we can fix bugs early, before they can get to stable.
This chapter will give an overview on how to use Clippy on different popular CI providers.
GitHub Actions
GitHub hosted runners using the latest stable version of Rust have Clippy pre-installed.
It is as simple as running cargo clippy
to run lints against the codebase.
on: push
name: Clippy check
# Make sure CI fails on all warnings, including Clippy lints
env:
RUSTFLAGS: "-Dwarnings"
jobs:
clippy_check:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
- name: Run Clippy
run: cargo clippy --all-targets --all-features
Travis CI
You can add Clippy to Travis CI in the same way you use it locally:
language: rust
rust:
- stable
- beta
before_script:
- rustup component add clippy
script:
- cargo clippy
# if you want the build job to fail when encountering warnings, use
- cargo clippy -- -D warnings
# in order to also check tests and non-default crate features, use
- cargo clippy --all-targets --all-features -- -D warnings
- cargo test
# etc.
Clippy Development
Hello fellow Rustacean! If you made it here, you're probably interested in making Clippy better by contributing to it. In that case, welcome to the project!
Note: If you're just interested in using Clippy, there's nothing to see from this point onward and you should return to one of the earlier chapters.
Getting started
If this is your first time contributing to Clippy, you should first read the Basics docs. This will explain the basics on how to get the source code and how to compile and test the code.
Writing code
If you have done the basic setup, it's time to start hacking.
The Adding lints chapter is a walk through on how to add a new lint to Clippy. This is also interesting if you just want to fix a lint, because it also covers how to test lints and gives an overview of the bigger picture.
If you want to add a new lint or change existing ones apart from bugfixing, it's also a good idea to give the stability guarantees and lint categories sections of the Clippy 1.0 RFC a quick read. The lint categories are also described earlier in this book.
Note: Some higher level things about contributing to Clippy are still covered in the
CONTRIBUTING.md
document. Some of those will be moved to the book over time, like:
- Finding something to fix
- IDE setup
- High level overview on how Clippy works
- Triage procedure
- Bors and Homu
Basics for hacking on Clippy
This document explains the basics for hacking on Clippy. Besides others, this includes how to build and test Clippy. For a more in depth description on the codebase take a look at Adding Lints or Common Tools.
Get the Code
First, make sure you have checked out the latest version of Clippy. If this is your first time working on Clippy, create a fork of the repository and clone it afterwards with the following command:
git clone git@github.com:<your-username>/rust-clippy
If you've already cloned Clippy in the past, update it to the latest version:
# If the upstream remote has not been added yet
git remote add upstream https://github.com/rust-lang/rust-clippy
# upstream has to be the remote of the rust-lang/rust-clippy repo
git fetch upstream
# make sure that you are on the master branch
git checkout master
# rebase your master branch on the upstream master
git rebase upstream/master
# push to the master branch of your fork
git push
Building and Testing
You can build and test Clippy like every other Rust project:
cargo build # builds Clippy
cargo test # tests Clippy
Since Clippy's test suite is pretty big, there are some commands that only run a subset of Clippy's tests:
# only run UI tests
cargo uitest
# only run UI tests starting with `test_`
TESTNAME="test_" cargo uitest
# only run dogfood tests
cargo dev dogfood
If the output of a UI test differs from the expected output, you can update the reference file with:
cargo dev bless
For example, this is necessary, if you fix a typo in an error message of a lint or if you modify a test file to add a test case.
Note: This command may update more files than you intended. In that case only commit the files you wanted to update.
cargo dev
Clippy has some dev tools to make working on Clippy more convenient. These tools
can be accessed through the cargo dev
command. Available tools are listed
below. To get more information about these commands, just call them with
--help
.
# formats the whole Clippy codebase and all tests
cargo dev fmt
# register or update lint names/groups/...
cargo dev update_lints
# create a new lint and register it
cargo dev new_lint
# deprecate a lint and attempt to remove code relating to it
cargo dev deprecate
# automatically formatting all code before each commit
cargo dev setup git-hook
# (experimental) Setup Clippy to work with IntelliJ-Rust
cargo dev setup intellij
# runs the `dogfood` tests
cargo dev dogfood
More about intellij command usage and reasons here
lintcheck
cargo lintcheck
will build and run clippy on a fixed set of crates and
generate a log of the results. You can git diff
the updated log against its
previous version and see what impact your lint made on a small set of crates.
If you add a new lint, please audit the resulting warnings and make sure there
are no false positives and that the suggestions are valid.
Refer to the tools README for more details.
PR
We follow a rustc no merge-commit policy. See https://rustc-dev-guide.rust-lang.org/contributing.html#opening-a-pr.
Common Abbreviations
Abbreviation | Meaning |
---|---|
UB | Undefined Behavior |
FP | False Positive |
FN | False Negative |
ICE | Internal Compiler Error |
AST | Abstract Syntax Tree |
MIR | Mid-Level Intermediate Representation |
HIR | High-Level Intermediate Representation |
TCX | Type context |
This is a concise list of abbreviations that can come up during Clippy development. An extensive general list can be found in the rustc-dev-guide glossary. Always feel free to ask if an abbreviation or meaning is unclear to you.
Install from source
If you are hacking on Clippy and want to install it from source, do the following:
First, take note of the toolchain
override in
/rust-toolchain
. We will use this override to install Clippy into the right
toolchain.
Tip: You can view the active toolchain for the current directory with
rustup show active-toolchain
.
From the Clippy project root, run the following command to build the Clippy binaries and copy them into the toolchain directory. This will override the currently installed Clippy component.
cargo build --release --bin cargo-clippy --bin clippy-driver -Zunstable-options --out-dir "$(rustc --print=sysroot)/bin"
Now you may run cargo clippy
in any project, using the toolchain where you
just installed Clippy.
cd my-project
cargo +nightly-2021-07-01 clippy
...or clippy-driver
clippy-driver +nightly-2021-07-01 <filename>
If you need to restore the default Clippy installation, run the following (from the Clippy project root).
rustup component remove clippy
rustup component add clippy
DO NOT install using
cargo install --path . --force
since this will overwrite rustup proxies. That is,~/.cargo/bin/cargo-clippy
and~/.cargo/bin/clippy-driver
should be hard or soft links to~/.cargo/bin/rustup
. You can repair these by runningrustup update
.
Adding a new lint
You are probably here because you want to add a new lint to Clippy. If this is the first time you're contributing to Clippy, this document guides you through creating an example lint from scratch.
To get started, we will create a lint that detects functions called foo
,
because that's clearly a non-descriptive name.
- Adding a new lint
- Setup
- Getting Started
- Testing
- Rustfix tests
- Testing manually
- Lint declaration
- Lint registration
- Lint passes
- Emitting a lint
- Adding the lint logic
- Specifying the lint's minimum supported Rust version (MSRV)
- Author lint
- Print HIR lint
- Documentation
- Running rustfmt
- Debugging
- PR Checklist
- Adding configuration to a lint
- Cheat Sheet
Setup
See the Basics documentation.
Getting Started
There is a bit of boilerplate code that needs to be set up when creating a new
lint. Fortunately, you can use the Clippy dev tools to handle this for you. We
are naming our new lint foo_functions
(lints are generally written in snake
case), and we don't need type information, so it will have an early pass type
(more on this later). If you're unsure if the name you chose fits the lint,
take a look at our lint naming guidelines.
Defining Our Lint
To get started, there are two ways to define our lint.
Standalone
Command: cargo dev new_lint --name=foo_functions --pass=early --category=pedantic
(category will default to nursery if not provided)
This command will create a new file: clippy_lints/src/foo_functions.rs
, as well
as register the lint.
Specific Type
Command: cargo dev new_lint --name=foo_functions --type=functions --category=pedantic
This command will create a new file: clippy_lints/src/{type}/foo_functions.rs
.
Notice how this command has a --type
flag instead of --pass
. Unlike a standalone
definition, this lint won't be registered in the traditional sense. Instead, you will
call your lint from within the type's lint pass, found in clippy_lints/src/{type}/mod.rs
.
A "type" is just the name of a directory in clippy_lints/src
, like functions
in
the example command. These are groupings of lints with common behaviors, so if your
lint falls into one, it would be best to add it to that type.
Tests Location
Both commands will create a file: tests/ui/foo_functions.rs
. For cargo lints,
two project hierarchies (fail/pass) will be created by default under tests/ui-cargo
.
Next, we'll open up these files and add our lint!
Testing
Let's write some tests first that we can execute while we iterate on our lint.
Clippy uses UI tests for testing. UI tests check that the output of Clippy is
exactly as expected. Each test is just a plain Rust file that contains the code
we want to check. The output of Clippy is compared against a .stderr
file.
Note that you don't have to create this file yourself, we'll get to generating
the .stderr
files further down.
We start by opening the test file created at tests/ui/foo_functions.rs
.
Update the file with some examples to get started:
#![warn(clippy::foo_functions)] // Impl methods struct A; impl A { pub fn fo(&self) {} pub fn foo(&self) {} pub fn food(&self) {} } // Default trait methods trait B { fn fo(&self) {} fn foo(&self) {} fn food(&self) {} } // Plain functions fn fo() {} fn foo() {} fn food() {} fn main() { // We also don't want to lint method calls foo(); let a = A; a.foo(); }
Now we can run the test with TESTNAME=foo_functions cargo uitest
, currently
this test is meaningless though.
While we are working on implementing our lint, we can keep running the UI test. That allows us to check if the output is turning into what we want.
Once we are satisfied with the output, we need to run cargo dev bless
to
update the .stderr
file for our lint. Please note that, we should run
TESTNAME=foo_functions cargo uitest
every time before running cargo dev bless
. Running TESTNAME=foo_functions cargo uitest
should pass then. When we
commit our lint, we need to commit the generated .stderr
files, too. In
general, you should only commit files changed by cargo dev bless
for the
specific lint you are creating/editing. Note that if the generated files are
empty, they should be removed.
Note: you can run multiple test files by specifying a comma separated list:
TESTNAME=foo_functions,test2,test3
.
Cargo lints
For cargo lints, the process of testing differs in that we are interested in the
Cargo.toml
manifest file. We also need a minimal crate associated with that
manifest.
If our new lint is named e.g. foo_categories
, after running cargo dev new_lint
we will find by default two new crates, each with its manifest file:
tests/ui-cargo/foo_categories/fail/Cargo.toml
: this file should cause the new lint to raise an error.tests/ui-cargo/foo_categories/pass/Cargo.toml
: this file should not trigger the lint.
If you need more cases, you can copy one of those crates (under
foo_categories
) and rename it.
The process of generating the .stderr
file is the same, and prepending the
TESTNAME
variable to cargo uitest
works too.
Rustfix tests
If the lint you are working on is making use of structured suggestions, the test
file should include a // run-rustfix
comment at the top. This will
additionally run rustfix for that test. Rustfix will apply the suggestions
from the lint to the code of the test file and compare that to the contents of a
.fixed
file.
Use cargo dev bless
to automatically generate the .fixed
file after running
the tests.
Testing manually
Manually testing against an example file can be useful if you have added some
println!
s and the test suite output becomes unreadable. To try Clippy with
your local modifications, run
cargo dev lint input.rs
from the working copy root. With tests in place, let's have a look at implementing our lint now.
Lint declaration
Let's start by opening the new file created in the clippy_lints
crate at
clippy_lints/src/foo_functions.rs
. That's the crate where all the lint code
is. This file has already imported some initial things we will need:
#![allow(unused)] fn main() { use rustc_lint::{EarlyLintPass, EarlyContext}; use rustc_session::{declare_lint_pass, declare_tool_lint}; use rustc_ast::ast::*; }
The next step is to update the lint declaration. Lints are declared using the
declare_clippy_lint!
macro, and we just need to update
the auto-generated lint declaration to have a real description, something like
this:
#![allow(unused)] fn main() { declare_clippy_lint! { /// ### What it does /// /// ### Why is this bad? /// /// ### Example /// ```rust /// // example code /// ``` #[clippy::version = "1.29.0"] pub FOO_FUNCTIONS, pedantic, "function named `foo`, which is not a descriptive name" } }
- The section of lines prefixed with
///
constitutes the lint documentation section. This is the default documentation style and will be displayed like this. To render and open this documentation locally in a browser, runcargo dev serve
. - The
#[clippy::version]
attribute will be rendered as part of the lint documentation. The value should be set to the current Rust version that the lint is developed in, it can be retrieved by runningrustc -vV
in the rust-clippy directory. The version is listed under release. (Use the version without the-nightly
) suffix. FOO_FUNCTIONS
is the name of our lint. Be sure to follow the lint naming guidelines here when naming your lint. In short, the name should state the thing that is being checked for and read well when used withallow
/warn
/deny
.pedantic
sets the lint level toAllow
. The exact mapping can be found here- The last part should be a text that explains what exactly is wrong with the code
The rest of this file contains an empty implementation for our lint pass, which
in this case is EarlyLintPass
and should look like this:
#![allow(unused)] fn main() { // clippy_lints/src/foo_functions.rs // .. imports and lint declaration .. declare_lint_pass!(FooFunctions => [FOO_FUNCTIONS]); impl EarlyLintPass for FooFunctions {} }
Lint registration
When using cargo dev new_lint
, the lint is automatically registered and
nothing more has to be done.
When declaring a new lint by hand and cargo dev update_lints
is used, the lint
pass may have to be registered manually in the register_plugins
function in
clippy_lints/src/lib.rs
:
#![allow(unused)] fn main() { store.register_early_pass(|| Box::new(foo_functions::FooFunctions)); }
As one may expect, there is a corresponding register_late_pass
method
available as well. Without a call to one of register_early_pass
or
register_late_pass
, the lint pass in question will not be run.
One reason that cargo dev update_lints
does not automate this step is that
multiple lints can use the same lint pass, so registering the lint pass may
already be done when adding a new lint. Another reason that this step is not
automated is that the order that the passes are registered determines the order
the passes actually run, which in turn affects the order that any emitted lints
are output in.
Lint passes
Writing a lint that only checks for the name of a function means that we only have to deal with the AST and don't have to deal with the type system at all. This is good, because it makes writing this particular lint less complicated.
We have to make this decision with every new Clippy lint. It boils down to using
either EarlyLintPass
or LateLintPass
.
In short, the LateLintPass
has access to type information while the
EarlyLintPass
doesn't. If you don't need access to type information, use the
EarlyLintPass
. The EarlyLintPass
is also faster. However linting speed
hasn't really been a concern with Clippy so far.
Since we don't need type information for checking the function name, we used
--pass=early
when running the new lint automation and all the imports were
added accordingly.
Emitting a lint
With UI tests and the lint declaration in place, we can start working on the implementation of the lint logic.
Let's start by implementing the EarlyLintPass
for our FooFunctions
:
#![allow(unused)] fn main() { impl EarlyLintPass for FooFunctions { fn check_fn(&mut self, cx: &EarlyContext<'_>, fn_kind: FnKind<'_>, span: Span, _: NodeId) { // TODO: Emit lint here } } }
We implement the check_fn
method from the
EarlyLintPass
trait. This gives us access to various
information about the function that is currently being checked. More on that in
the next section. Let's worry about the details later and emit our lint for
every function definition first.
Depending on how complex we want our lint message to be, we can choose from a
variety of lint emission functions. They can all be found in
clippy_utils/src/diagnostics.rs
.
span_lint_and_help
seems most appropriate in this case. It allows us to
provide an extra help message and we can't really suggest a better name
automatically. This is how it looks:
#![allow(unused)] fn main() { impl EarlyLintPass for FooFunctions { fn check_fn(&mut self, cx: &EarlyContext<'_>, fn_kind: FnKind<'_>, span: Span, _: NodeId) { span_lint_and_help( cx, FOO_FUNCTIONS, span, "function named `foo`", None, "consider using a more meaningful name" ); } } }
Running our UI test should now produce output that contains the lint message.
According to the rustc-dev-guide, the text should be matter of fact and avoid capitalization and periods, unless multiple sentences are needed. When code or an identifier must appear in a message or label, it should be surrounded with single grave accents `.
Adding the lint logic
Writing the logic for your lint will most likely be different from our example, so this section is kept rather short.
Using the check_fn
method gives us access to FnKind
that has the FnKind::Fn
variant. It provides access to the name of the
function/method via an Ident
.
With that we can expand our check_fn
method to:
#![allow(unused)] fn main() { impl EarlyLintPass for FooFunctions { fn check_fn(&mut self, cx: &EarlyContext<'_>, fn_kind: FnKind<'_>, span: Span, _: NodeId) { if is_foo_fn(fn_kind) { span_lint_and_help( cx, FOO_FUNCTIONS, span, "function named `foo`", None, "consider using a more meaningful name" ); } } } }
We separate the lint conditional from the lint emissions because it makes the code a bit easier to read. In some cases this separation would also allow to write some unit tests (as opposed to only UI tests) for the separate function.
In our example, is_foo_fn
looks like:
#![allow(unused)] fn main() { // use statements, impl EarlyLintPass, check_fn, .. fn is_foo_fn(fn_kind: FnKind<'_>) -> bool { match fn_kind { FnKind::Fn(_, ident, ..) => { // check if `fn` name is `foo` ident.name.as_str() == "foo" } // ignore closures FnKind::Closure(..) => false } } }
Now we should also run the full test suite with cargo test
. At this point
running cargo test
should produce the expected output. Remember to run cargo dev bless
to update the .stderr
file.
cargo test
(as opposed to cargo uitest
) will also ensure that our lint
implementation is not violating any Clippy lints itself.
That should be it for the lint implementation. Running cargo test
should now
pass.
Specifying the lint's minimum supported Rust version (MSRV)
Sometimes a lint makes suggestions that require a certain version of Rust. For
example, the manual_strip
lint suggests using str::strip_prefix
and
str::strip_suffix
which is only available after Rust 1.45. In such cases, you
need to ensure that the MSRV configured for the project is >= the MSRV of the
required Rust feature. If multiple features are required, just use the one with
a lower MSRV.
First, add an MSRV alias for the required feature in clippy_utils::msrvs
.
This can be accessed later as msrvs::STR_STRIP_PREFIX
, for example.
#![allow(unused)] fn main() { msrv_aliases! { .. 1,45,0 { STR_STRIP_PREFIX } } }
In order to access the project-configured MSRV, you need to have an msrv
field
in the LintPass struct, and a constructor to initialize the field. The msrv
value is passed to the constructor in clippy_lints/lib.rs
.
#![allow(unused)] fn main() { pub struct ManualStrip { msrv: Option<RustcVersion>, } impl ManualStrip { #[must_use] pub fn new(msrv: Option<RustcVersion>) -> Self { Self { msrv } } } }
The project's MSRV can then be matched against the feature MSRV in the LintPass
using the meets_msrv
utility function.
#![allow(unused)] fn main() { if !meets_msrv(self.msrv, msrvs::STR_STRIP_PREFIX) { return; } }
The project's MSRV can also be specified as an inner attribute, which overrides
the value from clippy.toml
. This can be accounted for using the
extract_msrv_attr!(LintContext)
macro and passing
LateContext
/EarlyContext
.
#![allow(unused)] fn main() { impl<'tcx> LateLintPass<'tcx> for ManualStrip { fn check_expr(&mut self, cx: &LateContext<'tcx>, expr: &'tcx Expr<'_>) { ... } extract_msrv_attr!(LateContext); } }
Once the msrv
is added to the lint, a relevant test case should be added to
tests/ui/min_rust_version_attr.rs
which verifies that the lint isn't emitted
if the project's MSRV is lower.
As a last step, the lint should be added to the lint documentation. This is done
in clippy_lints/src/utils/conf.rs
:
#![allow(unused)] fn main() { define_Conf! { /// Lint: LIST, OF, LINTS, <THE_NEWLY_ADDED_LINT>. The minimum rust version that the project supports (msrv: Option<String> = None), ... } }
Author lint
If you have trouble implementing your lint, there is also the internal author
lint to generate Clippy code that detects the offending pattern. It does not
work for all of the Rust syntax, but can give a good starting point.
The quickest way to use it, is the Rust playground:
play.rust-lang.org. Put the code you want to lint into the
editor and add the #[clippy::author]
attribute above the item. Then run Clippy
via Tools -> Clippy
and you should see the generated code in the output below.
Here is an example on the playground.
If the command was executed successfully, you can copy the code over to where you are implementing your lint.
Print HIR lint
To implement a lint, it's helpful to first understand the internal
representation that rustc uses. Clippy has the #[clippy::dump]
attribute that
prints the High-Level Intermediate Representation (HIR) of the item,
statement, or expression that the attribute is attached to. To attach the
attribute to expressions you often need to enable
#![feature(stmt_expr_attributes)]
.
Here you can find an example, just select Tools and run Clippy.
Documentation
The final thing before submitting our PR is to add some documentation to our lint declaration.
Please document your lint with a doc comment akin to the following:
#![allow(unused)] fn main() { declare_clippy_lint! { /// ### What it does /// Checks for ... (describe what the lint matches). /// /// ### Why is this bad? /// Supply the reason for linting the code. /// /// ### Example /// /// ```rust,ignore /// // A short example of code that triggers the lint /// ``` /// /// Use instead: /// ```rust,ignore /// // A short example of improved code that doesn't trigger the lint /// ``` #[clippy::version = "1.29.0"] pub FOO_FUNCTIONS, pedantic, "function named `foo`, which is not a descriptive name" } }
Once your lint is merged, this documentation will show up in the lint list.
Running rustfmt
Rustfmt is a tool for formatting Rust code according to style guidelines. Your
code has to be formatted by rustfmt
before a PR can be merged. Clippy uses
nightly rustfmt
in the CI.
It can be installed via rustup
:
rustup component add rustfmt --toolchain=nightly
Use cargo dev fmt
to format the whole codebase. Make sure that rustfmt
is
installed for the nightly toolchain.
Debugging
If you want to debug parts of your lint implementation, you can use the dbg!
macro anywhere in your code. Running the tests should then include the debug
output in the stdout
part.
PR Checklist
Before submitting your PR make sure you followed all of the basic requirements:
- [ ] Followed lint naming conventions
- [ ] Added passing UI tests (including committed
.stderr
file) - [ ]
cargo test
passes locally - [ ] Executed
cargo dev update_lints
- [ ] Added lint documentation
- [ ] Run
cargo dev fmt
Adding configuration to a lint
Clippy supports the configuration of lints values using a clippy.toml
file in
the workspace directory. Adding a configuration to a lint can be useful for
thresholds or to constrain some behavior that can be seen as a false positive
for some users. Adding a configuration is done in the following steps:
-
Adding a new configuration entry to
clippy_lints::utils::conf
like this:#![allow(unused)] fn main() { /// Lint: LINT_NAME. /// /// <The configuration field doc comment> (configuration_ident: Type = DefaultValue), }
The doc comment is automatically added to the documentation of the listed lints. The default value will be formatted using the
Debug
implementation of the type. -
Adding the configuration value to the lint impl struct:
-
This first requires the definition of a lint impl struct. Lint impl structs are usually generated with the
declare_lint_pass!
macro. This struct needs to be defined manually to add some kind of metadata to it:#![allow(unused)] fn main() { // Generated struct definition declare_lint_pass!(StructName => [ LINT_NAME ]); // New manual definition struct #[derive(Copy, Clone)] pub struct StructName {} impl_lint_pass!(StructName => [ LINT_NAME ]); }
-
Next add the configuration value and a corresponding creation method like this:
#![allow(unused)] fn main() { #[derive(Copy, Clone)] pub struct StructName { configuration_ident: Type, } // ... impl StructName { pub fn new(configuration_ident: Type) -> Self { Self { configuration_ident, } } } }
-
-
Passing the configuration value to the lint impl struct:
First find the struct construction in the
clippy_lints
lib file. The configuration value is now cloned or copied into a local value that is then passed to the impl struct like this:#![allow(unused)] fn main() { // Default generated registration: store.register_*_pass(|| box module::StructName); // New registration with configuration value let configuration_ident = conf.configuration_ident.clone(); store.register_*_pass(move || box module::StructName::new(configuration_ident)); }
Congratulations the work is almost done. The configuration value can now be accessed in the linting code via
self.configuration_ident
. -
Adding tests:
- The default configured value can be tested like any normal lint in
tests/ui
. - The configuration itself will be tested separately in
tests/ui-toml
. Simply add a new subfolder with a fitting name. This folder contains aclippy.toml
file with the configuration value and a rust file that should be linted by Clippy. The test can otherwise be written as usual.
- The default configured value can be tested like any normal lint in
Cheat Sheet
Here are some pointers to things you are likely going to need for every lint:
- Clippy utils - Various helper functions. Maybe the function you need
is already in here (
is_type_diagnostic_item
,implements_trait
,snippet
, etc) - Clippy diagnostics
- Let chains
from_expansion
andin_external_macro
Span
Applicability
- Common tools for writing lints helps with common operations
- The rustc-dev-guide explains a lot of internal compiler concepts
- The nightly rustc docs which has been linked to throughout this guide
For EarlyLintPass
lints:
For LateLintPass
lints:
While most of Clippy's lint utils are documented, most of rustc's internals lack documentation currently. This is unfortunate, but in most cases you can probably get away with copying things from existing similar lints. If you are stuck, don't hesitate to ask on Zulip or in the issue/PR.
Common tools for writing lints
You may need following tooltips to catch up with common operations.
Useful Rustc dev guide links:
Retrieving the type of an expression
Sometimes you may want to retrieve the type Ty
of an expression Expr
, for
example to answer following questions:
- which type does this expression correspond to (using its
TyKind
)? - is it a sized type?
- is it a primitive type?
- does it implement a trait?
This operation is performed using the expr_ty()
method from the
TypeckResults
struct, that gives you access to the underlying
structure Ty
.
Example of use:
#![allow(unused)] fn main() { impl LateLintPass<'_> for MyStructLint { fn check_expr(&mut self, cx: &LateContext<'_>, expr: &Expr<'_>) { // Get type of `expr` let ty = cx.typeck_results().expr_ty(expr); // Match its kind to enter its type match ty.kind { ty::Adt(adt_def, _) if adt_def.is_struct() => println!("Our `expr` is a struct!"), _ => () } } } }
Similarly in TypeckResults
methods, you have the
pat_ty()
method to retrieve a type from a pattern.
Two noticeable items here:
cx
is the lint contextLateContext
. The two most useful data structures in this context aretcx
and theTypeckResults
returned byLateContext::typeck_results
, allowing us to jump to type definitions and other compilation stages such as HIR.typeck_results
's return value isTypeckResults
and is created by type checking step, it includes useful information such as types of expressions, ways to resolve methods and so on.
Checking if an expr is calling a specific method
Starting with an expr
, you can check whether it is calling a specific method
some_method
:
#![allow(unused)] fn main() { impl<'tcx> LateLintPass<'tcx> for MyStructLint { fn check_expr(&mut self, cx: &LateContext<'tcx>, expr: &'tcx hir::Expr<'_>) { // Check our expr is calling a method if let hir::ExprKind::MethodCall(path, _, _self_arg, ..) = &expr.kind // Check the name of this method is `some_method` && path.ident.name == sym!(some_method) // Optionally, check the type of the self argument. // - See "Checking for a specific type" { // ... } } } }
Checking for a specific type
There are three ways to check if an expression type is a specific type we want to check for. All of these methods only check for the base type, generic arguments have to be checked separately.
#![allow(unused)] fn main() { use clippy_utils::ty::{is_type_diagnostic_item, is_type_lang_item}; use clippy_utils::{paths, match_def_path}; use rustc_span::symbol::sym; use rustc_hir::LangItem; impl LateLintPass<'_> for MyStructLint { fn check_expr(&mut self, cx: &LateContext<'_>, expr: &Expr<'_>) { // Getting the expression type let ty = cx.typeck_results().expr_ty(expr); // 1. Using diagnostic items // The last argument is the diagnostic item to check for if is_type_diagnostic_item(cx, ty, sym::Option) { // The type is an `Option` } // 2. Using lang items if is_type_lang_item(cx, ty, LangItem::RangeFull) { // The type is a full range like `.drain(..)` } // 3. Using the type path // This method should be avoided if possible if match_def_path(cx, def_id, &paths::RESULT) { // The type is a `core::result::Result` } } } }
Prefer using diagnostic items and lang items where possible.
Checking if a type implements a specific trait
There are three ways to do this, depending on if the target trait has a diagnostic item, lang item or neither.
#![allow(unused)] fn main() { use clippy_utils::{implements_trait, is_trait_method, match_trait_method, paths}; use rustc_span::symbol::sym; impl LateLintPass<'_> for MyStructLint { fn check_expr(&mut self, cx: &LateContext<'_>, expr: &Expr<'_>) { // 1. Using diagnostic items with the expression // we use `is_trait_method` function from Clippy's utils if is_trait_method(cx, expr, sym::Iterator) { // method call in `expr` belongs to `Iterator` trait } // 2. Using lang items with the expression type let ty = cx.typeck_results().expr_ty(expr); if cx.tcx.lang_items() // we are looking for the `DefId` of `Drop` trait in lang items .drop_trait() // then we use it with our type `ty` by calling `implements_trait` from Clippy's utils .map_or(false, |id| implements_trait(cx, ty, id, &[])) { // `expr` implements `Drop` trait } // 3. Using the type path with the expression // we use `match_trait_method` function from Clippy's utils // (This method should be avoided if possible) if match_trait_method(cx, expr, &paths::INTO) { // `expr` implements `Into` trait } } } }
Prefer using diagnostic and lang items, if the target trait has one.
We access lang items through the type context tcx
. tcx
is of type
TyCtxt
and is defined in the rustc_middle
crate. A list of defined
paths for Clippy can be found in paths.rs
Checking if a type defines a specific method
To check if our type defines a method called some_method
:
#![allow(unused)] fn main() { use clippy_utils::ty::is_type_diagnostic_item; use clippy_utils::return_ty; impl<'tcx> LateLintPass<'tcx> for MyTypeImpl { fn check_impl_item(&mut self, cx: &LateContext<'tcx>, impl_item: &'tcx ImplItem<'_>) { // Check if item is a method/function if let ImplItemKind::Fn(ref signature, _) = impl_item.kind // Check the method is named `some_method` && impl_item.ident.name == sym!(some_method) // We can also check it has a parameter `self` && signature.decl.implicit_self.has_implicit_self() // We can go further and even check if its return type is `String` && is_type_diagnostic_item(cx, return_ty(cx, impl_item.hir_id), sym!(string_type)) { // ... } } } }
Dealing with macros and expansions
Keep in mind that macros are already expanded and desugaring is already applied to the code representation that you are working with in Clippy. This unfortunately causes a lot of false positives because macro expansions are "invisible" unless you actively check for them. Generally speaking, code with macro expansions should just be ignored by Clippy because that code can be dynamic in ways that are difficult or impossible to see. Use the following functions to deal with macros:
-
span.from_expansion()
: detects if a span is from macro expansion or desugaring. Checking this is a common first step in a lint.#![allow(unused)] fn main() { if expr.span.from_expansion() { // just forget it return; } }
-
span.ctxt()
: the span's context represents whether it is from expansion, and if so, which macro call expanded it. It is sometimes useful to check if the context of two spans are equal.#![allow(unused)] fn main() { // expands to `1 + 0`, but don't lint 1 + mac!() }
#![allow(unused)] fn main() { if left.span.ctxt() != right.span.ctxt() { // the coder most likely cannot modify this expression return; } }
Note: Code that is not from expansion is in the "root" context. So any spans where
from_expansion
returnstrue
can be assumed to have the same context. And so just usingspan.from_expansion()
is often good enough. -
in_external_macro(span)
: detect if the given span is from a macro defined in a foreign crate. If you want the lint to work with macro-generated code, this is the next line of defense to avoid macros not defined in the current crate. It doesn't make sense to lint code that the coder can't change.You may want to use it for example to not start linting in macros from other crates
#![allow(unused)] fn main() { #[macro_use] extern crate a_crate_with_macros; // `foo` is defined in `a_crate_with_macros` foo!("bar"); // if we lint the `match` of `foo` call and test its span assert_eq!(in_external_macro(cx.sess(), match_span), true); }
-
span.ctxt()
: the span's context represents whether it is from expansion, and if so, what expanded itOne thing
SpanContext
is useful for is to check if two spans are in the same context. For example, ina == b
,a
andb
have the same context. In amacro_rules!
witha == $b
,$b
is expanded to some expression with a different context froma
.#![allow(unused)] fn main() { macro_rules! m { ($a:expr, $b:expr) => { if $a.is_some() { $b; } } } let x: Option<u32> = Some(42); m!(x, x.unwrap()); // These spans are not from the same context // x.is_some() is from inside the macro // x.unwrap() is from outside the macro assert_eq!(x_is_some_span.ctxt(), x_unwrap_span.ctxt()); }
Infrastructure
In order to deploy Clippy over rustup
, some infrastructure is necessary. This
chapter describes the different parts of the Clippy infrastructure that need to
be maintained to make this possible.
The most important part is the sync between the rust-lang/rust
repository and
the Clippy repository that takes place every two weeks. This process is
described in the Syncing changes between Clippy and rust-lang/rust
section.
A new Clippy release is done together with every Rust release, so every six weeks. The release process is described in the Release a new Clippy Version section. During a release cycle a changelog entry for the next release has to be written. The format of that and how to do that is documented in the Changelog Update section.
Note: The Clippy CI should also be described in this chapter, but for now is left as a TODO.
Syncing changes between Clippy and rust-lang/rust
Clippy currently gets built with a pinned nightly version.
In the rust-lang/rust
repository, where rustc resides, there's a copy of
Clippy that compiler hackers modify from time to time to adapt to changes in the
unstable API of the compiler.
We need to sync these changes back to this repository periodically, and the
changes made to this repository in the meantime also need to be synced to the
rust-lang/rust
repository.
To avoid flooding the rust-lang/rust
PR queue, this two-way sync process is
done in a bi-weekly basis if there's no urgent changes. This is done starting on
the day of the Rust stable release and then every other week. That way we
guarantee that we keep this repo up to date with the latest compiler API, and
every feature in Clippy is available for 2 weeks in nightly, before it can get
to beta. For reference, the first sync following this cadence was performed the
2020-08-27.
This process is described in detail in the following sections. For general
information about subtree
s in the Rust repository see Rust's
CONTRIBUTING.md
.
Patching git-subtree to work with big repos
Currently, there's a bug in git-subtree
that prevents it from working properly
with the rust-lang/rust
repo. There's an open PR to fix that, but it's
stale. Before continuing with the following steps, we need to manually apply
that fix to our local copy of git-subtree
.
You can get the patched version of git-subtree
from here.
Put this file under /usr/lib/git-core
(making a backup of the previous file)
and make sure it has the proper permissions:
sudo cp --backup /path/to/patched/git-subtree.sh /usr/lib/git-core/git-subtree
sudo chmod --reference=/usr/lib/git-core/git-subtree~ /usr/lib/git-core/git-subtree
sudo chown --reference=/usr/lib/git-core/git-subtree~ /usr/lib/git-core/git-subtree
Note: The first time running
git subtree push
a cache has to be built. This involves going through the complete Clippy history once. For this you have to increase the stack limit though, which you can do withulimit -s 60000
. Make sure to run theulimit
command from the same session you call git subtree.
Note: If you are a Debian user,
dash
is the shell used by default for scripts instead ofsh
. This shell has a hardcoded recursion limit set to 1000. In order to make this process work, you need to force the script to runbash
instead. You can do this by editing the first line of thegit-subtree
script and changingsh
tobash
.
Defining remotes
You may want to define remotes, so you don't have to type out the remote
addresses on every sync. You can do this with the following commands (these
commands still have to be run inside the rust
directory):
# Set clippy-upstream remote for pulls
$ git remote add clippy-upstream https://github.com/rust-lang/rust-clippy
# Make sure to not push to the upstream repo
$ git remote set-url --push clippy-upstream DISABLED
# Set a local remote
$ git remote add clippy-local /path/to/rust-clippy
Note: The following sections assume that you have set those remotes with the above remote names.
Performing the sync from rust-lang/rust
to Clippy
Here is a TL;DR version of the sync process (all of the following commands have
to be run inside the rust
directory):
-
Clone the
rust-lang/rust
repository or make sure it is up to date. -
Checkout the commit from the latest available nightly. You can get it using
rustup check
. -
Sync the changes to the rust-copy of Clippy to your Clippy fork:
# Make sure to change `your-github-name` to your github name in the following command. Also be # sure to either use a net-new branch, e.g. `sync-from-rust`, or delete the branch beforehand # because changes cannot be fast forwarded and you have to run this command again. git subtree push -P src/tools/clippy clippy-local sync-from-rust
Note: Most of the time you have to create a merge commit in the
rust-clippy
repo (this has to be done in the Clippy repo, not in the rust-copy of Clippy):git fetch upstream # assuming upstream is the rust-lang/rust remote git checkout sync-from-rust git merge upstream/master --no-ff
Note: This is one of the few instances where a merge commit is allowed in a PR.
-
Bump the nightly version in the Clippy repository by changing the date in the rust-toolchain file to the current date and committing it with the message:
git commit -m "Bump nightly version -> YYYY-MM-DD"
-
Open a PR to
rust-lang/rust-clippy
and wait for it to get merged (to accelerate the process ping the@rust-lang/clippy
team in your PR and/or ask them in the Zulip stream.)
Performing the sync from Clippy to rust-lang/rust
All of the following commands have to be run inside the rust
directory.
- Make sure you have checked out the latest
master
ofrust-lang/rust
. - Sync the
rust-lang/rust-clippy
master to the rust-copy of Clippy:git checkout -b sync-from-clippy git subtree pull -P src/tools/clippy clippy-upstream master
- Open a PR to
rust-lang/rust
Backport Changes
Sometimes it is necessary to backport changes to the beta release of Clippy. Backports in Clippy are rare and should be approved by the Clippy team. For example, a backport is done, if a crucial ICE was fixed or a lint is broken to a point, that it has to be disabled, before landing on stable.
Backports are done to the beta
branch of Clippy. Backports to stable Clippy
releases basically don't exist, since this would require a Rust point release,
which is almost never justifiable for a Clippy fix.
Backport the changes
Backports are done on the beta branch of the Clippy repository.
# Assuming the current directory corresponds to the Clippy repository
$ git checkout beta
$ git checkout -b backport
$ git cherry-pick <SHA> # `<SHA>` is the commit hash of the commit(s), that should be backported
$ git push origin backport
Now you should test that the backport passes all the tests in the Rust repository. You can do this with:
# Assuming the current directory corresponds to the Rust repository
$ git checkout beta
$ git subtree pull -p src/tools/clippy https://github.com/<your-github-name>/rust-clippy backport
$ ./x.py test src/tools/clippy
Should the test fail, you can fix Clippy directly in the Rust repository. This has to be first applied to the Clippy beta branch and then again synced to the Rust repository, though. The easiest way to do this is:
# In the Rust repository
$ git diff --patch --relative=src/tools/clippy > clippy.patch
# In the Clippy repository
$ git apply /path/to/clippy.patch
$ git add -u
$ git commit -m "Fix rustup fallout"
$ git push origin backport
After this, you can open a PR to the beta
branch of the Clippy repository.
Update Clippy in the Rust Repository
This step must be done, after the PR of the previous step was merged.
After the backport landed in the Clippy repository, the branch has to be synced back to the beta branch of the Rust repository.
# Assuming the current directory corresponds to the Rust repository
$ git checkout beta
$ git checkout -b clippy_backport
$ git subtree pull -p src/tools/clippy https://github.com/rust-lang/rust-clippy beta
$ git push origin clippy_backport
Make sure to test the backport in the Rust repository before opening a PR. This
is done with ./x.py test src/tools/clippy
. If that passes all tests, open a PR
to the beta
branch of the Rust repository. In this PR you should tag the
Clippy team member, that agreed to the backport or the @rust-lang/clippy
team.
Make sure to add [beta]
to the title of the PR.
Changelog Update
If you want to help with updating the changelog, you're in the right place.
When to update
Typos and other small fixes/additions are always welcome.
Special care needs to be taken when it comes to updating the changelog for a new Rust release. For that purpose, the changelog is ideally updated during the week before an upcoming stable release. You can find the release dates on the Rust Forge.
Most of the time we only need to update the changelog for minor Rust releases. It's been very rare that Clippy changes were included in a patch release.
Changelog update walkthrough
1. Finding the relevant Clippy commits
Each Rust release ships with its own version of Clippy. The Clippy subtree can
be found in the tools
directory of the Rust repository.
Depending on the current time and what exactly you want to update, the following bullet points might be helpful:
- When writing the release notes for the upcoming stable release you need to
check out the Clippy commit of the current Rust
beta
branch. Link - When writing the release notes for the upcoming beta release, you need to
check out the Clippy commit of the current Rust
master
. Link - When writing the (forgotten) release notes for a past stable release, you need to check out the Rust release tag of the stable release. Link
Usually you want to write the changelog of the upcoming stable release. Make
sure though, that beta
was already branched in the Rust repository.
To find the commit hash, issue the following command when in a rust-lang/rust
checkout:
git log --oneline -- src/tools/clippy/ | grep -o "Merge commit '[a-f0-9]*' into .*" | head -1 | sed -e "s/Merge commit '\([a-f0-9]*\)' into .*/\1/g"
2. Fetching the PRs between those commits
Once you've got the correct commit range, run
util/fetch_prs_between.sh commit1 commit2 > changes.txt
and open that file in your editor of choice.
When updating the changelog it's also a good idea to make sure that commit1
is
already correct in the current changelog.
3. Authoring the final changelog
The above script should have dumped all the relevant PRs to the file you specified. It should have filtered out most of the irrelevant PRs already, but it's a good idea to do a manual cleanup pass where you look for more irrelevant PRs. If you're not sure about some PRs, just leave them in for the review and ask for feedback.
With the PRs filtered, you can start to take each PR and move the changelog:
content to CHANGELOG.md
. Adapt the wording as you see fit but try to keep it
somewhat coherent.
The order should roughly be:
- New lints
- Moves or deprecations of lints
- Changes that expand what code existing lints cover
- False positive fixes
- Suggestion fixes/improvements
- ICE fixes
- Documentation improvements
- Others
As section headers, we use:
### New Lints
### Moves and Deprecations
### Enhancements
### False Positive Fixes
### Suggestion Fixes/Improvements
### ICE Fixes
### Documentation Improvements
### Others
Please also be sure to update the Beta/Unreleased sections at the top with the relevant commit ranges.
If you have the time, it would be appreciated if you double-check, that the
#[clippy::version]
attributes for the added lints contains the correct version.
Release a new Clippy Version
NOTE: This document is probably only relevant to you, if you're a member of the Clippy team.
Clippy is released together with stable Rust releases. The dates for these releases can be found at the Rust Forge. This document explains the necessary steps to create a Clippy release.
- Remerge the
beta
branch - Update the
beta
branch - Find the Clippy commit
- Tag the stable commit
- Update
CHANGELOG.md
NOTE: This document is for stable Rust releases, not for point releases. For point releases, step 1. and 2. should be enough.
Remerge the beta
branch
This step is only necessary, if since the last release something was backported to the beta Rust release. The remerge is then necessary, to make sure that the Clippy commit, that was used by the now stable Rust release, persists in the tree of the Clippy repository.
To find out if this step is necessary run
# Assumes that the local master branch of rust-lang/rust-clippy is up-to-date
$ git fetch upstream
$ git branch master --contains upstream/beta
If this command outputs master
, this step is not necessary.
# Assuming `HEAD` is the current `master` branch of rust-lang/rust-clippy
$ git checkout -b backport_remerge
$ git merge upstream/beta
$ git diff # This diff has to be empty, otherwise something with the remerge failed
$ git push origin backport_remerge # This can be pushed to your fork
After this, open a PR to the master branch. In this PR, the commit hash of the
HEAD
of the beta
branch must exists. In addition to that, no files should be
changed by this PR.
Update the beta
branch
This step must be done after the PR of the previous step was merged.
First, the Clippy commit of the beta
branch of the Rust repository has to be
determined.
# Assuming the current directory corresponds to the Rust repository
$ git fetch upstream
$ git checkout upstream/beta
$ BETA_SHA=$(git log --oneline -- src/tools/clippy/ | grep -o "Merge commit '[a-f0-9]*' into .*" | head -1 | sed -e "s/Merge commit '\([a-f0-9]*\)' into .*/\1/g")
After finding the Clippy commit, the beta
branch in the Clippy repository can
be updated.
# Assuming the current directory corresponds to the Clippy repository
$ git checkout beta
$ git reset --hard $BETA_SHA
$ git push upstream beta
Find the Clippy commit
The first step is to tag the Clippy commit, that is included in the stable Rust release. This commit can be found in the Rust repository.
# Assuming the current directory corresponds to the Rust repository
$ git fetch upstream # `upstream` is the `rust-lang/rust` remote
$ git checkout 1.XX.0 # XX should be exchanged with the corresponding version
$ SHA=$(git log --oneline -- src/tools/clippy/ | grep -o "Merge commit '[a-f0-9]*' into .*" | head -1 | sed -e "s/Merge commit '\([a-f0-9]*\)' into .*/\1/g")
Tag the stable commit
After finding the Clippy commit, it can be tagged with the release number.
# Assuming the current directory corresponds to the Clippy repository
$ git checkout $SHA
$ git tag rust-1.XX.0 # XX should be exchanged with the corresponding version
$ git push upstream rust-1.XX.0 # `upstream` is the `rust-lang/rust-clippy` remote
After this, the release should be available on the Clippy release page.
Update the stable
branch
At this step you should have already checked out the commit of the rust-1.XX.0
tag. Updating the stable branch from here is as easy as:
# Assuming the current directory corresponds to the Clippy repository and the
# commit of the just created rust-1.XX.0 tag is checked out.
$ git push upstream rust-1.XX.0:stable # `upstream` is the `rust-lang/rust-clippy` remote
NOTE: Usually there are no stable backports for Clippy, so this update should be possible without force pushing or anything like this. If there should have happened a stable backport, make sure to re-merge those changes just as with the
beta
branch.
Update CHANGELOG.md
For this see the document on how to update the changelog.
If you don't have time to do a complete changelog update right away, just update the following parts:
-
Remove the
(beta)
from the new stable version:## Rust 1.XX (beta) -> ## Rust 1.XX
-
Update the release date line of the new stable version:
Current beta, release 20YY-MM-DD -> Current stable, released 20YY-MM-DD
-
Update the release date line of the previous stable version:
Current stable, released 20YY-MM-DD -> Released 20YY-MM-DD
The Clippy Book
This document explains how to make additions and changes to the Clippy book, the guide to Clippy that you're reading right now. The Clippy book is formatted with Markdown and generated by mdbook.
Get mdbook
While not strictly necessary since the book source is simply Markdown text
files, having mdbook locally will allow you to build, test and serve the book
locally to view changes before you commit them to the repository. You likely
already have cargo
installed, so the easiest option is to simply:
cargo install mdbook
See the mdbook installation instructions for other options.
Make changes
The book's
src
directory contains all of the markdown files used to generate the book. If you
want to see your changes in real time, you can use the mdbook serve
command to
run a web server locally that will automatically update changes as they are
made. From the top level of your rust-clippy
directory:
mdbook serve book --open
Then navigate to http://localhost:3000
to see the generated book. While the
server is running, changes you make will automatically be updated.
For more information, see the mdbook guide.
Proposals
This chapter is about accepted proposals for changes that should be worked on in or around Clippy in the long run.
Besides adding more and more lints and improve the lints that Clippy already has, Clippy is also interested in making the experience of its users, developers and maintainers better over time. Projects that address bigger picture things like this usually take more time and it is useful to have a proposal for those first. This is the place where such proposals are collected, so that we can refer to them when working on them.
Roadmap 2021
Summary
This Roadmap lays out the plans for Clippy in 2021:
- Improving usability and reliability
- Improving experience of contributors and maintainers
- Develop and specify processes
Members of the Clippy team will be assigned tasks from one or more of these topics. The team member is then responsible to complete the assigned tasks. This can either be done by implementing them or by providing mentorship to interested contributors.
Motivation
With the ongoing growth of the Rust language and with that of the whole ecosystem, also Clippy gets more and more users and contributors. This is good for the project, but also brings challenges along. Some of these challenges are:
- More issues about reliability or usability are popping up
- Traffic is hard to handle for a small team
- Bigger projects don't get completed due to the lack of processes and/or time of the team members
Additionally, according to the Rust Roadmap 2021, clear processes should be defined by every team and unified across teams. This Roadmap is the first step towards this.
Explanation
This section will explain the things that should be done in 2021. It is important to note, that this document focuses on the "What?", not the "How?". The later will be addressed in follow-up tracking issue, with an assigned team member.
The following is split up in two major sections. The first section covers the user facing plans, the second section the internal plans.
User Facing
Clippy should be as pleasant to use and configure as possible. This section covers plans that should be implemented to improve the situation of Clippy in this regard.
Usability
In the following, plans to improve the usability are covered.
No Output After cargo check
Currently when cargo clippy
is run after cargo check
, it does not produce
any output. This is especially problematic since rust-analyzer
is on the rise
and it uses cargo check
for checking code. A fix is already implemented, but
it still has to be pushed over the finish line. This also includes the
stabilization of the cargo clippy --fix
command or the support of multi-span
suggestions in rustfix
.
lints.toml
Configuration
This is something that comes up every now and then: a reusable configuration file, where lint levels can be defined. Discussions about this often lead to nothing specific or to "we need an RFC for this". And this is exactly what needs to be done. Get together with the cargo team and write an RFC and implement such a configuration file somehow and somewhere.
Lint Groups
There are more and more issues about managing lints in Clippy popping up. Lints
are hard to implement with a guarantee of no/few false positives (FPs). One way
to address this might be to introduce more lint groups to give users the ability
to better manage lints, or improve the process of classifying lints, so that
disabling lints due to FPs becomes rare. It is important to note, that Clippy
lints are less conservative than rustc
lints, which won't change in the
future.
Reliability
In the following, plans to improve the reliability are covered.
False Positive Rate
In the worst case, new lints are only available in nightly for 2 weeks, before hitting beta and ultimately stable. This and the fact that fewer people use nightly Rust nowadays makes it more probable that a lint with many FPs hits stable. This leads to annoyed users, that will disable these new lints in the best case and to more annoyed users, that will stop using Clippy in the worst. A process should be developed and implemented to prevent this from happening.
Internal
(The end of) 2020 has shown, that Clippy has to think about the available resources, especially regarding management and maintenance of the project. This section address issues affecting team members and contributors.
Management
In 2020 Clippy achieved over 1000 open issues with regularly between 25-35 open PRs. This is simultaneously a win and a loss. More issues and PRs means more people are interested in Clippy and in contributing to it. On the other hand, it means for team members more work and for contributors longer wait times for reviews. The following will describe plans how to improve the situation for both team members and contributors.
Clear Expectations for Team Members
According to the Rust Roadmap 2021, a document specifying what it means to be a member of the team should be produced. This should not put more pressure on the team members, but rather help them and interested folks to know what the expectations are. With this it should also be easier to recruit new team members and may encourage people to get in touch, if they're interested to join.
Scaling up the Team
More people means less work for each individual. Together with the document about expectations for team members, a document defining the process of how to join the team should be produced. This can also increase the stability of the team, in case of current members dropping out (temporarily). There can also be different roles in the team, like people triaging vs. people reviewing.
Regular Meetings
Other teams have regular meetings. Clippy is big enough that it might be worth to also do them. Especially if more people join the team, this can be important for sync-ups. Besides the asynchronous communication, that works well for working on separate lints, a meeting adds a synchronous alternative at a known time. This is especially helpful if there are bigger things that need to be discussed (like the projects in this roadmap). For starters bi-weekly meetings before Rust syncs might make sense.
Triaging
To get a handle on the influx of open issues, a process for triaging issues and PRs should be developed. Officially, Clippy follows the Rust triage process, but currently no one enforces it. This can be improved by sharing triage teams across projects or by implementing dashboards / tools which simplify triaging.
Development
Improving the developer and contributor experience is something the Clippy team works on regularly. Though, some things might need special attention and planing. These topics are listed in the following.
Process for New and Existing Lints
As already mentioned above, classifying new lints gets quite hard, because the probability of a buggy lint getting into stable is quite high. A process should be implemented on how to classify lints. In addition, a test system should be developed to find out which lints are currently problematic in real world code to fix or disable them.
Processes
Related to the point before, a process for suggesting and discussing major changes should be implemented. It's also not clearly defined when a lint should be enabled or disabled by default. This can also be improved by the test system mentioned above.
Dev-Tools
There's already cargo dev
which makes Clippy development easier and more
pleasant. This can still be expanded, so that it covers more areas of the
development process.
Contributor Guide
Similar to a Clippy Book, which describes how to use Clippy, a book about how to
contribute to Clippy might be helpful for new and existing contributors. There's
already the doc
directory in the Clippy repo, this can be turned into a
mdbook
.
rustc
integration
Recently Clippy was integrated with git subtree
into the rust-lang/rust
repository. This made syncing between the two repositories easier. A
#[non_exhaustive]
list of things that still can be improved is:
- Use the same
rustfmt
version and configuration asrustc
. - Make
cargo dev
work in the Rust repo, just as it works in the Clippy repo. E.g.cargo dev bless
orcargo dev update_lints
. And even add more things to it that might be useful for the Rust repo, e.g.cargo dev deprecate
. - Easier sync process. The
subtree
situation is not ideal.
Prioritization
The most pressing issues for users of Clippy are of course the user facing issues. So there should be a priority on those issues, but without losing track of the internal issues listed in this document.
Getting the FP rate of warn/deny-by-default lints under control should have the highest priority. Other user facing issues should also get a high priority, but shouldn't be in the way of addressing internal issues.
To better manage the upcoming projects, the basic internal processes, like meetings, tracking issues and documentation, should be established as soon as possible. They might even be necessary to properly manage the projects, regarding the user facing issues.
Prior Art
Rust Roadmap
Rust's roadmap process was established by RFC 1728 in 2016. Since then every year a roadmap was published, that defined the bigger plans for the coming years. This years roadmap can be found here.
Drawbacks
Big Roadmap
This roadmap is pretty big and not all items listed in this document might be addressed during 2021. Because this is the first roadmap for Clippy, having open tasks at the end of 2021 is fine, but they should be revisited in the 2022 roadmap.