12 KiB
Contributing to coreutils
Contributions are very welcome via Pull Requests. If you don't know where to
start, take a look at the
good-first-issues
.
If you have any questions, feel free to ask them in the issues or on
Discord.
Best practices
- Follow what GNU is doing in terms of options and behavior. It is recommended
to look at the GNU Coreutils manual (on the
web, or
locally using
info <utility>
). It is more in depth than the man pages and provides a good description of available features and their implementation details. - If possible, look at the GNU test suite execution in the CI and make the test work if failing.
- Use clap for argument management.
- Make sure that the code coverage is covering all of the cases, including errors.
- The code must be clippy-warning-free and rustfmt-compliant.
- Don't hesitate to move common functions into uucore if they can be reused by other binaries.
- Unsafe code should be documented with Safety comments.
- uutils is original code. It cannot contain code from existing GNU or Unix-like utilities, nor should it link to or reference GNU libraries.
Platforms
We take pride in supporting many operating systems and architectures. Any code
you contribute must at least compile without warnings for all platforms in the
CI. However, you can use #[cfg(...)]
attributes to create platform dependent features.
Tip: For Windows, Microsoft provides some images (VMWare, Hyper-V, VirtualBox and Parallels) for development: https://developer.microsoft.com/windows/downloads/virtual-machines/
Tools
We have an extensive CI that will check your code before it can be merged. This section explains how to run those checks locally to avoid waiting for the CI.
pre-commit hooks
A configuration for pre-commit
is provided in the repository. It allows
automatically checking every git commit you make to ensure it compiles, and
passes clippy
and rustfmt
without warnings.
To use the provided hook:
- Install
pre-commit
- Run
pre-commit install
while in the repository directory
Your git commits will then automatically be checked. If a check fails, an error
message will explain why, and your commit will be canceled. You can then make
the suggested changes, and run git commit ...
again.
clippy
cargo clippy --all-targets --all-features
The msrv
key in the clippy configuration file clippy.toml
is used to disable
lints pertaining to newer features by specifying the minimum supported Rust
version (MSRV).
rustfmt
cargo fmt --all
cargo-deny
This project uses cargo-deny to detect duplicate dependencies, checks licenses, etc. To run it locally, first install it and then run with:
cargo deny --all-features check all
Markdown linter
We use markdownlint to lint the Markdown files in the repository.
Spell checker
We use cspell
as spell checker for all files in the project. If you are using
VS Code, you can install the
code spell checker
extension to enable spell checking within your editor. Otherwise, you can
install cspell separately.
If you want to make the spell checker ignore a word, you can add
// spell-checker:ignore word_to_ignore
at the top of the file.
Testing
Testing can be done using either Cargo or make
.
Testing with Cargo
Just like with building, we follow the standard procedure for testing using Cargo:
cargo test
By default, cargo test
only runs the common programs. To run also platform
specific tests, run:
cargo test --features unix
If you would prefer to test a select few utilities:
cargo test --features "chmod mv tail" --no-default-features
If you also want to test the core utilities:
cargo test -p uucore -p coreutils
Running the complete test suite might take a while. We use nextest in the CI and you might want to try it out locally. It can speed up the execution time of the whole test run significantly if the cpu has multiple cores.
cargo nextest run --features unix --no-fail-fast
To debug:
gdb --args target/debug/coreutils ls
(gdb) b ls.rs:79
(gdb) run
Testing with GNU Make
To simply test all available utilities:
make test
To test all but a few of the available utilities:
make SKIP_UTILS='UTILITY_1 UTILITY_2' test
To test only a few of the available utilities:
make UTILS='UTILITY_1 UTILITY_2' test
To include tests for unimplemented behavior:
make UTILS='UTILITY_1 UTILITY_2' SPEC=y test
To run tests with nextest
just use the nextest target. Note you'll need to
install nextest
first. The nextest
target accepts the
same arguments like the default test
target, so it's possible to pass arguments to nextest run
via CARGOFLAGS
:
make CARGOFLAGS='--no-fail-fast' UTILS='UTILITY_1 UTILITY_2' nextest
Run Busybox Tests
This testing functionality is only available on *nix operating systems and
requires make
.
To run busybox tests for all utilities for which busybox has tests
make busytest
To run busybox tests for a few of the available utilities
make UTILS='UTILITY_1 UTILITY_2' busytest
To pass an argument like "-v" to the busybox test runtime
make UTILS='UTILITY_1 UTILITY_2' RUNTEST_ARGS='-v' busytest
Comparing with GNU
To run uutils against the GNU test suite locally, run the following commands:
bash util/build-gnu.sh
# Build uutils without release optimizations
UU_MAKE_PROFILE=debug bash util/build-gnu.sh
bash util/run-gnu-test.sh
# To run a single test:
bash util/run-gnu-test.sh tests/touch/not-owner.sh # for example
# To run several tests:
bash util/run-gnu-test.sh tests/touch/not-owner.sh tests/rm/no-give-up.sh # for example
# If this is a perl (.pl) test, to run in debug:
DEBUG=1 bash util/run-gnu-test.sh tests/misc/sm3sum.pl
Note that it relies on individual utilities (not the multicall binary).
Improving the GNU compatibility
The Python script ./util/remaining-gnu-error.py
shows the list of failing
tests in the CI.
To improve the GNU compatibility, the following process is recommended:
- Identify a test (the smaller, the better) on a program that you understand or
is easy to understand. You can use the
./util/remaining-gnu-error.py
script to help with this decision. - Build both the GNU and Rust coreutils using:
bash util/build-gnu.sh
- Run the test with
bash util/run-gnu-test.sh <your test>
- Start to modify
<your test>
to understand what is wrong. Examples:- Add
set -v
to have the bash verbose mode - Add
echo $?
where needed - When the variable
fail
is used in the test,echo $fail
to see when the test started to fail - Bump the content of the output (ex:
cat err
) - ...
- Add
- Or, if the test is simple, extract the relevant information to create a new test case running both GNU & Rust implementation
- Start to modify the Rust implementation to match the expected behavior
- Add a test to make sure that we don't regress (our test suite is super quick)
Commit messages
To help the project maintainers review pull requests from contributors across numerous utilities, the team has settled on conventions for commit messages.
From https://git-scm.com/book/ch5-2.html:
Capitalized, short (50 chars or less) summary
More detailed explanatory text, if necessary. Wrap it to about 72
characters or so. In some contexts, the first line is treated as the
subject of an email and the rest of the text as the body. The blank
line separating the summary from the body is critical (unless you omit
the body entirely); tools like rebase will confuse you if you run the
two together.
Write your commit message in the imperative: "Fix bug" and not "Fixed bug"
or "Fixes bug." This convention matches up with commit messages generated
by commands like git merge and git revert.
Further paragraphs come after blank lines.
- Bullet points are okay, too
- Typically a hyphen or asterisk is used for the bullet, followed by a
single space, with blank lines in between, but conventions vary here
- Use a hanging indent
Furthermore, here are a few examples for a summary line:
- commit for a single utility
nohup: cleanup and refactor
- commit for a utility's tests
tests/rm: test new feature
Beyond changes to an individual utility or its tests, other summary lines for non-utility modules include:
README: add help
uucore: add new modules
uutils: add new utility
gitignore: add temporary files
Code coverage
Code coverage report can be generated using grcov.
Using Nightly Rust
To generate gcov-based coverage report
export CARGO_INCREMENTAL=0
export RUSTFLAGS="-Zprofile -Ccodegen-units=1 -Copt-level=0 -Clink-dead-code -Coverflow-checks=off -Zpanic_abort_tests -Cpanic=abort"
export RUSTDOCFLAGS="-Cpanic=abort"
cargo build <options...> # e.g., --features feat_os_unix
cargo test <options...> # e.g., --features feat_os_unix test_pathchk
grcov . -s . --binary-path ./target/debug/ -t html --branch --ignore-not-existing --ignore build.rs --excl-br-line "^\s*((debug_)?assert(_eq|_ne)?\#\[derive\()" -o ./target/debug/coverage/
# open target/debug/coverage/index.html in browser
if changes are not reflected in the report then run cargo clean
and run the above commands.
Using Stable Rust
If you are using stable version of Rust that doesn't enable code coverage instrumentation by default
then add -Z-Zinstrument-coverage
flag to RUSTFLAGS
env variable specified above.
Other implementations
The Coreutils have different implementations, with different levels of completions:
However, when reimplementing the tools/options in Rust, don't read their source codes when they are using reciprocal licenses (ex: GNU GPL, GNU LGPL, etc).
Licensing
uutils is distributed under the terms of the MIT License; see the LICENSE
file
for details. This is a permissive license, which allows the software to be used
with few restrictions.
Copyrights in the uutils project are retained by their contributors, and no copyright assignment is required to contribute.
If you wish to add or change dependencies as part of a contribution to the
project, a tool like cargo-license
can be used to show their license details.
The following types of license are acceptable:
- MIT License
- Dual- or tri-license with an MIT License option ("Apache-2.0 or MIT" is a popular combination)
- "MIT equivalent" license (2-clause BSD, 3-clause BSD, ISC)
- License less restrictive than the MIT License (CC0 1.0 Universal)
- Apache License version 2.0
Licenses we will not use:
- An ambiguous license, or no license
- Strongly reciprocal licenses (GNU GPL, GNU LGPL)
If you wish to add a reference but it doesn't meet these requirements, please raise an issue to describe the dependency.