Contribution Procedures

Bug reports

While bugs are unfortunate, they're a reality in software. We can't fix what we don't know about, so please report liberally. If you're not sure if something is a bug or not, feel free to file a bug anyway.

If you believe reporting your bug publicly represents a security risk to Rust users, please follow our instructions for reporting security vulnerabilities.

If you're using the nightly channel, please check if the bug exists in the latest toolchain before filing your bug. It might be fixed already.

If you have the chance, before reporting a bug, please search existing issues, as it's possible that someone else has already reported your error. This doesn't always work, and sometimes it's hard to know what to search for, so consider this extra credit. We won't mind if you accidentally file a duplicate report.

Similarly, to help others who encountered the bug find your issue, consider filing an issue with a descriptive title, which contains information that might be unique to it. This can be the language or compiler feature used, the conditions that trigger the bug, or part of the error message if there is any. An example could be: "impossible case reached" on lifetime inference for impl Trait in return position.

Opening an issue is as easy as following this link and filling out the fields in the appropriate provided template.

Bug fixes or "normal" code changes

For most PRs, no special procedures are needed. You can just open a PR, and it will be reviewed, approved, and merged. This includes most bug fixes, refactorings, and other user-invisible changes. The next few sections talk about exceptions to this rule.

Also, note that it is perfectly acceptable to open WIP PRs or GitHub Draft PRs. Some people prefer to do this so they can get feedback along the way or share their code with a collaborator. Others do this so they can utilize the CI to build and test their PR (e.g. when developing on a slow machine).

New features

Rust has strong backwards-compatibility guarantees. Thus, new features can't just be implemented directly in stable Rust. Instead, we have 3 release channels: stable, beta, and nightly.

  • Stable: this is the latest stable release for general usage.
  • Beta: this is the next release (will be stable within 6 weeks).
  • Nightly: follows the master branch of the repo. This is the only channel where unstable, incomplete, or experimental features are usable with feature gates.

See this chapter on implementing new features for more information.

Breaking changes

Breaking changes have a dedicated section in the dev-guide.

Major changes

The compiler team has a special process for large changes, whether or not they cause breakage. This process is called a Major Change Proposal (MCP). MCP is a relatively lightweight mechanism for getting feedback on large changes to the compiler (as opposed to a full RFC or a design meeting with the team).

Example of things that might require MCPs include major refactorings, changes to important types, or important changes to how the compiler does something, or smaller user-facing changes.

When in doubt, ask on zulip. It would be a shame to put a lot of work into a PR that ends up not getting merged! See this document for more info on MCPs.

Performance

Compiler performance is important. We have put a lot of effort over the last few years into gradually improving it.

If you suspect that your change may cause a performance regression (or improvement), you can request a "perf run" (and your reviewer may also request one before approving). This is yet another bot that will compile a collection of benchmarks on a compiler with your changes. The numbers are reported here, and you can see a comparison of your changes against the latest master.

For an introduction to the performance of Rust code in general which would also be useful in rustc development, see The Rust Performance Book.

Pull requests

Pull requests (or PRs for short) are the primary mechanism we use to change Rust. GitHub itself has some great documentation on using the Pull Request feature. We use the "fork and pull" model described here, where contributors push changes to their personal fork and create pull requests to bring those changes into the source repository. We have more info about how to use git when contributing to Rust under the git section.

r?

All pull requests are reviewed by another person. We have a bot, @rustbot, that will automatically assign a random person to review your request based on which files you changed.

If you want to request that a specific person reviews your pull request, you can add an r? to the pull request description or in a comment. For example, if you want to ask a review to @awesome-reviewer, add

r? @awesome-reviewer

to the end of the pull request description, and @rustbot will assign them instead of a random person. This is entirely optional.

You can also assign a random reviewer from a specific team by writing r? rust-lang/groupname. As an example, if you were making a diagnostics change, then you could get a reviewer from the diagnostics team by adding:

r? rust-lang/diagnostics

For a full list of possible groupnames, check the adhoc_groups section at the triagebot.toml config file, or the list of teams in the rust-lang teams database.

Waiting for reviews

NOTE

Pull request reviewers are often working at capacity, and many of them are contributing on a volunteer basis. In order to minimize review delays, pull request authors and assigned reviewers should ensure that the review label (S-waiting-on-review and S-waiting-on-author) stays updated, invoking these commands when appropriate:

  • @rustbot author: the review is finished, and PR author should check the comments and take action accordingly.

  • @rustbot review: the author is ready for a review, and this PR will be queued again in the reviewer's queue.

Please note that the reviewers are humans, who for the most part work on rustc in their free time. This means that they can take some time to respond and review your PR. It also means that reviewers can miss some PRs that are assigned to them.

To try to move PRs forward, the Triage WG regularly goes through all PRs that are waiting for review and haven't been discussed for at least 2 weeks. If you don't get a review within 2 weeks, feel free to ask the Triage WG on Zulip (#t-release/triage). They have knowledge of when to ping, who might be on vacation, etc.

The reviewer may request some changes using the GitHub code review interface. They may also request special procedures for some PRs. See Crater and Breaking Changes chapters for some examples of such procedures.

CI

In addition to being reviewed by a human, pull requests are automatically tested, thanks to continuous integration (CI). Basically, every time you open and update a pull request, CI builds the compiler and tests it against the compiler test suite, and also performs other tests such as checking that your pull request is in compliance with Rust's style guidelines.

Running continuous integration tests allows PR authors to catch mistakes early without going through a first review cycle, and also helps reviewers stay aware of the status of a particular pull request.

Rust has plenty of CI capacity, and you should never have to worry about wasting computational resources each time you push a change. It is also perfectly fine (and even encouraged!) to use the CI to test your changes if it can help your productivity. In particular, we don't recommend running the full ./x test suite locally, since it takes a very long time to execute.

r+

After someone has reviewed your pull request, they will leave an annotation on the pull request with an r+. It will look something like this:

@bors r+

This tells @bors, our lovable integration bot, that your pull request has been approved. The PR then enters the merge queue, where @bors will run all the tests on every platform we support. If it all works out, @bors will merge your code into master and close the pull request.

Depending on the scale of the change, you may see a slightly different form of r+:

@bors r+ rollup

The additional rollup tells @bors that this change should always be "rolled up". Changes that are rolled up are tested and merged alongside other PRs, to speed the process up. Typically only small changes that are expected not to conflict with one another are marked as "always roll up".

Be patient; this can take a while and the queue can sometimes be long. PRs are never merged by hand.

Opening a PR

You are now ready to file a pull request? Great! Here are a few points you should be aware of.

All pull requests should be filed against the master branch, unless you know for sure that you should target a different branch.

Make sure your pull request is in compliance with Rust's style guidelines by running

$ ./x test tidy --bless

We recommend to make this check before every pull request (and every new commit in a pull request); you can add git hooks before every push to make sure you never forget to make this check. The CI will also run tidy and will fail if tidy fails.

Rust follows a no merge-commit policy, meaning, when you encounter merge conflicts you are expected to always rebase instead of merging. E.g. always use rebase when bringing the latest changes from the master branch to your feature branch. If your PR contains merge commits, it will get marked as has-merge-commits. Once you have removed the merge commits, e.g., through an interactive rebase, you should remove the label again:

@rustbot label -has-merge-commits

See this chapter for more details.

If you encounter merge conflicts or when a reviewer asks you to perform some changes, your PR will get marked as S-waiting-on-author. When you resolve them, you should use @rustbot to mark it as S-waiting-on-review:

@rustbot ready

GitHub allows closing issues using keywords. This feature should be used to keep the issue tracker tidy. However, it is generally preferred to put the "closes #123" text in the PR description rather than the issue commit; particularly during rebasing, citing the issue number in the commit can "spam" the issue in question.

However, if your PR fixes a stable-to-beta or stable-to-stable regression and has been accepted for a beta and/or stable backport (i.e., it is marked beta-accepted and/or stable-accepted), please do not use any such keywords since we don't want the corresponding issue to get auto-closed once the fix lands on master. Please update the PR description while still mentioning the issue somewhere. For example, you could write Fixes (after beta backport) #NNN..

As for further actions, please keep a sharp look-out for a PR whose title begins with [beta] or [stable] and which backports the PR in question. When that one gets merged, the relevant issue can be closed. The closing comment should mention all PRs that were involved. If you don't have the permissions to close the issue, please leave a comment on the original PR asking the reviewer to close it for you.

External dependencies

This section has moved to "Using External Repositories".

Writing documentation

Documentation improvements are very welcome. The source of doc.rust-lang.org is located in src/doc in the tree, and standard API documentation is generated from the source code itself (e.g. library/std/src/lib.rs). Documentation pull requests function in the same way as other pull requests.

To find documentation-related issues, sort by the A-docs label.

You can find documentation style guidelines in RFC 1574.

To build the standard library documentation, use x doc --stage 0 library --open. To build the documentation for a book (e.g. the unstable book), use x doc src/doc/unstable-book. Results should appear in build/host/doc, as well as automatically open in your default browser. See Building Documentation for more information.

You can also use rustdoc directly to check small fixes. For example, rustdoc src/doc/reference.md will render reference to doc/reference.html. The CSS might be messed up, but you can verify that the HTML is right.

Contributing to rustc-dev-guide

Contributions to the rustc-dev-guide are always welcome, and can be made directly at the rust-lang/rustc-dev-guide repo. The issue tracker in that repo is also a great way to find things that need doing. There are issues for beginners and advanced compiler devs alike!

Just a few things to keep in mind:

  • Please limit line length to 100 characters. This is enforced by CI, and you can run the checks locally with ci/lengthcheck.sh.

  • When contributing text to the guide, please contextualize the information with some time period and/or a reason so that the reader knows how much to trust or mistrust the information. Aim to provide a reasonable amount of context, possibly including but not limited to:

    • A reason for why the data may be out of date other than "change", as change is a constant across the project.

    • The date the comment was added, e.g. instead of writing "Currently, ..." or "As of now, ...", consider adding the date, in one of the following formats:

      • Jan 2021
      • January 2021
      • jan 2021
      • january 2021

      There is a CI action (in ~/.github/workflows/date-check.yml) that generates a monthly issue with any of these that are over 6 months old.

      For the action to pick the date, add a special annotation before specifying the date:

      <!-- date-check --> Jan 2023
      

      Example:

      As of <!-- date-check --> Jan 2023, the foo did the bar.
      

      For cases where the date should not be part of the visible rendered output, use the following instead:

      <!-- date-check: Jan 2023 -->
      
    • A link to a relevant WG, tracking issue, rustc rustdoc page, or similar, that may provide further explanation for the change process or a way to verify that the information is not outdated.

  • If a text grows rather long (more than a few page scrolls) or complicated (more than four subsections) it might benefit from having a Table of Contents at the beginning, which you can auto-generate by including the <!-- toc --> marker.

Issue triage

Sometimes, an issue will stay open, even though the bug has been fixed. And sometimes, the original bug may go stale because something has changed in the meantime.

It can be helpful to go through older bug reports and make sure that they are still valid. Load up an older issue, double check that it's still true, and leave a comment letting us know if it is or is not. The least recently updated sort is good for finding issues like this.

Thanks to @rustbot, anyone can help triage issues by adding appropriate labels to issues that haven't been triaged yet:

LabelsColorDescription
A- YellowThe area of the project an issue relates to.
B- MagentaIssues which are blockers.
beta- Dark BlueTracks changes which need to be backported to beta
C- Light PurpleThe category of an issue.
D- Mossy GreenIssues for diagnostics.
E- GreenThe experience level necessary to fix an issue.
F- PeachIssues for nightly features.
I- RedThe importance of the issue.
I-*-nominated RedThe issue has been nominated for discussion at the next meeting of the corresponding team.
I-prioritize RedThe issue has been nominated for prioritization by the team tagged with a T-prefixed label.
metabug PurpleBugs that collect other bugs.
O- Purple GreyThe operating system or platform that the issue is specific to.
P- OrangeThe issue priority. These labels can be assigned by anyone that understand the issue and is able to prioritize it, and remove the I-prioritize label.
regression- PinkTracks regressions from a stable release.
relnotes Light OrangeChanges that should be documented in the release notes of the next release.
S- GrayTracks the status of pull requests.
S-tracking- Steel BlueTracks the status of tracking issues.
stable- Dark BlueTracks changes which need to be backported to stable in anticipation of a point release.
T- BlueDenotes which team the issue belongs to.
WG- GreenDenotes which working group the issue belongs to.

Rfcbot labels

rfcbot uses its own labels for tracking the process of coordinating asynchronous decisions, such as approving or rejecting a change. This is used for RFCs, issues, and pull requests.

LabelsColorDescription
proposed-final-comment-period GrayCurrently awaiting signoff of all team members in order to enter the final comment period.
disposition-merge GreenIndicates the intent is to merge the change.
disposition-close RedIndicates the intent is to not accept the change and close it.
disposition-postpone GrayIndicates the intent is to not accept the change at this time and postpone it to a later date.
final-comment-period BlueCurrently soliciting final comments before merging or closing.
finished-final-comment-period Light YellowThe final comment period has concluded, and the issue will be merged or closed.
postponed YellowThe issue has been postponed.
closed RedThe issue has been rejected.
to-announce GrayIssues that have finished their final-comment-period and should be publicly announced. Note: the rust-lang/rust repository uses this label differently, to announce issues at the triage meetings.

This section has moved to the "About this guide" chapter.