Always show courtesy to individuals submitting issues and pull requests. Be welcoming to first-time contributors, identified by the GitHub badge.
For first-time contributors, check if the commit author is the same as the pull request author. This way, once their pull request lands, GitHub will show them as a Contributor. Ask if they have configured their git username and email to their liking.
Collaborators may close any issue or pull request that is not relevant to the future of the Node.js project. Where this is unclear, leave the issue or pull request open for several days to allow for discussion. Where this does not yield evidence that the issue or pull request has relevance, close it. Remember that issues and pull requests can always be re-opened if necessary.
A pull request is author ready when:
Please always add the
author ready label to the pull request in that case. Please always remove it again as soon as the conditions are not met anymore.
When you open a pull request, start a CI right away. Later, after new code changes or rebasing, start a new CI.
As soon as the pull request is ready to land, please do so. This allows other Collaborators to focus on other pull requests. If your pull request is not ready to land but is author ready, add the
author ready label. If you wish to land the pull request yourself, use the “assign yourself” link to self-assign it.
Contributors propose modifications to Node.js using GitHub pull requests. This includes modifications proposed by TSC members and other Collaborators. A pull request must pass code review and CI before landing into the codebase.
At least two Collaborators must approve a pull request before the pull request lands. One Collaborator approval is enough if the pull request has been open for more than seven days.
Approving a pull request indicates that the Collaborator accepts responsibility for the change.
Approval must be from Collaborators who are not authors of the change.
In some cases, it may be necessary to summon a GitHub team to a pull request for review by @-mention. See Who to CC in the issue tracker.
If you are the first Collaborator to approve a pull request that has no CI yet, please start one. Please also start a new CI if the PR creator pushed new code since the last CI run.
If there are no objecting Collaborators, a pull request may land if it has the needed approvals, CI, and wait time. If a pull request meets all requirements except the wait time, please add the
author ready label.
Where there is disagreement among Collaborators, consensus should be sought if possible. If reaching consensus is not possible, a Collaborator may escalate the issue to the TSC.
Collaborators should not block a pull request without providing a reason. Another Collaborator may ask an objecting Collaborator to explain their objection. If the objector is unresponsive, another Collaborator may dismiss the objection.
Breaking changes must receive TSC review. If two TSC members approve the pull request and no Collaborators object, then it may land. If there are objections, a Collaborator may apply the
tsc-agenda label. That will put the pull request on the TSC meeting agenda.
Before landing pull requests, allow 48 hours for input from other Collaborators. Certain types of pull requests can be fast-tracked and may land after a shorter delay. For example:
code-and-learntasks often fall into this category.
good-first-issuepull requests may also be suitable.
To propose fast-tracking a pull request, apply the
fast-track label. Then add a comment that Collaborators may upvote.
If someone disagrees with the fast-tracking request, remove the label. Do not fast-track the pull request in that case.
The pull request may be fast-tracked if two Collaborators approve the fast-tracking request. To land, the pull request itself still needs two Collaborator approvals and a passing CI.
Collaborators may request fast-tracking of pull requests they did not author. In that case only, the request itself is also one fast-track approval. Upvote the comment anyway to avoid any doubt.
All fixes must have a test case which demonstrates the defect. The test should fail before the change, and pass after the change.
All pull requests must pass continuous integration tests. Code changes must pass on project CI server. Pull requests that only change documentation and comments can use GitHub Actions results.
Do not land any pull requests without passing (green or yellow) CI runs. If there are CI failures unrelated to the change in the pull request, try “Resume Build”. It is in the left navigation of the relevant
node-test-pull-request job. It will preserve all the green results from the current job but re-run everything else. Start a fresh CI if more than seven days have elapsed since the original failing CI as the compiled binaries for the Windows and ARM platforms are only kept for seven days.
node-test-pull-request is the CI job to test pull requests. It runs the
test-ci targets on all supported platforms.
node-stress-single-test can run a group of tests over and over on a specific platform. Use it to check that the tests are reliable.
node-test-commit-v8-linux runs the standard V8 tests. Run it when updating V8 in Node.js or floating new patches on V8.
node-test-commit-custom-suites-freestyle enables customization of test suites and parameters. It can execute test suites not used in other CI test runs (such as tests in the
pummel directories). It can also make sure tests pass when provided with a flag not used in other CI test runs (such as
From the CI Job page, click “Build with Parameters” on the left side.
You generally need to enter only one or both of the following options in the form:
GIT_REMOTE_REF: Change to the remote portion of git refspec. To specify the branch this way,
refs/heads/BRANCHis used (e.g. for
refs/heads/master). For pull requests, it will look like
refs/pull/PR_NUMBER/head(e.g. for PR#42 ->
REBASE_ONTO: Change that to
origin/masterso the pull request gets rebased onto master. This can especially be important for pull requests that have been open a while.
Look at the list of jobs on the left hand side under “Build History” and copy the link to the one you started (which will be the one on top, but click through to make sure it says something like “Started 5 seconds ago” (top right) and “Started by user ...”.
Copy/paste the URL for the job into a comment in the pull request.
node-test-pull-request is an exception where the GitHub bot will automatically post for you.
All functionality in the official Node.js documentation is part of the public API. Any undocumented object, property, method, argument, behavior, or event is internal. There are exceptions to this rule. Node.js users have come to rely on some undocumented behaviors. Collaborators treat many of those undocumented behaviors as public.
All undocumented functionality exposed via
process.binding(...) is internal.
All undocumented functionality in
lib/internal/**/*.js is internal. It is public, though, if it is re-exported by code in
Symbol properties and methods are internal.
Any undocumented object property or method that begins with
_ is internal.
Any native C/C++ APIs/ABIs requiring the
NODE_WANT_INTERNALS flag are internal.
Sometimes, there is disagreement about whether functionality is internal or public. In those cases, the TSC makes a determination.
For undocumented APIs that are public, open a pull request documenting the API.
At least two TSC members must approve backward-incompatible changes to the master branch.
Examples of breaking changes include:
Existing stable public APIs that change in a backward-incompatible way must undergo deprecation. The exceptions to this rule are:
For more information, see Deprecations.
Breaking changes to internal elements may occur in semver-patch or semver-minor commits. Take significant care when making and reviewing such changes. Make an effort to determine the potential impact of the change in the ecosystem. Use Canary in the Goldmine to test such changes. If a change will cause ecosystem breakage, then it is semver-major. Consider providing a Public API in such cases.
Sometimes, a change intended to be non-breaking turns out to be a breaking change. If such a change lands on the master branch, a Collaborator may revert it. As an alternative to reverting, the TSC may apply the semver-major label after-the-fact.
Revert commits with
git revert <HASH> or
git revert <FROM>..<TO>. The generated commit message will not have a subsystem and may violate line length rules. That is OK. Append the reason for the revert and any
Fixes metadata. Raise a Pull Request like any other change.
Treat commits that introduce new core modules with extra care.
Check if the module's name conflicts with an existing ecosystem module. If it does, choose a different name unless the module owner has agreed in writing to transfer it.
If the new module name is free, register a placeholder in the module registry as soon as possible. Link to the pull request that introduces the new core module in the placeholder's
For pull requests introducing new core modules:
N-API provides an ABI-stable API guaranteed for future Node.js versions. N-API additions call for unusual care and scrutiny. If a change adds to
js_native_api_types.h, consult the relevant guide.
Node.js uses three Deprecation levels. For all deprecated APIs, the API documentation must state the deprecation status.
--throw-deprecationflag, will throw a runtime error.
notable change label to all pull requests that introduce Documentation-Only Deprecations. Such deprecations have no impact on code execution. Thus, they are not breaking changes (
Runtime Deprecations and End-of-Life APIs (internal or public) are breaking changes (
semver-major). The TSC may make exceptions, deciding that one of these deprecations is not a breaking change.
Avoid Runtime Deprecations when an alias or a stub/no-op will suffice. An alias or stub will have lower maintenance costs for end users and Node.js core.
All deprecations receive a unique and immutable identifier. Documentation, warnings, and errors use the identifier when referring to the deprecation. The documentation for the deprecation identifier must always remain in the API documentation. This is true even if the deprecation is no longer in use (for example, due to removal of an End-of-Life deprecated API).
A deprecation cycle is a major release during which an API has been in one of the three Deprecation levels. Documentation-Only Deprecations may land in a minor release. They may not change to a Runtime Deprecation until the next major release.
No API can change to End-of-Life without going through a Runtime Deprecation cycle. There is no rule that deprecated code must progress to End-of-Life. Documentation-Only and Runtime Deprecations may remain in place for an unlimited duration.
Communicate pending deprecations and associated mitigations with the ecosystem as soon as possible. If possible, do it before the pull request adding the deprecation lands on the master branch.
notable-change label on pull requests that add or change the deprecation level of an API.
Collaborators may opt to elevate pull requests or issues to the TSC. Do this if a pull request or issue:
@nodejs/tsc GitHub team if you want to elevate an issue to the TSC. Do not use the GitHub UI on the right-hand side to assign to
@nodejs/tsc or request a review from
The TSC should serve as the final arbiter where required.
For pull requests from first-time contributors, be welcoming. Also, verify that their git settings are to their liking.
All commits should be self-contained, meaning every commit should pass all tests. This makes it much easier when bisecting to find a breaking change.
$ npm install -g node-core-utils $ git node land $PRID
node-core-utils, you will need a GitHub access token. If you do not have one,
node-core-utils will create one for you the first time you use it. To do this, it will ask for your GitHub password and two-factor authentication code. If you wish to create the token yourself in advance, see the
rebase that may already be underway:
$ git am --abort $ git rebase --abort
Checkout proper target branch:
$ git checkout master
Update the tree (assumes your repo is set up as detailed in CONTRIBUTING.md):
$ git fetch upstream $ git merge --ff-only upstream/master
Apply external patches:
$ curl -L https://github.com/nodejs/node/pull/xxx.patch | git am --whitespace=fix
If the merge fails even though recent CI runs were successful, try a 3-way merge:
$ git am --abort $ curl -L https://github.com/nodejs/node/pull/xxx.patch | git am -3 --whitespace=fix
If the 3-way merge succeeds, check the results against the original pull request. Build and test on at least one platform before landing.
If the 3-way merge fails, then it is most likely that a conflicting pull request has landed since the CI run. You will have to ask the author to rebase.
Check and re-review the changes:
$ git diff upstream/master
Check the number of commits and commit messages:
$ git log upstream/master...master
Squash commits and add metadata:
$ git rebase -i upstream/master
This will open a screen like this (in the default shell editor):
pick 6928fc1 crypto: add feature A pick 8120c4c add test for feature A pick 51759dc crypto: feature B pick 7d6f433 test for feature B # Rebase f9456a2..7d6f433 onto f9456a2 # # Commands: # p, pick = use commit # r, reword = use commit, but edit the commit message # e, edit = use commit, but stop for amending # s, squash = use commit, but meld into previous commit # f, fixup = like "squash", but discard this commit's log message # x, exec = run command (the rest of the line) using shell # # These lines can be re-ordered; they are executed from top to bottom. # # If you remove a line here THAT COMMIT WILL BE LOST. # # However, if you remove everything, the rebase will be aborted. # # Note that empty commits are commented out
Replace a couple of
fixup to squash them into a previous commit:
pick 6928fc1 crypto: add feature A fixup 8120c4c add test for feature A pick 51759dc crypto: feature B fixup 7d6f433 test for feature B
reword to change the commit message:
reword 6928fc1 crypto: add feature A fixup 8120c4c add test for feature A reword 51759dc crypto: feature B fixup 7d6f433 test for feature B
Save the file and close the editor. When prompted, enter a new commit message for that commit. This is an opportunity to fix commit messages.
The commit message text must conform to the commit message guidelines.
Change the original commit message to include metadata. (The
git node metadata command can generate the metadata for you.)
PR-URL:line that references the full GitHub URL of the pull request. This makes it easy to trace a commit back to the conversation that led up to that change.
Fixes: Xline, where X is the full GitHub URL for an issue. A commit message may include more than one
Refs:lines referencing a URL for any relevant background.
Reviewed-By: Name <email>line for each Collaborator who reviewed the change.
Other changes may have landed on master since the successful CI run. As a precaution, run tests (
make -j4 test or
Confirm that the commit message format is correct using core-validate-commit.
$ git rev-list upstream/master...HEAD | xargs core-validate-commit
Optional: For your own commits, force push the amended commit to the pull request branch. If your branch name is
git push --force-with-lease origin master:bugfix. Don't close the PR. It will close after you push it upstream. It will have the purple merged status rather than the red closed status. If you close the PR before GitHub adjusts its status, it will show up as a 0 commit PR with no changed files. The order of operations is important. If you push upstream before you push to your branch, GitHub will close the issue with the red closed status.
Time to push it:
$ git push upstream master
Close the pull request with a “Landed in
<commit hash>” comment. If your pull request shows the purple merged status then you should still add the “Landed in <commit hash>..<commit hash>” comment if you added more than one commit.
Sometimes, when running
git push upstream master, you may get an error message like this:
To https://github.com/nodejs/node ! [rejected] master -> master (fetch first) error: failed to push some refs to 'https://github.com/nodejs/node' hint: Updates were rejected because the tip of your current branch is behind hint: its remote counterpart. Integrate the remote changes (e.g. hint: 'git pull ...') before pushing again. hint: See the 'Note about fast-forwards' in 'git push --help' for details.
That means a commit has landed since your last rebase against
upstream/master. To fix this, pull with rebase from upstream, run the tests again, and (if the tests pass) push again:
git pull upstream master --rebase make -j4 test git push upstream master
git, there‘s a way to override remote trees by force pushing (
git push -f). This is generally forbidden as it creates conflicts in other people’s forks. It is permissible for simpler slip-ups such as typos in commit messages. You are only allowed to force push to any Node.js branch within 10 minutes from your original push. If someone else pushes to the branch or the 10-minute period passes, consider the commit final.
--force-with-leaseto reduce the chance of overwriting someone else's change.
#node-dev(IRC) if you force push.
Long Term Support (LTS) guarantees 30-month support cycles for specific Node.js versions. You can find more information in the full release plan. Once a branch enters LTS, the release plan limits the types of changes permitted in the branch.
Each LTS release has a corresponding branch (v10.x, v8.x, etc.). Each also has a corresponding staging branch (v10.x-staging, v8.x-staging, etc.).
Commits that land on master are cherry-picked to each staging branch as appropriate. If a change applies only to the LTS branch, open the PR against the staging branch. Commits from the staging branch land on the LTS branch only when a release is being prepared. They may land on the LTS branch in a different order than they were in staging.
Only members of @nodejs/backporters should land commits onto LTS staging branches.
When you send your pull request, please state if your change is breaking. Also state if you think your patch is a good candidate for backporting. For more information on backporting, please see the backporting guide.
There are several LTS-related labels:
lts-watch- labels are for pull requests to consider for landing in staging branches. For example,
lts-watch-v10.x would be for a change to consider for the
land-on- are for pull requests that should land in a future v*.x release. For example,
land-on-v10.x would be for a change to land in Node.js 10.x.
Any Collaborator can attach these labels to any pull request/issue. As commits land on the staging branches, the backporter removes the
lts-watch- label. Likewise, as commits land in an LTS release, the releaser removes the
Attach the appropriate
lts-watch- label to any PR that may impact an LTS release.
|@nodejs/async_hooks for bugs/reviews (+ @nodejs/diagnostics for API)|
|@bnoordhuis, @indutny, @nodejs/streams|
|upgrading http-parser||@nodejs/http, @nodejs/http2|
|upgrading npm||@fishrock123, @MylesBorins|
|upgrading V8||@nodejs/V8, @nodejs/post-mortem|
|Embedded use or delivery of Node.js||@nodejs/delivery-channels|
When things need extra attention, are controversial, or
If you cannot find who to cc for a file,
git shortlog -n -s <file> may help.