8.5 KiB
pre-setup
Ensure everyone is added to https://github.com/orgs/nodejs/teams/collaborators
onboarding to nodejs
intros
thank you for doing this
- going to cover four things:
- local setup
- some project goals & values
- issues, labels, and reviewing code
- merging code
setup:
-
notifications setup
- use https://github.com/notifications or set up email
- watching the main repo will flood your inbox, so be prepared
-
git:
- make sure you have whitespace=fix:
git config --global --add core.whitespace fix
- usually PR from your own github fork
- See "Updating Node.js from Upstream"
- make new branches for all commits you make!
- make sure you have whitespace=fix:
-
#node-dev
onchat.freenode.net
is the best place to interact with the CTC / other collaborators
a little deeper about the project
-
collaborators are effectively part owners
- the project has the goals of its contributors
-
but, there are some higher-level goals and values
- not everything belongs in core (if it can be done reasonably in userland, let it stay in userland)
- empathy towards users matters (this is in part why we onboard people)
- generally: try to be nice to people
managing the issue tracker
-
you have (mostly) free rein – don't hesitate to close an issue if you are confident that it should be closed
- this will come more naturally over time
- IMPORTANT: be nice about closing issues, let people know why, and that issues and PRs can be reopened if necessary
- Still need to follow the Code of Conduct.
-
labels:
- generally sort issues by a concept of "subsystem" so that we know what part(s) of the codebase it touches, though there are also other useful labels.
- See "Labels"
ctc-agenda
if a topic is controversial or isn't coming to a conclusion after an extended time.semver-{minor,major}
:- be conservative – that is, if a change has the remote chance of breaking something, go for
semver-major
- when adding a semver label, add a comment explaining why you're adding it
- it's cached locally in your brain at that moment!
- be conservative – that is, if a change has the remote chance of breaking something, go for
-
Notifying humans
- See "Who to CC in issues"
- will also come more naturally over time
-
reviewing:
- primary goal is for the codebase to improve
- secondary (but not far off) is for the person submitting code to succeed
- helps grow the community
- and draws new people into the project
- Review a bit at a time. It is very important to not overwhelm newer people.
- it is tempting to micro-optimize / make everything about relative perf, don't succumb to that temptation. we change v8 a lot more often now, contortions that are zippy today may be unnecessary in the future
- be aware: your opinion carries a lot of weight!
- nits are fine, but try to avoid stalling the PR
- note that they are nits when you comment
- if they really are stalling nits, fix them yourself on merge (but try to let PR authors know they can fix these)
- improvement doesn't have to come all at once
- minimum wait for comments time
- There is a minimum waiting time which we try to respect for non-trivial changes, so that people who may have important input in such a distributed project are able to respond.
- It may help to set time limits and expectations:
- the collaborators are very distributed so it is unlikely that they will be looking at stuff the same time as you are.
- before merging code: give folks at least one working day to respond: "If no one objects, tomorrow at
- check in on the code every once in a while (set reminders!)
- It may help to set time limits and expectations:
- 48 hours for non-trivial changes, and 72 hours on weekends.
- if a PR is abandoned, check if they'd mind if you took it over (especially if it just has nits left)
- you have the power to
LGTM
another collaborator or TSC / CTC members' work
- There is a minimum waiting time which we try to respect for non-trivial changes, so that people who may have important input in such a distributed project are able to respond.
-
what belongs in node:
- opinions vary, but I find the following helpful:
- if node itself needs it (due to historic reasons), then it belongs in node
- that is to say, url is there because of http, freelist is there because of http, et al
- also, things that cannot be done outside of core, or only with significant pain (example: async-wrap)
-
CI testing:
- lives here: https://ci.nodejs.org/
- not automatically run - some of the platforms we test do not have full sandboxing support so we need to ensure what we run on it isn't potentially malicious
- make sure to log in – we use github authentication so it should be seamless
- go to "node-test-pull-request" and "Build with parameters"
- fill in the pull request number without the
#
, and check the verification that you have reviewed the code for potential malice- The other options shouldn't need to be adjusted in most cases.
- link to the CI run in the PR by commenting "CI: "
- lives here: https://ci.nodejs.org/
process for getting code in:
-
the collaborator guide is a great resource: https://github.com/nodejs/node/blob/master/COLLABORATOR_GUIDE.md#technical-howto
-
no one (including TSC or CTC members) pushes directly to master without review
- an exception is made for release commits only
-
one "LGTM" is usually sufficient, except for semver-major changes
- the more the better
- semver-major (breaking) changes must be reviewed in some form by the CTC
-
be sure to wait before merging non-trivial changes
- 48 hours for non-trivial changes, and 72 hours on weekends.
-
make sure to run the PR through CI before merging! (Except for documentation PRs)
-
once code is ready to go in:
- See "Landing PRs" below
-
what if something goes wrong?
- ping a CTC member
#node-dev
on freenode- force-pushing to fix things after is allowed for ~10 minutes, be sure to notify people in IRC if you need to do this, but avoid it
- Info on PRs that don't like to apply found under "If
git am
fails".
Landing PRs
- Please never use GitHub's green "Merge Pull Request" button.
- If you do, please force-push removing the merge.
Update your master
branch (or whichever branch you are landing on, almost always master
)
Landing a PR
- if it all looks good,
curl -L 'url-of-pr.patch' | git am
git rebase -i upstream/master
- squash into logical commits if necessary
./configure && make -j8 test
(-j8
builds node in parallel with 8 threads. adjust to the number of cores (or processor-level threads) your processor has (or slightly more) for best results.)- Amend the commit description
- commits should follow
subsystem[,subsystem]: small description\n\nbig description\n\n<metadata>
- first line 50 columns, all others 72
- add metadata:
Fixes: <full-issue-url>
Reviewed-By: human <email>
- Easiest to use
git log
then do a search - (
/Name
+enter
(+n
as much as you need to) in vim)
- Easiest to use
PR-URL: <full-pr-url>
- commits should follow
git push upstream master
- close the original PR with "Landed in
<commit hash>
".
- close the original PR with "Landed in
exercise: make PRs adding yourselves to the README.
- Example:
7b09aade84
- to see full URL:
git log 7b09aade8468e1c930f36b9c81e6ac2ed5bc8732 -1
- to see full URL:
- Collaborators in alphabetical order by username
- Label your pull request with the
doc
subsystem label - If you would like to run CI on your PR, feel free to
- Make sure to added the
PR-URL: <full-pr-url>
!
final notes:
- don't worry about making mistakes: everybody makes them, there's a lot to internalize and that takes time (and we recognize that!)
- very few (no?) mistakes are unrecoverable
- the existing node committers trust you and are grateful for your help!
- other repos: