blob: 0e323d54779a7c680ee8471330c85609f4644d43 [file] [log] [blame]
Reviewing Patches in the Git Project
====================================
Introduction
------------
The Git development community is a widely distributed, diverse, ever-changing
group of individuals. Asynchronous communication via the Git mailing list poses
unique challenges when reviewing or discussing patches. This document contains
some guiding principles and helpful tools you can use to make your reviews both
more efficient for yourself and more effective for other contributors.
Note that none of the recommendations here are binding or in any way a
requirement of participation in the Git community. They are provided as a
resource to supplement your skills as a contributor.
Principles
----------
Selecting patch(es) to review
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
If you are looking for a patch series in need of review, start by checking
latest "What's cooking in git.git" email
(https://lore.kernel.org/git/xmqqilm1yp3m.fsf@gitster.g/[example]). The "What's
cooking" emails & replies can be found using the query `s:"What's cooking"` on
the https://lore.kernel.org/git/[`lore.kernel.org` mailing list archive];
alternatively, you can find the contents of the "What's cooking" email tracked
in `whats-cooking.txt` on the `todo` branch of Git. Topics tagged with "Needs
review" and those in the "[New Topics]" section are typically those that would
benefit the most from additional review.
Patches can also be searched manually in the mailing list archive using a query
like `s:"PATCH" -s:"Re:"`. You can browse these results for topics relevant to
your expertise or interest.
If you've already contributed to Git, you may also be CC'd in another
contributor's patch series. These are topics where the author feels that your
attention is warranted. This may be because their patch changes something you
wrote previously (making you a good judge of whether the new approach does or
doesn't work), or because you have the expertise to provide an exceptionally
helpful review. There is no requirement to review these patches but, in the
spirit of open source collaboration, you should strongly consider doing so.
Reviewing patches
~~~~~~~~~~~~~~~~~
While every contributor takes their own approach to reviewing patches, here are
some general pieces of advice to make your reviews as clear and helpful as
possible. The advice is broken into two rough categories: high-level reviewing
guidance, and concrete tips for interacting with patches on the mailing list.
==== High-level guidance
- Remember to review the content of commit messages for correctness and clarity,
in addition to the code change in the patch's diff. The commit message of a
patch should accurately and fully explain the code change being made in the
diff.
- Reviewing test coverage is an important - but easy to overlook - component of
reviews. A patch's changes may be covered by existing tests, or new tests may
be introduced to exercise new behavior. Checking out a patch or series locally
allows you to manually mutate lines of new & existing tests to verify expected
pass/fail behavior. You can use this information to verify proper coverage or
to suggest additional tests the author could add.
- When providing a recommendation, be as clear as possible about whether you
consider it "blocking" (the code would be broken or otherwise made worse if an
issue isn't fixed) or "non-blocking" (the patch could be made better by taking
the recommendation, but acceptance of the series does not require it).
Non-blocking recommendations can be particularly ambiguous when they are
related to - but outside the scope of - a series ("nice-to-have"s), or when
they represent only stylistic differences between the author and reviewer.
- When commenting on an issue, try to include suggestions for how the author
could fix it. This not only helps the author to understand and fix the issue,
it also deepens and improves your understanding of the topic.
- Reviews do not need to exclusively point out problems. Feel free to "think out
loud" in your review: describe how you read & understood a complex section of
a patch, ask a question about something that confused you, point out something
you found exceptionally well-written, etc. In particular, uplifting feedback
goes a long way towards encouraging contributors to participate more actively
in the Git community.
==== Performing your review
- Provide your review comments per-patch in a plaintext "Reply-All" email to the
relevant patch. Comments should be made inline, immediately below the relevant
section(s).
- You may find that the limited context provided in the patch diff is sometimes
insufficient for a thorough review. In such cases, you can review patches in
your local tree by either applying patches with linkgit:git-am[1] or checking
out the associated branch from https://github.com/gitster/git once the series
is tracked there.
- Large, complicated patch diffs are sometimes unavoidable, such as when they
refactor existing code. If you find such a patch difficult to parse, try
reviewing the diff produced with the `--color-moved` and/or
`--ignore-space-change` options.
- If a patch is long, you are encouraged to delete parts of it that are
unrelated to your review from the email reply. Make sure to leave enough
context for readers to understand your comments!
- If you cannot complete a full review of a series all at once, consider letting
the author know (on- or off-list) if/when you plan to review the rest of the
series.
Completing a review
~~~~~~~~~~~~~~~~~~~
Once each patch of a series is reviewed, the author (and/or other contributors)
may discuss the review(s). This may result in no changes being applied, or the
author will send a new version of their patch(es).
After a series is rerolled in response to your or others' review, make sure to
re-review the updates. If you are happy with the state of the patch series,
explicitly indicate your approval (typically with a reply to the latest
version's cover letter). Optionally, you can let the author know that they can
add a "Reviewed-by: <you>" trailer if they resubmit the reviewed patch verbatim
in a later iteration of the series.
Finally, subsequent "What's cooking" emails may explicitly ask whether a
reviewed topic is ready for merging to the `next` branch (typically phrased
"Will merge to \'next\'?"). You can help the maintainer and author by responding
with a short description of the state of your (and others', if applicable)
review, including the links to the relevant thread(s).
Terminology
-----------
nit: ::
Denotes a small issue that should be fixed, such as a typographical error
or mis-alignment of conditions in an `if()` statement.
aside: ::
optional: ::
non-blocking: ::
Indicates to the reader that the following comment should not block the
acceptance of the patch or series. These are typically recommendations
related to code organization & style, or musings about topics related to
the patch in question, but beyond its scope.
s/<before>/<after>/::
Shorthand for "you wrote <before>, but I think you meant <after>," usually
for misspellings or other typographical errors. The syntax is a reference
to "substitute" command commonly found in Unix tools such as `ed`, `sed`,
`vim`, and `perl`.
cover letter::
The "Patch 0" of a multi-patch series. This email describes the
high-level intent and structure of the patch series to readers on the
Git mailing list. It is also where the changelog notes and range-diff of
subsequent versions are provided by the author.
+
On single-patch submissions, cover letter content is typically not sent as a
separate email. Instead, it is inserted between the end of the patch's commit
message (after the `---`) and the beginning of the diff.
#leftoverbits::
Used by either an author or a reviewer to describe features or suggested
changes that are out-of-scope of a given patch or series, but are relevant
to the topic for the sake of discussion.
See Also
--------
link:MyFirstContribution.html[MyFirstContribution]