This document provides information and guidelines for working with fixup commits:
This blog post is also a good resource on the subject.
At their core, fixup commits are just regular commits with a special commit message: The first line of their commit message starts with "fixup! " (notice the space after "!") followed by the first line of the commit message of an earlier commit (it doesn't have to be the immediately preceding one).
The purpose of a fixup commit is to modify an earlier commit.
I.e. it allows adding more changes in a new commit, but "marking" them as belonging to an earlier
commit.
Git provides tools to make it easy to squash fixup commits into the original commit at a later
time (see below for details).
For example, let's assume you have added the following commits to your branch:
feat: first commit
fix: second commit
If you want to add more changes to the first commit, you can create a new commit with the commit
message:
fixup! feat: first commit:
feat: first commit
fix: second commit
fixup! feat: first commit
So, when are fixup commits useful?
During the life of a Pull Request, a reviewer might request changes. The Pull Request author can make the requested changes and submit them for another review. Normally, these changes should be part of one of the original commits of the Pull Request. However, amending an existing commit with the changes makes it difficult for the reviewer to know exactly what has changed since the last time they reviewed the Pull Request.
Here is where fixup commits come in handy. By addressing review feedback in fixup commits, you make it very straight forward for the reviewer to see what are the new changes that need to be reviewed and verify that their earlier feedback has been addressed. This can save a lot of effort, especially on larger Pull Requests (where having to re-review all the changes is pretty wasteful).
When the time comes to merge the Pull Request into the repository, the merge script knows how to automatically squash fixup commits with the corresponding regular commits.
As mentioned above, the only thing that differentiates a fixup commit from a
regular commit is the commit message.
You can create a fixup commit by specifying an appropriate commit message (
i.e. fixup! <original-commit-message-subject>).
In addition, the git command-line tool provides an easy way to create a fixup commit
via git commit --fixup:
# Create a fixup commit to fix up the last commit on the branch:
git commit --fixup HEAD ...
# Create a fixup commit to fix up commit with SHA <COMMIT_SHA>:
git commit --fixup <COMMIT_SHA> ...As mentioned above, the merge script will automatically squash fixup commits. However, sometimes you might want to manually squash a fixup commit.
The easiest way to re-order and squash any commit is
via rebasing interactively. You move a
commit right after the one you want to squash it into in the rebase TODO list and change the
corresponding action from pick to fixup.
Git can do all these automatically for you if you pass the --autosquash option to git rebase.
See the git docs
for more details.
You may like to consider some optional configurations:
By default, git will not automatically squash fixup commits when interactively rebasing.
If you prefer to not have to pass the --autosquash option every time, you can change the default
behavior by setting the rebase.autoSquash git config option to true.
See
the git docs
for more details.
If you have rebase.autoSquash set to true, you can pass the --no-autosquash option
to git rebase to override and disable this setting.