Start from a clean checkout at main
.
Before running through the release it's good to run the build and the tests locally, and make sure CI is passing. You can also test-drive the commit in an existing Bazel workspace to sanity check functionality.
- Determine the next semantic version number.
- Update CHANGELOG.md: replace the
v0-0-0
and0.0.0
withX.Y.0
. - Replace
VERSION_NEXT_*
strings withX.Y.0
. - Send these changes for review and get them merged.
- Create a branch for the new release, named
release/X.Y
git branch --no-track release/X.Y upstream/main && git push upstream release/X.Y
- Create a tag and push:
NOTE: Pushing the tag will trigger release automation.
git tag X.Y.0 upstream/release/X.Y && git push upstream --tags
- Release automation will create a GitHub release and BCR pull request.
rules_python uses semantic version, so releases with API changes and new features bump the minor, and those with only bug fixes and other minor changes bump the patch digit.
To find if there were any features added or incompatible changes made, review
CHANGELOG.md and the commit history. This can be done using
github by going to the url:
https://github.com/bazelbuild/rules_python/compare/<VERSION>...main
.
If a patch release from head would contain changes that aren't appropriate for a patch release, then the patch release needs to be based on the original release tag and the patch changes cherry-picked into it.
In this example, release 0.37.0
is being patched to create release 0.37.1
.
The fix being included is commit deadbeef
.
git checkout release/0.37
git cherry-pick -x deadbeef
- Fix merge conflicts, if any.
git cherry-pick --continue
(if applicable)git push upstream
If multiple commits need to be applied, repeat the git cherry-pick
step for
each.
Once the release branch is in the desired state, use git tag
to tag it, as
done with a release from head. Release automation will do the rest.
- Announce the release in the #python channel in the Bazel slack (bazelbuild.slack.com).
Part of the release process uploads packages to PyPI as the user rules-python
.
This account is managed by Google; contact [email protected] if
something needs to be done with the PyPI account.