Skip to content

Latest commit

 

History

History
437 lines (306 loc) · 19.3 KB

RELEASE.md

File metadata and controls

437 lines (306 loc) · 19.3 KB

Release Process

Note

Throughout this document are references to the version number as YY.M.[$patch_number], this should be replaced with the correct version number. Do not prefix the version with a lowercase v.

1. Open the release issue and cut a release branch. (do this ~1 week prior to release)

Note

The new release branch should adhere to the naming convention of YY.M.x (make sure to put the .x at the end!). In the case of patch/hotfix releases, however, do NOT cut a new release branch; instead, use the previously-cut release branch with the appropriate YY.M.x version numbers.

Use the issue template below to create the release issue. After creating the release issue, pin it for easy access.

Release Template

### Summary

Placeholder for `conda-build YY.M.x` release.

| Pilot | <pilot> |
|---|---|
| Co-pilot | <copilot> |

### Tasks

[milestone]: https://github.com/conda/conda-build/milestone/<milestone>
[process]: https://github.com/conda/conda-build/blob/main/RELEASE.md
[releases]: https://github.com/conda/conda-build/releases
[main]: https://github.com/AnacondaRecipes/conda-build-feedstock
[conda-forge]: https://github.com/conda-forge/conda-build-feedstock
[ReadTheDocs]: https://readthedocs.com/projects/continuumio-conda-build/

<details open>  <!-- feel free to remove the open attribute once this section is completed -->
<summary><h4>The week before release week</h4></summary>

- [ ] Create release branch (named `YY.M.x`)
- [ ] Ensure release candidates are being successfully built (see `conda-canary/label/rc-conda-build-YY.M.x`)
- [ ] [Complete outstanding PRs][milestone]
- [ ] Test release candidates
    <!-- indicate here who has signed off on testing -->

</details>

<details open>  <!-- feel free to remove the open attribute once this section is completed -->
<summary><h4>Release week</h4></summary>

- [ ] Create release PR (see [release process][process])
- [ ] [Publish release][releases]
- [ ] Merge `YY.M.x` back into `main`
- [ ] Activate the `YY.M.x` branch on [ReadTheDocs][ReadTheDocs]
- [ ] Feedstocks
    - [ ] Bump version & update dependencies/tests in [Anaconda, Inc.'s feedstock][main]
    - [ ] Bump version & update dependencies/tests in [conda-forge feedstock][conda-forge]
    <!-- link any other feedstock PRs here -->
- [ ] Hand off to the Anaconda packaging team
- [ ] Announce release
    - Blog Post (optional)
        - [ ] conda.org (link to pull request)
    - Long form
        - [ ] Create release [announcement draft](https://github.com/conda/communications)
        - [ ] [Discourse](https://conda.discourse.group/)
        - [ ] [Matrix (conda/conda)](https://matrix.to/#/#conda_conda:gitter.im) (this auto posts from Discourse)
    - Summary
        - [ ] [Twitter](https://twitter.com/condaproject)

</details>

If a patch release is necessary, reopen the original release issue and append the following template to the release issue summary.

Patch Release Template

<details open>  <!-- feel free to remove the open attribute once this section is completed -->
<summary><h4>Patch YY.M.[$patch_number]</h4></summary>

- [ ] <!-- list issues & PRs that need to be resolved here -->
- [ ] Create release PR (see [release process][process])
- [ ] [Publish release][releases]
- [ ] Merge `YY.M.x` back into `main`
- [ ] Feedstocks
    - [ ] Bump version & update dependencies/tests in [Anaconda, Inc.'s feedstock][main]
    - [ ] Bump version & update dependencies/tests in [conda-forge feedstock][conda-forge]
- [ ] Hand off to the Anaconda packaging team

</details>

Note

The epic template is perfect for this; remember to remove the epic label.

2. Alert various parties of the upcoming release. (do this ~1 week prior to release)

Let various interested parties know about the upcoming release; at minimum, conda-forge maintainers should be informed. For major features, a blog post describing the new features should be prepared and posted once the release is completed (see the announcements section of the release issue).

3. Manually test canary build(s).

Canary Builds for Manual Testing

Once the release PRs are filed, successful canary builds will be available on https://anaconda.org/conda-canary/conda/files?channel=rc-conda-build-YY.M.x for manual testing.

Note

You do not need to apply the build::review label for release PRs; every commit to the release branch builds and uploads canary builds to the respective rc- label.

4. Ensure rever.xsh and news/TEMPLATE are up to date.

These are synced from conda/infrastructure.

5. Run rever. (ideally done on the Monday of release week)

Currently, there are only 2 activities we use rever for, (1) aggregating the authors and (2) updating the changelog. Aggregating the authors can be an error-prone process and also suffers from builtin race conditions (i.e., to generate an updated .authors.yml we need an updated .mailmap but to have an updated .mailmap we need an updated .authors.yml). This is why the following steps are very heavy-handed (and potentially repetitive) in running rever commands, undoing commits, squashing/reordering commits, etc.

  1. Install rever and activate the environment:

    $ conda create -n rever conda-forge::rever
    $ conda activate rever
    (rever) $
  2. Clone and cd into the repository if you haven't done so already:

    (rever) $ git clone [email protected]:/conda-build.git
    (rever) $ cd conda
  3. Fetch the latest changes from the remote and checkout the release branch created a week ago:

    (rever) $ git fetch upstream
    (rever) $ git checkout YY.M.x
  4. Create a versioned branch, this is where rever will make its changes:

    (rever) $ git checkout -b changelog-YY.M.[$patch_number]
  5. Run rever --activities authors <VERSION>:

    Note: Include --force when re-running any rever commands for the same <VERSION>, otherwise, rever will skip the activity and no changes will be made (i.e., rever remembers if an activity has been run for a given version).

    (rever) $ rever --activities authors --force <VERSION>
    • If rever finds that any of the authors are not correctly represented in .authors.yml it will produce an error. If the author that the error pertains to is:

      • a new contributor: the snippet suggested by rever should be added to the .authors.yml file.
      • an existing contributor, a result of using a new name/email combo: find the existing author in .authors.yml and add the new name/email combo to that author's aliases and alterative_emails.
    • Once you have successfully run rever --activities authors with no errors, review the commit made by rever. This commit will contain updates to one or more of the author files (.authors.yml, .mailmap, and AUTHORS.md). Due to the race condition between .authors.yml and .mailmap, we want to extract changes made to any of the following keys in .authors.yml and commit them separately from the other changes in the rever commit:

      • name
      • email
      • github
      • aliases
      • alternate_emails

      Other keys (e.g., num_commits and first_commit) do not need to be included in this separate commit as they will be overwritten by rever.

    • Here's a sample run where we undo the commit made by rever in order to commit the changes to .authors.yml separately:

      (rever) $ rever --activities authors --force YY.M.[$patch_number]
      
      # changes were made to .authors.yml as per the prior bullet
      (rever) $ git diff --name-only HEAD HEAD~1
      .authors.yml
      .mailmap
      AUTHORS.md
      
      # undo commit
      (rever) $ git reset --soft HEAD~1
      
      # undo changes made to everything except .authors.yml
      (rever) $ git restore --staged --worktree .mailmap AUTHORS.md
    • Commit these changes to .authors.yml:

      (rever) $ git add .
      (rever) $ git commit -m "Update .authors.yml"
    • Rerun rever --activities authors --force <VERSION> and finally check that your .mailmap is correct by running:

      git shortlog -se

      Compare this list with AUTHORS.md. If they have any discrepancies, additional modifications to .authors.yml is needed, so repeat the above steps as needed.

    • Once you are pleased with how the author's file looks, we want to undo the rever commit and commit the .mailmap changes separately:

      # undo commit (but preserve changes)
      (rever) $ git reset --soft HEAD~1
      
      # undo changes made to everything except .mailmap
      (rever) $ git restore --staged --worktree .authors.yml AUTHORS.md
    • Commit these changes to .mailmap:

      (rever) $ git add .
      (rever) $ git commit -m "Update .mailmap"
    • Continue repeating the above processes until the .authors.yml and .mailmap are corrected to your liking. After completing this, you will have at most two commits on your release branch:

      (rever) $ git cherry -v <release branch>
      + 86957814cf235879498ed7806029b8ff5f400034 Update .authors.yml
      + 3ec7491f2f58494a62f1491987d66f499f8113ad Update .mailmap
  6. Review news snippets (ensure they are all using the correct Markdown format, not reStructuredText) and add additional snippets for undocumented PRs/changes as necessary.

    Note: We've found it useful to name news snippets with the following format: <PR #>-<DESCRIPTIVE SLUG>.

    We've also found that we like to include the PR #s inline with the text itself, e.g.:

    ## Enhancements
    
    * Add `win-arm64` as a known platform (subdir). (#11778)
    • You can utilize GitHub's compare view to review what changes are to be included in this release. Make sure you compare the current release branch against the previous one (e.g., 24.5.x would be compared against 24.3.x)

    • Add a new news snippet for any PRs of importance that are missing.

    • Commit these changes to news snippets:

      (rever) $ git add .
      (rever) $ git commit -m "Update news"
    • After completing this, you will have at most three commits on your release branch:

      (rever) $ git cherry -v <release branch>
      + 86957814cf235879498ed7806029b8ff5f400034 Update .authors.yml
      + 3ec7491f2f58494a62f1491987d66f499f8113ad Update .mailmap
      + 432a9e1b41a3dec8f95a7556632f9a93fdf029fd Update news
  7. Run rever --activities changelog:

    Note: This has previously been a notoriously fickle step (likely due to incorrect regex patterns in the rever.xsh config file and missing github keys in .authors.yml) so beware of potential hiccups. If this fails, it's highly likely to be an innocent issue.

    (rever) $ rever --activities changelog --force <VERSION>
    • Any necessary modifications to .authors.yml, .mailmap, or the news snippets themselves should be amended to the previous commits.

    • Once you have successfully run rever --activities changelog with no errors simply revert the last commit (see the next step for why):

      # undo commit (and discard changes)
      (rever) $ git reset --hard HEAD~1
    • After completing this, you will have at most three commits on your release branch:

      (rever) $ git cherry -v <release branch>
      + 86957814cf235879498ed7806029b8ff5f400034 Update .authors.yml
      + 3ec7491f2f58494a62f1491987d66f499f8113ad Update .mailmap
      + 432a9e1b41a3dec8f95a7556632f9a93fdf029fd Update news
  8. Now that we have successfully run the activities separately, we wish to run both together. This will ensure that the contributor list, a side-effect of the authors activity, is included in the changelog activity.

    (rever) $ rever --force <VERSION>
    • After completing this, you will have at most five commits on your release branch:

      (rever) $ git cherry -v <release branch>
      + 86957814cf235879498ed7806029b8ff5f400034 Update .authors.yml
      + 3ec7491f2f58494a62f1491987d66f499f8113ad Update .mailmap
      + 432a9e1b41a3dec8f95a7556632f9a93fdf029fd Update news
      + a5c0db938893d2c12cab12a1f7eb3e646ed80373 Update authorship for YY.M.[$patch_number]
      + 5e95169d0df4bcdc2da9a6ba4a2561d90e49f75d Update CHANGELOG for YY.M.[$patch_number]
  9. Since rever does not include stats on first-time contributors, we will need to add this manually.

    • Use GitHub's auto-generated release notes to get a list of all new contributors (and their first PR) and manually merge this list with the contributor list in CHANGELOG.md. See GitHub docs for how to auto-generate the release notes.

    • Commit these final changes:

      (rever) $ git add .
      (rever) $ git commit -m "Add first-time contributions"
    • After completing this, you will have at most six commits on your release branch:

      (rever) $ git cherry -v <release branch>
      + 86957814cf235879498ed7806029b8ff5f400034 Update .authors.yml
      + 3ec7491f2f58494a62f1491987d66f499f8113ad Update .mailmap
      + 432a9e1b41a3dec8f95a7556632f9a93fdf029fd Update news
      + a5c0db938893d2c12cab12a1f7eb3e646ed80373 Update authorship for YY.M.[$patch_number]
      + 5e95169d0df4bcdc2da9a6ba4a2561d90e49f75d Update CHANGELOG for YY.M.[$patch_number]
      + 93fdf029fd4cf235872c12cab12a1f7e8f95a755 Add first-time contributions
  10. Push this versioned branch.

    (rever) $ git push -u upstream changelog-YY.M.[$patch_number]
  11. Open the Release PR targing the YY.M.x branch.

    GitHub PR Template
    ## Description
    
    ✂️ snip snip ✂️ the making of a new release.
    
    Xref #<RELEASE ISSUE>
  12. Update release issue to include a link to the release PR.

  13. Create the release and SAVE AS A DRAFT with the following values:

    Note: Only publish the release after the release PR is merged, until then always save as draft.

    Field Value
    Choose a tag YY.M.[$patch_number]
    Target YY.M.x
    Body copy/paste blurb from CHANGELOG.md

6. Wait for review and approval of release PR.

7. Merge release PR and publish release.

To publish the release, go to the project's release page (e.g., https://github.com/conda/conda/releases) and add the release notes from CHANGELOG.md to the draft release you created earlier. Then publish the release.

Note

Release notes can be drafted and saved ahead of time.

8. Merge/cherry pick the release branch over to the main branch.

Internal process
  1. From the main "< > Code" page of the repository, select the drop down menu next to the main branch button and then select "View all branches" at the very bottom.

  2. Find the applicable YY.M.x branch and click the "New pull request" button.

  3. "Base" should point to main while "Compare" should point to YY.M.x.

  4. Ensure that all of the commits being pulled in look accurate, then select "Create pull request".

[!NOTE] Make sure NOT to push the "Update Branch" button. If there are merge conflicts, create a temporary "connector branch" dedicated to fixing merge conflicts separately from the YY.M.x and main branches.

  1. Review and merge the pull request the same as any code change pull request.

[!NOTE] The commits from the release branch need to be retained in order to be able to compare individual commits; in other words, a "merge commit" is required when merging the resulting pull request vs. a "squash merge". Protected branches will require permissions to be temporarily relaxed in order to enable this action.

9. Open PRs to bump Anaconda Recipes and conda-forge feedstocks to use YY.M.[$patch_number].

Note

Conda-forge's PRs will be auto-created via the regro-cf-autotick-bot. Follow the instructions below if any changes need to be made to the recipe that were not automatically added (these instructions are only necessary for anyone who is not a conda-forge feedstock maintainer, since maintainers can push changes directly to the autotick branch):

  • Create a new branch based off of autotick's branch (autotick's branches usually use the regro-cf-autotick-bot:XX.YY.[$patch_number]_[short hash] syntax)
  • Add any changes via commits to that new branch
  • Open a new PR and push it against the main branch

Make sure to include a comment on the original autotick-bot PR that a new pull request has been created, in order to avoid duplicating work! regro-cf-autotick-bot will close the auto-created PR once the new PR is merged.

For more information about this process, please read the "Pushing to regro-cf-autotick-bot branch" section of the conda-forge documentation.

10. Hand off to Anaconda's packaging team.

Note

This step should NOT be done past Thursday morning EST; please start the process on a Monday, Tuesday, or Wednesday instead in order to avoid any potential debugging sessions over evenings or weekends.

Internal process
  1. Open packaging request in #package_requests Slack channel, include links to the Release PR and feedstock PRs.

  2. Message packaging team/PM to let them know that a release has occurred and that you are the release manager.

11. Continue championing and shepherding.

Remember to make all relevant announcements and continue to update the release issue with the latest details as tasks are completed.