Skip to content

How to Publish a New Version

The code to be published should be in the main branch. Make sure that all the PRs to go in the release are merged, and decide on the release tag. Should it be a release candidate or the final tag, and should it be a major, minor or patch release, per semver rules.

Once ready to do a release, create a local branch that includes the following updates:

  1. Create a PR branch from an updated main branch.

  2. Update the CHANGELOG.md to add the new release. Only create a new section when working on the first release candidate for a new release. When transitioning from one release candidate to the next, or to an official release, just update the title and date of the change log section.

  3. Include details of the merged PRs included in this release. General process to follow:

  4. Gather the set of PRs since the last release and put them into a list. A good tool to use for this is the github-changelog-generator. Steps:

  5. Create a read only GitHub token for your account on this page: https://github.com/settings/tokens with a scope of repo / public_repo.
  6. Use a command like the following, adjusting the tag parameters as appropriate. docker run -it --rm -v "$(pwd)":/usr/local/src/your-app githubchangeloggenerator/github-changelog-generator --user hyperledger --project aries-cloudagent-python --output 0.11.0rc2.md --since-tag 0.10.4 --future-release 0.11.1rc2 --release-branch main --token <your-token>
  7. In the generated file, use only the PR list -- we don't include the list of closed issues in the Change Log.

In some cases, the approach above fails because of too many API calls. An alternate approach to getting the list of PRs in the right format is to use OpenAI ChatGPT.

Prepare the following ChatGPT request. Don't hit enter yet--you have to add the data.

Generate from this the github pull request number, the github id of the author and the title of the pull request in a tab-delimited list

Get a list of the merged PRs since the last release by displaying the PR list in the GitHub UI, highlighting/copying the PRs and pasting them below the ChatGPT request, one page after another. Hit <Enter>, let the AI magic work, and you should have a list of the PRs in a nice table with a Copy link that you should click.

Once you have that, open this Google Sheet and highlight the A1 cell and paste in the ChatGPT data. A formula in column E will have the properly formatted changelog entries. Double check the list with the GitHub UI to make sure that ChatGPT isn't messing with you and you have the needed data.

If using ChatGPT doesn't appeal to you, try this scary sed/command line approach:

  • Put the following commands into a file called changelog.sed
/Approved/d
/updated /d
/^$/d
/^ [0-9]/d
s/was merged.*//
/^@/d
s# by \(.*\) # [\1](https://github.com/\1)#
s/^ //
s#  \#\([0-9]*\)# [\#\1](https://github.com/hyperledger/aries-cloudagent-python/pull/\1) #
s/  / /g
/^Version/d
/tasks done/d
s/^/- /
  • Navigate in your browser to the paged list of PRs merged since the last release (using in the GitHub UI a filter such as is:pr is:merged sort:updated merged:>2022-04-07) and for each page, highlight, and copy the text of only the list of PRs on the page to use in the following step.
  • For each page, run the command sed -e :a -e '$!N;s/\n#/ #/;ta' -e 'P;D' <<EOF | sed -f changelog.sed, paste in the copied text and then type EOF. Redirect the output to a file, appending each page of output to the file.
  • The first sed command in the pipeline merges the PR title and PR number plus author lines onto a single line. The commands in the changelog.sed file just clean up the data, removing unwanted lines, etc.
  • At the end of that process, you should have a list of all of the PRs in a form you can use in the CHANGELOG.md file.
  • To verify you have right number of PRs, you can do a wc of the file and there should be one line per PR. You should scan the file as well, looking for anomalies, such as missing \s before # characters. It's a pretty ugly process.
  • Using a curl command and the GitHub API is probably a much better and more robust way to do this, but this was quick and dirty...

Once you have the list of PRs:

  • Organize the list into suitable categories, update (if necessary) the PR description and add notes to clarify the changes. See previous release entries to understand the style -- a format that should help developers.
  • Add a narrative about the release above the PR that highlights what has gone into the release.

  • Check to see if there are any other PRs that should be included in the release.

  • Update the ReadTheDocs in the /docs folder by following the instructions in the ./UpdateRTD.md file. That will likely add a number of new and modified files to the PR. Eliminate all of the errors in the generation process, either by mocking external dependencies or by fixing ACA-Py code. If necessary, create an issue with the errors and assign it to the appropriate developer. Experience has demonstrated to use that documentation generation errors should be fixed in the code.

  • Search across the repository for the previous version number and update it everywhere that makes sense. The CHANGELOG.md entry for the previous release is a likely exception, and the pyproject.toml in the root MUST be updated. You can skip (although it won't hurt) to update the files in the open-api folder as they will be automagically updated by the next step in publishing. The incremented version number MUST adhere to the Semantic Versioning Specification based on the changes since the last published release. For Release Candidates, the form of the tag is "0.11.0rc2". As of release 0.11.0 we have dropped the previously used - in the release candidate version string to better follow the semver rules.

  • Regenerate openapi.json and swagger.json by running ../scripts/generate-open-api-spec from within the aries_cloudagent folder.

Command: cd aries_cloudagent;../scripts/generate-open-api-spec;cd ..

  1. Double check all of these steps above, and then submit a PR from the branch. Add this new PR to CHANGELOG.md so that all the PRs are included. If there are still further changes to be merged, mark the PR as "Draft", repeat ALL of the steps again, and then mark this PR as ready and then wait until it is merged. It's embarrassing when you have to do a whole new release just because you missed something silly...I know!

  2. Immediately after it is merged, create a new GitHub tag representing the version. The tag name and title of the release should be the same as the version in pyproject.toml. Use the "Generate Release Notes" capability to get a sequential listing of the PRs in the release, to complement the manually curated Changelog. Verify on PyPi that the version is published.

  3. New images for the release are automatically published by the GitHubAction Workflows: publish.yml and publish-indy.yml. The actions are triggered when a release is tagged, so no manual action is needed. The images are published in the Hyperledger Package Repository under aries-cloudagent-python and a link to the packages added to the repositories main page (under "Packages").

Additional information about the container image publication process can be found in the document Container Images and Github Actions.

  1. Update the ACA-Py Read The Docs site by building the new "latest" (main branch) and activating and building the new release. Appropriate permissions are required to publish the new documentation version.

  2. Update the https://aca-py.org website with the latest documentation by creating a PR and tag of the latest documentation from this site. Details are provided in the aries-acapy-docs repository.