Release process

Tagged:

This page aims to document our release process. It's a multi-step process:

For the specifics of release naming conventions and the cycle, see the branch naming convention.

1. Make sure Jenkins is all green

Look into Jenkins to see if all tasks have been performed without errors since the last commit. If there is an error, fix it before the release.

2. Running the release.sh script

Each time we make a new release, we run a script called release.sh in provision.

This script should only be used by the core dev team when doing an official release. If you are not one of those people, you probably shouldn't be running this.

This script does all the 'hard' work in that it doesn't forget all the very many places to edit version numbers etc of relevant documentation and other scripts. This includes install.sh.txt and upgrade.sh.txt.

Paraphrasing from the script itself:

The following operations will be done:
1. change the makefile to download tarball
2. change the upgrade.sh.txt version
3. display the resulting diff
4. commit those changes to git
5. lay down the tag (prompting you for a changelog)
6. revert the commit
7. (optionally) push those changes

The operation can be aborted before step 4 and 7. Don't forget that as long as changes are not pushed upstream, this can all be reverted (see git-reset(1) and git-revert(1) ).

So in short, this sums up as:

cd provision
sh release.sh 6.x-1.2
cd ../hostmaster
git tag 6.x-1.2
git push --tags

3. Test the manual install in Jenkins

Before making a full release, test the release in Jenkins. To do so, create a new build based on the 6.x-1.4 install build. Change the name of your new job to match the release version, and change the default value of the AEGIR_VERSION and DRUSH_VERSION to match what is required for this release.

Start a new build using 'git' as the AEGIR_FETCH_MODE.

If the build fails, delete the remote tags, fix the bugs and start again.

4. Generating the release notes

We build complete release notes for every release. Those are made up of a summary of the release, an outline of key changes, of known issues, install and upgrade instructions and a full list of bugfixes and new features.

We have a handy bookmarklet that takes the list of issues and build a release-notes-ready list:

javascript:(function(){x%20=%20open().document;$(".views-field-title a").each(%20function()%20{%20x.write('* [' + $(this).attr('href').replace(/\/node\//, '#') + ': ' + $(this).text()+'](http://drupal.org'+$(this).attr('href')+ ')<br%20/>')%20})})()

The developers then proceed to format/edit the list of fixes as well as list other significant information/changes for this release. These notes end up becoming the Release Notes for the release.

5. Creating release nodes on Drupal.org

Once the tags are pushed and release notes published, we create a release node with an excerpt of (and a link to) the release notes so that tarballs are created and issue queue versions updated.

This needs to be done in the hostmaster and provision projects on Drupal.org.

6. Build the release in Jenkins again

At this point, it's possible that the tarballs on Drupal.org were not created properly. We want to test the real procedure, so run a your build again, but choose 'package' as the AEGIR_FETCH_MODE.

7. Build the Debian packages

Build the package and upload to http://debian.aegirproject.org/, as explained in those detailed instructions. Also test the aegir-build-official job.

8. (eventually) Build the release in Jenkins yet again

... once we have such a script (ref.: http://drupal.org/node/1128614)

9. Publish the release notes widely

Once all this is done and the tarballs are generated, the release notes are published in:

Optionally, blog posts on koumbit.org, greenbeedigital.com.au, and elsewhere may go into further detail about significant changes, screencasts etc.