creative-commoners / cow

Release tool for SilverStripe

Cow

The ineptly named tool which may one day supercede the older build tools.

Install

Requires Docker.

You do not need to know how Docker works, but you need to have it installed and running.

For production purposes Cow should be run via the Docker-based scripts included in the ./docker/bin folder, rather than via direct invocation on your machine. That ensures use of consistent versions of all 3rd party libraries and tools and is the only supported installation option.

Tested on Linux and macOS, but should work on Windows too (with either Cygwin or WSL).

When it's installed, to run Cow launch ./docker/bin/run script. That will automatically download the latest released version of Cow docker image and launch it transparently in your terminal.

When publishing a release, use the ./docker/bin/release script. This is similar to run, but runs an SSH-Agent and performs some extra checks.

At the time of writing this readme (2020-02-03) GitHub Packages require you to have docker logged in with a GITHUB_TOKEN (scopes: repo and read:packages) even for accessing public images. This is a usability issue and they've said it's going to be fixed in the future (no login will be required for public images).

In short you need to issue a GITHUB_TOKEN and log in as follows (only once):

docker login -u USERNAME -p TOKEN docker.pkg.github.com

Then you need to configure composer of your Cow distribution by running

./docker/bin/composer config -g github-oauth.github.com TOKEN

For more details on these and other available scripts, see docker/README.md.

Adding Cow to your $PATH (optional)

If you want to run Cow globally, you can create a symlink to docker/bin/release in a folder that is included in your $PATH.

Here's an example (feel free to reuse any of your existing folders from yor $PATH)

mkdir -p ~/.local/bin;  # ensure existing ~/.local/bin
ln -s ./docker/bin/release ~/.local/bin/cow  # create symlink to the launcher and name it cow
echo 'export PATH=$PATH:~/.local/bin' >> ~/.bashrc  # add ~/.local/bin to the $PATH

After you reload your session (e.g. reopen the terminal) you may launch cow from anywhere on your system:

cd ./my-project;
cow moo

Native

To run Cow without Docker, reference the Dockerfile for system requirements.

Extra setup

• For translation steps you may need the Transifex API Key and ~/.transifexrc (run ./docker/bin/tx init to generate it).
• When publishing to GitHub make sure you have ssh-agent running and your SSH keys loaded (run ssh-add)

Commands

Cow is a collection of different tools (steps) grouped by top level commands. It is helpful to think about not only the commands available but each of the steps each command contains.

It is normally recommended that you run with -vvv verbose flag so that errors can be viewed during release.

For example, this is what I would run to release 3.1.14-rc1.

cow release 3.1.14-rc1 -vvv

And once I've checked that all is fine, and am 100% sure that this code is ready to go.

cow release:publish 3.1.14-rc1 -vvv

Release

cow release <version> <recipe> will perform the first part of the release tasks.

• <version> is mandatory and must be the exact tag name to release.
• <recipe> will allow you to release a recipe other than 'silverstripe/installer'

This command has these options:

• -vvv to ensure all underlying commands are echoed
• --directory <directory> to specify the folder to create or look for this project in. If you don't specify this, it will install to the path specified by ./release-<version> in the current directory.
• --repository <repository> will allow a custom composer package url to be specified. E.g. http://packages.cwp.govt.nz Note: If you specify the repository during setup it will be re-used for subsquent commands unless the .cow.repository file is deleted.
• --branching <type> will specify a branching strategy. This allows these options:
  • auto - Default option, will branch to the minor version (e.g. 1.1) unless doing a non-stable tag (e.g. rc1)
  • major - Branch all repos to the major version (e.g. 1) unless already on a more-specific minor version.
  • minor - Branch all repos to the minor semver branch (e.g. 1.1)
  • none - Release from the current branch and do no branching.
• --skip-tests to skip tests
• --skip-i18n to skip updating localisations

release actually has several sub-commands which can be run independently. These are as below:

• release:create creates the project folder
• release:plan Initiates release planning tool to preview release dependency versions
• release:branch Will (if needed) branch all modules
• release:translate Updates translations and commits this to source control
• release:test Run unit tests
• release:changelog Just generates the changelog and commits this to source control.

Publishing releases

cow release will only build the release itself. Once all of the above steps are complete, it is necessary to take the finished release and push it out to the open source community. A second major command cow release:publish is necessary to perform the final steps. The format for this command is:

cow release:publish <version>

This command has these options:

• -vvv to ensure all underlying commands are echoed
• --directory <directory> to specify the folder to look for the project created in the prior step. As with above, it will be guessed if omitted. You can run this command in the ./release-<version> directory and omit this option.

The release process, as with the initial cow release command, will actually be composed of several sub-commands, each of which could be run separately.

• release:tag Add annotated tags to each module and pushes

After the push step, release:publish will automatically wait for this version to be available in packagist.org before continuing.

Creating changelogs

cow release:changelog will create a changelog which is categorised into various sets of change types, e.g. enhancements, bug fixes, API changes and security fixes.

The changelog command takes the follow arguments and options:

• version The version you're releasing the project as
• recipe The recipe you're releasing
• --include-other-changes If provided, uncategorised commits will also be included in an "Other changes" section. Note that commits which match ChangelogItem::isIgnored() will still be excluded, e.g. merge commits.
• --changelog--use-legacy-format If provided, falls back to the old changelog format (used before Oct 2020)
• --changelog--audit-mode swaps changelog format to Audit Mode, which turns on include-upgrade-only flag and uses audit template for changelog logs generation (including every single change).

Pro-tip: Part of this command involves plan generation and/or confirmation, and you can provide the --skip-fetch-tags option to prevent Cow from re-fetching all tags from origin if you have already done this and only want to make a quick change.

Changelog Templates

You can specify a file to use as the template for generating fresh changelogs via the changelog-template configuration in .cow.json. This template can use Twig syntax to inject relevant information:

• {{ version }} will inject the version that the changelog is being generated for (e.g. 1.2.3)
• {{ logs }} will inject the commit logs with before/after delimiters, so they can be updated later without destroying any other changes to the contents.

Synchronising data to supported modules

Cow includes commands to help synchronise standardised data to all commercially supported modules:

• cow github:synclabels Pushes a centralised list of labels to all supported module GitHub repositories
• cow github:ratelimit Check your current GitHub API rate limiting status (sync commands can use this up quickly)

Note: All GitHub API commands require a GITHUB_ACCESS_TOKEN environment variable to be set before they can be used.

Labels

Centralised label configuration can be pushed out to all supported modules using the cow github:synclabels command.

This command takes an optional argument to specify which module(s) to update:

• modules Optionally sync to specific modules (comma delimited)

If the modules argument is not provided, the list of supported modules will be loaded from the supported-modules GitHub repository. You can then confirm the list, before you will be shown a list of the labels that will be sync'd and finally all labels will either be created or updated on the target repositories.

This command can max out your GitHub API rate limiting credits, so use it sparingly. If you exceed the limit you may need to go and make a coffee and come back in an hour (check current rate limits with cow github:ratelimit).

Metadata files

File templates for supported modules can be synchronised out to all supported modules using the module:sync:metadata command.

This command will pull the latest version from the supported-modules repository, write the contents to each repository then stage, commit and push directly to the default branch.

This command takes an optional argument to skip the clone/pull of each repository beforehand:

• --skip-update Optionally skip the clone/fetch/pull for each repository before running the sync

You will need git available in your system path, as well as write permission to push to each repository.

Schema

The cow schema file is in the root of this project.

You can run cow schema:validate to check the .cow.json configuration file in your project or module to ensure it matches against the Cow schema.