Forgejo Runner

WARNING: this is alpha release quality code and should not be considered secure enough to deploy in production.

A daemon that connects to a Forgejo instance and runs jobs for continous integration. The installation and usage instructions are part of the Forgejo documentation.

Hacking

The Forgejo runner depends on a fork of ACT and is a dependency of the setup-forgejo action. Together they provide a development environment with end to end testing. Each repository also has some unit testing that can be used to quickly detect the simplest mistakes such as a failure to compile or static code checking failures (vulnerability, lint, etc.).

Assuming the modifications to the Forgejo runner are pushed to a fork in a branch named wip-runner-change, a pull request will verify it compiles and the binary is sane (running forgejo-runner --version). It will not verify that it is able to properly run jobs when connected to a live Forgejo instance.

For end to end testing, a workflow will create a cascading-pr in setup-forgejo and wait for its success.

The runner can be released by merging the wip-runner-change branch and by pushing a new tag, for instance v10.2.3. For more information see the documentation that details this release process in the Forgejo infrastructure. Once published, the setup-forgejo action can be updated to default to this latest version knowing it already passed integration tests.

ACT

Assuming the modifications to ACT are pushed to a fork in a branch named wip-act-change, a pull request will verify it compiles. It will not verify that the Forgejo runner can compile with it.

For verifying it is compatible with the Forgejo runner, a branch should be pushed to a fork of the Forgejo runner (for instance wip-runner-change) that uses the ACT version under test in wip-act-change by modifying go.mod to contain something like the following and running go mod tidy:

replace github.com/nektos/act => code.forgejo.org/earl-warren/act wip-act-change

Where https://code.forgejo.org/earl-warren/act is the URL of the ACT fork and wip-act-change is the branch where the changes under test were pushed. It will not verify that it is able to properly run jobs when connected to a live Forgejo instance. The wip-runner-change branch must, in turn, be tested as explained above. When the Forgejo runner modified to include the changes in the wip-act-change branch pass the end to end test of the setup-forgejo action, it is ready to be released.

ACT can be released by merging the wip-act-change branch and by pushing a new tag, for instance v48.8.20. Once published, the Forgejo runner can be updated to default to this latest version knowing it already passed end to end tests with something like:

replace github.com/nektos/act => code.forgejo.org/forgejo/act v48.8.20

Local debug

The repositories are checked out in the same directory:

• runner: Forgejo runner
• act: ACT
• setup-forgejo: setup-forgejo

Install dependencies

The dependencies are installed manually or with:

setup-forgejo/forgejo-dependencies.sh

Build the Forgejo runner with the local ACT

The Forgejo runner is rebuilt with the ACT directory by changing the runner/go.mod file to:

replace github.com/nektos/act => ../act

Running:

cd runner ; go mod tidy

Building:

cd runner ; rm -f forgejo-runner ; make forgejo-runner

Launch Forgejo and the runner

A Forgejo instance is launched with:

cd setup-forgejo ; ./forgejo.sh setup
firefox http://$(cat forgejo-ip):3000

The user is root with password admin1234. The runner is registered with:

cd setup-forgejo
docker exec --user 1000 forgejo forgejo actions generate-runner-token > forgejo-runner-token
../runner/forgejo-runner register --no-interactive --instance "http://$(cat forgejo-ip):3000/" --name runner --token $(cat forgejo-runner-token) --labels docker:docker://node:16-bullseye,self-hosted

And launched with:

cd setup-forgejo ; ../runner/forgejo-runner --config runner-config.yml daemon

Note that the runner-config.yml is required in that particular case to configure the network in bridge mode, otherwise the runner will create a network that cannot reach the forgejo instance.

Try a sample workflow

From the Forgejo web interface, create a repository and add the following to .forgejo/workflows/try.yaml. It will launch the job and the result can be observed from the actions tab.

on: [push]
jobs:
  ls:
    runs-on: docker
    steps:
      - uses: actions/checkout@v3
      - run: |
          ls ${{ github.workspace }}