• I want to run 3 steps, e.g. lint, type-check, unit tests.
• I want the unit tests to run even if the linting or type-checks fail, and I want the type-check to run even if the linting fails (e.g. continue-on-error).
• If any step fails, I want the job to fail (e.g. if: failure() and run: exit 1).

The reason I want this is because I want to see a complete list of static code failures. I could separate them into individual jobs and a job dependent on their completions, but I don't need them run in parallel and it seems like a lot of wasteful repeated identical setups.

What I have so far is this:

jobs:
  checks:
    name: Run Checks
    runs-on: ubuntu-22.04
    steps:

      # ... Various setup code...

      - name: Lint
        uses: borales/actions-yarn@v4
        continue-on-error: true
        with:
          cmd: lint --max-warnings=0

      - name: Run the type check
        uses: borales/actions-yarn@v4
        continue-on-error: true
        with:
          cmd: type-check

      - name: Run the unit tests
        uses: borales/actions-yarn@v4
        continue-on-error: true
        with:
          cmd: test

      - name: A step has failed
        if: failure()
        run: exit 1

As may be evident, this will not work. continue-on-error means the step is never marked as a failure, and without it I never see what else is going to fail. This means I need a different approach to the pipeline configuration to achieve the outcome.

Does anyone have any recommendations?

• I want to run 3 steps, e.g. lint, type-check, unit tests.
• I want the unit tests to run even if the linting or type-checks fail, and I want the type-check to run even if the linting fails (e.g. continue-on-error).
• If any step fails, I want the job to fail (e.g. if: failure() and run: exit 1).

The reason I want this is because I want to see a complete list of static code failures. I could separate them into individual jobs and a job dependent on their completions, but I don't need them run in parallel and it seems like a lot of wasteful repeated identical setups.

What I have so far is this:

jobs:
  checks:
    name: Run Checks
    runs-on: ubuntu-22.04
    steps:

      # ... Various setup code...

      - name: Lint
        uses: borales/actions-yarn@v4
        continue-on-error: true
        with:
          cmd: lint --max-warnings=0

      - name: Run the type check
        uses: borales/actions-yarn@v4
        continue-on-error: true
        with:
          cmd: type-check

      - name: Run the unit tests
        uses: borales/actions-yarn@v4
        continue-on-error: true
        with:
          cmd: test

      - name: A step has failed
        if: failure()
        run: exit 1

As may be evident, this will not work. continue-on-error means the step is never marked as a failure, and without it I never see what else is going to fail. This means I need a different approach to the pipeline configuration to achieve the outcome.

Does anyone have any recommendations?

• github-actions

• 1 To rephrase, you want all steps to always run but the job only passes if all pass? – Schwern Commented Mar 17 at 18:24

1 Answer 1

Make each check its own job, then write one job which needs all the check jobs. All the jobs will execute in parallel (bonus, this will be faster), but the top level job will only succeed if they all succeed.

jobs:
  checks:
    name: Run Checks
    needs: [lint, type-checks, test]
  lint:
    - name: Lint
      uses: borales/actions-yarn@v4
      with:
        cmd: lint --max-warnings=0
  type-checks:
    - name: Run the type check
      uses: borales/actions-yarn@v4
      with:
        cmd: type-check
  test:
    - name: Run the unit tests
      uses: borales/actions-yarn@v4
      with:
        cmd: test