.github: reuse hardware-test workflow

[ozan956]

Use the shared Linux hardware-test workflow for Bootstrap validation instead of maintaining a br2-external wrapper.

Run Bootstrap on both SC598 EZKIT and EZLITE. Remove the br2-external U-Boot test since it is now covered by the U-Boot repository’s workflow.

This is to be merged after Linux side gets merged.

[pamolloy]

Both Buildroot and Yocto should run all tests and not depend on tests in the U-Boot and Linux repos. Both combine numerous pieces of software and it is important to test the the full system when they are all combined.

The bootstrap image is a br2_external concept so it is confusing why we'd use a shared action from the Linux repository. What logic exactly is being shared here?

[ozan956]

Both Buildroot and Yocto should run all tests and not depend on tests in the U-Boot and Linux repos. Both combine numerous pieces of software and it is important to test the the full system when they are all combined.

The bootstrap image is a br2_external concept so it is confusing why we'd use a shared action from the Linux repository. What logic exactly is being shared here?

Yep, I should have added more context to that PR.

So this hw-test.yaml in br2 was old version, clearly not really scaleable.

But the new ones as in U-Boot: https://github.com/analogdevicesinc/u-boot/blob/ci/.github/workflows/hw-test.yml

They are just generic things to convert tests to hw-test format. So they do not include and specific test names, board names, etc.

Then you can call that in top-level yaml as:

  hardware-test:
    needs: [build_aarch64]
    if: ${{ github.event_name != 'release' }}
    strategy:
      fail-fast: false
      matrix:
        include:
          - needs: '["sc598", "ezkit"]'
          - needs: '["sc598", "ezlite"]'
          - needs: '["sc846", "ezkit"]'
    uses: analogdevicesinc/u-boot/.github/workflows/hw-test.yml@ci
    with:
      test-set: 'adsp/u-boot'
      needs: ${{ matrix.needs }}
    secrets: inherit
    permissions:
      actions: read
      contents: read
      id-token: write

I picked Linux repo as base so I can just change one file and distribute to everywhere.

[gastmaier]

Hi, from u-boot

            ${{ inputs.needs != '' && format(', "needs": {0}', inputs.needs) || '' }}
            ${{ inputs.labgrid-target != '' && format(', "labgrid_target": "{0}"', inputs.labgrid-target) || '' }}

Couldn't empty string means any? so this becomes

          set: >
            {
              "name": "${{ inputs.test-set }}",
              "needs": "${{ inputs.needs }}",
              "labgrid_target": "${{ inputs.labgrid-target }}"
            }

[ozan956]

Hi, from u-boot

            ${{ inputs.needs != '' && format(', "needs": {0}', inputs.needs) || '' }}
            ${{ inputs.labgrid-target != '' && format(', "labgrid_target": "{0}"', inputs.labgrid-target) || '' }}

Couldn't empty string means any? so this becomes

          set: >
            {
              "name": "${{ inputs.test-set }}",
              "needs": "${{ inputs.needs }}",
              "labgrid_target": "${{ inputs.labgrid-target }}"
            }

You are right to call this out. An empty needs does not mean “any” for these tests, the test fails when it is omitted.

In this workflow, needs is always provided through the matrix so it should never be empty. I think we should make the input required and remove the empty-string/default handling.

I can fix that on linux side before merge this, then I can also have similar PR in u-boot to let U-Boot repo also use the linux side workflow as well.