A Skimmer Higgs

Overview

Teaching: 5 min
Exercises: 10 min
Questions
  • How can I run my skimming code in the GitHub Actions?

Objectives
  • Learn how to skim code and set up artifacts.

The First Naive Attempt

Let’s just attempt to try and get the code working as it is. Since it worked for us already locally, surely the CI/CD must be able to run it???

As a reminder of what we’ve ended with from the last session:

jobs:
  greeting:
    runs-on: ubuntu-latest
    steps:
      - run: echo hello world
 
  build_skim:
    needs: greeting
    runs-on: ubuntu-latest
    container: rootproject/root-conda:${{ matrix.version }}
    strategy:
      matrix:
        version: [6.18.04, latest]
    steps:
      - name: checkout repository
        uses: actions/checkout@v2

      - name: build
        run: |
          COMPILER=$(root-config --cxx)
          FLAGS=$(root-config --cflags --libs)
          $COMPILER -g -O3 -Wall -Wextra -Wpedantic -o skim skim.cxx $FLAGS

Since the skim binary is built, let’s see if we can run it. We need to add a job , with name skim.

skim is meant to process data (skimming) we are going to run on.

Let’s go ahead and figure out how to define a run job. Seems too easy to be true?

skim:
  needs: build_skim
  runs-on: ubuntu-latest
  container: rootproject/root-conda:6.18.04
  steps:
      - name: checkout repository
        uses: actions/checkout@v2

      - name: skim
        run: ./skim

Let’s do a quick check

act -l
ID          Stage  Name        
greeting    0      greeting    
build_skim  1      build_skim  
skim        2      skim

We now have 3 stages.

After you’ve added the skim job you can push your changes to GitHub:

git add .github/workflows/main.yml
git commit -m "add skim job"
git push -u origin feature/add-actions

Skim waiting

Skim waiting

Failed???

Let’s have a look at the log message

./skim: not found

We’re too naive

Ok, fine. That was way too easy. It seems we have a few issues to deal with. The skim binary in the build_skim job isn’t in the skim job by default. We need to use GitHub artifacts to copy over this from the right job.

Artifacts

Artifacts are used to upload (upload-artifact) and download (download-artifact) files and directories which should be attached to the job after this one has completed. That way it can share those files with another job in the same workflow.

More Reading

Passing data between two jobs in a workflow

job_1:
  - uses: actions/upload-artifact@v2
    with:
      name: <name>
      path: <file>
job_2:
  - uses: actions/download-artifact@v2
    with:
      name: <name>

Note that the artifact name should not contain any of the following characters ",:,<,>,|,*,?,\,/.

In order to take advantage of passing data between two jobs, one combines download-artifact with needs.

Combining download-artifact with needs

Let’s do it.

Solution

...
...
build_skim:
  needs: greeting
  runs-on: ubuntu-latest
  container: rootproject/root-conda:${{ matrix.version }}
  strategy:
    matrix:
      version: [6.18.04, latest]
  steps:
    - name: checkout repository
      uses: actions/checkout@v2

    - name: build
      run: |
        COMPILER=$(root-config --cxx)
        FLAGS=$(root-config --cflags --libs)
        $COMPILER -g -O3 -Wall -Wextra -Wpedantic -o skim skim.cxx $FLAGS

    - uses: actions/upload-artifact@v2
      with:
        name: skim${{ matrix.version }}
        path: skim

skim:
  needs: build_skim
  runs-on: ubuntu-latest
  container: rootproject/root-conda:6.18.04
  steps:
    - name: checkout repository
      uses: actions/checkout@v2

    - uses: actions/download-artifact@v2
      with:
        name: skim6.18.04

    - name: skim
      run: ./skim

What happened?

./skim: Permission denied

Permissions can be changed using the chmod command.

Solution

run: |
  chmod +x ./skim
  ./skim

From the log message, one can see that we are missing arguments.

Use executable with following arguments: ./skim input output cross_section integrated_luminosity scale

We will deal with that in the next lesson.

Key Points

  • Making jobs aware of each other is pretty easy.

  • Artifacts are pretty neat.

  • We’re too naive.