Building with Images


Teaching: 10 min
Exercises: 5 min
  • Can we use docker images to ease our setup?

  • Use docker images

  • Making reusable/flexible CI/CD jobs

Say “Docker” 🐳

While we won’t be going into detail about containers (for that check our Docker lesson), we’ve been using them all this time with Gitlab. Gitlab runners are working within a barebones virtual environment that runs Linux, and is itself an image.

Naturally, we can leverage the fact that Gitlab runners can run Docker to further simplify setting up the working environment. This is done using the image keyword. The input for image is the name of the image, including the registry path if needed, in one of these formats:

Here’s an example:

  image: $IMAGE
    - python3 --version
      - IMAGE: "python:3.7-buster"
      - IMAGE: "python:3.8-buster"
      - IMAGE: "python:3.9-buster"
      - IMAGE: "python:3.10-buster"
      - IMAGE: "python:3.11-buster"
#   You could also do
# - IMAGE: ["python:3.7-buster","python:3.8-buster","python:3.9-buster","python:3.10-buster","python:3.11-buster"]

Back to our CI file

Go to the ROOT docker hub page and choose a version any version you wish to try.

Let’s add image: $ROOT_IMAGE because we can still use parallel:matrix: to make various builds easily. Since we’re going to use a docker image to have a working version of ROOT, we can omit the lines that install and set up conda and ROOT. Again, taking the yml file we’ve been working on, we can further reduce the text using Docker images as follows.

    - echo "Hello World"

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

  extends: .template_build
  image: $ROOT_IMAGE
      - ROOT_IMAGE: ["rootproject/root:6.28.10-ubuntu22.04","rootproject/root:latest"]


We used the latest docker image and an ubuntu image in this particular example but the script remains the same regardless if you wish to use the conda build or an ubuntu build of ROOT.

Make sure your image works with the CI, not all images listed in the rootproject’s docker hub work 100% of the time.

Key Points

  • We can shorten a lot of the setup with Docker images