Speed Up Industrial CI and Codecov by Using a Prebuilt ROS Humble Base Image

[kluge7]

Motivation

Many of our CI workflows (Industrial CI, Codecov, and others) repeatedly run the same dependency-installation script on every job execution.
For example, vortex-auv runs scripts/ci_install_dependencies.sh before its builds.
This includes installing Python packages, compiler toolchains, and other system dependencies — all of which significantly increase CI runtime and resource usage.

By prebuilding a Docker image that already contains:

• ros:humble-ros-base
• All CI-required dependencies
• The logic currently inside ci_install_dependencies.sh

…we can remove these heavy setup steps entirely from our CI runs.
This will noticeably speed up CI pipelines, reduce duplicated work across repositories, and improve overall reliability and reproducibility.

Description

Introduce a new Docker image (hosted on GHCR under vortexntnu) that extends ros:humble-ros-base and pre-installs all common CI dependencies used across our repositories.
The image should mirror what ci_install_dependencies.sh currently performs (Python deps, GCC toolchain setup, etc.).

A new GitHub Actions workflow should:

• Detect changes to CI-related scripts (e.g., scripts/ci_install_dependencies.sh)
• Automatically build and publish a new version of the Docker image when these files change
• Tag and push the image to GHCR (e.g., ghcr.io/vortexntnu/ci-base:latest)

Industrial CI and other workflows (Codecov, testing workflows, etc.) should then accept a new docker_image input and use this prebuilt image as their base, removing the need to run manual install steps during the job. This will greatly reduce CI duration and improve reproducibility.

Task

1. In vortexntnu/vortex-ci

   • Add a GitHub Actions workflow that:
     • Uses ros:humble-ros-base as the base image.
     • Installs all dependencies currently defined in vortex-auv/scripts/ci_install_dependencies.sh.
     • Publishes the image to GHCR (ghcr.io/vortexntnu/ci-base:latest).
     • Triggers on:
       • Changes to dependency installation scripts
       • Manual dispatch
       • Push to main
   • Document how to use DOCKER_IMAGE in Industrial CI workflows.

2. Modify the reusable Industrial CI workflow

   • Add a new input: docker_image.
   • If provided, pass it to Industrial CI via the DOCKER_IMAGE environment variable.
   • Update the example workflows to demonstrate using the new shared CI base image.

3. Update vortex-auv and other repos

   • Remove (or bypass) the dependency-install step when a custom DOCKER_IMAGE is supplied.
   • Update existing CI workflows (Industrial CI, Codecov, etc.) to use the shared base image.

Deliverables

• A published Docker image (ghcr.io/vortexntnu/ci-base:latest) with:
  • ROS Humble base
  • All CI-required dependencies pre-installed
• A GitHub Actions workflow that auto-builds the image when dependency scripts change.
• Updated vortex-ci reusable workflows supporting a docker_image input.
• Updated documentation describing:
  • Build triggers
  • Tagging conventions
  • How to use the image in CI
• Updated CI workflows in vortex-auv (and optionally other repos) to use the new base image.
• Demonstration that CI time is reduced (e.g., before/after comparison).

References

• Dependency script: https://github.com/vortexntnu/vortex-auv/blob/main/scripts/ci_install_dependencies.sh
• ROS base image: https://hub.docker.com/_/ros
• Industrial CI docs: https://github.com/ros-industrial/industrial_ci
• GHCR documentation: https://docs.github.com/en/packages/working-with-a-github-packages-registry
• Current Docker build workflow example: https://github.com/vortexntnu/vortex-auv/blob/main/.github/workflows/docker-export.yml

Contacts

@kluge7
@Andeshog