atcoder-docker-images

Docker images for AtCoder language environments.

Use a published image

1. Pick the selector for the language you want. The safest choice is the info/*.toml filename stem.
2. Pull the image:

docker pull ghcr.io/cputils/atcoder-docker-images/<normalized-stem>:2025-10

3. Run it with your source mounted into /judge:

docker run --rm -it -v "$PWD:/judge" ghcr.io/cputils/atcoder-docker-images/<normalized-stem>:2025-10 bash

Inside the container, work in /judge and use the language tools directly.

What is included

Each image is built by running the install step in its metadata file. Common pieces include:

• Base image: ubuntu:24.04
• Shared packages: bash, curl, git, python3, build-essential, and more
• Working directory: /judge
• Metadata: /opt/atcoder/info.toml
• Install script: /opt/atcoder/atcoder-install.sh
• Default shell at runtime: bash

Metadata file

Each image is defined by one info/*.toml file. That file is the source of truth for the image and usually contains:

• language: a short language name
• display: the user-facing name shown in selectors
• install: the build-time setup script
• compile: the compile check or command, when present
• execution: the runtime command, when present
• filename: the main source filename expected by the runtime, when present
• license and library: dependency metadata, when present

If you are choosing an image, start with the filename stem under info/, then check the display field when needed.

About patches/

patches/ contains short notes for images that need extra handling. The patch filename matches the image stem, and each file names its target language:

If a language has a patch file, read it before using that image.

Build locally (optional)

uv run build.py python-cpython-3-13-7

selector can be any of:

• the info/*.toml filename stem
• the language field
• the display field

Examples:

uv run build.py "Python"
uv run build.py "Python (CPython 3.13.7)"
uv run build.py python-cpython-3-13-7

The local tag after building is "<normalized-stem>:2025-10". For example, python-cpython-3-13-7 becomes python-cpython-3-13-7:2025-10.

Run the container after building

docker run --rm -it python-cpython-3-13-7:2025-10 bash

The working directory is /judge. Put your source files there and use tools like python and g++ directly.

Publish to GHCR

This is only needed if you are publishing a new image.

uv run upload.py python-cpython-3-13-7

upload.py resolves the GHCR namespace in this order:

1. GITHUB_REPOSITORY
2. remote.origin.url

It accepts credentials from either:

• GHCR_USERNAME + GHCR_TOKEN
• GITHUB_ACTOR + GITHUB_TOKEN

If neither is set, it falls back to the gh CLI.

The full pushed image reference is:

ghcr.io/cputils/atcoder-docker-images/<normalized-stem>:2025-10

GitHub Actions

Two workflows are included:

• build-and-upload-language.yml
  • builds and uploads one selector
• build-and-upload-all-languages.yml
  • builds and uploads all languages
  • supports skip_selectors to exclude specific images