CI🔗︎

gitlabci🔗︎

bt gitlabici prints configuration for Gitlab Continuous Integration to the terminal. This can be piped into the .gitlab-ci.yml file in the root of the repository. When there are multiple contests, just append the bt gitlabci of each of them, but deduplicate the top level image: and default: keys.

Use the --latest-bt flag to pull the latest version of BAPCtools before each run. By default, the version in the docker image is used.

Example output:

~nwerc2020 % bt gitlabci
image: bapctools

default:
  before_script:
    - git -C /cache/BAPCtools pull || git clone https://github.com/RagnarGrootKoerkamp/BAPCtools.git /cache/BAPCtools
    - ln -s /cache/BAPCtools/bin/tools.py bt

contest_pdf_nwerc2020:
  script:
      - ./bt pdf --cp --no-bar --contest nwerc2020
      - ./bt solutions --cp --no-bar --contest nwerc2020
  only:
    changes:
      - nwerc2020/testproblem/statement/**/*

  artifacts:
    expire_in: 1 week
    paths:
      - nwerc2020/solution*.pdf
      - nwerc2020/contest*.pdf

verify_testproblem:
  script:
      - ./bt all --cp --no-bar --problem nwerc2020/testproblem
  only:
    changes:
      - nwerc2020/testproblem/**/*
  artifacts:
    expire_in: 1 week
    paths:
      - nwerc2020/testproblem/problem*.pdf

The default behaviour is:

• Use the bapctools Docker image. This has to be installed manually from the Dockerfile found in the root of the repository.
• Before each stage, pull BAPCtools to the /cache partition. This makes sure to always use the latest version of BAPCtools.
• For contests: build the problem and solutions pdf and cache these artefacts 1 week.
• For problems: run bt all on the problem and keep the problem pdf for 1 week.

We use the following configuration for the gitlab runners:

[[runners]]
  name = "BAPC group runner"
  url = "<redacted>"
  token = "<redacted>"
  executor = "docker"
  [runners.custom_build_dir]
  [runners.docker]
    tls_verify = false
    image = "bapctools"
    privileged = false
    disable_entrypoint_overwrite = false
    oom_kill_disable = false
    disable_cache = false
    volumes = ["/cache"]
    shm_size = 0
    pull_policy = "never"
    memory = "2g"
    memory_swap = "2g"
  [runners.cache]
    [runners.cache.s3]
    [runners.cache.gcs]
  [runners.docker.tmpfs]
    "/tmp" = "rw,exec"

forgejo_actions🔗︎

bt forgejo_actions writes Forgejo Actions workflows for the current contest to the .forgejo directory in the root of the git repository. When there are multiple contests, run bt forgejo_actions once for each contest (either in the contest directory, or by passing --contest <contest>).

Use the --latest-bt flag to pull the latest version of BAPCtools before each run. By default, the version in the docker image is used.

The generated workflows are similar to those for bt gitlabci described above.

For smooth operation, use the following in the forgejo runner config.yaml to increase the memory limit of the container and mount /tmp to memory.

container:
  options: --memory=4g --memory-swap=4g --tmpfs /tmp:exec

{
  "labels": [
    "bapctools-docker:docker://ragnargrootkoerkamp/bapctools"
  ]
}

github_actions🔗︎

bt github_actions writes Github Actions workflows for the current contest to the .github directory in the root of the git repository. When there are multiple contests, run bt github_actions once for each contest (either in the contest directory, or by passing --contest <contest>).

The generated workflows are similar to those for bt gitlabci described above.