CAVISE Pipelines

Reusable GitHub Actions workflows and actions shared by CAVISE repositories.

Contract

Reusable workflows accept only predefined typed inputs. They must not accept shell scripts, arbitrary commands, arbitrary command-line arguments, or environment export blocks from consumer repositories.

Generic Python Workflows

• .github/workflows/python-black.yml runs black --check --diff with a pinned Black version.
• .github/workflows/python-pytest.yml runs pytest, optional protobuf generation, optional JUnit upload.
• .github/workflows/python-ruff.yml runs ruff check and/or ruff format --check --diff.
• .github/workflows/python-mypy.yml installs mypy tooling, optionally generates protobuf modules, and runs mypy.
• .github/workflows/python-cmake-cuda.yml builds CMake CUDA extensions, validates the generated artifact manifest, and optionally imports installed Python modules.
• .github/workflows/python-deadcode.yml installs and runs deadcode.
• .github/workflows/python-pre-commit.yml runs pre-commit with a controlled --all-files flag and optional SKIP hook list.

Generic C++ Workflows

• .github/workflows/cpp-clang-format.yml runs a Meson project's clang-format-check target inside a caller-provided GHCR image.
• .github/workflows/cpp-meson-clang-tidy.yml builds a Meson project inside a caller-provided GHCR image, optionally prepares ns-3, and runs Meson's clang-tidy target.

The caller must grant packages: read to GITHUB_TOKEN. The container image must provide Bash, Git, a C++ compiler, Meson, Ninja, the requested Clang tool, and the project's build dependencies.

Composite Actions

• actions/setup-python-pip sets up Python, enables pip cache, upgrades pip, and optionally installs a requirements file.
• actions/generate-protobufs generates Python protobuf modules using protoc, grpc_tools.protoc, or named CMake presets.

Child repositories should keep their own trigger and composition files, then call these reusable workflows with jobs.<job_id>.uses.

Before publishing a stable release, internal action references in workflows should be pinned to the same release tag, for example @v1.