Make in CI and team workflows
Same commands everywhere
The pipeline should call the same targets developers use:
stage('Test') { steps { sh 'make lint test' } }
stage('Image') { steps { sh 'make docker-push TAG=$GIT_COMMIT' } }
# GitHub Actions
- run: make lint test
- run: make docker-push TAG=${{ github.sha }}
Benefit: CI logic lives in the repo, is testable locally, and switching CI systems is easy.
"Works locally, fails in CI"
Usual causes: different tool versions (Go, Node, kubectl, terraform), different shells (/bin/sh is dash on Ubuntu agents vs zsh/bash on a Mac), environment variables present locally but not in CI (AWS_PROFILE, PATH entries), BSD vs GNU tools on macOS (sed -i '' vs sed -i), and stale local files make thinks are up to date.
Fixes: pin tools and run targets inside a toolchain container (the TOOLS_IMG pattern: docker run --rm -v $PWD:/w -w /w $(TOOLS_IMG) make test), or use asdf/mise/.tool-versions, set SHELL := bash, and make CI start from a clean checkout.
Windows developers
GNU make isn't native on Windows: use WSL2, run targets in a dev container, or provide equivalent scripts. This is one reason some teams choose cross-platform task runners.
Alternatives
| Tool | Notes |
|---|---|
just (justfile) |
Make-like syntax without file timestamps or tab pitfalls; command runner only |
Task (Taskfile.yml) |
YAML, cross-platform, supports checksums for up-to-date checks |
| npm scripts | Natural for JS repos |
| mage | Targets written in Go |
| Bazel / Nx / Pants | Large monorepos with remote caching |
Make's advantages remain: it's preinstalled almost everywhere, universally understood, and does incremental builds.