tf-drift

Scheduled Release Automation Workflow

Context

tf-drift already had a tag-triggered GoReleaser workflow that builds binaries and updates the Homebrew tap. The missing piece was an automatic release check so merged commits do not wait on a manual tag.

Intended Outcome

Every day at midnight UTC, the repository should publish a release only when the default branch has commits after the latest stable release tag. Manual tag releases and manual workflow dispatch should remain available.

Failure Modes Avoided

• A separate workflow that creates tags with GITHUB_TOKEN would not trigger the tag-push release workflow.
• A top-of-hour schedule can be delayed by GitHub Actions load, so manual dispatch remains documented as recovery.
• A release without tests could publish broken binaries, so the release job now runs the race test suite before GoReleaser.
• Parallel release attempts could collide on tags or uploaded assets, so the workflow serializes release runs.

Evidence

Validation used go test -race -v ./..., actionlint, git diff --check, and a local SemVer simulation. The current tag history would compute v1.1.0 from unreleased feat: commits after v1.0.0.

Reusable Pattern

For GitHub Actions release automation, keep tag calculation and publishing in the same workflow when using GITHUB_TOKEN. Use a local release tag plus GoReleaser target_commitish instead of depending on a second workflow run.

Limits

The Homebrew formula path still uses GoReleaser’s deprecated brews section. That preserves the existing brew install tf-drift install flow, but it should be revisited before moving to a future GoReleaser major version that removes formula support.

This site is open source. Improve this page.