Skip to content

0006-solo-maintainer-review-count-exception.md

Latest commit

 

History

History
53 lines (43 loc) · 2.56 KB

0006-solo-maintainer-review-count-exception.md

File metadata and controls

53 lines (43 loc) · 2.56 KB
Raw

6. Required PR review count is 0, by explicit waiver, not a silent gap

  • Status: Accepted
  • Date: 2026-07-05

Context

The code-quality and CI/CD standards call for branch protection requiring at least one approving review before merge (CQ-37/38, CICD-11..16). habitable today has one active maintainer. A hard required_approving_review_count: 1 on a solo-maintained repo does not add a second pair of eyes — it either deadlocks every PR (nobody else can approve) or forces the maintainer to add a bypass actor for themselves, which is functionally identical to having no review requirement, except now it is hidden inside a bypass-actor exception instead of stated as policy.

The audit (2026-07-05, CQ/CICD family) is correct that "documented intent, not enforced" is a real defect distinct from "enforced." The wrong fix is to pretend a review gate exists when it cannot, in practice, gate anything.

Decision

.github/rulesets/main-branch.json's pull_request rule sets required_approving_review_count: 0 explicitly. Pull requests themselves, current-branch status checks, review-thread resolution, no force-push, and no deletion are enforced with no bypass actor, including for the repository owner.

require_code_owner_review is deliberately false while the sole CODEOWNER is also the author of every maintainer PR. GitHub treats code-owner review as an independent approval requirement even when the general approval count is zero; turning it on now would deadlock every change, not add a second pair of eyes. When a second maintainer joins, this ADR requires enabling code-owner review and raising the approval count together.

This is a dated, explicit waiver, not an absent gate: the ruleset artifact records the real posture (zero-review, solo-maintainer) instead of a branch-protection setting that claims "1 review required" while the only human who can approve is also the only human who can push.

Consequences

  • All currently configured AUTO-checkable merge gates are enforced without exception, including against the maintainer; required checks must be rerun against current main before merge.
  • The moment a second regular contributor joins, this ADR is superseded: bump required_approving_review_count to 1 and set require_code_owner_review to true in .github/rulesets/main-branch.json and re-apply the ruleset. That is the trigger condition — not a calendar date.
  • Until then, the review-count control is scored as an honest, recorded exception (per the standard's own waiver mechanism) rather than a silent FAIL or a misleading PASS.