molecule-core/tools/branch-protection/drift_check.sh
Hongming Wang 2e505e7748 fix(branch-protection): apply.sh respects live state + full-payload drift
Multi-model review of #2827 caught: the script as-shipped would have
silently weakened branch protection on EVERY non-checks dimension
the moment anyone ran it. Live staging had

  enforce_admins=true, dismiss_stale_reviews=false, strict=true,
  allow_fork_syncing=false, bypass_pull_request_allowances={
    HongmingWang-Rabbit + molecule-ai app
  }

Script wrote the opposite for all five. Per memory
feedback_dismiss_stale_reviews_blocks_promote.md, the
dismiss_stale_reviews flip alone is the load-bearing one — would
silently re-block every auto-promote PR (cost user 2.5h once).

This PR:

1. apply.sh: per-branch payloads (build_staging_payload /
   build_main_payload) that codify the deliberate per-branch policy
   already on the repo, with the script's net contribution being
   ONLY the new check names (Canvas tabs E2E + E2E API Smoke on
   staging, Canvas tabs E2E on main).

2. apply.sh: R3 preflight that hits /commits/{sha}/check-runs and
   asserts every desired check name has at least one historical run
   on the branch tip. Catches typos like "Canvas Tabs E2E" vs
   "Canvas tabs E2E" — pre-fix a typo would silently block every PR
   forever waiting for a context that never emits. Skip via
   --skip-preflight for genuinely-new workflows whose first run
   hasn't fired.

3. drift_check.sh: compares the FULL normalised payload (admin,
   review, lock, conversation, fork-syncing, deletion, force-push)
   not just the checks list. Pre-fix the drift gate would have
   missed a UI click that flipped enforce_admins or
   dismiss_stale_reviews. Drops app_id from the comparison since
   GH auto-resolves -1 to a specific app id post-write.

4. branch-protection-drift.yml: per memory
   feedback_schedule_vs_dispatch_secrets_hardening.md — schedule +
   pull_request triggers HARD-FAIL when GH_TOKEN_FOR_ADMIN_API is
   missing (silent skip masks the gate disappearing).
   workflow_dispatch keeps soft-skip for one-off operator runs.

Verified by running drift_check against live state: pre-fix would
have shown 5 destructive drifts on staging + 5 on main. Post-fix
shows ONLY the 2 intended additions on staging + 1 on main, which
go away after `apply.sh` runs.
2026-05-04 20:52:11 -07:00

158 lines
5.8 KiB
Bash
Executable File

#!/usr/bin/env bash
# tools/branch-protection/drift_check.sh — compare the live branch
# protection on staging + main against what apply.sh would set. Used
# by branch-protection-drift.yml (cron) to catch out-of-band UI edits.
#
# Pre-2026-05-05 version diffed only required_status_checks.checks —
# would have missed a UI click that flipped enforce_admins or
# dismiss_stale_reviews. Now compares the full normalized payload so
# any silent rewrite of admin/review/lock/deletion settings trips the
# drift gate.
#
# Exit codes:
# 0 — live state matches the script
# 1 — drift detected (output shows the diff)
# 2 — gh API call failed
set -euo pipefail
REPO="Molecule-AI/molecule-core"
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
EXIT_CODE=0
# Normalise the GET /branches/:b/protection response so we can compare
# against apply.sh's payload. The GET response inflates booleans into
# {url, enabled} sub-objects and bypass list users/apps into full
# user/app objects with avatar_url etc — strip those down to match
# the input shape.
NORMALISE_LIVE='{
required_status_checks: (
.required_status_checks
| { strict: .strict,
checks: (.checks | map({context}) | sort_by(.context)) }
),
enforce_admins: (
if (.enforce_admins | type) == "object"
then .enforce_admins.enabled
else .enforce_admins end
),
required_pull_request_reviews: (
.required_pull_request_reviews
| if . == null then null else
{ required_approving_review_count,
dismiss_stale_reviews,
require_code_owner_reviews,
require_last_push_approval,
bypass_pull_request_allowances: (
if .bypass_pull_request_allowances == null then null
else {
users: (.bypass_pull_request_allowances.users // [] | map(.login) | sort),
teams: (.bypass_pull_request_allowances.teams // [] | map(.slug) | sort),
apps: (.bypass_pull_request_allowances.apps // [] | map(.slug) | sort)
} end
)
}
end
),
restrictions: (
if .restrictions == null then null
else { users: (.restrictions.users | map(.login) | sort),
teams: (.restrictions.teams | map(.slug) | sort),
apps: (.restrictions.apps | map(.slug) | sort) }
end
),
allow_deletions: (
if (.allow_deletions | type) == "object" then .allow_deletions.enabled
else (.allow_deletions // false) end
),
allow_force_pushes: (
if (.allow_force_pushes | type) == "object" then .allow_force_pushes.enabled
else (.allow_force_pushes // false) end
),
block_creations: (
if (.block_creations | type) == "object" then .block_creations.enabled
else (.block_creations // false) end
),
required_conversation_resolution: (
if (.required_conversation_resolution | type) == "object"
then .required_conversation_resolution.enabled
else (.required_conversation_resolution // false) end
),
required_linear_history: (
if (.required_linear_history | type) == "object" then .required_linear_history.enabled
else (.required_linear_history // false) end
),
lock_branch: (
if (.lock_branch | type) == "object" then .lock_branch.enabled
else (.lock_branch // false) end
),
allow_fork_syncing: (
if (.allow_fork_syncing | type) == "object" then .allow_fork_syncing.enabled
else (.allow_fork_syncing // false) end
)
}'
# Apply.sh's payload is already in the input shape; we just need to
# canonicalise the checks order and fill in optional fields with their
# defaults so the comparison aligns.
NORMALISE_SCRIPT='{
required_status_checks: {
strict: .required_status_checks.strict,
checks: (.required_status_checks.checks | map({context}) | sort_by(.context))
},
enforce_admins: .enforce_admins,
required_pull_request_reviews: (
if .required_pull_request_reviews == null then null else
{ required_approving_review_count: .required_pull_request_reviews.required_approving_review_count,
dismiss_stale_reviews: .required_pull_request_reviews.dismiss_stale_reviews,
require_code_owner_reviews: (.required_pull_request_reviews.require_code_owner_reviews // false),
require_last_push_approval: (.required_pull_request_reviews.require_last_push_approval // false),
bypass_pull_request_allowances: (
if .required_pull_request_reviews.bypass_pull_request_allowances == null then null
else {
users: (.required_pull_request_reviews.bypass_pull_request_allowances.users // [] | sort),
teams: (.required_pull_request_reviews.bypass_pull_request_allowances.teams // [] | sort),
apps: (.required_pull_request_reviews.bypass_pull_request_allowances.apps // [] | sort)
} end
)
}
end
),
restrictions: .restrictions,
allow_deletions: (.allow_deletions // false),
allow_force_pushes: (.allow_force_pushes // false),
block_creations: (.block_creations // false),
required_conversation_resolution: (.required_conversation_resolution // false),
required_linear_history: (.required_linear_history // false),
lock_branch: (.lock_branch // false),
allow_fork_syncing: (.allow_fork_syncing // false)
}'
check_branch() {
local branch="$1"
local want
want=$(bash "$SCRIPT_DIR/apply.sh" --dry-run --branch "$branch" 2>&1 |
sed -n '/^{$/,/^}$/p' |
jq -S "$NORMALISE_SCRIPT")
local have_raw
if ! have_raw=$(gh api "repos/$REPO/branches/$branch/protection" 2>/dev/null); then
echo "drift_check: FAIL to fetch $branch protection (gh API error)"
return 2
fi
local have
have=$(echo "$have_raw" | jq -S "$NORMALISE_LIVE")
if [[ "$want" != "$have" ]]; then
echo "=== DRIFT on $branch ==="
diff <(echo "$want") <(echo "$have") || true
return 1
fi
echo "OK: $branch matches desired state"
}
for b in staging main; do
if ! check_branch "$b"; then
EXIT_CODE=1
fi
done
exit "$EXIT_CODE"