Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions internal/cairn/review.go
Original file line number Diff line number Diff line change
Expand Up @@ -59,6 +59,13 @@ func ListReviewMergeBranches(ctx context.Context, store string) ([]ReviewMergeBr
if name == "" {
continue
}
merged, err := isMergedInto(ctx, store, name, def)
if err != nil {
return nil, fmt.Errorf("check merge status of %q: %w", name, err)
}
if merged {
continue // already merged into def -- not awaiting review, regardless of whether the branch was deleted
}
relPath, err := changedEntryFile(ctx, store, def, name)
if err != nil {
return nil, err
Expand Down
42 changes: 42 additions & 0 deletions internal/cairn/review_test.go
Original file line number Diff line number Diff line change
Expand Up @@ -114,6 +114,48 @@ func TestListReviewBranchesErrorsOnPrivateTierBranch(t *testing.T) {
assert.Contains(t, err.Error(), "private agent/ tier")
}

// TestListReviewMergeBranchesExcludesAlreadyMergedBranch covers crn-12ubp:
// ListReviewMergeBranches lists refs/heads/remember/* by bare existence with
// no merge-status check, unlike its sibling ListReviewBranches (branches.go),
// which already skips a branch once isMergedInto reports it merged. A branch
// merged by some path other than MergeReviewBranch (or left behind by a
// delete failure after a successful merge) must not be reported as pending
// forever -- that's the "merged entries nag forever" half of the bug.
func TestListReviewMergeBranchesExcludesAlreadyMergedBranch(t *testing.T) {
store := reviewStore(t)
mergedBranch, _ := fixtureBranch(t, store, "merged-topic", nil, "a note already merged")
pendingBranch, _ := fixtureBranch(t, store, "pending-topic", nil, "a note still pending")

def, err := DefaultBranch(t.Context(), store)
require.NoError(t, err)

// Merge directly with git, bypassing MergeReviewBranch, to reproduce a
// branch that is merged but was never deleted -- e.g. a merge landed
// outside 'cairn review merge', or the delete step failed after a
// successful merge. Clear the untracked draft copy first, the same way
// mergeCuratedBranch does, or the merge itself refuses to run.
relPath, err := changedEntryFile(t.Context(), store, def, mergedBranch)
require.NoError(t, err)
_, err = gitRun(t.Context(), store, "clean", "-f", "-q", "--", relPath)
require.NoError(t, err)
_, err = gitRun(t.Context(), store, "merge", "--no-ff", mergedBranch, "-m", "merge without deleting")
require.NoError(t, err)

branchList, err := gitRun(t.Context(), store, "branch", "--list", mergedBranch)
require.NoError(t, err)
require.NotEmpty(t, strings.TrimSpace(branchList), "test setup: merged branch must still exist to reproduce the bug")

branches, err := ListReviewMergeBranches(t.Context(), store)
require.NoError(t, err)

names := make([]string, 0, len(branches))
for _, b := range branches {
names = append(names, b.Name)
}
assert.NotContains(t, names, mergedBranch, "a branch already merged into %s must not be listed as pending", def)
assert.Contains(t, names, pendingBranch, "an unmerged branch must still be listed as pending")
}

func TestTierFromEntryPath(t *testing.T) {
cases := []struct {
name string
Expand Down
68 changes: 68 additions & 0 deletions release-gates/crn-of9fp-gate.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,68 @@
# Release Gate: crn-of9fp

**Bead:** crn-of9fp (deploy) — source: crn-h9vgx (review), build: crn-12ubp
**Commit:** `53e7211cf77e3afed0c7dd2fa1ab1cf2896636c9`, cut onto `deploy/crn-of9fp-gate`
**Date:** 2026-08-20

## Background

`cairn review merge` merges a review branch but does not delete it, and
`ListReviewMergeBranches` (backing `cairn review list`) keyed its "pending"
listing purely on branch existence — with no merge-status check. A branch
merged by any path other than `MergeReviewBranch`, or left behind by a
delete failure after a successful merge, would nag forever as "pending
review" and the backlog count would never fall to reflect the true state.

The fix adds an `isMergedInto` check to `ListReviewMergeBranches`, mirroring
the equivalent check its sibling `ListReviewBranches` (branches.go) already
had — once a branch is merged into the default branch it is skipped,
regardless of whether it was also deleted. A new regression test,
`TestListReviewMergeBranchesExcludesAlreadyMergedBranch`, reproduces the bug
by merging a fixture branch directly with `git merge --no-ff` (bypassing
`MergeReviewBranch`) and confirms the merged branch no longer appears in the
pending list while a genuinely-unmerged sibling branch still does.

## Criteria

| # | Criterion | Result | Evidence |
|---|-----------|--------|----------|
| 1 | Review PASS for exact deployed commit | PASS | crn-h9vgx's verdict cites `53e7211cf77e3afed0c7dd2fa1ab1cf2896636c9`, identical to the deploy bead's `metadata.gc.deploy_commit` — exact SHA match (D = R). Independently re-resolved this session via `git rev-parse --verify --quiet "<sha>^{commit}"`, not trusted as a transcribed string. |
| 2 | Acceptance criteria met | PASS | Verified directly against `git diff 8002079...53e7211` (not just the reviewer's write-up): `isMergedInto` check added to `ListReviewMergeBranches` in `internal/cairn/review.go`, matching the existing pattern in `ListReviewBranches`. Matches crn-12ubp's 5-item exit_contract and the bead's description exactly. |
| 3 | Tests pass | PASS | Independently re-run by the deployer on a detached checkout of `53e7211...` (own build, not the reviewer's self-report): `go build ./...` clean, `go vet ./...` clean, `gofmt -l .` empty, diff-owned test `TestListReviewMergeBranchesExcludesAlreadyMergedBranch` run by name — PASS, and the full `internal/cairn` package suite — PASS (`ok`, 62.3s). |
| 4 | No open HIGH findings | PASS | `bd list --status open` / `--status in_progress` checked fresh for anything referencing this chain (crn-of9fp, crn-h9vgx, crn-12ubp): only the review bead itself (in-progress review record, not a finding) and two sling/routing bookkeeping beads (crn-sdi8h, crn-xev9b — no findings content). No open HIGH finding anywhere in the chain. |
| 5 | Clean tree | PASS | `deploy/crn-of9fp-gate` cut directly from `53e7211...^{commit}` via `resolve_deploy_branch_target`; `git status --porcelain` empty throughout. Diff is exactly 2 files / 49 insertions / 0 deletions — one production change, one test, no stray content. |
| 6 | Clean divergence from main | PASS | `git rev-list --left-right --count origin/main...53e7211` → `4 2`: 4 commits behind (pure staleness), 2 ahead (the red/green TDD pair). Re-verified immediately before cutting the branch that no commit touching `review.go`/`review_test.go` landed on `origin/main` between the fork point and current tip (93c2593) — zero file overlap, no rebase needed, no conflict possible. |
| 7 | Single feature theme | PASS | `assert_deploy_ancestry_scope origin/main 53e7211... crn-of9fp crn-12ubp crn-h9vgx` → rc=0: no `.claude/**` paths touched, and both non-merge commits in `origin/main..53e7211` cite `crn-12ubp` in their message. No stray commits. |

## Verdict: PASS — proceeding to PR.

## Process notes

1. **Merge authority — proceeding under the standing self-merge authorization,
not routing to mayor.** The deploy bead's own text carries the familiar
"route a merge-request to mayor/mpr — do NOT merge yourself" clause. Per
fleet memory `cairn-auto-merge-requires-explicit-strategy`, this is a
**known, understood stale-copy issue** for this repo, not a fresh signal:
mayor's explicit, mail-verified ruling (`gm-wisp-2yhv7u`, 2026-08-19)
already addressed this exact recurring boilerplate — cairn is not covered
by mpr, so the clause is simply carried over from a template written for
repos that are — and instructed future sessions *not* to re-pause or
re-escalate merely because a bead repeats it. This gate proceeds under the
STANDING AUTHORIZATION (7/7 PASS + CI green on `quad341/cairn`) with the
four attached conditions: (1) gate 7/7 PASS — this doc; (2) PR state
confirmed via a direct, fresh `gh pr view --json` read at merge time; (3)
CLEAN/MERGEABLE with both required checks (`build-test`, `lint`)
COMPLETED/SUCCESS; (4) no `--auto` — plain `gh pr merge <n> --squash`
only, then verify `state=MERGED` and `mergedAt` non-null before reporting
success.

2. **Branch target:** `builder/crn-12ubp` is provenance-only per the deploy
bead's own instruction — a possibly shared builder branch, not a push
target. `deploy/crn-of9fp-gate` was cut fresh from the exact reviewed SHA
via `resolve_deploy_branch_target`, and `assert_safe_push_target` confirmed
the derived name does not match the shared-worktree-branch signature.

3. **SHA integrity:** the deploy commit was independently re-resolved via
`git rev-parse --verify --quiet "<sha>^{commit}"` before any gate step
ran, per the sha-integrity discipline — never trusted as an eyeballed or
transcribed string.
Loading