fix(cli): harden PR merge path against stale-base dead-ends (empty-diff no-op + conflict escape)#1943
Conversation
Two failure modes wedged the PR-based merge flow (dormant under the
current direct-merge config, but the rollback target):
1. Empty-diff branches ("No commits between ...") were marked `failed`,
pinning the serial merge slot + file leases on a task that is a
legitimate no-op. Now finalized as a terminal DONE via
`finalizeNoOpMergeTask`, mirroring the engine's canonical
`noOpResult` decision (merger-ai.ts).
2. A stale-base PR that GitHub reports CONFLICTING never becomes
mergeable on its own (nothing in the PR path rebases the head), so
`awaiting-pr-checks` waited unbounded — no escape, because the PR
path never incremented `mergeRetries`. Now each conflicting poll
counts against `mergeRetries`, so the existing
`getInReviewStallReason` "merge-retries-exhausted" escape disposes
the task after `maxAutoMergeRetries` cycles. Pending/behind PRs
still wait (checks legitimately run; "behind" is fixed by rebase).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@packages/cli/src/commands/__tests__/task-lifecycle.test.ts`:
- Around line 671-736: The current tests cover only the per-task PR path, but
the same invariant must also be verified on the shared-branch-group path. Add
regression coverage in the task lifecycle tests for the shared-group surface by
exercising the shared-group finalization flow that handles the “No commits
between” case and asserting it resolves as a no-op done outcome, and also cover
the shared-group not-ready/conflicting handling if the task-lifecycle logic now
routes there. Use the existing task lifecycle helpers and the shared-group
branch-group entrypoints in task-lifecycle.test.ts to locate the relevant flows.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 7a14e2bf-7565-434f-8341-e82c522611b5
📒 Files selected for processing (2)
packages/cli/src/commands/__tests__/task-lifecycle.test.tspackages/cli/src/commands/task-lifecycle.ts
| it("counts a conflicting stale-base PR against mergeRetries so the stall escape fires", async () => { | ||
| const task: MockTask = { | ||
| id: "FN-9020", | ||
| title: "test", | ||
| description: "desc", | ||
| column: "in-review", | ||
| mergeRetries: 1, | ||
| }; | ||
| const branch = getTaskBranchName(task.id); | ||
| const store = makeStore(task); | ||
| execMock.mockImplementation(() => ""); | ||
|
|
||
| const existingPr = { number: 9, url: "https://github.com/x/y/pull/9", status: "open" as const, headBranch: branch, baseBranch: "main" }; | ||
| const github = { | ||
| findPrForBranch: vi.fn(async () => existingPr), | ||
| createPr: vi.fn(), | ||
| getPrMergeStatus: vi.fn(async () => ({ | ||
| prInfo: { ...existingPr, mergeable: "conflicting" as const }, | ||
| reviewDecision: null, | ||
| checks: [], | ||
| mergeReady: false, | ||
| blockingReasons: [], | ||
| })), | ||
| mergePr: vi.fn(), | ||
| }; | ||
|
|
||
| const result = await processPullRequestMergeTask(store as never, "/repo", task.id, github as never, () => undefined); | ||
|
|
||
| expect(result).toBe("waiting"); | ||
| expect(store.updateTask).toHaveBeenCalledWith(task.id, { | ||
| status: "awaiting-pr-checks", | ||
| mergeRetries: 2, | ||
| }); | ||
| }); | ||
|
|
||
| it("does not bump mergeRetries for a non-conflicting not-ready PR", async () => { | ||
| const task: MockTask = { | ||
| id: "FN-9021", | ||
| title: "test", | ||
| description: "desc", | ||
| column: "in-review", | ||
| mergeRetries: 1, | ||
| }; | ||
| const branch = getTaskBranchName(task.id); | ||
| const store = makeStore(task); | ||
| execMock.mockImplementation(() => ""); | ||
|
|
||
| const existingPr = { number: 10, url: "https://github.com/x/y/pull/10", status: "open" as const, headBranch: branch, baseBranch: "main" }; | ||
| const github = { | ||
| findPrForBranch: vi.fn(async () => existingPr), | ||
| createPr: vi.fn(), | ||
| getPrMergeStatus: vi.fn(async () => ({ | ||
| prInfo: { ...existingPr, mergeable: "unknown" as const }, | ||
| reviewDecision: null, | ||
| checks: [], | ||
| mergeReady: false, | ||
| blockingReasons: [], | ||
| })), | ||
| mergePr: vi.fn(), | ||
| }; | ||
|
|
||
| const result = await processPullRequestMergeTask(store as never, "/repo", task.id, github as never, () => undefined); | ||
|
|
||
| expect(result).toBe("waiting"); | ||
| expect(store.updateTask).toHaveBeenCalledWith(task.id, { status: "awaiting-pr-checks" }); | ||
| }); |
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
Add regression coverage for the shared-branch-group surface.
Both new tests (and the reworked no-delta test) exercise only the per-task PR creation/merge path. The same two behaviors also have a shared-branch-group surface: no-op finalization on "No commits between" (Line 799-812) and the not-ready/conflicting handling (Line 858-861). Please enumerate and assert the invariant on the shared-group surface too — at minimum a shared-group "No commits between" → no-op done case, and (if the gap flagged in task-lifecycle.ts is addressed) a shared-group conflicting case.
As per coding guidelines: "When fixing a bug, regression tests must assert the general invariant across all known surfaces, and bug-fix work must include symptom verification and surface enumeration rather than only the single reproduction."
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@packages/cli/src/commands/__tests__/task-lifecycle.test.ts` around lines 671
- 736, The current tests cover only the per-task PR path, but the same invariant
must also be verified on the shared-branch-group path. Add regression coverage
in the task lifecycle tests for the shared-group surface by exercising the
shared-group finalization flow that handles the “No commits between” case and
asserting it resolves as a no-op done outcome, and also cover the shared-group
not-ready/conflicting handling if the task-lifecycle logic now routes there. Use
the existing task lifecycle helpers and the shared-group branch-group
entrypoints in task-lifecycle.test.ts to locate the relevant flows.
Source: Coding guidelines
|
This looks good please update to ready for review when comfortable |
Greptile SummaryThis PR updates the CLI PR merge lifecycle for stale-base edge cases. The main changes are:
Confidence Score: 4/5The shared branch-group no-op path can leave sibling tasks in an invalid lifecycle state.
packages/cli/src/commands/task-lifecycle.ts; add a changeset for the published CLI fix. Important Files Changed
Reviews (1): Last reviewed commit: "fix(cli): harden dormant PR merge path a..." | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
| await store.updateBranchGroup(branchGroup.id, { prState: "none", prNumber: null, prUrl: null }); | ||
| await store.updateTask(task.id, { status: "failed", error: `No pull request created for ${branchGroup.branchName}: no commits relative to ${projectDefaultBranch}.` }); | ||
| await store.logEntry(task.id, "No group pull request created", message); | ||
| await finalizeNoOpMergeTask(store, cwd, task, "No group pull request created (no commits vs base) — finalizing as no-op", pool); |
There was a problem hiding this comment.
When a shared branch-group PR has no commits relative to the base, this branch clears the group PR state but finalizes only the current task. Sibling members can remain in in-review with the shared group reset to prState: "none", so later lifecycle passes can recreate or split the group flow instead of completing the whole no-op group together.
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
| * slot + file leases indefinitely; a no-op branch is not a failure — there was | ||
| * simply nothing to merge. | ||
| */ | ||
| async function finalizeNoOpMergeTask( |
There was a problem hiding this comment.
This CLI change affects published @runfusion/fusion behavior by moving empty-diff PR tasks to no-op done and by bounding conflicting PR waits. The repository release rules require a patch changeset for published package fixes, so without one this operator-facing fix can be missed by versioning and release notes.
Context Used: AGENTS.md (source)
2 participants
Problem
Two stale-base failure modes wedge the PR-based merge flow (
processPullRequestMergeTaskinpackages/cli/src/commands/task-lifecycle.ts). Both pin the serial merge slot + file-scope leases on a dead-end task, starving ready work behind it.Empty-diff branch —
createPrthrowsNo commits between <base> and <head>. The task was markedfailed, even though a zero-commit branch is a legitimate no-op, not a failure. This is the PR-path analog of the engine-path fix in fix(engine): short-circuit zero-commits-ahead branch before AI-merge clean-room churn #1920 (noOpResult→ done).Conflicting stale-base PR — a PR GitHub reports as
CONFLICTINGnever becomes mergeable on its own (nothing in the PR path rebases the head branch). The task sat inawaiting-pr-checksunbounded:awaiting-pr-checksis not inTRANSIENT_MERGE_STATUSES, so the age-based stall escape never catches it, and the only other escape —getInReviewStallReason'smerge-retries-exhausted— never fired because the PR path never incrementedmergeRetries.Fix
finalizeNoOpMergeTaskfinalizes empty-diff branches as a terminal DONE, mirroring the engine's canonicalnoOpResultdecision (merged:false, noOp:true). Both the group and per-task empty-diff sites route through it.Each
CONFLICTINGpoll now counts againstmergeRetries, so the existingmerge-retries-exhaustedescape disposes the task aftermaxAutoMergeRetriescycles. Detection is gated onprInfo.mergeable === "conflicting"(the genuine unresolvable signal), notblockingReasons(status/review/checks only). Pending /behindPRs still wait indefinitely — checks legitimately run, andbehindis resolved by the pre-merge rebase.No new machinery — fix #2 reuses the existing stall-escape counter rather than a hand-rolled fail branch.
Tests
packages/cli/src/commands/__tests__/task-lifecycle.test.ts:moveTask(id,"done"),updateTask({status:null,mergeRetries:0}), neverfailed)mergeRetriesbumpedmergeRetriesnot bumpedpnpm --filter @runfusion/fusion test(task-lifecycle suite 39/39),typecheck, andbuildall green.🤖 Generated with Claude Code
Summary by CodeRabbit