test: isolate and clean temporary test roots

[hxy91819]

Summary

Repeated test runs can leave fixture directories behind in the shared system temporary directory. Some fixtures intentionally contain non-regular files such as FIFOs, which can also cause content-inspecting host cleanup tools to block indefinitely.

This change keeps the FIFO safety coverage while tightening the test lifecycle:

• track and remove validation fixtures after each test
• run standard test commands inside a unique per-run temporary root
• clean the run root after success, assertion failure, and termination signals
• reap abandoned run roots after a TTL while protecting active concurrent runs with a heartbeat
• remove roots recursively without opening fixture contents

The tracker follows the established OpenClaw auto-cleanup temp-directory pattern, adapted locally so ClawSweeper does not depend on another repository's test helpers.

Review follow-up

Validated at head a602f2ba67.

• P2 fixed at the supported test-entry boundary: target-validation is now a canonical run-node-tests.mjs target. Validation fixtures created through the supported narrow-test command inherit the run root, so a SIGKILL, crash, or reboot leaves a clawsweeper-test-run-* root that the existing heartbeat/TTL reaper recognizes. The raw node --test ... target-validation.test.ts command is no longer used as the recommended evidence path.
• P3 fixed: the Windows CI job no longer passes test/codex-process.test.ts twice; the named target remains the single source of file selection.
• A new heartbeat-specific unit test was not added because this follow-up does not change heartbeat semantics; the existing lifecycle suite continues to cover stale-root reaping, assertion failure, SIGTERM, and successful cleanup.

Local black-box validation

Linux, Node v24.15.0. Corepack resolved the repository pin to pnpm 11.10.0.

Runner and lifecycle behavior

node --test test/run-node-tests.test.ts test/test-temp-root-runner.test.ts

Result: 10 passed, 0 failed. This covers named-target selection plus successful FIFO cleanup, assertion-failure cleanup, stale-root TTL cleanup, and SIGTERM cleanup through the real CLI runner.

Incident-equivalent FIFO fixture through the canonical target

node scripts/run-node-tests.mjs target-validation --test-concurrency=1 -- \
  --test-name-pattern "workspace preflight rejects non-regular package metadata without blocking"

Result: 1 passed, 0 failed; FIFO behavior completed in 19.75 ms. Matching counts were unchanged:

clawsweeper-test-run-*   0 -> 0
clawsweeper-validation-* 0 -> 0

Repeated full target validation through the canonical target

Executed twice:

node scripts/run-node-tests.mjs target-validation --test-concurrency=1 -- \
  --test-reporter=tap

Results:

• run 1: 198 passed, 3 skipped, 0 failed; 133.8 s test duration / 134 s wall clock
• run 2: 198 passed, 3 skipped, 0 failed; 131.6 s test duration / 131 s wall clock
• clawsweeper-test-run-*: 0 -> 0 -> 0
• clawsweeper-validation-*: 0 -> 0 -> 0
• no run root or validation fixture remained after either run

Repository checks

• corepack pnpm run check: static checks, formatting, builds, lint, and changed coverage passed; the final full-coverage run failed only in three test/review-blob-hydration.test.ts partial-clone hydration cases, which returned hydrated: false / blobs: 0 in this environment. The same failures reproduce when that file is run directly outside the new runner; this PR does not modify that test or hydration code. The current-head hosted pnpm check passed: https://github.com/openclaw/clawsweeper/actions/runs/31464912192/job/93695706262.
• focused autoreview before commit: clean; TruffleHog clean
• branch autoreview against origin/main after commit: clean; TruffleHog clean

Scope boundary

This change affects only test fixture and test-runner lifecycle. It does not change OpenClaw Bay state, queues, sweep workflows, publication behavior, or dashboard contracts. It does not downgrade pnpm and does not delete host-level pnpm, npm, or Docker caches.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[clawsweeper]

Codex review: needs maintainer review before merge. Reviewed August 30, 2026, 5:02 AM ET / 09:02 UTC.

ClawSweeper review

What this changes

The PR isolates named Node test runs in a per-run temporary root, removes that root after completion or signals, reaps stale roots, and converts validation-test fixtures to automatic cleanup.

Merge readiness

⚠️ Ready for maintainer review - 3 items remain

Keep open for maintainer handling: this MEMBER-authored PR remains a coherent, unmerged test-lifecycle change, but GitHub reports a dirty merge state. Current main still has the pre-isolation test runner, so the central change is not already implemented there; rebase and refresh the exact-head proof before landing.

Priority: P3
Reviewed head: a602f2ba67f9db0b25c604324cacd307a1075ed8
Owner decision: Required. See Decision needed.

Review scores

Measure	Result	What it means
Overall readiness	🐚 platinum hermit (4/6)	The patch is coherent and accompanied by useful lifecycle evidence, with merge readiness limited by the unresolved dirty merge state rather than an identified code defect.
Proof confidence	🌊 off-meta tidepool	Not applicable: This MEMBER-authored PR is exempt from the external-contributor proof gate; nevertheless, the reviewed body supplies head-specific real CLI evidence for FIFO, assertion-failure, stale-root, SIGTERM, and repeated canonical validation cleanup. That evidence must be refreshed after rebase because GitHub reports a dirty merge state.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Verification

Check	Result	Evidence
Real behavior	Not applicable	Not applicable: This MEMBER-authored PR is exempt from the external-contributor proof gate; nevertheless, the reviewed body supplies head-specific real CLI evidence for FIFO, assertion-failure, stale-root, SIGTERM, and repeated canonical validation cleanup. That evidence must be refreshed after rebase because GitHub reports a dirty merge state.
Evidence reviewed	6 items	Current-main comparison: The fetched current-main runner still uses the earlier direct child-process invocation and has no per-run temporary-root lifecycle; the proposed runner isolation is therefore not already on main. Introduced implementation: The pinned PR delta changes the runner, its lifecycle tests, validation-fixture cleanup, package command routing, and the Windows CI invocation. Behavior evidence supplied at the reviewed head: The PR body documents a real CLI lifecycle run, FIFO cleanup, stale-root cleanup, SIGTERM cleanup, and two canonical target-validation runs at the reviewed head; it also records no OpenClaw Bay impact.
Findings	None	None.
Security	None	None.

How this fits together

ClawSweeper’s Node test runner starts unit and repair tests that validate automation and repair-lane behavior. Named commands enter the runner, which launches child test processes that create temporary filesystem fixtures and feeds their result back to CI or local validation.

flowchart LR
  A[Named test command] --> B[Test runner]
  B --> C[Per-run temporary root]
  C --> D[Node test process]
  D --> E[Temporary fixtures]
  D --> F[Exit or termination signal]
  F --> G[Cleanup and stale-root reaper]
  G --> H[Clean temporary area]

Loading

Decision needed

Question	Recommendation
Should this MEMBER-authored test-runner cleanup be rebased onto current main and revalidated as a landing candidate?	Rebase and refresh proof: Resolve the current-main conflicts, then rerun the lifecycle and canonical target-validation proof against the resulting head.

Why: The patch is coherent, but the dirty merge state prevents this review from determining the behavior of the resolved merge, and repository policy reserves maintainer-authored work for explicit judgment.

Before merge

• Resolve merge risk (P1) - Merging cannot currently be evaluated as the tested branch because GitHub reports a dirty merge state; conflict resolution could alter the runner or validation-fixture behavior.
• Complete next step (P2) - No discrete repair finding was identified; a maintainer must decide whether to rebase and retain this MEMBER-authored change, then refresh proof on the resolved branch.

Agent review details

Security

None.

Review metrics

Metric	Value	Why it matters
Patch scope	8 files affected; 342 added, 111 removed	The change spans the shared test runner, CI command routing, and high-volume validation fixtures, so the resolved merge needs focused lifecycle proof.

Merge-risk options

Maintainer options:

1. Rebase and revalidate the resolved runner (recommended)
   Resolve the dirty merge state and repeat the documented FIFO, signal, stale-root, and canonical validation-target proof on the resulting head before merge.

Technical review

Best possible solution:

Rebase the focused test-lifecycle change onto current main, preserve the per-run cleanup contract, and refresh the documented lifecycle proof against the resolved head before merge.

Do we have a high-confidence way to reproduce the issue?

Yes, source and the submitted lifecycle fixtures define a focused reproduction path for temporary-root cleanup, but no new execution was performed in this read-only review and the current merge result is unavailable.

Is this the best way to solve the issue?

Unclear until rebase: isolating the shared runner and using per-test cleanup is a maintainable direction, but the dirty merge must be resolved before the exact landing behavior can be confirmed.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against fa535d54a032.

Labels

Label changes:

• add rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
• remove rating: 🦐 gold shrimp: Current PR rating is rating: 🐚 platinum hermit, so this older rating label is no longer current.

Label justifications:

• P3: This is test and CI hygiene with no demonstrated production-user regression.
• merge-risk: 🚨 automation: The PR changes the shared Node test command and a Windows CI invocation, so a resolved merge could affect validation automation despite passing checks on the old base.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🌊 off-meta tidepool and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Not applicable: This MEMBER-authored PR is exempt from the external-contributor proof gate; nevertheless, the reviewed body supplies head-specific real CLI evidence for FIFO, assertion-failure, stale-root, SIGTERM, and repeated canonical validation cleanup. That evidence must be refreshed after rebase because GitHub reports a dirty merge state.

Evidence

What I checked:

• Current-main comparison: The fetched current-main runner still uses the earlier direct child-process invocation and has no per-run temporary-root lifecycle; the proposed runner isolation is therefore not already on main. (scripts/run-node-tests.mjs:132, fa535d54a032)
• Introduced implementation: The pinned PR delta changes the runner, its lifecycle tests, validation-fixture cleanup, package command routing, and the Windows CI invocation. (scripts/run-node-tests.mjs:144, a602f2ba67f9)
• Behavior evidence supplied at the reviewed head: The PR body documents a real CLI lifecycle run, FIFO cleanup, stale-root cleanup, SIGTERM cleanup, and two canonical target-validation runs at the reviewed head; it also records no OpenClaw Bay impact. (test/test-temp-root-runner.test.ts:12, a602f2ba67f9)
• Current merge state: GitHub reports the PR as dirty, so the supplied old-base proof cannot establish behavior of the actual merged result until the branch is rebased and re-reviewed.
• Feature history: The runner’s earlier implementation dates to the dedicated heavy-runner commit; subsequent repair-validation maintenance is concentrated in the same area. The PR’s two commits are branch-only and are not used for historical ownership attribution. (scripts/run-node-tests.mjs:20, 6d5ce41fc489)
• Repository policy: The full target policy requires focused behavior proof and a fresh review for the current head before landing code-bearing changes; it also treats maintainer-authored items as non-closeable. (AGENTS.md:58, a602f2ba67f9)

Likely related people:

• steipete: Suggested for follow-up; no historical authorship or introduction is verified. (role: unverified routing candidate; confidence: low)

Rank-up moves

Optional improvements that raise the rating; they are not merge blockers.

• Rebase onto current main and update the lifecycle proof for the resolved head.

Rating scale

Score	Internal tier	Crab rank	Meaning
6/6	S	🦀 challenger crab	Exceptional readiness
5/6	A	🦞 diamond lobster	Very strong readiness
4/6	B	🐚 platinum hermit	Good normal PR; ordinary maintainer review
3/6	C	🦐 gold shrimp	Useful, but confidence is limited
2/6	D	🦪 silver shellfish	Proof or implementation needs work
1/6	F	🧂 unranked krab	Not merge-ready
N/A	NA	🌊 off-meta tidepool	Rating does not apply

Overall follows the weaker of proof and patch quality.
Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

Workflow

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

History

Review history (9 earlier review cycles; latest 8 shown)

• reviewed 2026-08-10T15:50:37.034Z sha 1aa53a6 :: needs maintainer review before merge. :: none
• reviewed 2026-08-11T00:28:19.431Z sha 1aa53a6 :: needs maintainer review before merge. :: none
• reviewed 2026-08-11T06:29:52.041Z sha a602f2b :: found issues before merge. :: [P2] Refresh proof for the latest runner invocation
• reviewed 2026-08-11T06:36:47.348Z sha a602f2b :: needs maintainer review before merge. :: none
• reviewed 2026-08-11T12:11:07.285Z sha a602f2b :: needs maintainer review before merge. :: none
• reviewed 2026-08-12T06:24:31.715Z sha a602f2b :: needs maintainer review before merge. :: none
• reviewed 2026-08-12T23:32:24.400Z sha a602f2b :: needs maintainer review before merge. :: none
• reviewed 2026-08-23T15:00:03.905Z sha a602f2b :: needs maintainer review before merge. :: none

[hxy91819]

Local black-box validation completed against commit 1aa53a6a09.

Environment:

• Linux, Node v24.15.0
• Corepack resolved the repository pin to pnpm 11.10.0

Evidence:

1. Lifecycle and runner behavior

   Command:

   node --test test/run-node-tests.test.ts test/test-temp-root-runner.test.ts
   

   Result: 10 passed, 0 failed. This exercises successful FIFO cleanup, assertion-failure cleanup, stale-root TTL cleanup, and SIGTERM cleanup through the real CLI runner. Matching run-root and validation-directory counts were unchanged before and after.

2. Incident-equivalent FIFO fixture

   Command:

   node --test --test-name-pattern "workspace preflight rejects non-regular package metadata without blocking" test/repair/target-validation.test.ts
   

   Result: 1 passed, 0 failed. The FIFO preflight itself completed in 28.8 ms, below its 1-second behavior bound, and left no matching validation directory behind.

3. Repeated full target-validation runs

   Command, executed twice:

   node --test --test-reporter=tap test/repair/target-validation.test.ts
   

   Results:

   • run 1: 198 passed, 3 skipped, 0 failed; 132.9 s
   • run 2: 198 passed, 3 skipped, 0 failed; 130.6 s
   • the total count of matching top-level ClawSweeper temporary directories was identical before and after each run
   • no clawsweeper-validation-* directory remained after either run

No pre-existing historical temporary directories were deleted during this validation.

The PR CI pnpm check also passed: https://github.com/openclaw/clawsweeper/actions/runs/31405434604/job/93510465585