[HAYDEN-OAI]

Summary

Do not run the test wrapper's port-wide cleanup when the mock launcher returns an ordinary startup failure.

scripts/test installs an EXIT trap that finds processes on port 4010 and terminates them. If an unrelated local service owns that port but fails the health probe, the attempted mock startup fails with EADDRINUSE. The launcher correctly cleans up its own child, but the wrapper then terminates the unrelated service.

Keep the existing EXIT trap armed during startup and successful test execution. On a returned launcher failure, capture its exit status and disarm the wrapper's trap only for ordinary statuses up to 128, then exit with the same status. The launcher retains responsibility for its ordinary failure cleanup. For signal-style statuses above 128, retain the wrapper's fallback because a killed launcher may not have run its own EXIT trap.

The production change is limited to the handwritten test wrapper. The mock launcher, health checks, successful-run cleanup, suite routing, SDK source, dependencies, and generation metadata remain unchanged.

Regression coverage

Seven POSIX subprocess cases execute a byte-copied scripts/test with real fixture-owned processes and isolated health/port-lookup commands:

• Failed startup preserves the unrelated process and exact exit status 23, without running Jest or port cleanup.
• Successful startup terminates its mock after both successful and failing Jest runs, preserving exit statuses 0 and 17.
• A pre-existing healthy mock is reused and preserved after either Jest result.
• SIGTERM sent only to the wrapper during startup still terminates its owned mock.
• SIGKILL sent only to the launcher still lets the wrapper terminate the server and preserve exit status 137.

Fixtures record and clean up both server and launcher PIDs; these unit tests do not bind the shared mock port or act on real user processes.

Verification

• The original code fails the unrelated-process regression. A first trap-move implementation failed the new startup-interruption regression; the final failure-only disarm passes both.
• All seven final cases pass on exact Node 22.0.0, 24.19.0, and 26.7.0.
• The new suite and neighboring test-wrapper, mock-launcher, and mock-fetch suites pass 46/46 on Node 24.19.0.
• The full canonical generated suite passes on Node 24.19.0 with the installed, verified Steady 0.22.2 / Deno 2.7.11 toolchain: 82 suites and 556 tests pass, with one suite and two tests skipped; all 12 snapshots pass. The wrapper exits 0, its owned server PID exits, and port 4010 is released afterward.
• Independent native Bash/curl/lsof fixtures use the actual test and mock scripts with synthetic loopback services. On all three runtimes, failed startup now leaves the unrelated listener alive and healthy-server reuse remains unchanged. A fresh before/after wrapper-only SIGTERM comparison confirms startup cleanup is preserved.
• An additional native baseline/reviewed/final comparison reproduces the launcher-only SIGKILL leak, verifies its correction, and checks ordinary statuses 23 and 128, exact exit 137, wrapper SIGTERM, successful/failing Jest runs, and healthy reuse. These fixtures use owned ephemeral ports and clean up their exact PIDs.
• Normal full-project tsc --noEmit, pinned Oxfmt 0.62.0 and Oxlint 1.79.0, Bash 3 syntax, and whitespace checks pass.

Local handwritten tests use Vitest 4.1.10; repository CI runs the pinned 4.1.11 environment. The new process-signal regressions are POSIX-only; existing cross-platform coverage is unchanged.

Adversarial review and scope

Adversarial review caught that simply moving the EXIT trap after startup regressed interruption cleanup. The implementation was tightened to disarm cleanup only on a returned startup failure, a dedicated regression was added, and independent reviews and native-process checks were repeated against the final bytes.

PR review then identified that a launcher killed with SIGKILL cannot perform its own cleanup. A seventh regression failed on that revision. The follow-up retains the wrapper's fallback for signal-style statuses and repeats the lifecycle, native-process, full generated-suite, type, and lint checks.

This is a finite ordinary-startup-failure correction, not a general process-ownership redesign. A status above 128 is treated conservatively, not as proof of a signal; an explicit high exit code also retains the legacy fallback. Existing port-based cleanup after successful startup and existing signal behavior remain in place. The PR does not claim to solve every possible later port-reuse race.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-05T02:23:26.286535Z	9e9feb8	New commits
🔒 Security Review	✅ Completed 2026-09-05T02:24:22.351712Z	9e9feb8	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[openai-sdks]

OkTest Summary

✅ 237/237 SDK tests passed in 10.618s for Node SDK PR #2622.

Test results — 42 files

Test	Result	Time
tests/chat-completions-complex-body.test.ts	✅ Passed	222ms
tests/chat-completions-create.test.ts	✅ Passed	206ms
tests/chat-completions-stream.test.ts	✅ Passed	226ms
tests/files-content-binary.test.ts	✅ Passed	147ms
tests/files-create-multipart.test.ts	✅ Passed	157ms
tests/files-list-pagination.test.ts	✅ Passed	201ms
tests/initialize-config.test.ts	✅ Passed	234ms
tests/instance-isolation.test.ts	✅ Passed	142ms
tests/models-list.test.ts	✅ Passed	140ms
tests/responses-background-lifecycle.test.ts	✅ Passed	225ms
tests/responses-body-method-errors.test.ts	✅ Passed	311ms
tests/responses-cancel-timeout.test.ts	✅ Passed	203ms
tests/responses-cancel.test.ts	✅ Passed	245ms
tests/responses-compact-retries.test.ts	✅ Passed	281ms
tests/responses-compact.test.ts	✅ Passed	250ms
tests/responses-create-advanced-stream.test.ts	✅ Passed	153ms
tests/responses-create-advanced.test.ts	✅ Passed	184ms
tests/responses-create-disconnect.test.ts	✅ Passed	1.055s
tests/responses-create-errors.test.ts	✅ Passed	234ms
tests/responses-create-malformed-api-responses.test.ts	✅ Passed	184ms
tests/responses-create-retries.test.ts	✅ Passed	324ms
tests/responses-create-stream-failures.test.ts	✅ Passed	205ms
tests/responses-create-stream-timeout.test.ts	✅ Passed	2.156s
tests/responses-create-stream-wire.test.ts	✅ Passed	2.465s
tests/responses-create-stream.test.ts	✅ Passed	101ms
tests/responses-create-terminal-states.test.ts	✅ Passed	213ms
tests/responses-create-timeout.test.ts	✅ Passed	229ms
tests/responses-create.test.ts	✅ Passed	293ms
tests/responses-delete.test.ts	✅ Passed	238ms
tests/responses-input-items-errors.test.ts	✅ Passed	238ms
tests/responses-input-items-list.test.ts	✅ Passed	193ms
tests/responses-input-items-options.test.ts	✅ Passed	130ms
tests/responses-input-tokens-count-timeout.test.ts	✅ Passed	236ms
tests/responses-input-tokens-count.test.ts	✅ Passed	262ms
tests/responses-malformed-inputs.test.ts	✅ Passed	2.219s
tests/responses-not-found-errors.test.ts	✅ Passed	282ms
tests/responses-parse.test.ts	✅ Passed	198ms
tests/responses-retrieve-retries.test.ts	✅ Passed	300ms
tests/responses-retrieve.test.ts	✅ Passed	192ms
tests/responses-stored-method-errors.test.ts	✅ Passed	604ms
tests/retry-behavior.test.ts	✅ Passed	2.91s
tests/sdk-error-shape.test.ts	✅ Passed	334ms

View OkTest run #33938856545

_{SDK merge (83f57c9963b1) · head (9e9feb892dc8) · base (1850d58b0f0c) · OkTest (2b1bdfd25e98)}

[github-actions]

Castiron custom code

✅ No new custom-code files detected.

32 mixed files remain; 0 existing customizations changed.

Compared ef39397e33c5 → 9e9feb892dc8. Generated baselines verified.

32 existing customizations unchanged

• api.md
• scripts/castiron/README.md
• scripts/castiron/custom_code_report.py
• scripts/castiron/test_custom_code_report.py
• src/client.ts
• src/resources/audio/transcriptions.ts
• src/resources/audio/translations.ts
• src/resources/beta/assistants.ts
• src/resources/beta/beta.ts
• src/resources/beta/index.ts
• src/resources/beta/responses/internal-base.ts
• src/resources/beta/responses/responses.ts
• src/resources/beta/threads/index.ts
• src/resources/beta/threads/runs/index.ts
• src/resources/beta/threads/runs/runs.ts
• src/resources/beta/threads/threads.ts
• src/resources/chat/completions/completions.ts
• src/resources/chat/completions/index.ts
• src/resources/conversations/index.ts
• src/resources/embeddings.ts
• src/resources/files.ts
• src/resources/fine-tuning/checkpoints/permissions.ts
• src/resources/images.ts
• src/resources/responses/internal-base.ts
• src/resources/responses/responses.ts
• src/resources/skills/skills.ts
• src/resources/skills/versions/versions.ts
• src/resources/vector-stores/file-batches.ts
• src/resources/vector-stores/files.ts
• src/resources/webhooks/index.ts
• src/resources/webhooks/webhooks.ts
• tests/lib/data-residency.test.ts

A changed generated baseline means this report cannot reliably identify which handwritten lines changed.

Inspect the custom-code diff

Download the exact patch produced by this run (requires repository access):

gh run download 33938883542 --repo openai/openai-node \
  --name castiron-custom-code-33938883542-1 --dir /tmp/castiron-custom-code-33938883542-1
git apply --stat /tmp/castiron-custom-code-33938883542-1/custom-code.patch
cat /tmp/castiron-custom-code-33938883542-1/custom-code.patch

Or reproduce it from an SDK checkout containing the vendored reporter:

git fetch --no-tags origin 1850d58b0f0cf669aa7f8cb8ba8c06a9ea857fbd 9e9feb892dc8ee33d06ca052d6f58e2357dff636
python3 scripts/castiron/custom_code_report.py report \
  --base 1850d58b0f0cf669aa7f8cb8ba8c06a9ea857fbd \
  --head 9e9feb892dc8ee33d06ca052d6f58e2357dff636 --fetch --require-head-hash --public \
  --out /tmp/castiron-custom-code-9e9feb892dc8
cat /tmp/castiron-custom-code-9e9feb892dc8/custom-code.patch

This is the current full custom patch for mixed files, not an attribution of only the handwritten lines changed by this PR.

Full report and patch

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 12ab10c39d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[jbeckwith-oai]

Reviewed exact head 12ab10c against the requested outcome: a returned mock-launcher failure must preserve the unrelated port owner and the launcher status, while successful runs and startup interruption retain the existing cleanup trap. The failure-only trap disarm is narrowly placed after the launcher returns, preserves set -e behavior and the exact status, and does not alter healthy-server reuse or later generated-test cleanup. I ran the new six-case process suite on local Node 22.15.1, 24.16.0, and 26.2.0; the four related wrapper/launcher/fetch suites pass 45/45, and Oxfmt, Oxlint, and bash -n pass. No blocking findings.