[HAYDEN-OAI]

Problem

The compiler CLI resolves configured project globs from the selected tsc-multi.json directory, but passes the resulting relative paths to workers running in the caller's working directory. With a nested or separately located config, a build can fail to find its project or silently compile a same-named project from the wrong directory.

The regression reproduces the silent case through the real CLI and TypeScript compiler. A nested config selects its own project, while a different project with the same relative name exists beside the caller. On main, the build exits successfully but emits the caller-directory decoy instead of the configured project.

Change

Request absolute output paths from the existing fast-glob call. Each match then retains the directory it was resolved against when it reaches the compiler worker.

Configured projects remain relative to the selected config file. Positional CLI projects remain relative to the caller and continue to override configured projects. Worker working directories, compiler resolution, glob patterns, scheduling, and configuration defaults are unchanged.

The production change is one line in the handwritten config loader, with a separate four-case real-CLI regression file. Both paths are absent from the verified pure-generated snapshot d223f5ab382c6a7d64f3863e75865624cbefad21; SDK maintainers own this vendored build tool. No dependency, generated helper, provenance, license, SDK source, or published SDK API changes are included. No open PR covers this issue.

Verification

• The exact final four-case regression on unchanged main 124b6525497a93315f8563d8e4e3d552cbb0518f has two expected wrong-project failures and two passing controls. Both nested directory and explicit tsconfig-file selections emit the decoy before the fix.
• All four cases pass after the fix on Node 22.22.3, 24.19.0, and 26.7.0. Controls verify default same-directory configuration and positional-project precedence.
• The new regression and existing compiler-worker exit tests pass all 10 cases on Node 24.19.0.
• All 7,671 handwritten tests pass across 199 files on Node 24.19.0 through the canonical test script with two workers.
• Canonical lint, full repository TypeScript checking, targeted checking and formatting of the vendored config loader, and the canonical build pass with the pinned tools.
• The complete default dist output is byte-for-byte identical to the verified build of unchanged main, including CommonJS/ESM output and source maps.

Adversarial review

Reviewed config-directory versus caller-directory ownership, directory and file project patterns, positional precedence, worker consumers, fixture cleanup, and generated ownership. Independent review found no remaining required changes. A test-only typing adjustment was rechecked against both main and the fix without weakening its assertions.

The fix uses the existing glob library's output option and preserves process working directories. All compiler fixtures are synthetic; no live API calls or dependency installations are needed by the regressions.

[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 2026-09-05T19:08:24.899520Z	659a306	PR opened
🔒 Security Review	✅ Completed 2026-09-05T19:09:42.134941Z	659a306	PR opened

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 8.485s for Node SDK PR #2637.

Test results — 42 files

Test	Result	Time
tests/chat-completions-complex-body.test.ts	✅ Passed	92ms
tests/chat-completions-create.test.ts	✅ Passed	122ms
tests/chat-completions-stream.test.ts	✅ Passed	84ms
tests/files-content-binary.test.ts	✅ Passed	97ms
tests/files-create-multipart.test.ts	✅ Passed	214ms
tests/files-list-pagination.test.ts	✅ Passed	103ms
tests/initialize-config.test.ts	✅ Passed	91ms
tests/instance-isolation.test.ts	✅ Passed	88ms
tests/models-list.test.ts	✅ Passed	87ms
tests/responses-background-lifecycle.test.ts	✅ Passed	152ms
tests/responses-body-method-errors.test.ts	✅ Passed	247ms
tests/responses-cancel-timeout.test.ts	✅ Passed	202ms
tests/responses-cancel.test.ts	✅ Passed	176ms
tests/responses-compact-retries.test.ts	✅ Passed	174ms
tests/responses-compact.test.ts	✅ Passed	149ms
tests/responses-create-advanced-stream.test.ts	✅ Passed	112ms
tests/responses-create-advanced.test.ts	✅ Passed	106ms
tests/responses-create-disconnect.test.ts	✅ Passed	1.025s
tests/responses-create-errors.test.ts	✅ Passed	250ms
tests/responses-create-malformed-api-responses.test.ts	✅ Passed	99ms
tests/responses-create-retries.test.ts	✅ Passed	188ms
tests/responses-create-stream-failures.test.ts	✅ Passed	162ms
tests/responses-create-stream-timeout.test.ts	✅ Passed	2.115s
tests/responses-create-stream-wire.test.ts	✅ Passed	1.965s
tests/responses-create-stream.test.ts	✅ Passed	58ms
tests/responses-create-terminal-states.test.ts	✅ Passed	142ms
tests/responses-create-timeout.test.ts	✅ Passed	184ms
tests/responses-create.test.ts	✅ Passed	124ms
tests/responses-delete.test.ts	✅ Passed	167ms
tests/responses-input-items-errors.test.ts	✅ Passed	126ms
tests/responses-input-items-list.test.ts	✅ Passed	131ms
tests/responses-input-items-options.test.ts	✅ Passed	84ms
tests/responses-input-tokens-count-timeout.test.ts	✅ Passed	196ms
tests/responses-input-tokens-count.test.ts	✅ Passed	166ms
tests/responses-malformed-inputs.test.ts	✅ Passed	1.502s
tests/responses-not-found-errors.test.ts	✅ Passed	163ms
tests/responses-parse.test.ts	✅ Passed	118ms
tests/responses-retrieve-retries.test.ts	✅ Passed	162ms
tests/responses-retrieve.test.ts	✅ Passed	170ms
tests/responses-stored-method-errors.test.ts	✅ Passed	398ms
tests/retry-behavior.test.ts	✅ Passed	3.065s
tests/sdk-error-shape.test.ts	✅ Passed	245ms

View OkTest run #33986106256

_{SDK merge (e7a19cdeed4a) · head (659a30620f65) · base (124b6525497a) · OkTest (2b1bdfd25e98)}

[github-actions]

Castiron custom code

✅ No new custom-code files detected.

32 mixed files remain; 0 existing customizations changed.

Compared 124b6525497a → 659a30620f65. 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 33986122541 --repo openai/openai-node \
  --name castiron-custom-code-33986122541-1 --dir /tmp/castiron-custom-code-33986122541-1
git apply --stat /tmp/castiron-custom-code-33986122541-1/custom-code.patch
cat /tmp/castiron-custom-code-33986122541-1/custom-code.patch

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

git fetch --no-tags origin 124b6525497a93315f8563d8e4e3d552cbb0518f 659a30620f655809df4f222ed94d51cf2055ca4a
python3 scripts/castiron/custom_code_report.py report \
  --base 124b6525497a93315f8563d8e4e3d552cbb0518f \
  --head 659a30620f655809df4f222ed94d51cf2055ca4a --fetch --require-head-hash --public \
  --out /tmp/castiron-custom-code-659a30620f65
cat /tmp/castiron-custom-code-659a30620f65/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

[markstuart-oai]

Reviewed the full change and the config loader, CLI precedence, worker launch, and TypeScript builder/transpile consumers. No substantive findings. Requesting absolute paths from the existing glob resolver puts the invariant at the right boundary and avoids adding compensating path logic to workers. The focused regression covers both directory and explicit tsconfig selections, default configuration, and positional-project precedence.

Independently verified on Node 24.19.0: all 10 path-resolution and worker-exit tests pass through ./scripts/test. The exact four-case regression on freshly fetched main (aa08b449) reproduces both wrong-project failures, with both controls passing. git diff --check passes. I did not rerun the full build, complete suite, or cross-runtime matrix.

[jbeckwith-oai]

Reviewed the exact two-file head and traced project paths from config selection through fast-glob, CLI positional override, and the compiler worker boundary. Returning absolute matches preserves the config directory as the resolution owner without changing glob patterns, worker cwd, positional-project precedence, or default same-directory behavior. The real CLI regression specifically guards the silent same-named decoy case for both directory and tsconfig-file selections and includes the relevant controls. Review threads are clear and the full Node test matrix is green. No blocking findings.