[dpiet-oai]

Changes being requested

Run the three fractional source-deadline regressions in one child interpreter per sync/async mode, reducing interpreter launches from six to two. Each scenario gets a fresh transcript and event loop; scenario labels and tracebacks remain visible, and the parent retains a five-second timeout.

Additional context & links

• Preserves all three scenarios and their segment and closure assertions. Pytest collects two nodes instead of six; an ordinary failure does not skip later scenarios, but a hung child cannot reach later scenarios in that mode.
• Local Python 3.10.16 / Pydantic 2.12.5, DEFER_PYDANTIC_BUILD=false, serial pytest, three runs per revision: median process wall 10.72s → 6.62s; median summed test-call time 8.63s → 3.89s. Local cache conditions do not measure CI gate latency or billed runner time.
• Validation: full file 104 passed on Python 3.10 with Pydantic v1 and v2 and on Python 3.14 with Pydantic v2; Ruff/format passed. Python 3.14 required an explicit repository PYTHONPATH because this host sets PYTHONSAFEPATH=1. Injected failure continued through later scenarios; injected hang reported the scenario and stopped after 5.02s.

[github-actions]

Castiron custom code

Evaluated main: 43045fcaf91aec08571e54c57706597cefd32d45.

Trusted main policy: 43045fcaf91aec08571e54c57706597cefd32d45.

✅ No new custom-code files detected.

0 mixed files remain; 0 existing customizations changed.

Compared 43045fcaf91a → 8844c138e045. Generated baselines verified.

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 40029439510 --repo openai/openai-python \
  --name castiron-custom-code-40029439510-1 --dir /tmp/castiron-custom-code-40029439510-1
git apply --stat /tmp/castiron-custom-code-40029439510-1/custom-code.patch
cat /tmp/castiron-custom-code-40029439510-1/custom-code.patch

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

git fetch --no-tags origin 43045fcaf91aec08571e54c57706597cefd32d45 8844c138e045a490cec5277a31444912d16a6fd4
python3 scripts/castiron/custom_code_report.py report \
  --base 43045fcaf91aec08571e54c57706597cefd32d45 \
  --head 8844c138e045a490cec5277a31444912d16a6fd4 --target-branch main --fetch --require-head-hash --public \
  --out /tmp/castiron-custom-code-8844c138e045
cat /tmp/castiron-custom-code-8844c138e045/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

[marcuswood-oai]

LGTM 👍 No changes requested.

Failure-continuation and five-second timeout probes passed with a stub transcript. The SDK test and lint jobs were skipped in this draft’s CI run; I did not run the full SDK suite locally.

---

Reviewed: 8844c138 · Base: 43045fca · sdk-pr-review

[markstuart-oai]

Reviewed 8844c138e045a490cec5277a31444912d16a6fd4. No blocking findings. All six sync/async scenarios keep their segment and closure assertions. Each scenario gets a fresh transcript and event loop, failures remain visible, and the parent still bounds a hung child.

Source-only review; I did not run candidate code or tests. Hosted checks: 13 skipped, 22 success.