test(mcp-gateway): scope stop and kill paths by platform

[bolichen97]

Summary

• scope the two raw POSIX killpg / init-PID assertions to POSIX runners
• keep the cross-platform quarantine and async tree-kill mechanism assertions running on Windows
• exercise abort acknowledgements through the production transport so the same tests use Unix sockets or Windows named pipes

Determinism evidence

• explicit selection of the 8 affected nodes on Windows failed 8/8 because it bypassed the repository's whole-file Windows collection exclusion
• after the fix, the combined strict selection passes 15 tests and skips only the 2 explicitly POSIX assertions
• Windows alternatives independently cover taskkill error mapping, async offload, missing-taskkill failure, and real named-pipe round trips
• no retries, sleeps, timeout changes, xfails, or warning filters were added

Validation

• targeted strict RuntimeWarning/PytestUnraisable run: 15 passed, 2 POSIX-only skips
• flake8, isort check, and git diff --check passed
• the whole historical test file remains in the Black baseline on unmodified main, so it was not reformatted

Overlap audit

• refactor: extract subagent lifecycle boundaries #5278 touches the same test file only in TestConservativeShutdown; the changed regions and semantics are disjoint
• open platform-compat PRs do not own these test-scoping or transport-roundtrip changes

[github-actions]

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of cfab972025d9392aebc7a5e8c9462bd900a948cb and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] cfab972

False positive or not applicable? A repository writer can comment:
/ai-review override gpt cfab972025d9392aebc7a5e8c9462bd900a948cb: <one-sentence reason>

[github-actions]

Design Review (Fable 5) — ✅ PASS

Design-level review of cfab972025d9392aebc7a5e8c9462bd900a948cb — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Test-only platform scoping that follows the repo's platform_compat/transport conventions exactly; assertions now track the production kill path and transport seam.

[DESIGN-REVIEWED] cfab972

[github-actions]

Opus 4.8 Review — ✅ no blocking findings

Reviewed cfab972025d9392aebc7a5e8c9462bd900a948cb — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] cfab972

Verdict parsed from the review's SHA-scoped output markers for commit cfab972025d9392aebc7a5e8c9462bd900a948cb.

False positive or not applicable? A repository writer can comment:
/ai-review override fable cfab972025d9392aebc7a5e8c9462bd900a948cb: <one-sentence reason>

[dwu96]

Tier 1 auto-approve: test (1 file). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: test — test/test_stop_kill_cancel.py only; POSIX-specific killpg/PID-1 assertions are now skipped on Windows via platform_compat.IS_WINDOWS, the kill-tree patch target and SIGKILL constant are read through platform_compat, and the two fake gatewayd servers go through transport.serve instead of asyncio.start_unix_server. No production code touched.

[chenmingwei23]

Tier 1 auto-approve: test (1 file). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: test-only, scopes stop and kill paths by platform in test_stop_kill_cancel.py.

[iamwhatever]

Tier 1 auto-approve: test (1 file). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: test-only change to test/test_stop_kill_cancel.py -- skips the POSIX killpg/PID-1 assertions on Windows and patches the platform_compat kill seam instead of os.killpg, so no production code is touched.