[Chisanan232]

Description

test/integration/test_native_core_runtime.py skips itself unless AAASM_RUN_NATIVE_CORE_TESTS=1 and no workflow set it, so the native-core runtime tests never ran in CI (AAASM-6359; the two stale tests were already fixed in #363). This wires them into the existing native-core-build job, which already builds the PyO3 core with maturin.

• 🔧 (ci): the job sets the gate variable and runs only that file. -o addopts="" drops pytest.ini's --reruns 1 (a retry would hide a flaky timing assertion). A skip exits 0, so the job also fails if the junit report shows any skipped test; otherwise a removed or typo'd variable would be a silent pass. The file is added to the PR path filter.
• ✅ (test): test_send_event_is_non_blocking flaked while validating this: 2 of 14 runs under CPU contention measured 255-286 ms against its 250 ms budget (steady about 54 ms idle). Budget is now 1 s, with the comment updated. This is a weak oracle for "does not block" (it is a wall-clock bound); a structural test with a deliberately slow runtime would be stronger and is out of scope here.

CI-definition change: minimal diff, existing action pins and permissions: contents: read untouched, no new secrets.

Type of Change

• ✨ New feature
• 🔧 Bug fix
• ♻️ Refactoring
• 🍀 Performance improvement
• 📚 Documentation update
• 🚀 Release

(Configuration / CI change; the template has no such box.)

Breaking Changes

• No
• Yes (please describe below)

Related Issues

• Related JIRA ticket: AAASM-6359
• Related GitHub issues: follows [AAASM-6359] ✅ (test): Fix stale env-gated native core runtime tests #363

Testing

• Unit tests added/updated
• Integration tests added/updated
• Manual testing performed
• No tests required (explain why)

Local, with the exact job steps against a freshly maturin-built core (venv and target dir under /private/tmp): 13 passed, junit skipped="0" grep exit 0. Gate off: 13 skipped and the grep exits 1, so the job would fail rather than pass silently. Mutation: changing the expected decision in test_query_policy_returns_deny_when_runtime_denies from deny to allow fails that test (1 failed, 12 passed); restored byte-identical. After the budget change, 10 of 10 gated runs passed. Not verified: that the new CI step is green on a GitHub runner (checked after this PR is up), and that uv run --with pytest keeps the maturin-installed _core in the CI venv.

Checklist

• Code follows project style guidelines
• Self-review completed
• Comments added for complex logic
• Documentation updated if needed
• All tests passing

🤖 Generated with Claude Code

https://claude.ai/code/session_01JvXQ9VcS72E7DREhsgjyHV

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[Chisanan232]

Review record (independent AI review, author-account posting) — NOT a GitHub approval.

Reviewed head: e591122b5d9dfda644d41a4de40534b63f827d03 (AAASM-6359). Independent review of 74752f6 = APPROVE-WITH-NOTES; I read the one later commit (e591122b5, +2 lines: agent_assembly/core/** added to the pull_request paths), which addresses the review's trigger-gap note.

Verified (independent reviewer, from CI log and a local pytest 9.1.1 run):

• Permissions stay contents: read; no secrets; the new run block has no ${{ }}; action pins unchanged.
• The skipped-count guard fails closed: a missing junit attribute, a removed gate variable, a missing _core, a collection error and zero collected tests all fail the job. (An xfail would false-fail; none exist in the file.)
• uv run --with pytest did not disturb the maturin-installed _core: CI log shows collected 13 items, 13 passed, --reruns dropped.
• On my side: gate off gives 13 skipped and the grep exits 1; a mutated deny→allow assertion fails that test; 10 of 10 gated runs pass after the budget change.

CI on this head: build-native-core = success (new step included), CI Success, CodeQL, unit/integration, pip-audit, SonarCloud, codecov/patch all success; no failing check-run.

Notes (none blocking): (1) no push to main trigger, so nothing re-runs this after merge; (2) the 1 s budget only catches a regression to blocking if a blocking send costs more than about 2 ms per event, and the PR says so; (3) --with pytest is redundant (harmless); (4) the 2-of-14 flake statistics are my local measurements.
Not verified: behaviour on a runner under heavy load.
Merge: author is the owner account, so no formal self-approval; needs the owner's merge (head-pinned merge commit).