test: shave loopback timing overhead from remaining slow tests

[bluetoothbot]

Summary

Issue #1707 listed 20 tests over 0.5s. Several of them weren't pinning RFC behavior — they were paying production-grade delays on a 127.0.0.1 fake network. This PR cuts the ones that can be safely shortened and leaves the RFC-pinned ones (response aggregation, duplicate-question suppression, TC-deferral, 4-retry budget) alone.

Fixes #1707

Changes

• Extract the 150-250ms probe collision-avoidance delay in Zeroconf._async_check_service into a new module-level _PROBE_RANDOM_DELAY_INTERVAL constant (src/zeroconf/_core.py). The existing quick_timing fixture now patches it to (1, 5), dropping ~200ms per register_service call on loopback. Production behaviour unchanged.
• Add quick_request_timing to:
  • tests/services/test_info.py::TestServiceInfo::test_get_info_partial
  • tests/test_services.py::ListenerTest::test_integration_with_listener_class
  • tests/test_asyncio.py::test_service_info_async_request
    …so get_service_info calls no longer pay the 200ms _LISTENER_TIME + 20-120ms jitter on loopback.
• Tighten the deliberate forced-timeout in test_service_info_async_request from 1500ms to 300ms — under quick_request_timing the first QM lands in <20ms, so 300ms is comfortable.
• Patch _TASK_AWAIT_TIMEOUT to 50ms in test_shutdown_loop so the inner asyncio.wait on the deliberately never-completing _still_running() task returns quickly instead of blocking the default 1s.

Slow test before / after (local measurements, top of --durations=20 for the issue's list)

Test	Before	After
test_service_info_async_request	2.19s	0.78s
test_get_info_partial	1.50s	1.05s
test_integration_with_listener_class	1.68s	1.16s
test_shutdown_loop	1.15s	0.15s
test_get_info_single	1.01s	0.03s
test_async_zeroconf_service_types	0.52s	0.29s
test_async_service_registration_name_conflict	0.68s	0.09s
test_verify_name_change_with_lots_of_names	0.80s	0.36s
test_name_conflicts	0.47s	0.09s
test_integration_with_listener (types)	1.28s	0.19s
test_integration_with_subtype_and_listener	1.27s	0.20s
test_integration_with_listener_v6_records	1.27s	0.20s
test_asking_qu_questions	0.52s	0.06s

Tests intentionally left at their original runtime because they pin RFC 6762 timing:

• test_response_aggregation_timings / _multiple (§14)
• test_we_try_four_times_with_random_delay (3000ms / 4 retries)
• test_get_info_suppressed_by_question_history (§7.1, _DUPLICATE_QUESTION_INTERVAL)
• test_tc_bit_defers_last_response_missing (§18.5, _TC_DELAY_RANDOM_INTERVAL)

Test plan

• poetry run pytest --timeout=60 -q — 385 passed, 2 skipped, no regressions
• poetry run ruff check + poetry run ruff format --check clean on touched files
• Targeted re-run of the 20 tests from the issue shows the expected drops

---

Generated by Kōan /fix

---

Quality Report

Changes: 6 files changed, 32 insertions(+), 17 deletions(-)

Code scan: clean

Tests: passed (4 PASSED)

Branch hygiene: clean

Generated by Kōan post-mission quality pipeline

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.77%. Comparing base (5444495) to head (c8e442d).
⚠️ Report is 2 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1760   +/-   ##
=======================================
  Coverage   99.77%   99.77%           
=======================================
  Files          33       33           
  Lines        3508     3509    +1     
  Branches      493      493           
=======================================
+ Hits         3500     3501    +1     
  Misses          5        5           
  Partials        3        3           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[codspeed-hq]

Merging this PR will not alter performance

✅ 14 untouched benchmarks

---

_{Comparing bluetoothbot:koan/fix-issue-1707 (c8e442d) with master (5444495)¹}

Footnotes

1. No successful run was found on master (e95297f) during the generation of this report, so 5444495 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Reduces end-to-end runtime of several loopback-based tests by patching production-grade timing delays (probe/jitter/timeouts) down to test-friendly values while keeping RFC-timing tests unchanged.

Changes:

• Extracted the probe collision-avoidance jitter into _PROBE_RANDOM_DELAY_INTERVAL and patched it in the quick_timing fixture.
• Added quick_request_timing to selected tests and shortened an intentional timeout in test_service_info_async_request.
• Reduced shutdown-loop waiting overhead by patching _TASK_AWAIT_TIMEOUT in a loop shutdown test.

Reviewed changes

Copilot reviewed 6 out of 6 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tests/utils/test_asyncio.py	Speeds up shutdown-loop test by shrinking sleeps/timeouts and patching _TASK_AWAIT_TIMEOUT.
tests/test_services.py	Applies quick_request_timing fixture to a listener integration test.
tests/test_asyncio.py	Applies quick_request_timing and reduces a forced timeout in an async service-info request test.
tests/services/test_info.py	Applies quick_request_timing to a partial info lookup test.
tests/conftest.py	Updates quick_timing fixture to also patch the new probe-delay interval.
src/zeroconf/_core.py	Introduces _PROBE_RANDOM_DELAY_INTERVAL and uses it for pre-probe random wait.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[bdraco]

@bluetoothbot rr

[bluetoothbot]

PR Review — test: shave loopback timing overhead from remaining slow tests

Clean test-speedup PR. The _PROBE_RANDOM_DELAY_INTERVAL extraction is the right shape — keeps the production tuple in _core.py, lets quick_timing patch it, and the .pxd doesn't need to change because _core.py isn't in TO_CYTHONIZE. The quick_request_timing rollouts at test_get_info_partial, test_integration_with_listener_class, and test_service_info_async_request are well-justified, and the matching 1500ms→300ms timeout tightening in the TOCTOU race-condition block has a clear new comment explaining the new arithmetic. test_shutdown_loop rewrite is correct: moving runcoro_thread_ready.set() after run_coroutine_threadsafe and replacing the time.sleep(0.1) buffer with a patched _TASK_AWAIT_TIMEOUT is a sound trade. RFC-pinned tests are correctly left alone. Two suggestion-level notes inline; neither blocks merge.

---

🟢 Suggestions

1. Task-scheduling race in _run_coro thread (`tests/utils/test_asyncio.py`, L89-95)

Good fix moving runcoro_thread_ready.set() after run_coroutine_threadsafe(...) — that closes the worst of the race Copilot flagged. One residual concern: run_coroutine_threadsafe only queues call_soon_threadsafe(callback), where callback is what actually does ensure_future(coro, loop=loop). So when the worker thread calls .set(), the concurrent.futures.Future exists but the asyncio Task may not be on the loop yet. If the main thread races into shutdown_loop(loop) before the loop processes the call_soon callback, asyncio.all_tasks(loop) will be empty and the test silently passes a different code path than intended.

If you want to make this fully deterministic, have _still_running() set a second threading.Event at its first yield point and wait on that from the main thread:

task_started = threading.Event()

async def _still_running():
    task_started.set()
    await asyncio.sleep(5)

# main thread
runcoro_thread_ready.wait()
task_started.wait()
with patch.object(aioutils, "_TASK_AWAIT_TIMEOUT", 0.05):
    aioutils.shutdown_loop(loop)

Not a blocker — the original code had the same class of race buffered by time.sleep(0.1), and in practice call_soon_threadsafe wakes the loop fast enough that this almost never fires. Worth a follow-up if the test starts flaking on the 3.14t / PyPy matrix entries.

future = asyncio.run_coroutine_threadsafe(_still_running(), loop)
runcoro_thread_ready.set()

2. Comment can be tightened to one line (`src/zeroconf/_core.py`, L107-110)

Per CLAUDE.md (Code style / Comments): "Default to writing no comments. RFC citations are useful when the reason for a timing constant or framing decision is 'the spec says so' — leave those in." The RFC citation is the load-bearing bit here; the rest restates quick_timing's job. Consider collapsing to:

# RFC 6762 §8.1: random 0-250ms pre-probe delay against thundering herd.
_PROBE_RANDOM_DELAY_INTERVAL = (150, 250)  # ms

The (150, 250) value already speaks for itself — readers don't need a paragraph to know that 150 is a more conservative floor than the spec's 0, nor that test fixtures can patch.object a module constant. Not a blocker.

# RFC 6762 §8.1 thundering-herd avoidance: wait a random
# 0-250ms before the first probe so simultaneously-started
# responders don't collide. We default to 150-250ms to
# preserve existing timing assumptions; tests on loopback
# may patch this lower via the `quick_timing` fixture.
_PROBE_RANDOM_DELAY_INTERVAL = (150, 250)  # ms

---

Checklist

• No hardcoded secrets or unsafe deserialization
• Error paths handle cleanup correctly
• No untested branches introduced
• Tests verify behavior, not source-code presence
• No race conditions in test setup — suggestion ＃1
• Comment style matches CLAUDE.md (terse, why-only) — suggestion ＃2
• Cython .pxd sync — not required (constant lives in non-Cythonized _core.py)
• RFC-pinned timing tests untouched
• Free-threaded (3.14t) safety — no new module-level mutable state

---

Summary

Clean test-speedup PR. The _PROBE_RANDOM_DELAY_INTERVAL extraction is the right shape — keeps the production tuple in _core.py, lets quick_timing patch it, and the .pxd doesn't need to change because _core.py isn't in TO_CYTHONIZE. The quick_request_timing rollouts at test_get_info_partial, test_integration_with_listener_class, and test_service_info_async_request are well-justified, and the matching 1500ms→300ms timeout tightening in the TOCTOU race-condition block has a clear new comment explaining the new arithmetic. test_shutdown_loop rewrite is correct: moving runcoro_thread_ready.set() after run_coroutine_threadsafe and replacing the time.sleep(0.1) buffer with a patched _TASK_AWAIT_TIMEOUT is a sound trade. RFC-pinned tests are correctly left alone. Two suggestion-level notes inline; neither blocks merge.

---

Automated review by Kōan038579d
a30eb88

[bluetoothbot]

Rebase with requested adjustments

Branch koan/fix-issue-1707 was rebased onto master and review feedback was applied.

Changes applied

• Task small, no tracking needed.
• Summary of changes addressing review comments:
• src/zeroconf/_core.py — Rephrased _PROBE_RANDOM_DELAY_INTERVAL comment per Copilot: removed "enforces 150ms floor" language (code doesn't enforce, it defaults), now says "We default to 150-250ms to preserve existing timing assumptions; tests on loopback may patch this lower."
• tests/conftest.py — Updated both quick_timing and quick_request_timing docstrings per Copilot to document both opt-in mechanisms (argument injection and @pytest.mark.usefixtures(...)).
• tests/utils/test_asyncio.py — Fixed race per Copilot: moved runcoro_thread_ready.set() to after asyncio.run_coroutine_threadsafe(...) submits the future, removed the now-unneeded time.sleep(0.05) since ordering is guaranteed. Also reworded the _TASK_AWAIT_TIMEOUT patch comment from "returns immediately" to "returns within 50ms" to match the patched value.

Stats

6 files changed, 40 insertions(+), 21 deletions(-)

Actions performed

• Already-solved check: negative (confidence=high, reasoning=No recent master commit shortens loopback test timings via _PROBE_RANDOM_DELAY_INTERVAL or quick_req)
• Rebased koan/fix-issue-1707 onto upstream/master
• Applied review feedback
• Pre-push CI check: previous run passed
• Force-pushed koan/fix-issue-1707 to origin
• CI check enqueued in ## CI (async)

CI status

CI will be checked asynchronously.

---

Automated by Kōan

[bdraco]

@bluetoothbot rr

[bluetoothbot]

Rebase with requested adjustments

Branch koan/fix-issue-1707 was rebased onto master and review feedback was applied.

Stats

6 files changed, 40 insertions(+), 21 deletions(-)

Actions performed

• Already-solved check: negative (confidence=high, reasoning=No recent master commit addresses loopback test timing overhead via _PROBE_RANDOM_DELAY_INTERVAL ext)
• Rebased koan/fix-issue-1707 onto upstream/master
• Pre-push CI check: previous run passed
• Force-pushed koan/fix-issue-1707 to origin
• CI check enqueued in ## CI (async)

CI status

CI will be checked asynchronously.

---

Automated by Kōan