tests: drop unused 'event_loop' fixture arg from test_resolver.py (3 collection errors) #214

Problem

Three tests in tests/test_resolver.py declare event_loop as a fixture argument but never use it:

async def test_get_real_ext_ip(event_loop, mocker, resolver):  # line 63
async def test_resolve(event_loop, mocker, resolver):          # line 101
async def test_resolve_cache(event_loop, mocker, resolver):    # line 224

pytest-asyncio deprecated the auto-provided event_loop fixture in 0.21 and removed it in 0.23. This project is on 0.26.0 (per pyproject.toml), so all three tests error at collection with:

fixture 'event_loop' not found

Root cause

These declarations date back to 2017 (commit d5602973, Constverum). At that time, declaring event_loop as an arg ensured pytest-asyncio set up a loop before the test body ran. Modern pytest-asyncio handles loop setup automatically via @pytest.mark.asyncio (or auto mode), so the arg is redundant.

The arg is never used inside any of the three function bodies — grep -c "event_loop" tests/test_resolver.py would show 0 references inside the bodies.

Fix (15-minute change)

Remove the unused event_loop arg from the three function signatures:

async def test_get_real_ext_ip(mocker, resolver):
async def test_resolve(mocker, resolver):
async def test_resolve_cache(mocker, resolver):

That's the entire change. No other code modifications.

Acceptance criteria

• All 3 tests collect cleanly (no fixture 'event_loop' not found).
• All 3 tests pass against the existing mocks.
• No new fixtures introduced; no conftest.py change needed.
• pytest tests/ reports zero collection errors.

Why this is safe

• The arg has been unused for years; removing it can't change test semantics.
• pytest-asyncio's automatic loop management (already enabled per pyproject.toml) handles what event_loop arg used to handle.
• Existing @pytest.mark.asyncio decorator stays.

Why this matters

These 3 errors mask real test failures — every full-suite run reports "3 errors" alongside actual passes/failures, so a contributor scanning CI output for new breakage has to mentally subtract these. PR #212 already trained me to filter them out, which is exactly the sort of "noise drowns signal" anti-pattern that makes test suites lose their value over time.

Out of scope

• Pinning pytest-asyncio<0.23 (regression to old behavior, not the modern path)
• Defining a custom event_loop fixture in conftest.py (overkill; the arg isn't even used)