invisible_playwright/tests/conftest.py
feder-cr 5dac302938 test: activate the full e2e (browser-driving) suite + add fetch --force
The 138 @pytest.mark.e2e tests were doubly inactive: deselected by addopts AND
skipped without a cached binary — and 3 of the 6 per-file firefox_binary
fixtures silently ignored INVPW_BINARY_PATH, so they'd test whatever was cached
even when you pointed the suite elsewhere (a false-confidence trap).

- Centralize firefox_binary into conftest.py (env INVPW_BINARY_PATH → cache →
  skip); delete the 6 duplicates. Unify test_webrtc_realness onto the same env.
- scripts/run_e2e.py: one command that runs ALL e2e against a chosen binary,
  with reruns so an under-load interaction flake (dblclick/hover pass 3/3 in
  isolation) self-heals while a real break fails every attempt. The webrtc e2e
  fake a TCP-only SOCKS locally, so the suite is offline. This is the MANDATORY
  pre-release browser gate (local — hosted runners are too interaction-flaky).
- Running the suite against firefox-9 surfaced a real gap: `invisible_playwright
  fetch --force` was unrecognized (the subparser took no args) though the e2e
  test + docstring expect it. Implement it: drop the cached version dir, refetch.
- Add pytest-rerunfailures + playwright to the dev extras.

Baseline against firefox-9: 136 passed, 1 skipped (linux_only on win host),
1 was the --force gap now fixed.
2026-06-09 15:40:02 +02:00

54 lines
1.9 KiB
Python

import os
import random
import sys
from pathlib import Path
import pytest
from invisible_playwright._fpforge import generate_profile
from invisible_playwright.constants import BINARY_ENTRY_REL
@pytest.fixture
def deterministic_rng():
"""Seeded RNG for reproducible tests."""
return random.Random(42)
@pytest.fixture
def sample_profile():
"""A Profile generated from seed=42 for reuse across tests."""
return generate_profile(seed=42)
@pytest.fixture(scope="session")
def firefox_binary():
"""Locate the patched Firefox binary for E2E tests, or skip cleanly.
Single source of truth for every E2E test (previously each test file had its
own copy — and three of them silently ignored INVPW_BINARY_PATH, so they kept
testing whatever was in the cache even when you pointed the suite at a
specific build: a false-confidence trap). Lookup order:
1. ``INVPW_BINARY_PATH`` env var — point the whole suite at a local build
or a freshly-extracted release (this is how the full-suite gate runs).
2. Cached binary under ``cache_dir_for_version()`` (post ``fetch``).
3. Skip — we never trigger an implicit multi-hundred-MB network download
inside a test run.
"""
env_path = os.environ.get("INVPW_BINARY_PATH")
if env_path:
if Path(env_path).exists():
return env_path
pytest.skip(f"INVPW_BINARY_PATH={env_path!r} does not exist")
if sys.platform not in BINARY_ENTRY_REL:
pytest.skip(f"unsupported platform: {sys.platform}")
from invisible_playwright.download import cache_dir_for_version
entry = cache_dir_for_version() / BINARY_ENTRY_REL[sys.platform]
if not entry.exists():
pytest.skip(
"patched Firefox binary not cached and INVPW_BINARY_PATH unset; "
"set INVPW_BINARY_PATH=<firefox binary> or run `invisible-playwright fetch`"
)
return str(entry)