hermes-agent/tests/test_profile_isolation_runtime.py
Teknium 6b81590c55
test: prune low-value tests suite-wide (wave 1) — 46,820 → 28,106 test functions
Systematic prune per AGENTS.md test policy, one pass over every major
test tree (gateway, hermes_cli, tools, agent, run_agent, plugins, cli,
cron, tui_gateway, honcho/openviking, root-level):

- DELETE: source-reading tests (read_text/getsource on prod files),
  change-detector tests (exact catalog counts, model-name snapshots,
  config version literals), mock-echo tests (assert a mock returns what
  it was told), assertion-free/trivial tests, near-duplicate
  parametrizations (boundaries + one representative kept), async/sync
  twin duplicates, cosmetic within-file variations.
- KEEP (mandatory): security/redaction/approval guards, message-role
  alternation invariants, prompt-caching/deterministic-call-id
  invariants, issue-number regression tests (deduped), E2E tests.
- 6 test files deleted outright (script-style/no-assert or fully
  redundant); conftest.py, fakes/, fixtures/ untouched.
- tests/acp/conftest.py added: autouse fixture stubs the live
  models.dev/GitHub/Copilot/Anthropic inventory fetches that ACP server
  tests performed on every session create — test_server.py 147s → 3.4s,
  and the tests are now genuinely hermetic.
- Sleep-based slowness shrunk where safe (codex_ttfb_watchdog,
  compression_concurrent_fork, etc.); no wall-clock assertion tightened.

Verification: full hermetic suite via scripts/run_tests.sh —
2439 files, 31,130 tests passed, 0 failed, 0 flaky retries, 315s wall
(baseline: 583s wall, 13,564s subprocess CPU).
2026-07-29 13:10:23 -07:00

159 lines
5.7 KiB
Python

"""Profile-isolation regression tests for single-process multi-profile runtimes.
In runtimes that serve every profile from one OS process (the desktop
``tui_gateway``), the profile boundary is the context-local
``_HERMES_HOME_OVERRIDE`` ContextVar, not the process environment. State that
escapes the request call stack — import-time-frozen path constants, direct
``os.environ`` reads, or worker threads that don't inherit the request context —
silently reverts to the launch/default profile and leaks one profile's data
into another.
These tests drive each previously-leaking site under override A then override B
with real temp HERMES_HOME directories (no mocks) and assert the *active*
profile's path is used. They are the productionized form of the manual smoke
probes used to confirm the bug class.
"""
import threading
from pathlib import Path
import pytest
from hermes_constants import (
get_hermes_home,
reset_hermes_home_override,
set_hermes_home_override,
)
@pytest.fixture
def two_profiles(tmp_path):
"""Two distinct profile HERMES_HOME dirs with the dir skeleton created."""
prof_a = tmp_path / "profA"
prof_b = tmp_path / "profB"
for p in (prof_a, prof_b):
(p / "skills").mkdir(parents=True, exist_ok=True)
(p / "state").mkdir(parents=True, exist_ok=True)
(p / "cache").mkdir(parents=True, exist_ok=True)
return prof_a, prof_b
def _under_override(home: Path, fn):
"""Run ``fn`` with the profile override set to ``home`` and reset after."""
token = set_hermes_home_override(str(home))
try:
return fn()
finally:
reset_hermes_home_override(token)
# ---------------------------------------------------------------------------
# M1 — import-time path globals / direct os.environ reads
# ---------------------------------------------------------------------------
class TestSkillsHubPathResolution:
"""tools/skills_hub.py path constants must reflect the active profile."""
def test_skills_dir_follows_override(self, two_profiles):
prof_a, prof_b = two_profiles
import tools.skills_hub as sh
# Importing/touching under A must NOT pin the path for B.
a_seen = _under_override(prof_a, lambda: Path(sh.SKILLS_DIR))
b_seen = _under_override(prof_b, lambda: Path(sh.SKILLS_DIR))
assert a_seen == prof_a / "skills"
assert b_seen == prof_b / "skills"
assert a_seen != b_seen
def test_hub_derived_paths_follow_override(self, two_profiles):
prof_a, prof_b = two_profiles
import tools.skills_hub as sh
b_lock = _under_override(prof_b, lambda: Path(sh.LOCK_FILE))
b_audit = _under_override(prof_b, lambda: Path(sh.AUDIT_LOG))
b_index = _under_override(prof_b, lambda: Path(sh.INDEX_CACHE_DIR))
assert b_lock == prof_b / "skills" / ".hub" / "lock.json"
assert b_audit == prof_b / "skills" / ".hub" / "audit.log"
assert b_index == prof_b / "skills" / ".hub" / "index-cache"
class TestGatewayCacheDirResolution:
"""gateway/platforms/base.py cache getters must follow the active profile."""
def test_image_cache_dir_follows_override(self, two_profiles):
prof_a, prof_b = two_profiles
import gateway.platforms.base as gb
a_seen = _under_override(prof_a, lambda: gb.get_image_cache_dir())
b_seen = _under_override(prof_b, lambda: gb.get_image_cache_dir())
assert str(a_seen).startswith(str(prof_a))
assert str(b_seen).startswith(str(prof_b))
assert a_seen != b_seen
class TestRichSentStorePathResolution:
"""gateway/rich_sent_store.py must honor the override, not read os.environ."""
def test_store_path_follows_override(self, two_profiles, monkeypatch):
prof_a, prof_b = two_profiles
# Ensure no ambient HERMES_HOME env masks the test.
monkeypatch.delenv("HERMES_HOME", raising=False)
import gateway.rich_sent_store as rss
b_seen = _under_override(prof_b, lambda: rss._store_path())
assert b_seen.startswith(str(prof_b))
assert b_seen.endswith("state/rich_sent_index.json")
# ---------------------------------------------------------------------------
# M2 — thread / executor context propagation
# ---------------------------------------------------------------------------
class TestThreadContextPropagation:
"""Worker threads must inherit the spawning turn's profile override."""
def test_raw_thread_loses_override(self, two_profiles):
"""Document the underlying hazard: a bare thread does NOT inherit it."""
_prof_a, prof_b = two_profiles
seen = {}
def worker():
seen["home"] = str(get_hermes_home())
def run():
t = threading.Thread(target=worker)
t.start()
t.join()
_under_override(prof_b, run)
# A bare thread falls back to the process default — this is WHY the fix
# primitive is needed. (Asserted as the hazard, not the desired state.)
assert seen["home"] != str(prof_b)
def test_run_async_worker_preserves_override(self, two_profiles):
"""model_tools._run_async's worker-thread branch must keep the override.
This is the generic sync->async bridge for every async tool; if it
leaks, every async tool that resolves get_hermes_home() leaks.
"""
import asyncio
_prof_a, prof_b = two_profiles
import model_tools
async def reads_home():
return str(get_hermes_home())
async def driver():
# Inside a running loop, _run_async spawns a worker thread + loop.
return model_tools._run_async(reads_home())
seen = _under_override(prof_b, lambda: asyncio.run(driver()))
assert seen == str(prof_b)