hermes-agent/tests/plugins/platforms/photon/test_overflow_recovery.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

495 lines
17 KiB
Python

"""Photon adapter resilience to transient Spectrum/Envoy upstream overflow.
Covers the three behaviors that let the adapter ride through a Photon
"reset reason: overflow" event instead of degrading delivery and silently
dying (issue #50185):
1. ``_is_retryable_error`` classifies the Envoy/sidecar overflow strings as
retryable so ``_send_with_retry`` actually engages its backoff loop.
2. ``send_typing`` is rate-gated per chat, and ``stop_typing`` resets the
gate so the next turn's typing indicator fires immediately.
3. ``_supervise_sidecar`` detects an unexpected sidecar exit and raises a
``retryable=True`` fatal so the gateway reconnect watcher revives the
platform — instead of returning silently and leaving ``_inbound_loop``
spinning against a dead port.
4. ``_monitor_sidecar_health`` promotes degraded upstream stream health
reported by ``/healthz`` into the same retryable reconnect path.
No Node sidecar is spawned and no ports are bound.
"""
from __future__ import annotations
import asyncio
from typing import Any, Dict
import pytest
from gateway.config import PlatformConfig
from gateway.platforms.base import SendResult
from plugins.platforms.photon import adapter as photon_adapter
from plugins.platforms.photon.adapter import PhotonAdapter
def _make_adapter(monkeypatch: pytest.MonkeyPatch) -> PhotonAdapter:
monkeypatch.setenv("PHOTON_PROJECT_ID", "test-project-id")
monkeypatch.setenv("PHOTON_PROJECT_SECRET", "test-project-secret")
cfg = PlatformConfig(enabled=True, token="", extra={})
return PhotonAdapter(cfg)
# -- Gap 1: retryable classification of overflow errors ---------------------
@pytest.mark.parametrize(
"error",
[
"UNAVAILABLE: internal sidecar error",
"upstream connect error or disconnect/reset before headers",
"reset reason: overflow",
# Case-insensitive: real strings arrive with mixed case.
"Internal Sidecar Error",
],
)
def test_overflow_strings_classified_retryable(error: str) -> None:
assert PhotonAdapter._is_retryable_error(error) is True
def test_unrelated_error_not_retryable() -> None:
# A genuine permanent failure must NOT be retried.
assert PhotonAdapter._is_retryable_error("400 bad request: invalid spaceId") is False
assert PhotonAdapter._is_retryable_error(None) is False
def test_base_network_patterns_still_match() -> None:
# The override delegates to the base classifier first, so generic
# network strings keep working.
assert PhotonAdapter._is_retryable_error("ConnectError: connection refused") is True
def test_structured_non_retryable_sidecar_error_not_legacy_retried() -> None:
error = str(
photon_adapter.PhotonSidecarError(
path="/send",
status_code=500,
error="internal sidecar error",
error_class="auth_or_config",
retryable=False,
)
)
assert PhotonAdapter._is_retryable_error(error) is False
@pytest.mark.asyncio
async def test_send_with_retry_uses_structured_retryable_flag(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
calls = 0
sleeps: list[float] = []
async def _fake_sleep(delay: float) -> None:
sleeps.append(delay)
async def _fake_sidecar_call(path: str, body: Dict[str, Any]) -> Dict[str, Any]:
nonlocal calls
calls += 1
if calls == 1:
raise photon_adapter.PhotonSidecarError(
path=path,
status_code=500,
error="temporary upstream failure",
error_class="upstream_transient",
retryable=True,
)
return {"ok": True, "messageId": "m-2"}
monkeypatch.setattr(photon_adapter.asyncio, "sleep", _fake_sleep)
monkeypatch.setattr(adapter, "_sidecar_call", _fake_sidecar_call)
result = await adapter._send_with_retry(
"space-1", "hello", max_retries=1, base_delay=0.25
)
assert result.success is True
assert result.message_id == "m-2"
assert calls == 2
assert sleeps == [0.25]
# -- Gap 2: typing-indicator cooldown ---------------------------------------
@pytest.mark.asyncio
async def test_typing_cooldown_suppresses_rapid_repeats(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
calls: list[Dict[str, Any]] = []
async def _fake_call(path: str, payload: Dict[str, Any]) -> Any:
calls.append(payload)
return {"ok": True}
monkeypatch.setattr(adapter, "_sidecar_call", _fake_call)
# First call fires; immediate repeats are suppressed by the cooldown.
await adapter.send_typing("chat-1")
await adapter.send_typing("chat-1")
await adapter.send_typing("chat-1")
assert len(calls) == 1
@pytest.mark.asyncio
async def test_stop_typing_resets_cooldown(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
starts = 0
async def _fake_call(path: str, payload: Dict[str, Any]) -> Any:
nonlocal starts
if payload.get("state") == "start":
starts += 1
return {"ok": True}
monkeypatch.setattr(adapter, "_sidecar_call", _fake_call)
# A start, then a stop (end of turn), then a start for the next turn must
# fire immediately — the cooldown only suppresses rapid consecutive starts
# without an intervening stop.
await adapter.send_typing("chat-1")
await adapter.stop_typing("chat-1")
await adapter.send_typing("chat-1")
assert starts == 2
# -- Gap 3: sidecar crash detection -----------------------------------------
class _EofStdout:
"""A proc.stdout whose readline() reports immediate EOF (dead sidecar)."""
def readline(self) -> bytes:
return b""
class _DeadProc:
"""Minimal subprocess.Popen stand-in for a sidecar that has exited."""
def __init__(self, exit_code: int = 1) -> None:
self.stdout = _EofStdout()
self.stdin = None
self._exit_code = exit_code
def poll(self) -> int:
return self._exit_code
@pytest.mark.asyncio
async def test_unexpected_sidecar_exit_raises_retryable_fatal(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
# Simulate a live session whose sidecar then dies underneath it.
adapter._inbound_running = True
notified: list[bool] = []
async def _fake_notify() -> None:
notified.append(True)
monkeypatch.setattr(adapter, "_notify_fatal_error", _fake_notify)
await adapter._supervise_sidecar(_DeadProc(exit_code=137)) # type: ignore[arg-type]
assert adapter.has_fatal_error is True
assert adapter.fatal_error_code == "SIDECAR_CRASHED"
# retryable=True routes the platform into the reconnect watcher rather
# than crashing the whole gateway.
assert adapter.fatal_error_retryable is True
assert adapter._running is False
# The notification is dispatched onto its own task rather than awaited on
# the supervisor's stack, so that disconnect() cancelling the supervisor
# cannot kill the handoff. Let that task run before asserting delivery.
await _drain_pending_tasks()
assert notified == [True]
@pytest.mark.asyncio
async def test_clean_shutdown_does_not_raise_fatal(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
# disconnect() sets _inbound_running = False before stopping the sidecar,
# so the detection block must NOT fire on a clean shutdown.
adapter._inbound_running = False
notified: list[bool] = []
async def _fake_notify() -> None:
notified.append(True)
monkeypatch.setattr(adapter, "_notify_fatal_error", _fake_notify)
await adapter._supervise_sidecar(_DeadProc(exit_code=0)) # type: ignore[arg-type]
assert adapter.has_fatal_error is False
assert notified == []
@pytest.mark.asyncio
async def test_degraded_stream_health_raises_retryable_fatal(
monkeypatch: pytest.MonkeyPatch,
) -> None:
adapter = _make_adapter(monkeypatch)
adapter._inbound_running = True
adapter._sidecar_health_interval = 0.0
async def _fake_call(path: str, payload: Dict[str, Any]) -> Any:
assert path == "/healthz"
return {
"ok": True,
"stream": {
"ok": False,
"state": "degraded",
"degradedForMs": 120000,
"lastIssue": "[spectrum.stream] stream interrupted; reconnecting",
},
}
notified: list[bool] = []
async def _fake_notify() -> None:
notified.append(True)
adapter._inbound_running = False
monkeypatch.setattr(adapter, "_sidecar_call", _fake_call)
monkeypatch.setattr(adapter, "_notify_fatal_error", _fake_notify)
await adapter._monitor_sidecar_health()
assert adapter.has_fatal_error is True
assert adapter.fatal_error_code == "UPSTREAM_STREAM_DEGRADED"
assert adapter.fatal_error_retryable is True
# Dispatched detached (see _dispatch_fatal_notification) so the health
# task's own teardown cannot cancel the handoff; drain before asserting.
await _drain_pending_tasks()
assert notified == [True]
async def _drain_pending_tasks(limit: int = 50) -> None:
"""Let detached fatal-notification tasks finish before asserting on them.
``_dispatch_fatal_notification`` deliberately does not await the
notification (that is what kept ``disconnect()`` from cancelling its own
caller), so a test that drives ``_monitor_sidecar_health`` /
``_supervise_sidecar`` directly returns before the notification has run.
"""
for _ in range(limit):
pending = [
t for t in asyncio.all_tasks()
if t is not asyncio.current_task() and not t.done()
]
if not pending:
return
await asyncio.wait(pending, timeout=1.0)
# -- Gap 5: self-cancellation race in _stop_sidecar() (issue #73159) --------
#
# The tests above mock out _notify_fatal_error() entirely, so the real
# integration chain (supervisor task -> _notify_fatal_error() ->
# disconnect() -> _stop_sidecar() -> cancel the supervisor task) is never
# exercised end to end. That chain is exactly where the bug lives: when
# _notify_fatal_error() is a real callback that calls adapter.disconnect(),
# _stop_sidecar() is invoked FROM WITHIN the currently-running supervisor
# task, and cancelling self._sidecar_supervisor_task there cancels the very
# task executing the fatal-error handler -- aborting it (via CancelledError,
# a BaseException the handler's `except Exception` guards don't catch)
# before the Gateway's reconnect-queue step ever runs.
class _FakeStoppedProc:
"""Minimal proc stand-in so _stop_sidecar() reaches its finally block
(it early-returns entirely when self._sidecar_proc is None) without
spawning a real subprocess or doing any real I/O."""
def __init__(self) -> None:
self.stdin = None
def wait(self, timeout: float | None = None) -> int:
return 0
@pytest.mark.asyncio
async def test_supervisor_task_survives_self_triggered_disconnect(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""The real chain: _supervise_sidecar() (running as the actual
self._sidecar_supervisor_task) detects the crash and calls a REAL
_notify_fatal_error() that calls adapter.disconnect() -- which reaches
_stop_sidecar() from inside the task it's about to try to cancel.
Before the fix: this raises CancelledError out of _supervise_sidecar(),
so the task ends up in the "cancelled" state and reconnect_queued below
is never set (mirroring how the real Gateway's fatal-error handler,
which runs the reconnect-queue logic AFTER disconnect() returns, never
gets there either).
After the fix: disconnect() completes normally, _supervise_sidecar()
returns normally, and the task is NOT cancelled.
"""
adapter = _make_adapter(monkeypatch)
adapter._inbound_running = True
# A minimal fake proc so _stop_sidecar() reaches its finally block
# (the real process-management behavior is covered separately by
# test_sidecar_lifecycle.py) -- this test is purely about the
# self-cancellation race.
adapter._sidecar_proc = _FakeStoppedProc()
reconnect_queued: list[bool] = []
async def _real_notify_fatal_error() -> None:
# Stand-in for the Gateway's actual fatal-error handler: it calls
# adapter.disconnect() (real chain: disconnect -> _stop_sidecar,
# which used to self-cancel), then -- only if that completes
# without the CancelledError escaping -- proceeds to queue the
# platform for background reconnection.
await adapter.disconnect()
reconnect_queued.append(True)
monkeypatch.setattr(adapter, "_notify_fatal_error", _real_notify_fatal_error)
async def _run_supervisor():
await adapter._supervise_sidecar(_DeadProc(exit_code=75))
task = asyncio.ensure_future(_run_supervisor())
adapter._sidecar_supervisor_task = task
# Must complete cleanly -- must NOT raise CancelledError out to us.
await task
assert task.cancelled() is False, (
"The supervisor task must not end up cancelled by its own "
"fatal-error handling chain"
)
assert adapter.has_fatal_error is True
assert adapter.fatal_error_code == "SIDECAR_CRASHED"
assert reconnect_queued == [True], (
"The reconnect-queue step (everything after disconnect() returns "
"in the real Gateway handler) must actually run -- this is the "
"exact step issue #73159 reports as silently skipped"
)
# _stop_sidecar() must have cleared the task reference either way.
assert adapter._sidecar_supervisor_task is None
@pytest.mark.asyncio
async def test_external_disconnect_still_cancels_supervisor(
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""The OTHER call path -- external cleanup (Gateway shutdown, an
explicit /platform disconnect) -- calls _stop_sidecar() from a
DIFFERENT task than the supervisor. That legitimate case must still
cancel a still-running supervisor task exactly as before."""
adapter = _make_adapter(monkeypatch)
adapter._sidecar_proc = _FakeStoppedProc()
supervisor_ran_forever = asyncio.Event()
async def _hangs_forever():
try:
supervisor_ran_forever.set()
await asyncio.sleep(3600)
except asyncio.CancelledError:
raise
task = asyncio.ensure_future(_hangs_forever())
adapter._sidecar_supervisor_task = task
await supervisor_ran_forever.wait()
# Called from THIS (different) task -- the external-cleanup case.
await adapter._stop_sidecar()
# cancel() only schedules the CancelledError; await it so the task
# actually settles into the cancelled state before asserting.
try:
await task
except asyncio.CancelledError:
pass
assert task.cancelled() is True, (
"External cleanup must still cancel a running supervisor task"
)
assert adapter._sidecar_supervisor_task is None
# -- target_not_allowed: shared/free-tier outbound-send restriction ----------
#
# Spectrum throws AuthenticationError("Target not allowed for this project")
# from space.send when a shared/free-tier line initiates an outbound send to
# a new target. The sidecar classifies it as the structured code
# `target_not_allowed`; the adapter must treat it as permanent in BOTH
# _send_with_retry and _standalone_send, surfacing the canonical user-facing
# message instead of raw upstream error text (issues #50971 / #51897).
def test_target_not_allowed_maps_to_canonical_message() -> None:
err = photon_adapter._sidecar_error_from_response(
"/send",
500,
'{"ok":false,"error":"internal sidecar error",'
'"error_class":"target_not_allowed","retryable":false}',
)
assert err.error_class == "target_not_allowed"
assert err.retryable is False
assert err.error == photon_adapter._TARGET_NOT_ALLOWED_MESSAGE
# No raw upstream text may leak through the structured code path.
assert "Target not allowed for this project" not in str(err)
@pytest.mark.asyncio
async def test_standalone_send_classifies_target_not_allowed(
monkeypatch: pytest.MonkeyPatch,
) -> None:
monkeypatch.setenv("PHOTON_SIDECAR_TOKEN", "token")
class _Resp:
status_code = 500
text = (
'{"ok":false,"error":"internal sidecar error",'
'"error_class":"target_not_allowed","retryable":false}'
)
@staticmethod
def json() -> Dict[str, Any]:
return {
"ok": False,
"error": "internal sidecar error",
"error_class": "target_not_allowed",
"retryable": False,
}
class _FakeClient:
def __init__(self, *a: Any, **k: Any) -> None:
pass
async def __aenter__(self) -> "_FakeClient":
return self
async def __aexit__(self, *a: Any) -> bool:
return False
async def post(self, *a: Any, **k: Any) -> _Resp:
return _Resp()
monkeypatch.setattr(photon_adapter.httpx, "AsyncClient", _FakeClient)
result = await photon_adapter._standalone_send(
PlatformConfig(enabled=True, extra={}), "space-1", "hello",
)
assert result.get("error") == photon_adapter._TARGET_NOT_ALLOWED_MESSAGE
assert result.get("error_class") == "target_not_allowed"
assert result.get("retryable") is False
assert "Target not allowed for this project" not in str(result)