test(gateway): accept arbitrary kwargs in _install_fake_aiohttp to absorb trust_env

[briandevans]

Summary

Widen the _install_fake_aiohttp helper lambdas in tests/gateway/test_google_chat.py and tests/gateway/test_teams.py to accept arbitrary keyword arguments. Currently four tests in TestGoogleChatStandaloneSend and TestTeamsStandaloneSend fail on every PR's CI run with TypeError: lambda() got an unexpected keyword argument 'trust_env'.

The bug

Commit c1ae18ee8 (fix(gateway): add trust_env=True to aiohttp sessions in SMS, Slack, Teams, Google Chat adapters, merged 2026-05-17 via #27308) added a new keyword argument to the production aiohttp.ClientSession(...) calls in both adapters:

• plugins/platforms/google_chat/adapter.py:3249 — async with _aiohttp.ClientSession(timeout=..., trust_env=True)
• plugins/platforms/teams/adapter.py:569 — async with _aiohttp.ClientSession(trust_env=True)

The fake aiohttp.ClientSession defined inside _install_fake_aiohttp in each test file is lambda timeout=None: session — it accepts only the timeout kwarg. When production now also passes trust_env=True, the lambda raises TypeError. The adapters catch the error in their broad try/except and surface it as the assertion mismatch in the CI log:

{'error': 'Google Chat standalone send failed: '
          '_install_fake_aiohttp.<locals>.<lambda>() got an unexpected keyword '
          "argument 'trust_env'"}

Failing tests (all four reproduce on clean origin/main):

• tests/gateway/test_google_chat.py::TestGoogleChatStandaloneSend::test_standalone_send_refreshes_token_and_posts_message
• tests/gateway/test_google_chat.py::TestGoogleChatStandaloneSend::test_standalone_send_propagates_api_failure
• tests/gateway/test_teams.py::TestTeamsStandaloneSend::test_standalone_send_acquires_token_and_posts_activity
• tests/gateway/test_teams.py::TestTeamsStandaloneSend::test_standalone_send_propagates_token_failure

The fix

Change both helper lambdas to accept and ignore arbitrary args/kwargs:

ClientSession=lambda *args, **kwargs: session,
ClientTimeout=lambda *args, **kwargs: None,

The fake's job is just to return the scripted session/timeout regardless of how the production code invokes it. This makes the helper resilient to future kwargs added to the production call sites without further test churn.

Test plan

• Confirmed c1ae18ee8 lands trust_env=True in both adapters on current origin/main.
• Confirmed only tests/gateway/test_google_chat.py and tests/gateway/test_teams.py define a _install_fake_aiohttp helper of this shape (rg "fake_aiohttp|ClientSession=lambda" tests/ finds no other call sites).
• Verified all 6 call sites in both files pass exactly (monkeypatch, session) — no caller depends on the lambda signature shape.
• CI repro on every recent PR (e.g. #27316 run 25983241078) — four _install_fake_aiohttp.<locals>.<lambda>() got an unexpected keyword argument 'trust_env' failures.

Local repro is partially blocked because my dev install lacks google-cloud-pubsub and googleapiclient, so the test hits service_account is None before reaching the lambda. The fix is verifiable by inspection: a lambda with *args, **kwargs accepts every shape (timeout=...) was previously called with, plus the new (trust_env=...) call.

Related

• Production change: c1ae18ee8 (merged in #27308)
• SMS and Slack adapters also gained trust_env=True in the same commit but neither has a _install_fake_aiohttp helper of this shape, so no companion fix is needed for them.

[Copilot]

Pull request overview

Note

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

Loosens the signatures of fake aiohttp.ClientSession and ClientTimeout constructors in two test helpers so they accept arbitrary positional and keyword arguments.

Changes:

• Accept *args, **kwargs in fake ClientSession lambda.
• Accept *args, **kwargs in fake ClientTimeout lambda.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
tests/gateway/test_teams.py	Make _install_fake_aiohttp tolerant of extra args.
tests/gateway/test_google_chat.py	Same change applied to Google Chat test helper.

---

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

[briandevans]

CI audit — the four trust_env failures this PR targets (TestGoogleChatStandaloneSend::test_standalone_send_refreshes_token_and_posts_message, ::test_standalone_send_propagates_api_failure, TestTeamsStandaloneSend::test_standalone_send_acquires_token_and_posts_activity, ::test_standalone_send_propagates_token_failure) all pass in this run. The 14 remaining failures are pre-existing baselines unrelated to this PR's scope (tests/gateway/test_google_chat.py, tests/gateway/test_teams.py).

10 reproduce on clean origin/main at 519657aa9 with identical error text; 2 are CI-only (PermissionError on /root/.hermes/... which only triggers under root); 2 (test_transcription_dotenv_fallback) fail identically across other open PRs and don't reproduce locally — environment artifact, not in this diff.

Test	Symptom	Root cause on main
tests/acp/test_registry_manifest.py::test_agent_json_version_matches_pyproject	'0.13.0' == '0.14.0'	plugins/acp/manifest/agent.json version field lags pyproject.toml bump to 0.14.0
tests/acp/test_registry_manifest.py::test_agent_json_pins_uvx_package_to_pyproject_version	'hermes-agent[acp]==0.13.0' == '...==0.14.0'	same manifest staleness
tests/gateway/test_background_command.py::TestRunBackgroundTask::test_telegram_dm_topic_completion_preserves_reply_anchor_metadata	{'direct_message...} == {'telegram_dm...}	metadata-key shape mismatch after recent reply-anchor refactor
tests/gateway/test_telegram_thread_fallback.py::test_gateway_runner_busy_ack_replies_to_triggering_message_for_telegram_dm_topic	same	same
tests/gateway/test_voice_command.py::TestSendVoiceReply::test_auto_voice_reply_uses_thread_metadata_helper	same	same
tests/hermes_cli/test_model_switch_custom_providers.py::test_list_groups_same_name_custom_providers_into_one_row	hardcoded model list diverges from current catalog	catalog grew but test expectations weren't refreshed
tests/hermes_cli/test_startup_plugin_gating.py::test_builtin_set_covers_every_registered_subcommand	_BUILTIN_SUBCOMMANDS is missing ['send']	send subcommand was registered without being added to hermes_cli/main.py::_BUILTIN_SUBCOMMANDS
tests/run_agent/test_deepseek_reasoning_content_echo.py::TestNeedsKimiToolReasoning::test_non_kimi_provider	assert True is False	_needs_kimi_tool_reasoning heuristic widened past test's expectations
tests/run_agent/test_provider_parity.py::TestAuxiliaryClientProviderPriority::test_openrouter_always_wins	'google/gemini-2.5-flash' == 'google/gemini-2.5-flash-preview'	OpenRouter default model id moved; assertion not refreshed
tests/tools/test_voice_cli_integration.py::TestVprintForceParameter::test_error_messages_use_force_in_run_agent	Expected at least one _vprint with force=True for error messages	run_agent.py error-message branch no longer routes through _vprint(..., force=True)
tests/hermes_cli/test_gateway_service.py::TestSystemUnitHermesHome::test_system_unit_uses_target_user_home_not_calling_user	PermissionError: '/root/.hermes/node/bin'	CI-only — test attempts a real mkdir outside HERMES_HOME when running as root
tests/hermes_cli/test_gateway_service.py::TestSystemUnitHermesHome::test_system_unit_remaps_profile_to_target_user	PermissionError: '/root/.hermes/profiles/coder/node/bin'	same
tests/tools/test_transcription_dotenv_fallback.py::TestProviderSelectionGate::test_explicit_xai_sees_dotenv	'none' == 'xai'	passes locally on clean origin/main; CI-only environment difference (not in this PR's scope)
tests/tools/test_transcription_dotenv_fallback.py::TestEndToEndRegressionGuard::test_xai_key_only_in_dotenv_before_fix	assert False is True	same — passes locally on clean origin/main

Local repro command:

uv run --python 3.11 --with pytest --with pytest-xdist --with pytest-asyncio python3 -m pytest <test_path> -v

[briandevans]

Closing — teknium1 landed the same fix directly on main as 06924e8 (test(gateway): accept trust_env in fake aiohttp ClientSession lambdas) on 2026-05-17. Thanks for picking it up.