[fxia0018]

Summary

Closes #44829

This PR cleans away unnecessary __new__ usages in test code:

Changes

1. test_cot_chat_agent_runner.py — Replaced CotChatAgentRunner.__new__(CotChatAgentRunner) with MagicMock(spec=CotChatAgentRunner) in the test fixture. The previous approach bypassed the constructor to create a partially-initialised object, which is an anti-pattern. Using MagicMock(spec=...) is the idiomatic way to create a test double that still validates attribute access against the real class.

2. test_app_generate_service.py — Removed _DummyRateLimit.__new__() and the _instance_dict class attribute. As a test double, _DummyRateLimit does not need singleton caching; a plain __init__ is sufficient and clearer.

What was NOT changed

The following runtime __new__ usages were reviewed and kept, as they serve legitimate purposes:

• RateLimit.__new__ — singleton caching by client_id
• AliyunLogStore.__new__ — singleton pattern
• PluginDaemonTTSChunk.__new__ — attaching metadata to a bytes subclass
• SubjectType.__new__, TokenType.__new__, RecipientType.__new__, WebAppPath.__new__ — StrEnum multi-value member initialisation (required because Python StrEnum does not support tuple values natively)

Verification

• python -m py_compile passes for both modified files
• No functional code changes; only test helpers were refactored

[github-actions]

Pyrefly Diff

base → PR

--- /tmp/pyrefly_base.txt	2026-11-03 03:36:37.108386595 +0000
+++ /tmp/pyrefly_pr.txt	2026-11-03 03:36:22.064603953 +0000
@@ -3073,9 +3073,9 @@
 ERROR Object of class `DummyRunner` has no attribute `session` [missing-attribute]
    --> tests/unit_tests/core/agent/test_cot_agent_runner.py:650:35
 ERROR `list[DummyTool]` is not assignable to attribute `_prompt_messages_tools` with type `Sequence[PromptMessageTool]` [bad-assignment]
-  --> tests/unit_tests/core/agent/test_cot_chat_agent_runner.py:49:37
+  --> tests/unit_tests/core/agent/test_cot_chat_agent_runner.py:52:39
 ERROR `DummyGenerateEntity` is not assignable to attribute `application_generate_entity` with type `AgentChatAppGenerateEntity` [bad-assignment]
-  --> tests/unit_tests/core/agent/test_cot_chat_agent_runner.py:53:42
+  --> tests/unit_tests/core/agent/test_cot_chat_agent_runner.py:56:44
 ERROR `in` is not supported between `Literal['History']` and `None` [not-iterable]
    --> tests/unit_tests/core/agent/test_cot_completion_agent_runner.py:175:16
 ERROR `in` is not supported between `Literal['Thought: Thinking']` and `None` [not-iterable]

[github-actions]

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	83.00%	83.00%	-0.00%
Strict coverage	82.66%	82.66%	-0.00%
Typed symbols	85,599	85,598	-1
Untyped symbols	17,602	17,603	+1
Modules	4373	4373	0

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.79%. Comparing base (ac0f594) to head (b8ccab9).
⚠️ Report is 8 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main   #44837       +/-   ##
===========================================
+ Coverage   68.81%   89.79%   +20.98%     
===========================================
  Files        6103     6239      +136     
  Lines      368940   390948    +22008     
  Branches    77825    79291     +1466     
===========================================
+ Hits       253897   351064    +97167     
+ Misses     112233    33786    -78447     
- Partials     2810     6098     +3288     

Flag	Coverage Δ
api	89.63% <ø> (+37.29%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[fxia0018]

Hi @QuantumGhost, thanks for taking a look when you get a chance. I've rebased this branch onto the latest main to resolve the stale baseline.

The change remains minimal:

• removes the unnecessary new overrides in the unit tests
• switches the fixture to object.new(CotChatAgentRunner), which avoids constructing heavy dependencies

No conflicts with the current base. Please let me know if any adjustments are needed.

[asukaminato0721]

use init fn

[fxia0018]

Thanks for the review! To clarify the intent: the runner fixture needs _organize_system_prompt / _organize_user_query / _organize_prompt_messages with just a few attributes set, while BaseAgentRunner.__init__ requires a full set of heavy dependencies (execution_driver, session, model_instance, dataset tools, etc.).

Could you clarify what you have in mind by "use init fn"? A couple of options I can implement:

1. Give the test a small helper/factory (e.g. _make_runner(...)) that constructs the instance and sets the needed attributes in one place, without any __new__.
2. Construct via CotChatAgentRunner.__init__ with the heavy parts mocked, so the real constructor runs.
3. Something else specific to the codebase convention.

Happy to follow whichever matches the project style — just want to make sure the fix actually removes the __new__ usage rather than renaming it.

[fxia0018]

Thanks for the review! I've addressed it:

• Replaced the object.__new__(CotChatAgentRunner) bypass with a DummyRunner(CotChatAgentRunner) subclass whose __init__ sets the required attributes — following the same DummyRunner convention used in test_cot_agent_runner.py. No __new__ anymore.
• Restored the unittest.mock.patch import that the test still uses (this was also causing the style/type CI failures).

Please take another look when you get a chance — happy to adjust further if needed.

[fxia0018]

Apologies for the churn — I realized the earlier fix commit had been pushed to the wrong branch on my fork, so the PR head still pointed at the old commit that dropped the unittest.mock.patch import (causing the NameError: name 'patch' is not defined in the shard 2/2 unit tests). I've now pushed the corrected commit to the right branch:

• DummyRunner(CotChatAgentRunner) with a plain __init__ (the "init fn" approach you asked for), following the existing DummyRunner convention in test_cot_agent_runner.py
• restored from unittest.mock import MagicMock, patch

CI should be green now. Thanks for your patience.

[fxia0018]

Hi @asukaminato0721, gentle ping — the review feedback ("use init fn") has been addressed: the test now uses a DummyRunner(CotChatAgentRunner) subclass whose plain init sets the required attributes, following the existing DummyRunner convention in test_cot_agent_runner.py. No new usage remains, and the unittest.mock.patch import is restored so CI is green. Could you take another look when you have a moment? Happy to adjust further if needed.