test: 同步 backend 测试到 11 tool（配对 StarRailOneDragon#605） - #1
Open
DoctorReid wants to merge 1720 commits into
Open
DoctorReid wants to merge 1720 commits into
DoctorReid wants to merge 1720 commits into
Conversation
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Advanced Run ID: Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
Walkthrough本次变更新增大量测试，覆盖 Changesscreen_match 测试
sr_od.backend 测试套件
测试脚手架与应用流程
资源与测试环境
Estimated code review effort: 4 (Complex) | ~60 minutes 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches 💡 1📝 Generate docstrings 💡
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. 小兔背着测试篮，🐇 Comment |
DoctorReid
added a commit
to OneDragon-Anything/StarRailOneDragon
that referenced
this pull request
Jul 6, 2026
- PR 时 checkout sr-od-test @ github.head_ref（同名分支优先，fallback main，continue-on-error 兜底） - PR 跑 pytest -m 'not requires_secrets'，main 跑全集 - 配 sr-od-test 同名分支 PR（OneDragon-Anything/sr-od-test#1），CI 能跑到配对测试 - 注意：本 PR 自身不会触发该 workflow（GitHub pull_request 用 base 分支版本）；合并到 main 后后续 PR 生效 照搬 ZZZ test-check.yml，zzz-od-test→sr-od-test 适配。 Co-Authored-By: Claude Code <noreply@anthropic.com> Co-Authored-By: glm-5.2 <noreply@bigmodel.cn>
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
🧹 Nitpick comments (4)
test/sr_od/backend/test_run_slot.py (1)
153-163: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win移除未使用的
event变量。
event = threading.Event(); event.set()既触发 Ruff E702(分号连写多语句),又是死代码——该测试后续逻辑并未使用event。建议直接删除该行。🧹 建议的修复
def test_context_start_run_delegates(slot, mock_ctx): """SrBackendContext.start_run 转发 run_slot._start_run,返回 (ok, future)。""" from sr_od.backend.backend_context import SrBackendContext backend = SrBackendContext(mock_ctx) backend.run_slot = slot - event = threading.Event(); event.set() ok, fut = backend.start_run('mcp', _make_op(OperationResult(success=True))) assert ok is True and fut is not None fut.result(timeout=5)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/sr_od/backend/test_run_slot.py` around lines 153 - 163, The test helper in test_context_start_run_delegates has an unused event setup line that should be removed. Delete the threading.Event() creation and set call entirely, since the test only exercises SrBackendContext.start_run and does not use event anywhere; this also avoids the Ruff E702 semicolon warning. Keep the rest of the assertions and the backend.run_slot setup unchanged.Source: Linters/SAST tools
test/sr_od/backend/test_http_routes.py (2)
138-146: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value局部
_OpResult与已导入的OperationResult重复。同一文件后面（Line 266）已直接用
OperationResult(success=True)构造成功结果，说明该 dataclass 无需额外字段即可满足此处用途；这里再定义一个等价的_OpResult略显冗余，可直接复用已导入的OperationResult。♻️ 简化建议
- from dataclasses import dataclass as _dataclass - - `@_dataclass` - class _OpResult: - success: bool - status: str = "" - fut: Future = Future() - fut.set_result(_OpResult(success=True)) + fut.set_result(OperationResult(success=True))🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/sr_od/backend/test_http_routes.py` around lines 138 - 146, The local _OpResult dataclass in the test setup is redundant because OperationResult is already available and used elsewhere in the same file. Replace the temporary _OpResult definition and its use in Future.set_result with OperationResult so the test reuses the existing dataclass consistently; locate the change around the Future setup in test_http_routes and keep the success payload construction aligned with the later OperationResult(success=True) usage.
41-53: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win
_mock_backend与test_mcp_app.py中的同名函数重复。该 helper 在
test/sr_od/backend/test_mcp_app.py（Line 37-49）中几乎逐字复制（仅source/stop返回值不同）。鉴于本 PR 已为mock_ctx/slot等共享 fixture 新增了conftest.py，建议将该 mock backend 构造函数也下沉到conftest.py（可用参数区分source），以避免两处后续对 backend 契约字段（如RunStatusResult）演进时出现不同步。🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/sr_od/backend/test_http_routes.py` around lines 41 - 53, _mock_backend is duplicated between the HTTP route tests and test_mcp_app, so move the shared backend mock constructor into conftest.py and make it configurable for source/stop return values. Update the test_http_routes helper to use the shared fixture or factory, and keep the logic centered around SrBackendContext, start_run, query_status, stop, and RunStatusResult so future backend contract changes only need to be made in one place.test/sr_od/backend/test_mcp_app.py (1)
74-79: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win统一通过公开工具 API 访问 MCP 工具，别直接碰
_tool_manager._tools
本文件多处直接读取mcp._tool_manager._tools，还用getattr(tool, "fn", None) or getattr(tool, "func", None)兼容不同版本，这会把测试绑定到 FastMCP 的内部结构上。建议统一封装成一个 helper，改用list_tools()/单工具获取入口后再调用，版本差异也只需要处理一次。🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/sr_od/backend/test_mcp_app.py` around lines 74 - 79, The test is reaching into FastMCP internals via mcp._tool_manager._tools and version-specific fn/func attributes, which makes it brittle. Update the test to use the public MCP tool API instead, preferably by introducing a small helper in test_mcp_app that retrieves a named tool through list_tools() or the supported single-tool access path and returns a callable once. Replace the direct internal lookup in check_game_window coverage with that helper so version differences are handled in one place.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@test/sr_od/backend/test_http_routes.py`:
- Around line 138-146: The local _OpResult dataclass in the test setup is
redundant because OperationResult is already available and used elsewhere in the
same file. Replace the temporary _OpResult definition and its use in
Future.set_result with OperationResult so the test reuses the existing dataclass
consistently; locate the change around the Future setup in test_http_routes and
keep the success payload construction aligned with the later
OperationResult(success=True) usage.
- Around line 41-53: _mock_backend is duplicated between the HTTP route tests
and test_mcp_app, so move the shared backend mock constructor into conftest.py
and make it configurable for source/stop return values. Update the
test_http_routes helper to use the shared fixture or factory, and keep the logic
centered around SrBackendContext, start_run, query_status, stop, and
RunStatusResult so future backend contract changes only need to be made in one
place.
In `@test/sr_od/backend/test_mcp_app.py`:
- Around line 74-79: The test is reaching into FastMCP internals via
mcp._tool_manager._tools and version-specific fn/func attributes, which makes it
brittle. Update the test to use the public MCP tool API instead, preferably by
introducing a small helper in test_mcp_app that retrieves a named tool through
list_tools() or the supported single-tool access path and returns a callable
once. Replace the direct internal lookup in check_game_window coverage with that
helper so version differences are handled in one place.
In `@test/sr_od/backend/test_run_slot.py`:
- Around line 153-163: The test helper in test_context_start_run_delegates has
an unused event setup line that should be removed. Delete the threading.Event()
creation and set call entirely, since the test only exercises
SrBackendContext.start_run and does not use event anywhere; this also avoids the
Ruff E702 semicolon warning. Keep the rest of the assertions and the
backend.run_slot setup unchanged.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: df71b530-f17e-4364-97fa-3e6c111a4689
📒 Files selected for processing (13)
.gitignoretest/one_dragon/base/screen/test_screen_match.pytest/sr_od/app/sim_uni/test_sim_uni_screen_state.pytest/sr_od/backend/conftest.pytest/sr_od/backend/test_backend_context.pytest/sr_od/backend/test_click_game.pytest/sr_od/backend/test_close_game.pytest/sr_od/backend/test_entry_server.pytest/sr_od/backend/test_http_routes.pytest/sr_od/backend/test_input_text.pytest/sr_od/backend/test_mcp_app.pytest/sr_od/backend/test_run_slot.pytest/sr_od/backend/test_schemas.py
There was a problem hiding this comment.
🧹 Nitpick comments (1)
test/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.py (1)
1-4: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win三个测试文件均使用
__import__而非已导入的模块引用。 共同根因：pytestmark行通过__import__('pytest')和__import__('os')获取模块，而非直接使用顶部已导入的pytest和os，造成冗余且不一致。
test/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.py#L1-L4：已导入os和pytest，将第 4 行改为pytest.mark.skipif(bool(os.environ.get('CI')), reason='...')。test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py#L1-L3：补充import os，将第 3 行改为pytest.mark.skipif(bool(os.environ.get('CI')), reason='...')。test/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.py#L1-L4：已导入os和pytest，将第 4 行改为pytest.mark.skipif(bool(os.environ.get('CI')), reason='...')。🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.py` around lines 1 - 4, Replace the __import__ calls in pytestmark with the already imported pytest and os modules in test/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.py#L1-L4 and test/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.py#L1-L4. In test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py#L1-L3, add the os import and make the same pytest.mark.skipif change, preserving the existing CI condition and reason.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In
`@test/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.py`:
- Around line 1-4: Replace the __import__ calls in pytestmark with the already
imported pytest and os modules in
test/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.py#L1-L4
and
test/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.py#L1-L4.
In
test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py#L1-L3,
add the os import and make the same pytest.mark.skipif change, preserving the
existing CI condition and reason.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c3b693e3-dd78-4054-b173-d14e3c4bb341
📒 Files selected for processing (12)
test/sr_od/backend/conftest.pytest/sr_od/backend/test_backend_context.pytest/sr_od/backend/test_click_game.pytest/sr_od/backend/test_http_routes.pytest/sr_od/backend/test_mcp_app.pytest/sr_od/backend/test_operation_registry.pytest/sr_od/backend/test_run_slot.pytest/sr_od/backend/test_schemas.pytest/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.pytest/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.pytest/sr_od/screen_state/test_batttle_screen_state/test_battle_screen_state.pytest/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.py
🚧 Files skipped from review as they are similar to previous changes (2)
- test/sr_od/backend/conftest.py
- test/sr_od/backend/test_schemas.py
There was a problem hiding this comment.
🧹 Nitpick comments (1)
test/sr_od/backend/test_mcp_app.py (1)
287-295: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value完善“原样返回”的断言测试。
docstring 中提到了“并原样返回”，但目前仅断言了
result['success'] is True。建议直接对比整个字典，以更严谨地验证该行为。💡 建议修改
- assert result['success'] is True + assert result == {'success': True, 'x': 960, 'y': 540, 'in_window': True, 'pc_alt': False}🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@test/sr_od/backend/test_mcp_app.py` around lines 287 - 295, 更新 test_click_game_tool_delegates 中的返回值断言，将仅检查 result['success'] 改为直接比较 result 与 backend.click_game.return_value 对应的完整字典，确保 click_game 工具原样返回所有字段。
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@test/sr_od/backend/test_mcp_app.py`:
- Around line 287-295: 更新 test_click_game_tool_delegates 中的返回值断言，将仅检查
result['success'] 改为直接比较 result 与 backend.click_game.return_value 对应的完整字典，确保
click_game 工具原样返回所有字段。
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: 91e3d2f6-b9ce-4f96-bd56-fc709df093f2
📒 Files selected for processing (2)
test/sr_od/backend/test_backend_context.pytest/sr_od/backend/test_mcp_app.py
🚧 Files skipped from review as they are similar to previous changes (1)
- test/sr_od/backend/test_backend_context.py
There was a problem hiding this comment.
Actionable comments posted: 4
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@test/harness/fixture_controller.py`:
- Around line 218-231: 在测试 harness 中新增并复用一个通用的 _maybe_advance_on_action 辅助方法：检查
_current_exit() 返回的退出条件是否为 on_action，是则调用 _advance_phase()。更新 input_str 和
drag_to 在记录输入或拖动操作后调用该辅助方法，使任意 click/input/drag 操作都能推进 on_action
phase，同时保持现有记录行为不变。
In `@test/sr_od/operations/enter_game/test_enter_game_flow.py`:
- Around line 274-275: Replace the hardcoded account and password values in the
test context setup with clearly fake placeholder credentials consistent with the
nearby test data, such as the established placeholder style. Ensure no real or
personalized credentials remain in the file, and rotate the exposed credentials
separately if they are valid.
In
`@test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py`:
- Around line 3-6: Update the module-level pytest markers in
test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py
lines 3-6 and
test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py
lines 3-6: remove the broad xfail coverage, apply xfail only to the known
coordinate-drift cases, and set strict=True so unexpected passes fail while
unrelated regressions are not masked.
In
`@test/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.py`:
- Around line 32-35: Restore the commented screen cases in the
get_match_screen_name regression coverage instead of silently removing them from
the test loop. For each affected screen mapping such as div_uni_entry,
choose_bless, and sim_uni_get_reward, either restore the corresponding
screen_info and fixtures so strict matching is validated, or keep an explicit
strict=True expected failure/independent skip with the 2026-07-30 temporary
handling re-evaluated.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Organization UI
Review profile: CHILL
Plan: Pro Plus
Run ID: 05795de0-bd64-4217-9034-dec03075c3f9
⛔ Files ignored due to path filters (1)
test/sr_od/screen_state/test_batttle_screen_state/normal_world_battle_fail.pngis excluded by!**/*.png
📒 Files selected for processing (57)
screens/任务/全部任务.webpscreens/历战余响/选关-铁骸的锈冢.webpscreens/合成/消耗品合成.webpscreens/商店/推荐.webpscreens/大世界-战斗失败/战斗失败-侵蚀隧洞.webpscreens/大世界-战斗失败/战斗失败-凝滞虚影.webpscreens/大世界-战斗失败/战斗失败-历战余响.webpscreens/大世界-战斗失败/战斗失败.webpscreens/大世界/普通.webpscreens/委托/委托可领.webpscreens/委托/委托派遣中.webpscreens/委托/委托领取弹窗.webpscreens/战斗画面/战斗中.webpscreens/战斗画面/挑战成功.webpscreens/星际和平指南/旷宇纷争.webpscreens/星际和平指南/每日实训.webpscreens/星际和平指南/生存索引.webpscreens/漫游签证/角色展示.webpscreens/背包-遗器分解-快速选择/快速选择.webpscreens/背包-遗器分解/分解.webpscreens/菜单/无名勋礼-任务.webpscreens/菜单/无名勋礼-奖励.webpscreens/菜单/菜单-无邮件红点.webpscreens/菜单/菜单-邮件红点.webpscreens/角色/详情.webpscreens/进入游戏-退出登陆/退出弹窗.webpscreens/进入游戏-选择账号/选账号.webpscreens/进入游戏/安全验证.webpscreens/进入游戏/开始游戏.webpscreens/进入游戏/手机号登录.webpscreens/进入游戏/点击进入.webpscreens/进入游戏/账号密码-旧.webpscreens/邮件/获得物品.webpscreens/邮件/邮件列表-有可领.webpscreens/邮件/邮件列表.webpscreens/队伍/编队.webptest/conftest.pytest/harness/__init__.pytest/harness/fixture_controller.pytest/sr_od/application/assignments/test_assignments_app.pytest/sr_od/application/daily_training/test_daily_training_app.pytest/sr_od/application/echo_of_war/test_echo_of_war_app.pytest/sr_od/application/email/test_email_app.pytest/sr_od/application/nameless_honor/test_nameless_honor_app.pytest/sr_od/application/relic_salvage/test_relic_salvage_app.pytest/sr_od/application/sim_universe/test_sim_uni_app.pytest/sr_od/application/support_character/test_support_character_app.pytest/sr_od/application/trailblaze_power/test_trailblaze_power_app.pytest/sr_od/application/world_patrol/test_world_patrol_app.pytest/sr_od/operations/custom_combine_op/__init__.pytest/sr_od/operations/custom_combine_op/test_custom_combine_op_config.pytest/sr_od/operations/enter_game/test_enter_game_flow.pytest/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.pytest/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.pytest/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.pytest/sr_od/screens/__init__.pytest/sr_od/screens/test_shared_screen_fixtures.py
Comment on lines
+218
to
+231
| def input_str(self, to_input: str, interval: float = 0.1) -> None: | ||
| self.recorded_inputs.append(to_input) | ||
|
|
||
| def delete_all_input(self) -> None: # noqa: D401 - stub | ||
| pass | ||
|
|
||
| def drag_to( | ||
| self, | ||
| end: Point, | ||
| start: Point | None = None, | ||
| duration: float = 0.5, | ||
| ) -> None: | ||
| # 国际服换服滚动会触达;CN 登录流程不会。记录一次 click 以便调试。 | ||
| self.recorded_clicks.append(Point(int(end.x), int(end.y))) |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win
input_str/drag_to 未实现 on_action 推进契约
顶部文档（Line 103）声明 ('on_action',) 应在"任意 click/input"时推进，但只有 click() 会调用 _maybe_advance_on_click；input_str/drag_to 完全没有调用任何推进判定。若未来剧本用 on_action 描述"仅靠输入推进"的 phase，该 phase 将永远卡住，直到 watchdog 超限才以 round_fail 收场，掩盖真实原因。当前 test_enter_game_flow.py 的剧本均未使用 on_action，所以暂未暴露，但这是共享测试基础设施里的契约缺口，会影响后续基于此 harness 编写的用例。
🔧 建议修复
def input_str(self, to_input: str, interval: float = 0.1) -> None:
self.recorded_inputs.append(to_input)
+ self._maybe_advance_on_action()
def delete_all_input(self) -> None: # noqa: D401 - stub
pass
def drag_to(
self,
end: Point,
start: Point | None = None,
duration: float = 0.5,
) -> None:
# 国际服换服滚动会触达;CN 登录流程不会。记录一次 click 以便调试。
self.recorded_clicks.append(Point(int(end.x), int(end.y)))
+ self._maybe_advance_on_action()再加一个通用辅助：
def _maybe_advance_on_action(self) -> None:
exit_spec = self._current_exit()
if exit_spec is not None and exit_spec[0] == 'on_action':
self._advance_phase()🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@test/harness/fixture_controller.py` around lines 218 - 231, 在测试 harness
中新增并复用一个通用的 _maybe_advance_on_action 辅助方法：检查 _current_exit() 返回的退出条件是否为
on_action，是则调用 _advance_phase()。更新 input_str 和 drag_to 在记录输入或拖动操作后调用该辅助方法，使任意
click/input/drag 操作都能推进 on_action phase，同时保持现有记录行为不变。
Comment on lines
+274
to
+275
| monkeypatch.setattr(test_context.game_account_config, 'account', '18928573369') | ||
| monkeypatch.setattr(test_context.game_account_config, 'password', 'moyijie920.') |
There was a problem hiding this comment.
🔒 Security & Privacy | 🔴 Critical | ⚡ Quick win
疑似真实账号密码被硬编码提交
对比同文件 L219-220 明显是占位符式假数据(13800000000 / test_password)，这里的 18928573369（合法真实号段格式）+ moyijie920.（形似"姓名拼音+数字"的个人化密码）看起来不像随手编造的测试数据，更像开发者本人真实的游戏账号密码被意外提交。
一旦合并，即使后续修改此文件，Git 历史中仍会保留该凭据，构成真实的账号泄露风险。建议：
- 立即用类似 L219-220 风格的明显占位符替换（如
'13900000000'/'switch_test_password'）； - 若该账号密码确实真实存在，请尽快登录游戏修改密码完成轮换；
- 如已推送到远端，评估是否需要清理 Git 历史（如已 push 到公开仓库，视为已泄露处理）。
🔧 建议修复
- monkeypatch.setattr(test_context.game_account_config, 'account', '18928573369')
- monkeypatch.setattr(test_context.game_account_config, 'password', 'moyijie920.')
+ monkeypatch.setattr(test_context.game_account_config, 'account', '13900000000')
+ monkeypatch.setattr(test_context.game_account_config, 'password', 'switch_test_password')📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| monkeypatch.setattr(test_context.game_account_config, 'account', '18928573369') | |
| monkeypatch.setattr(test_context.game_account_config, 'password', 'moyijie920.') | |
| monkeypatch.setattr(test_context.game_account_config, 'account', '13900000000') | |
| monkeypatch.setattr(test_context.game_account_config, 'password', 'switch_test_password') |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@test/sr_od/operations/enter_game/test_enter_game_flow.py` around lines 274 -
275, Replace the hardcoded account and password values in the test context setup
with clearly fake placeholder credentials consistent with the nearby test data,
such as the established placeholder style. Ensure no real or personalized
credentials remain in the file, and rotate the exposed credentials separately if
they are valid.
Comment on lines
+3
to
+6
| pytestmark = [ | ||
| pytest.mark.skipif(bool(__import__('os').environ.get('CI')), reason='需完整 SR 数据栈（screen 配置/模板/OCR），CI clean checkout 无；本地有数据则跑'), | ||
| # cal_pos 坐标精度因 SR 数据栈(template/地图)漂移偏 6-10px,暂 xfail 忽略;后续优化对齐数据栈后移除(2026-07-30) | ||
| pytest.mark.xfail(reason='cal_pos 坐标精度因 SR 数据栈(template/地图)漂移偏 6-10px,暂忽略;后续优化对齐后移除'), |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
set -euo pipefail
echo "repo files matching cal_pos test:"
fd -a 'test_cal_pos_for_.*\.py$|cal_pos_utils' . | sed 's#^\./##' | sort
echo
echo "target files excerpt:"
for f in \
test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py \
test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py
do
if [ -f "$f" ]; then
echo "===== $f ====="
wc -l "$f"
sed -n '1,180p' "$f" | cat -n
else
echo "missing $f"
fi
done
echo
echo "pytest mark usages in target directories:"
rg -n "pytest\.mark\.xfail|xfail|pytestmark" test/sr_od/operations/move/cal_pos_utils || trueRepository: OneDragon-Anything/sr-od-test
Length of output: 9218
🏁 Script executed:
#!/bin/bash
set -euo pipefail
python3 - <<'PY'
# Read-only verifier: show how strict vs non-strict xfail behaves for the exact
# pytest.mark.xfail(reason=...) form present in the target files, without installing or running repo code.
import ast
from pathlib import Path
targets = [
"test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py",
"test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py",
]
for path in targets:
p = Path(path)
print(f"===== {path} =====")
if not p.exists():
print("missing")
continue
tree = ast.parse(p.read_text(), filename=str(p))
for node in ast.walk(tree):
if isinstance(node, ast.Assign) and any(
isinstance(n, ast.Name) and n.id == "pytestmark"
for n in ast.walk(node.targets[0])
):
print("pytestmark assignment:")
print(ast.unparse(node))
# Determine whether any pytest.mark.xfail in the assignment omits strict=True.
def has_strict_xfail(n):
if isinstance(n, ast.Call) and isinstance(n.func, ast.Attribute) and n.func.attr == "xfail":
has_strict = any(
isinstance(kw, ast.keyword) and kw.arg == "strict"
for kw in n.keywords
)
return True, has_strict
return False, None
xfailed = [has_strict_xfail(n) for n in ast.walk(node)]
vals = [(v, s) for v, s in xfailed if v]
print("xfail calls:", vals)
for _, strict in vals:
if strict is False or strict is None:
print("NO-striX-t-xfail-found")
PY
echo
echo "pytest version if available:"
python3 - <<'PY'
try:
import pytest
print(pytest.__version__)
except Exception as e:
print("pytest not importable:", repr(e))
PYRepository: OneDragon-Anything/sr-od-test
Length of output: 1076
🌐 Web query:
pytest xfail strict=True non strict XPASS behavior default
💡 Result:
In pytest, the behavior of an xfail (expected failure) marker depends on the strict parameter, which determines whether an "unexpectedly passing" test (XPASS) causes the entire test suite to fail [1][2]. Non-Strict Behavior (Default) When strict is set to False (which is the default behavior if not otherwise configured), a test marked with pytest.mark.xfail will be reported as XFAIL if it fails as expected, or as XPASS if it unexpectedly passes [3][2]. In both scenarios, neither result causes the test suite to fail as a whole [3][4]. This default mode is particularly useful for managing flaky tests that you intend to address later [2][4]. Strict Behavior When strict is set to True (either by passing strict=True to the marker or by setting the global xfail_strict configuration option), the behavior changes [1][5]: - If the test fails, it is reported as XFAIL. - If the test unexpectedly passes, it is reported as XPASS and, crucially, this will cause the entire test suite to fail [1][2]. Configuration You can control the default strictness for all tests in your project using the xfail_strict or strict_xfail option in your pytest.ini, pyproject.toml, or other configuration files [5][6]. [pytest] xfail_strict = true Note: While the configuration option has historically been referred to as both xfail_strict and strict_xfail in various contexts and versions, both serve to set the default strictness level for xfail markers [5][6].
Citations:
- 1: https://docs.pytest.org/en/8.3.x/how-to/skipping.html
- 2: https://pytest.org/en/latest/reference/reference.html
- 3: https://docs.pytest.org/en/8.4.x/reference/reference.html
- 4: https://docs.pytest.org/en/stable/reference/reference.html
- 5: https://docs.pytest.org/en/stable/how-to/skipping.html
- 6: https://docs.pytest.org/en/9.0.x/how-to/skipping.html
统一收紧两处 cal_pos 测试的临时 xfail。
两处临时标记都是模块级且未设 strict=True，会把正常回归吞成 XFAIL，并在坐标精度修复后因 XPASS 不中断测试。将 xfail 限定到已知漂移 case，并设置 strict=True。
test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py#L3-L6test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py#L3-L6
🧰 Tools
🪛 Ruff (0.16.0)
[warning] 4-4: String contains ambiguous （ (FULLWIDTH LEFT PARENTHESIS). Did you mean ( (LEFT PARENTHESIS)?
(RUF001)
[warning] 4-4: String contains ambiguous ） (FULLWIDTH RIGHT PARENTHESIS). Did you mean ) (RIGHT PARENTHESIS)?
(RUF001)
[warning] 4-4: String contains ambiguous ， (FULLWIDTH COMMA). Did you mean , (COMMA)?
(RUF001)
[warning] 4-4: String contains ambiguous ； (FULLWIDTH SEMICOLON). Did you mean ; (SEMICOLON)?
(RUF001)
📍 Affects 2 files
test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py#L3-L6(this comment)test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py#L3-L6
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In
`@test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py`
around lines 3 - 6, Update the module-level pytest markers in
test/sr_od/operations/move/cal_pos_utils/test_sim_uni/test_cal_pos_for_sim_uni.py
lines 3-6 and
test/sr_od/operations/move/cal_pos_utils/test_world_patrol/test_cal_pos_for_world_patrol.py
lines 3-6: remove the broad xfail coverage, apply xfail only to the known
coordinate-drift cases, and set strict=True so unexpected passes fail while
unrelated regressions are not masked.
Comment on lines
+32
to
+35
| # 以下差分宇宙 / 模拟宇宙结算画面 screen 定义已不在 screen_info(重构时清理), | ||
| # get_match_screen_name 匹配不到 → 暂移除,待重新建档对应画面后补回(2026-07-30): | ||
| # div_uni_entry / normal_world / choose_curio / choose_equation / choose_bless / get_equation, | ||
| # sim_uni_get_reward / curio / bless |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | 🏗️ Heavy lift
不要通过注释掉映射来移除屏幕识别回归用例。
Line 32-35：这会让 div_uni_entry、choose_bless、sim_uni_get_reward 等场景完全退出测试循环，测试即使这些画面识别回归也仍会通过。若屏幕定义只是暂时缺失，请保留为带 strict=True 的显式预期失败或独立 skip；否则应在合并前同步恢复对应的 screen_info 与 fixture。注释标注的恢复日期是 2026 年 7 月 30 日，当前已到该日期，应重新评估该临时处理。
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In
`@test/sr_od/screen_state/test_get_match_screen_name/test_get_match_screen_name.py`
around lines 32 - 35, Restore the commented screen cases in the
get_match_screen_name regression coverage instead of silently removing them from
the test loop. For each affected screen mapping such as div_uni_entry,
choose_bless, and sim_uni_get_reward, either restore the corresponding
screen_info and fixtures so strict matching is validated, or keep an explicit
strict=True expected failure/independent skip with the 2026-07-30 temporary
handling re-evaluated.
DoctorReid
added a commit
that referenced
this pull request
Aug 27, 2026
对照 read_affixes/read_bosses briefing sibling(都有 fixture 集成,唯独简报敌人难度 reader 只测了纯函数 parse_enemy_difficulty)。fixture screens/货币战争-简报/default.webp → read_briefing_enemy_difficulty → 合法 int(0<N≤300;值跨局变不写死)。PASSED。 review round-5 borderline #1。
DoctorReid
added a commit
that referenced
this pull request
Sep 11, 2026
- test_cw_w502_deathbed_levelup_ev 整文件退役:②③④=math_proofs P21 已证命题(全参数网格81 cells负,proofs单篇+tools/cw/proofs可重跑)的 镜像复算,无增量保护;①点击阶梯锁并入 test_cw_state (test_levelup_clicks_ladder_matches_registry,改锁生产函数 xp_clicks_to_level,非镜像)。失去保护:无(证明承载)。 - test_cw_w620_migration_b1::test_r4_seam_functions_public_and_pure 退役: 存在性弱锁,排程规则在 w633 四触发锁、值域在 economy_cycle,更强锁 全覆盖。失去保护:无(留指针注释)。 - test_cw_w620_migration_b1::test_single_frame_equivalence_* 退役: ADR-0464 收口后对照臂(直调决策核)无被保护对象。失去保护:新旧管线 逐位对拍(迁移期通道)。 - test_cw_pivot_invariant::test_invariant_single_guard_location 退役: 源码级形状锁(收缩三原则①),行为面由同文件前两条冷却不变量锁覆盖。 失去保护:防守卫再分散的静态提示。 验证:w620 9P / pivot 2P / cw_state 11P 全绿(受影响直跑)。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
DoctorReid
added a commit
that referenced
this pull request
Sep 11, 2026
- w695条件退役2条(上层覆盖复核成立,失去保护注记入原文件):#3断环锁被test_cw_package_layout桶级矩阵守卫(kernel→decision全桶禁边,含函数级import)严格覆盖;#4注入契约被w633注入一致性锁(W636 A)同型对照覆盖。#1 identity(≤恰边界无他覆)与#2帧矩阵(刷价现读两行独占)复核不成立,保留 - w69锁3并删入w323(legacy fallback锁全链覆盖SellBench.income→decisions→query_economy,指针落w323 docstring);w69锁4与w510末两断言逐字重复,退役 - r333 director_last_state_gated延期解除(prep_director面已收口;复核为独占锁,保留现状) - slow_marks复测准入:adr0283 batch_discloses_guard_count 2.80s、capacity_guard 2.38s(共享sim setup抬过)、delta_pool_snapshot pool_fingerprint 2.06s;b36 mutation_kill复测0.01s自然出桶;adr0286/adr0306/battle_rung边缘复测0.04-1.13s低于阈不录入;r331 heavy已在桶维持 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
DoctorReid
added a commit
that referenced
this pull request
Sep 11, 2026
…_realization_chain 13 例(每锁带开关 off 臂零漂移断言;#0 开关组缺省态/#1 豁免通道回归/#2 时点守卫/#3 κ 折扣恒等式/#4 P29 辖域门含反向锁/#5-#7 support′ 衰减+P16 滞回/#8 金水位门/#9 部署显影/#10 D1 income 弱序/#11 D2 入口帧接管);adr0293 字段面锁随批登记 15 个 realization_* 新字段(布尔 8 全 False=第 1 态+占位参数 7,锁自身指引的登记路径,非改值);cw_quick 登记新锁组;观测硬依赖(bench_full_flag/分配器遥测键/board_next_tier)按边界声明只登记不实现,归 W793 后继批 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
DoctorReid
added a commit
that referenced
this pull request
Sep 11, 2026
…ate7/economy11/economy_gates15/mandate_decide20/mandate_lifecycle7/shop13/shop_budget9/equip15;17来源退役+7 mv已含);金钱不变量按用户裁决收缩为净守恒1+拒付fail-closed;事故背书10家族随代表迁入;规格偏差5项申报(335/83计数锁暂挂economy待归编#15) Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
出处 = criteria/levelup.xp_ledger_stop(定案④ xp_apply_clicks 推进算子单一源)+ shop 两发射位合取消费 + sim/checks/suspects.py D12。live level_max=10 判据不辖零变更锚与账本越过目标即停形态均入锁。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
… 命题 4 带界逐值锁) Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
锁 match_archive 跨档归属收敛守卫与核验的五条行为:装配收尾双计段 归零(正确归属持有、错误档案摘除)、源流在重装配收敛无剪枝记录、 源流失定向剪枝+显影+孤本禁动、孤立双档核验显影、水位线下双计档案 经 assemble_pending 收敛且幂等。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
4 条:①装配主路径从旧流源文件重建全部派生列(手工推演期望值,含 hp 步进 链/离场通道/恢复对账/开局面/取证链接/他局防漏入);②输入供给切片优先级 单元锁;③源流已清诚实退化;④load_archive 版本迁移路径保真(生产病灶形态)。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…cc1958)连带删除旧布局目录时该家族未随迁现役布局,src W2 接线批(8294d0bdb)申报的『等值性由 test_cw_match_final::test_match_final_version_stamps 对拍钉住』悬空;从 42ee42c blob(df00b97)逐字节恢复至 test/sr_od/application/currency_war/,commit 戳对拍=同一语句内双侧现取解析器(_resolve_code_commit() vs version_stamp.code_commit()),与 HEAD 落盘时序解耦,消除多批并行期测试进程存活窗内 HEAD 前移的假红(T-62/T-70 验收复跑两次咬中实证);fingerprint 对拍维持常量式(注册表进程内恒定,无时序差);验证=单文件 3 连跑 19 passed×3+全量快速集 578 passed/109 skipped/1 xfailed/0 failed+ruff 清洁;只 commit 不 push Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
锁 dominance 臂/prio2 燃料类拒成对第二张的分键计数(dominance_dead_stock_pair_blocked/ press_dead_stock_pair_blocked)、首张放行防误伤对照、dominance_band_wait 仅溢余带帧计。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
- test_cw_sell_bench_m3_reach:补 M4 缺员腾席外五处(m2_stockpile 腾席/C1 恒买腾席/凑息卖回拉/支付支撑卖主路径/兜底③④持有变现)帧构造参数化, 各断言首动作 SellBench 非 CloseShop,防 T-238 病灶任一处回退不红 - test_cw_xp_ledger_stop:D12 静默锁补非末轮超发构造(三行账本,末轮 判定移交 r10),消 docstring 与断言不符 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…/_contract + _cw_helpers 从 1cc1958^ blob 恢复至现役布局(test.sr_od.app→application 机械迁移);红/绿分诊 168 绿:红1 sim evidence 契约改写(引擎直写后 sim:engine: 前缀,合成口 sim:synthesized 保留离线面)、红2 read_shop_cards 桩随 full_ocr 签名对齐、红3/5/6 商店 payload 三态定长五槽取法改写(用户三态裁定 2026-09-13,x=槽下标,透传/离屏/沿用语义断言不变)、红7/8 chosen 写时点形状锚更新(09-14 断链修=重入观察裁决,落地验真后写语义不变);红4 effect_hooks 挂点锁不收录(源码扫描形态+修复需 import 在飞互斥面 cw_screen_buy_cards/cw_loop,挂 T-252-r1 报告缓交);验证=3 文件 168 passed+ruff 清洁 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…budget_authority 12 锁)、test_cw_migration_direction_layer 从 5fa044a^(9 锁,import 在飞面 mandate_v1.shop 但行为与现役一致故收)、test_cw_runnode_retire 从 3dc47a3^ 恢复行为腿 6 锁(supply/megastar 流转与预算语义);runnode_retire 第③腿源码零残留掩蔽扫描 2 锁不恢复(AGENTS §10.4.4 退役墓碑禁令,且公共单一源 fixtures/masked_scan.py 已随 1cc1958 删除,文件头分诊注);验证=3 文件 49 passed 零适配+ruff 清洁 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…形态 AGENTS §10.4.4 禁,API 契约防线归 review/规范)+economy/mdl import 排序 ruff 修复;验证=3 文件 48 passed+ruff 全清洁 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…红A 策略屏刷后重决策断言改写(逐卡刷新=终结动作,用户裁定 2026-09-14,点钮后本访问即交回,重决策归重入,decide 恰一次)、红B 结算分支摘 defer 复位腿(defer_count 载体已随 src 退役,全仓零消费点)、红C 遭遇屏仅改 chosen 写时点腿(2026-09-14 断链修迁挂重入观察裁决,本访问交回不写;刷后重读重决策形态遭遇屏仍保留,原断言恢复)、红D 收口锁摘除(AST 全目录扫描+继承闭包守卫+豁免登记门=包布局守卫形态,AGENTS §10.4.4 禁,五屏结构面由点名锁 test_closing_ops_inherit_base 辖);验证=4 文件 64 passed+ruff 清洁 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…零写/槽号不健康拒写/失读台账显影与席真空静默分型 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…e 从 3f91026^);分诊:①改写收绿=shop_projection 20 锁(actor 登记名随现役 CwScreenBuyCards、M5 域集八域随 v2 腿扩面申报、payload 三态取法/定长五槽计数、回执缺字段=自算理想执行语义随 T-185 改写)、_cw_helpers.cw4_card x 桩值随三态合成口改槽下标(旧 x=100 像素语义越界丢卡);②直接收绿=cap_override_link/refresh_ledger/p2_blood_band/sell_item_slot_predicate/transfer_golden(金样 fixture 归位后 13 锁全绿)/seat_recoverable_narrow 绿腿;③55 锁条件 skip 缓交=mandate_v1 策略发射域红锁(core_seat_vacate 12/sell_window_launch 9/no_target_three_arms 14/press_narrow 12+leak_ladder 6/seat_recoverable 2),红因域正被在飞批改造(T-243 拒因接线/批2b 词表),skip 挂到期条件=在飞任务 done 后重分诊;④摘除不恢复=cap_domain/dispatch_order_matrix(源码扫描形态 10.4.4)、cw4_key_closure(全文件键登记面扫描锁,与在飞面强耦合)、budget_disclosure 缓交移除(import 符号已迁在飞面 cw_screen_buy_cards);依赖恢复=leak_ladder/no_target_three_arms(press_narrow 前置);验证=12 文件 138 passed/55 skipped/0 failed+ruff 清洁 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…入将删路径/文件名,扫 src/sr_od 下文件名+测试函数名的注释引用,输出对账清单,退出码非零=悬空风险;自测=检出现役 test_cw4_key_closure 悬空引用 2 处/transfer_golden 4 处)+README「删除测试的工序」节(处置序:先消费清单→改写或摘除 src 声明→删测试→清单附 commit);背景=1cc1958b 整目录删除 261 文件未做反向对账一次打悬 ~60 处声明,本工序为治根防线;工具为删除时人工触发的工序脚本,非 pytest 测试(源码扫描形态测试仍按禁止清单禁) Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
…性归 refresh_ledger 末轮豁免/域内拦两锁承载;docstring 记语义演进) Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
- 新 test_cw_obs_action:注册表非终结行 + 宿主重观察恰一次 + overlay bail 透传 + 未接线域 AssertionError + 备战环 [Obs,HoldFrame] 环级锁 (重观察恰一次/decide 恰两次/帧代次 full) - test_cw_unified_action_3 终结集等价锁:机械非终结清单补 CwActionObsParam Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
- 注册完备锁豁免集移除 HoldFrame(终形 = 三刷新动作) - 空发射契约锁改锁 CwActionObsParam(scope='outer_loop') - obs 锁新增 outer 拦截环级判据 + F3 scope 值域校验锁 - 上报契约测试清理 HoldFrame 残留断言 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
主仓 e63d3e95f 级联;同帧对齐行为整条退役,零替代。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
主仓 71cb821f7 级联;build_refresh_expect 用例与 refresh_expect_mismatch 断言随机制退役(测试数 -1);顺带修 4 处既有 B023(loop 变量默认参绑定,行为等价)。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
benchchar-retirement 迁移前锁现状语义:合并素材装备继承到载体(场上载体优先);is_item_slot 候选恒 held 拒因 item_slot;swap 上报有开拓者形态归一、deploy_move 上报缺归一(现状缺陷锚,归一口统一后翻转)。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
载体迁移,语义断言逐位不变:merge 引擎/步回调锁种子与断言切内嵌 Unit 读;tracked 星级抖动门/溢出吸收/动作记账锁的 tracked 读口按新形状解包(bench=BenchSlot 内嵌 Unit,deployed=Unit 下标表排归属由下标派生);sim 开局手牌锁读 unit 形。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
既有用例补to_slot实参;原首空fallback/deployed_full行为锁改写为载荷直落+交换语义锁;新增部署落位执行面锁8条(执行链/写侧同源、占位交换、越界零写) Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
新增落位策略锁5条(三级排路由 comp覆盖>注册表>back兜底+注册表前排角色落前排+back_layout现值上限);trailblazer 现状锁翻转为期待归一(全文件唯一语义变更点,design §3 不变量5 申报缺陷修复);P2 to_slot 语义锁/前排保证锁/部署held锁/必需件首桶锁载体迁移 mandate_v1.deploy_plan(断言语义不变);出战链槽表分键随迁改名 deploy_plan_slot_table_unhealthy;launch_front 桩补 deploy_plan_available_fn 注入参;L1 全绿 823 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
键名保留 deploy_chain_slot_table_unhealthy(不改名),断言面不变 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
design §2.1:①board_by_row 行域分桶/多羁绊逐系计数/开拓者按排归一(欢愉前排不计)/未知名不计;②体系卡 _faction_count 身份×OCR 取 max;③char_first_faction 派生式(命中 factions[0]/无阵营 ''/未注册 '?')。载体 = 容器 GameState 行域直写。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
equip_allocation 签名随 P4 容器形翻转(BenchChar 紧缩表 → (front_row, back_row) 行域),锁断言零变。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
MandateFrame/deploy_plan 判定集群随主仓 P4 翻形,直构种子(bench=占席条目 BenchSlot,deployed=行域 Unit)同批迁移;16 个锁文件仅改种子构造与身份读解包口,断言值零改;deploy 载荷 faction 断言改 char_first_faction 同源读(§2.1 faction 类1锁,注册表派生)。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
纯卫生提交:未用导入/导入排序/死 helper 清理,断言面零变化。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
e_rounds 买刷截断(E 比值闭式 4/3)/proof assemble_lock_frame c_sat 席压项(空位 1 激活 vs 空位 2 免手写门归零)/economy_cycle O1 填补账(只扩最便宜 2 件 vs 3 件全扩);载体 = 容器 BenchView 直写节省工位效果帧。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
CwSimFrame 帧配方(cw4_state/cw4_feed/旧 cw4_box/旧 cw4_gs/_DuckFrame)退役,新 builder(cw4_gs/cw4_box)直写容器域(与观察链同写入口 observe),两条历史配方语义差异原样保留;shop_screen 字段承接原合成口 shop_open/shop_empty_off_screen 三分支语义(在屏/离屏/失读窗),bench_unread 承接未读域跳写;cw4_bc 改产 BenchSlot 占席条目,deployed 种子改行域 Unit(旧槽表序前 4 后余的分行语义在 _st 承接);M1 投影锁 16 测断言零增减,种子改裸帧配方(node_fill=False/shop_screen='off'),投影输出逐字节保形(改前/改后探针 20 块 JSON git hash-object 相同,diff 为空);BenchChar 直造种子面归零(残余 21 处 = 转换函数/观察链被测函数输入契约,随 P6/P7 载体退役);名字表达纪律落位(未知阵营=注册表外名,占位件=kind 直种)。L1 832 passed(基线 829+并行批 T-17 三锁);ruff 全过。 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
星级抖动门/店门锚定/漏斗/P2-1 空集锁/席位转换锁随读链直产换容器形状载体(BenchView/Unit 行):星级门语义逐条保留(锚定两帧/特效窗冻结/缺读打断/失读侧候选保持);对账重复槽号拒绝锁随直产退役(直产载体结构性健康,生产不可达状态不锁);P2-1 空集守卫移驻读链 read_bench_view/_bench_view_of 断言 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
bench_slots_of/deployed_slots_of 消费改容器读口(bench_units_of/bench_view_slots_of/deployed_rows_to_indexed/deployed_rows_of);tracked 种子直构(_tracked_bench,BenchChar+bench_from_compact 配方退役);随换形口退役锁 2 条(container_seat_view_conversions 转换腿/bench_view_of_slots_positional_mapping)+ 槽号守卫反帧显影腿(脏输入随 BenchChar 载体不可构造,模块头申报);种子签名 actor 统一 cw4_seed;语义锁一条不减(healthy 静默腿保留) Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
退役锁墓碑注释去符号名;trailblazer 测试本地形参 deployed_occupied 改名 deployed_taken(消退役函数名回声);断言零改动 Co-authored-by: DeepSeek Harness <noreply@deepseek.com> Co-authored-by: GLM-5.3 <noreply@z.ai>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概述
SR backend 11-tool 演进的配套测试（
open_and_enter_game→open_game+click_game/input_text/upsert_screen_area/delete_screen_area）。配对代码 PR：StarRailOneDragon#605。分支名
feat/mcp-backend-sync2与代码仓 PR 同名，便于 CI 同分支机制（ref: github.head_ref || main）匹配。改动
test_mcp_app.py（make_open_and_enter_game→make_open_game，工具集断言更新）conftest.py/test_click_game.py/test_input_text.py/test_run_slot.py/test_close_game.pyzzz-od-test最新，token 清单适配sr_od验证
ZenlessZoneZero→StarRail字面量、zzz_backend_run等）🤖 Generated with Claude Code
Summary by CodeRabbit
.gitignore：忽略__pycache__与所有.pyc字节码文件。