Add an option to forbid planner skips so coverage can be guaranteed
还没有人认领这个 Issue。
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 65/100
- Issue 类型
- 功能
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- typescript
- 领域
- tooling
调研方向
Start by tracing the [review] configuration, planner decisions, packet-review dispatch, and coverage records described in the issue. Add the option with its default behavior, ensure disabled skips are dispatched or cause the run to fail, and verify that deterministic exclusions remain unaffected. Done means callers can enforce complete review coverage without parsing model-written skip reasons.
由索引模型根据 Issue 内容生成。
描述
Problem
The planner may decide a changed hunk does not need review and record a model-written reason such as "mechanical". On a 36-file diff reviewed with 0.5.6, 22 hunks were skipped this way in a single run.
For interactive use this is a sensible default - it saves calls on noise. For an automated gate it is a hole, because the decision is made by a model and there is no setting that forbids it. A caller who wants "every changed hunk was actually read by the reviewer" cannot express that today, and cannot distinguish "reviewed and clean" from "never looked at" without parsing the coverage records and re-deriving the policy itself.
The deterministic exclusions are not the problem: configured path rules and generated-file markers are auditable, and a caller can verify both independently. The planner skip is the one exclusion whose justification cannot be checked.
Proposal
A configuration switch that turns planner skips off:
[review]
allowPlannerSkips = false # default true, current behaviour
With it disabled, every changed hunk that is not excluded by a deterministic rule is dispatched to packet review. A run that cannot do so fails rather than reporting partial coverage as complete.
An alternative, if the skip is valuable enough to keep by default: a strict coverage mode that keeps the planner's judgement but records it as an explicit, machine-readable exclusion class distinct from configured rules, so a caller can decide its own policy per class. The current coverage records are close to this already - the missing piece is the guarantee that the class is complete and the ability to refuse it.
Why it matters
Anyone using this tool as a merge gate has to answer "what did the reviewer not look at, and why". Deterministic exclusions answer that. A model-written "mechanical" does not, and it silently sets the coverage denominator. One flag makes the tool usable for gating without changing the default experience for everyone else.
Related: #29 (per-stage model selection) is the cheaper way to reach the same goal from the other side - if the packet review on trivial hunks costs little, skipping them buys less.
- 主要语言
- TypeScript
- 星标
- 8
- 派生
- 1
- 平均合并
- 5 分钟
- 30 天内合并 PR
- 2
环境准备
这个项目没有提供开发容器、Dockerfile 或贡献指南,环境需要你自己搭建:先看它的 README,通用步骤见我们的新手贡献指南。
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
0xPolygon/codegenie 的其他 Issue
-
bug
难度 2/5 1-3 小时 新手友好度 76/100
-
bug
难度 2/5 1-3 小时 新手友好度 76/100
-
bug
难度 2/5 1-3 小时 新手友好度 78/100
-
bug
难度 2/5 1-3 小时 新手友好度 82/100
-
bug
难度 2/5 1-3 小时 新手友好度 82/100
查看 0xPolygon/codegenie 的全部 Issue
相似的 Issue
-
community first-timers-only good first issue hacktoberfest help wanted low hanging fruit up-for-grabs
难度 1/5 1 小时以内 新手友好度 78/100
lingdojo/kana-dojo#34344 · 1 条评论 · 5 个 reaction ·
维护者通常 1 天内回复
-
documentation good first issue
难度 2/5 1-3 小时 新手友好度 78/100
维护者通常 1 天内回复
-
bug priority:low
难度 2/5 1-3 小时 新手友好度 75/100
Konfirm-xyz/konfirm-frontend#369 ·
维护者通常 1 天内回复