v1.5.2 · Batch preflight and repair feedback

v1.5.2 · 批次预检与修复反馈 / Batch preflight and repair feedback

中文

• 新增 G2-A 完整布局批次预检：读取新鲜源数据，计算所有目标对象和分区后的状态，检查几何及保护属性，再交给选定后端执行。
• 输出结构化修复报告：记录对象、预期值、实际值、失败检查及候选动作；身份、引脚、属性异常和重载失败不会直接生成移动指令。
• 新增持久化重试账本：阻止同计划与同适配器原样重试，识别重复失败，限制连续无进展，并支持幂等记录。
• 补充后端版本、原始观测和读回证据的对应要求；制造与供应信息保持可配置，不引入硬编码费用或库存门槛。
• 新增 11 项离线回归测试；复测相邻数据恢复和布局工具。更新中英文 README 与操作说明。

English

• Add full-target G2-A layout batch preflight using fresh source data, geometry and protected properties before handoff to the selected backend.
• Emit structured repair evidence with expected/actual values, objects, failed checks and scoped proposals. Identity, pin/property and reload failures do not authorize movement.
• Add a persistent, idempotent attempt ledger that blocks unchanged retries, detects repeated failures and bounds unsuccessful attempts.
• Document adapter revisions, raw captures and native readback evidence. Keep manufacturing/supply policies configurable; add no hard-coded fees or stock threshold.
• Add 11 offline regression tests and rerun adjacent data/recovery and layout tests. Update both READMEs and the execution guide.

范围 / Scope

本版本支持已测量的无连线分区布局。它不拦截任意官方 API、不自动执行修复、不求解电气意图、不回滚实时 EDA；真实 EDA 采集、写入和保存重载仍需后端联调。测试使用合成数据，未验证实板或 Token 节省比例。

This release covers measured unwired block layouts. It does not intercept arbitrary official API calls, execute repairs, solve electrical intent or roll back live EDA. Native capture, application and save/reload still require backend integration. Tests use synthetic data; no hardware or token-saving benchmark is claimed.