Pin fixes from independent U3 review

Why: Review found three code-level failures in approval phrasing, checkpoint counting and evaluator identity.

Evidence: Each targeted regression was RED then GREEN; Backend full Go, race and live fixture checks passed at 9126dae.

Impact: The Dev composition now includes the corrected Backend while real-model QA remains NOT_RUN.

Refs: U3 quality revision r0.2, independent final review
This commit is contained in:
2026-10-13 10:51:51 +00:00
parent 3437de97e5
commit 2b72f192df
2 changed files with 3 additions and 1 deletions
+2
View File
@@ -47,6 +47,8 @@ r0.2新隔离机制实例:Dev 85bed4d5d9bd8f014eea3e9bff4e2a3d2db40e13、Backe
r0.2正式组合:Dev 8c1fc390fec59bd80eb1224a1257daef32e95774 的 `make test && make build` exit0,完整 `SELF` 收据 `/tmp/fast-slow-u2-u3-r02-combination-20261013.json` 经工具导入下方;组合 Core c55ebbc、Backend 627b6e4、Frontend 329bdf5、UI cbbef50、Documents 2fdda63。独立临时 PostgreSQL 实例上 Backend runmgr/cardsvc/store 三包 PASS,`SELF` 收据 `/tmp/fast-slow-u2-u3-r02c/postgres-receipt.json` 已导入,实例已停止。本轮文档增补将推进 Documents/Dev SHA,最终精确HEAD组合门禁仍须重跑;独立代码复审和当前HEAD平台CI待执行。真实模型 r0.2 样本依旧NOT_RUN。
r0.2独立代码复审:`/root/u3_r02_final_review` 对 Dev 1fa58cb、Backend 627b6e4、UI cbbef50 的只读复审发现三项 Important:常见未批准/疑问误拒、幂等检查点重复计数、最终答复提醒改变受控评测run分组;无 Critical/Minor。Backend 9126daed9020ec98b9943f3e74d2af501ce0c6da 逐项 RED→GREEN 修复,完整 `go test ./... -count=1`、定向 race 与 live-tag 夹具均PASS(SELF)。复审者不评判新的真实模型语义和既有服务部署;前者仍为NOT_RUN且阻止合入,后者不在本轮范围。修复后的实际HTTP浏览器、Dev精确HEAD组合及平台CI仍待验证。
详见modules/documents/docs/reports/2026-10-13-fast-slow-u2-u3-validation.md。历史已验组合7fd5288(Backend6a2d85c、FE3087b06、Core6e38574、UI3b34dfd、Docs53a7d37)的make test/build通过(治理311、FE224、UI156、完整后端及构建);PG恢复和两项来源/阅读实际浏览器用例亦通过。完整SELF原始收据如下,独立代码复审已闭合,未把先前SHA冒充新对象。原六真实样本211974 reported tokens;A01/A02语义失败,A04预算耗尽保留部分发现。原始失败、引用核验与评測局限保留。
## 规划进展与下一步