| content |
{"Commits":[{"Sha1":"86a7e52c6 {"Commits":[{"Sha1":"86a7e52c6b5843851881a91625bc944a51df2205","Message":"chore(governance): 刷新门禁运行与消费两份报告\n\n本会话多轮运行 run-foundation-gates 后的最新一次结果：阻断性 7/7 通过，仅记录 3\n（audit-platform-openings / check-workspace-consistency / check-domain-layer，\n均已在门禁消费登记里具名接入条件）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-15T00:54:26-07:00"},{"Sha1":"de0ebe04b4a27a0bba7f8981c635193a19d8e902","Message":"docs(framework): 框架 0.19.0 的变更请求受理记录落盘（另一会话的工作区改动）\n\n本提交的内容不是本会话所写，是框架侧会话留在工作区的 5 个文档改动，按\"全面提交\"\n一并落盘。内容自洽且已静置 35 分钟（最后修改 00:19，提交于 00:54）：\n\n- 基础设施仓框架变更请求-2026-09-12.md：受理表由五项全\"待\"改为 FR-1 受理并落地\n (0.19.0)、FR-4 / FR-5 受理（实现在下一个提交）、FR-2 / FR-3 暂不受理并写明理由\n （FR-2 触及写链四同边界与 check:write-guard 规则形态，框架侧无法独立判绿；\n FR-3 的 Fact 信封 v1 是 additionalProperties: false，加字段不能自称向后兼容，\n 属契约裁决）\n- CLAUDE.md：状态行推进到 v0.19.0，补 0.17.0 / 0.18.0 两节\n- 开发计划.md：补 0.17.0（compose 身份与端口，F14）与 0.18.0（G20 运维平面）两段进展\n- README.md：补 compose 变更请求的受理结论与\"扩散未决\"纪律\n- 框架变更请求-compose身份与端口-2026-09-14.md：受理结论回填\n\n本会话未改动其中任何一个字，只做落盘。若框架侧会话对措辞或时机有异议，按本仓惯例\n以后续提交订正，不 rebase 已落盘内容。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-15T00:54:21-07:00"},{"Sha1":"3cffd5406c379159d0074622c550ce5d4a81b17a","Message":"docs(governance): 回灌台账 U-37 / U-38 —— 接真实 DB 一轮跑出的两条下层缺口\n\n本轮把四个领域仓的证据等级从\"静态\"往\"真实 DB\"推，基座（base-framework-postgres-1 /\nredis-1）已在运行，按本目录\"每轮新建空验收库\"的口径逐仓建独立库。结果：\n\nU-37 工单仓在文档所述基座上根本跑不起来。prisma migrate deploy 在\n20260827015000_ticket_ai_pgvector 失败（P3009）——基座镜像 postgres:16-alpine 的\npg_available_extensions 里没有 vector 扩展。知识云用 DOUBLE PRECISION[] 绕开了\n同一需求所以不受影响，工单需要真扩展。二选一：换镜像（影响所有共用该实例的项目）\n或改列形态（影响检索实现）。在此之前工单的证据等级封顶在\"静态\"。\n\nU-38 HR 的 check-outbox-multi-instance 接上真实 DB 后可复现失败：\ndispatched=120, observed=0, subscriberStatus=ready，双后端一致，连跑两次相同。\n门禁跑完了全程，不属\"缺环境变量跑不起来\"那一类。本仓 .env 没有任何 REDIS/DATABASE\n项，无法判断是缺少未文档化的 relay/worker 配置还是真实缺陷——不在治理层臆断，\n证据已如实落盘，根因归仓 Owner。\n\n顺带记下：G15 库名守卫工作正常。首次用 juhai_hr_parity_* 被三条门禁正确拒绝\n（\"Database … does not belong to juhai_ai_hr_system\"），前缀按 package.json 的 name\n推导，是库名起错了不是守卫误报。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-15T00:50:36-07:00"},{"Sha1":"b4c4c61f455cd46e6c13bb28beb6ac506fee936a","Message":"docs(governance): 回灌台账 U-36 —— 门禁在缺实例时把\"跑不起来\"写成\"0 例通过\"\n\n知识云 docs-truth 报 3 处文档漂移，根因不在文档：CLAUDE.md 记\nbackend-behavior-parity 为 28 例，而机器真源被写成 0。\ncheck-backend-behavior-parity.mjs:140 写 `raw?.metrics?.… ?? 0`——vitest 因缺\nDATABASE_URL / REDIS_URL 跑不起来时不产出报告，于是\"跑不起来\"被记成\"0 例通过\"，\n再经 check-docs-truth.mjs:292 变成文档漂移。\n\ndocs-truth 自己已经察觉（第 3 条违规：\"动态数字负向探针只命中 7/9，门禁可能是\n假绿\"），但它只能报征状。\n\n**没有直接修**：最小改法（指标记 null + docs-truth 跳过比较）会让 :381 的负向探针\n失效，docs-truth 反而变成真的假绿。这是\"证据不可用\"如何穿过门禁链表达的设计问题，\n不是两行修补；治理层正在做同构的事（4d3efd5 拆开「报告没写结论」与「结论不合格」），\n本条等该口径定稿后按同一形态落到各仓。\n\n§3 第 9 条补一行：四个仓的需实例门禁共 8 条有同类风险；口径定稿前这些报告不应\n提交——本轮 HR 3 份、知识云 2 份即按此处理。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-15T00:41:51-07:00"},{"Sha1":"812ed430b9066c1fedebd7dd3aa5b0138f2ed644","Message":"fix(governance): D8 的构造性缺陷 —— 提交证据这个动作本身会让证据失效\n\n实测发现：门禁在 clean HEAD 上生成报告，报告记下当时的 gitSha；把报告提交，HEAD\n随即前进到新提交，于是报告永远指向父提交。按 D8 原本\"必须绑定当前 HEAD\"的字面规则，\n**落盘那一刻证据就自动作废**——这条判据不可能被满足。四个领域仓刚提交完新鲜证据，\nD8 依旧报 4 条，就是这个缺陷的现场。\n\n正确口径是看被验收的内容有没有变：从证据所绑的提交走到 HEAD，若沿途每个提交只动\nreports/，则源码与配置一字未改，证据仍然当前。这与 report-provenance.mjs 把\nreports/*.latest.json 排除在 worktreeDirty 之外是同一条道理，一个作用于工作树、\n一个作用于提交历史。\n\n不放宽到\"内容相似\"：沿途只要有一个提交碰了 reports/ 以外的东西，证据即失效；\n查不出文件清单的提交（如合并提交）一律判失效。\n\n修正后 D8 由 4 条降为 **0 条**——四个领域仓的证据第一次全部当前。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-15T00:36:22-07:00"}],"HeadCommit":{"Sha1":"86a7e52c6b5843851881a91625bc944a51df2205","Message":"chore(governance): 刷新门禁运行与消费两份报告\n\n本会话多轮运行 run-foundation-gates 后的最新一次结果：阻断性 7/7 通过，仅记录 3\n（audit-platform-openings / check-workspace-consistency / check-domain-layer，\n均已在门禁消费登记里具名接入条件）。\n\nCo-Authored-By: Claude Opus 5 \u003cnoreply@anthropic.com\u003e\n","AuthorEmail":"hillao@juhailaoluodeMacBook-Pro.local","AuthorName":"juhailaoluo pro","CommitterEmail":"hillao@juhailaoluodeMacBook-Pro.local","CommitterName":"juhailaoluo pro","Timestamp":"2026-09-15T00:54:26-07:00"},"CompareURL":"luoanwu/platform-governance/compare/cdb7f45657484b0ee996962444a060ca1f33c305...86a7e52c6b5843851881a91625bc944a51df2205","Len":42}... |