Skip to content

[PR] fix: web blocks-pane message previews survive native client compaction (unified-storage fallback) | web 折叠块消息预览回查统一存储(原生压缩后不再全部 missing) - #4761

Open
ranxianglei wants to merge 15 commits into
masterfrom
wt-4738-ework-daemon
Open

ranxianglei wants to merge 15 commits into
masterfrom
wt-4738-ework-daemon

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

What & why

#4738:web 会话详情页折叠块里的消息预览只读 session.lastMessages(客户端最后一次请求窗口)。客户端做原生压缩(如 pi)把旧消息清出自己的窗口后,所有块消息预览瞬间变「该消息已不在会话里」——而统一存储里一行没丢。

修法:给存储栈加一条按身份定向查询,web 层对 missing 的叶子回查它:

  • UnifiedStore.getMessagesByIds(sessionId, msgIds) —— 走既有 (session_id, msg_id) 索引的分块 IN-list 查询(MESSAGE_ID_IN_CHUNK = 500),按 seq 顺序返回每个 id 的全部版本(split tool-call core 同 id 多行);
  • StorageBackend / SqliteBackend / MysqlBackend 增加同名方法;UnifiedShadow 加 sync/async 双生;SessionStore.unifiedMessageRowsByIds(...) 对外暴露:null = 统一存储不可用(保持 legacy 行为),[] = 可用但不存在(诚实 miss);
  • src/web/sessions-data.ts:detail render 时批量一次查询所有非驻留叶子 id,取 max-seq 行作为最新字节;messageLeaves() 与 buildMessageContent() 共用同一回退路径。

Behavior change (explicit)

Tests

  • 新增 tests/web-storage-leaf-fallback.test.ts:shadow 模式真实 persist 双写 + 模拟原生压缩剪窗;断言 leaves/click-through 回查存储生效、从未入库的 id 仍 missing、重发改字节时最新版本胜出;
  • 扩展 tests/storage-backend-contract.test.ts:getMessagesByIds 接口级契约(sqlite/mysql 双车道:子集过滤、seq 顺序、未知 id 空结果、多版本全返回);
  • 本地定向验证:typecheck clean;三个套件 16 pass / 0 fail / 1 skip(mysql 车道需 DSN);全量交给 CI。

无新配置面:未引入任何 env/config/flag。

Closes #4738

ci-check and others added 8 commits October 25, 2026 23:02
…ment

#4316 PR3/3 (restampRefsToLedger) re-homes citation numbers once on every
full-mode cold load; goldens pinning exact ref numbers are born-red. Re-pin
the 7 storage-lane tests (engine-proofs x6 + orphan-recovery x1) modulo that
renumbering — zero production code touched.

- golden-storage-drive.ts: per-turn citation-normalized digest (shaCiteNorm)
  and state digest modulo stripOneTimeAlignment (stateDigestAligned); the
  deep-sort helper is hoisted into golden-canonical.ts for sharing.
- corrupt-index-child.ts: new stripOneTimeAlignment normalizer — strips the
  refLedgerAligned stamp and re-homes refs onto canonical raw-id order so a
  clean full-mode read and the legacy fallback load compare equal.
- storage-engine-proofs.test.ts: wire legs assert shaCiteNorm equality vs
  the legacy restart leg and the committed goldens; state legs assert
  stateDigestAligned; rollback leg now spells BILI_STORAGE_UNIFIED=legacy
  explicitly (#3818 flip: absent resolves to full, deleting the env would
  vacate the legacy path).
- storage-orphan-recovery.test.ts: pins the content-addressable
  byRaw->byRef round-trip instead of a counter-era m00003 literal.
- golden/storage-render/*.json regenerated with shaCiteNorm/stateDigestAligned.

Converges the master-red storage-lane family (#4599 plan A, owner directive).
…dger parity masks

The #4316 one-time renumbering restamped messageRefs but left
state.tokenSnapshot keyed by the OLD refs, so post-restart bodies
carried stale per-message token counts (1.3K vs 1.4K) and the
full-mode-restart wire drifted off the legacy baseline (#4599).

persist.ts: two-pass remap (stage moves, delete old keys, then write
new keys) so entries whose new ref collides with another move's old
ref survive; fires only when superseded refs exist.

Aligned normalizer (corrupt-index-child.ts): the cacheLedger's
segment bookkeeping (be/bl/bs, unfinalized tail line), the derived
agg block, and per-line miss-attribution (prov/seam) are restart
experience, not content — no-restart and restart legs legitimately
differ there. Masked in the aligned view only; raw digests keep
full accounting. Goldens regenerated (stateDigest + aligned).

storage-engine-proofs 12/12, golden-storage-render 5/5 locally.
…persisted state (#4599)

The fork-survival test asserted immediately after the compress tool
response, but the fold lands in the live session state asynchronously
(#3573 replay semantics). Slow CI runners persisted the pre-fold state
('0 block(s) (compressions=0)' cold load) and the deepEqual diverged.
Bounded 5s wait on the in-memory fold; fast machines exit on the first
iteration. Product semantics unchanged.
The #4316 cold-load alignment reads the unified sidecar; its ingest of
just-flushed records completes asynchronously. Slow runners compared
the pre-ingest state and saw refLedgerAligned missing while the fold
itself had landed. Poll (bounded 5s) until the loaded state carries
the flag; fast machines exit on the first probe.
Git-for-Windows autocrlf checks the bundled .md posts out with CRLF;
the exact-match '---' fences then failed to parse and every post lost
its title/excerpt/date frontmatter (Windows CI: 'bundled meet-bili post
found' assertion). Normalize CRLF to LF inside parseFrontmatter when
the fence is CRLF-shaped; LF content is byte-identical as before.
Regression test with a CRLF fixture.
… storage mode (#4599)

The #4316 cold-load alignment sets refLedgerAligned only when the
unified store is available; config-clean machines (CI runners) run
legacy mode where it is an honest skip. The test asserted the flag
unconditionally, so every shard carrying this file was born-red on CI
while unified-config machines passed. Expectation now mirrors
store.replicasAvailable(); verified 49/49 in BOTH modes locally.
@ranxianglei ranxianglei changed the title fix: web blocks-pane message previews survive native client compaction (unified-storage fallback) | web 折叠块消息预览回查统一存储(原生压缩后不再全部 missing) [PR] fix: web blocks-pane message previews survive native client compaction (unified-storage fallback) | web 折叠块消息预览回查统一存储(原生压缩后不再全部 missing) Oct 25, 2026
…ter native compaction (#4738)

The detail-pane tree leaves and the click-through text read message bytes
only from session.lastMessages — the client's last request window. When a
client (pi/codex) natively compacts and prunes old messages out of that
window, every folded-message preview flipped to "missing" although the
unified storage log still holds every row.

Add a targeted identity read through the whole storage stack
(UnifiedStore.getMessagesByIds -> StorageBackend -> MysqlBackend ->
UnifiedShadow -> SessionStore.unifiedMessageRowsByIds): chunked
session-scoped IN-list queries on the existing (session_id, msg_id) index,
returning EVERY stored version per id in seq order. The web layer batches
one query per detail render for exactly the leaves absent from
lastMessages and falls back to the newest stored row (max seq); "missing"
now means absent from BOTH the client window and the unified log. Legacy
installs without a unified store keep the previous behavior unchanged.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

冲突已解除。把 PR 分支 wt-4738-ework-daemon rebase 到当前 master(bb7c24972),逐文件解掉 4 处文本冲突,force-push 为 89ddeebaf。GitHub 现报 mergeable: true、mergeable_state: blocked——blocked 只指等 CI/评审,不是冲突。合并仍由人执行。

按结构化八问:

  1. 是什么问题:本 PR(web 折叠块消息预览回查统一存储)的分支和目标分支 master 撞车,卡在 dirty 状态合不进去。
  2. 什么原因导致的:分支基于旧 base fe5ba5912,master 已前进 575 个提交,期间 PR 碰的 6 个存储文件被大改——尤其 store.ts 里 PR 依据的旧 getMessages 已被拆成 seqByRaw / cloneSessionRows / 带 retiredAt+opts 的新 getMessages,rebase 时 4 个文件(persist / store / mysql / sessions-data)文本冲突。
  3. 影响范围与严重程度:只影响该 PR 能否合入,不影响任何已发布行为;属调度阻塞类,非用户可见缺陷。解错才会引入回归,所以做了定向验证(见 5)。
  4. 怎么修复的:rebase 到 master,按「union of intent」解冲突——保留 master 新增的方法与功能(The "compressed structure" tree diagram for large sessions is scaled down to sub-pixel size (180 session blocks, fit ≈ 0 | 大会话的「压缩结构」树图被缩成亚像素(180块会话 fit≈0.05,方块只剩~6px) #4173 live-tree、#D5-5 render windowing、retired_at 语义),叠加 PR 净增的全栈读取链 getMessagesByIds(web-storage fallback)。store.ts 因 master 重写过大,改用「取 master 整版 + 手工补回 PR 三处净增(常量 + getMessagesByIds + mapMessageRow)」,避免巨型手拼出错。
  5. 是否有回归问题:审过——typecheck 全绿;PR 新回归测试 web-storage-leaf-fallback 2/2 过;storage-backend-contract + web-sessions + web-blocks-tree 合计 44 过 / 0 挂 / 1 skip(mysql lane 无 DSN)。高危模式逐项过:无阈值/时序/输出格式漂移,版本号未动(0.1.207 = master)。
  6. 是否破坏缓存:不涉及——纯读取路径新增,不碰前缀缓存/重试/游标。
  7. 是否改了设计方向:不涉及——沿用既有统一存储读取契约,未改架构/边界。
  8. 是否引入了新的配置:不涉及——无新 env/config/默认值。

行为变更披露(Deliverables #4):有一处非机械适配需要说明——给新增的 getMessagesByIds SQL 加了 AND retired_at IS NULL(sqlite store.ts + mysql.ts 两处)。原因:master 在 PR 之后引入了 retired_at 概念(#4390),且所有其它读取路径(getMessages/liveRowCount/seqByRaw/messageIdRows/uncoveredRows)都默认排除 retired 行;若不加,新的 web 显示读取会把已撤销/回退的消息内容也渲染出来,既与全部同级读取不一致,又会让 sqlite/mysql 两个后端行为分叉。旧值(PR 原样)= 不排除 retired;新值 = 排除,与 master 全体读取对齐。这是 union-of-intent 的必要修正,不是遗漏。

判定:冲突已解除、定向验证通过,进入正常 CI 把关即可;无设计取舍需拍板,无需开 decision-request issue。

@github-actions

github-actions Bot commented Oct 25, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: wt-4738-ework-daemon (ccd9071)

Latest build — install this exact version

npm install -g billion-context@0.0.5906-pr.4761.4578d0d

This exact version is immutable and reproducible. The rolling pr tag tracks the newest test build across all PRs.

Recent builds (commit → version)

commit version build
ccd9071 billion-context@0.0.5906-pr.4761.4578d0d build
89ddeeb billion-context@0.0.5843-pr.4761.15d9e60 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr4761.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

查重:不重复 —— 本 PR 是 #4738 的唯一实现(open+closed issue/PR 检索无并行修复;#4737 是折叠块面板功能本身,相关但机制不同)。

标签:bug pi scope/ui p2 —— web 会话详情页展示层缺陷(原生压缩后块预览全部 missing),数据未丢失、局部影响,故 p2 非 p1。

接单验证中(不重做):分支代码已逐项核对,CI 唯一红条 unused-exports 已定位为本 PR 引入(master 本地同门禁干净),正在分支上修复并重跑定向验证,完整八问结论等 CI 转绿后发。

ci-check added 2 commits October 26, 2026 04:58
…gate (#4761 review)

Review fix: StorageLeaf is only referenced within sessions-data.ts, so the
leading export tripped check-unused-exports (INTERNAL_ONLY). Drop the
keyword; behavior unchanged.
…#4761 review)

The chunked IN-list query orders each chunk by seq, so a batch of >500 ids
can return rows out of global seq order at chunk boundaries. Sort the
aggregated result once in both backends so the documented 'seq order'
contract holds for any batch size. Callers are order-independent today
(group-by-id / max-seq wins); this makes the interface invariant true.
@ranxianglei ranxianglei added bug Something isn't working p2 pi scope/ui labels Oct 25, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

验证完成(不重做)+ review 小修已推分支。八问:

  1. 是什么问题:客户端做原生压缩(如 pi)把旧消息清出自己的请求窗口后,web 会话详情页折叠块里的消息预览全部变「该消息已不在会话里」——而统一存储里一行没丢(Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738)。
  2. 什么原因导致的:叶子预览只从 session.lastMessages(客户端最后一次请求窗口的字节)取数;原生压缩剪窗后查找必然 miss,直接显示「已清理」。不是存储丢了数据,是展示层没有对 append-only log 的回查路径。
  3. 影响范围与严重程度:web 会话详情页的折叠块面板 + 叶子点看(renderDetail/buildMessageContent);触发条件 = 走过原生压缩的会话。p2 展示层缺陷:无数据丢失、不碰压缩路径与 wire、legacy 安装不受影响。
  4. 怎么修复的:存储栈加一条按身份定向查询 getMessagesByIds(sessionId, msgIds)——走既有 (session_id, msg_id) 索引、每批 ≤500 id 分块 IN-list、仅非 retired 行、每个 id 返回全部版本(split tool-call core 同 id 多行);web 层 detail 渲染时对全部非驻留叶子 id 批量一次查询,取 max-seq 行作为最新字节。null=统一存储不可用(保持 legacy 行为),[]=诚实 miss。
  5. 是否有回归问题:逐项审过既有行为——① legacy 安装(无统一存储)回查返回 null,渲染路径与旧行为逐字节一致(feat(web): session detail compact structure tab — tier tree, graph view and lazy originals | feat(web): session detail 压缩结构 tab — tier tree, graph view and lazy originals #3917 回归钉保持绿);② 无 missing 叶子的会话提前返回、零额外查询;③ retired 行维持不可见(retired_at IS NULL,与 Four storage-layer defects: identity hash contaminated with positional metadata (truncation triggers identity swap) / sw | 存储层四缺陷:身份哈希掺位置元数据(截断即换血)/清扫删除回退历史/渲染接缝图片只比长度/fork 谱系死列 #4390 一致);④ StorageBackend 接口纯追加(sqlite/mysql 两个实现均已补齐,全仓无其他实现者);⑤ 请求管道/压缩判定面/缓存前缀零触碰。review 发现并已在分支修掉两处(独立标注 commit):caa2ea41d 去掉 INTERNAL_ONLY 的 StorageLeaf export(原 head 唯一的 CI 红条 unused-exports,master 本地同门禁干净、非继承红);ccd907122 让 getMessagesByIds 在 >500 id 跨 chunk 时仍全局 seq 有序(文档契约「seq order」在 chunk 边界原本不成立;现调用方按 id 分组不依赖顺序,此修使接口不变量恒真)。
  6. 是否破坏缓存:不涉及(纯展示层读路径)。
  7. 是否改了设计方向:没有;展示层多一个只读回退源,写路径不变。
  8. 是否引入新配置:无(无 env/config/flag)。

行为变更披露(旧→新):旧=消息不在客户端最后窗口即显示「已清理」;新=先回查统一存储,两源皆无才显示「已清理」。legacy 安装(无统一存储)行为不变。

验证证据:新 head ccd907122 上本地 unused-exports 门禁 clean、typecheck clean、定向测试 8 pass/0 fail/1 skip(mysql 车道需 DSN);上一 head 89ddeebaf 的全量 CI 除 unused-exports 外 13 项全绿(含 test 车道与 mysql-contract)。当前 head 的 CI 在等 runner(舰队排队中),出结果我跟进一句。

结论:代码 ready-to-merge,待 CI 转绿确认后即可合并。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

CI 跟进:新 head ccd907122 的 workflow 已正常触发,全部 job 处于排队等 runner(约 35 分钟未开跑,与 #4599 所述舰队排队一致,非本分支触发问题)。上一 head 的全量结果(除 unused-exports 外全绿)+ 本地定向验证(门禁/typecheck/8 pass)已覆盖本 diff 的全部风险面,CI 出结果后我再报一句。

ci-check and others added 4 commits October 26, 2026 09:14
The ci-check lane's expectation change (c496cfb) exposed a real Windows
bug: stripInheritedProxy builds inlineEnv as a spread copy of process.env,
which keeps the original key casing (ComSpec) — Windows env vars are
case-insensitive but plain objects are not. planClientSpawn's exact
env.COMSPEC probe then missed it and silently degraded every wrapped spawn
to the bare-name 'cmd.exe' fallback (verified by repro on a real win32 box:
childEnv.COMSPEC=undefined while process.env.COMSPEC was set).

Fix: probe COMSPEC/ComSpec/comspec in planClientSpawn and planDshSpawn.
Verified on real Windows (Node 22): launcher-opencode-inline + fix-679 +
dsh-refresh 37/37; host tsc clean, 35 pass / 2 win32-skips.
…OMSPEC case-probe

c496cfb re-derived the win32 expectation from ambient process.env, which
still mismatches production: Node's process.env member access is
case-insensitive on Windows while the child env handed to the planner is a
plain object whose key keeps the OS casing (ComSpec), so the exact
env.COMSPEC probe fell through to the bare-name fallback. The assertion now
derives from the child's OWN captured env/options and probes
COMSPEC/ComSpec/comspec exactly like planClientSpawn does post-dfeea8a11 —
self-contained under any concurrent-test env churn in single-process CI
shards — and additionally pins windowsVerbatimArguments=true on the wrap.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

🔶 需要决策
(ai: vllm-qwen/qwen3.8-27b)

结论:PR #4761 与 master 冗余,建议关闭。

发现:PR 修的是 #4738(web 折叠块消息预览在原生客户端压缩后回查统一存储)。但 master 已独立实现同样的修复,且更完整:

  1. eb2fa5388e(2026-10-26 00:17)"web: hydrate folded-message previews from storage when the resident window prunes them (Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738)" — master 的 Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738 原始修复(messageContentByRefNums,ref 号键控)。比 PR 主修复 89ddeebaf0(2026-10-26 02:52)早 2.5 小时落地。
  2. 9e3d8e1cf8(2026-11-09 21:20)"fix: hydrate web block-detail leaves by stable msg_id, not restamped ref number"(Closes [Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029)— 键控从 ref 号改为稳定 msg_id(restamp-proof),并加 snapshotRefIndex 冻结快照回退。

master 的 web 层(src/web/sessions-data.ts)现在两条路径都有:

  • 叶子预览回水:messageLeaves → hydratedByMsgId(unifiedMessageContentByIds,msg_id 键控)+ snapshotRefIndex 快照回退
  • 点击全文回水:buildMessageContent → unifiedMessageContentByIds + storedMessageText

PR 的平行实现(getMessagesByIds/messageRowsByIds/unifiedMessageRowsByIds/storageLeavesFor/storedMessageById)与 master 的 messageContentByMsgIds/unifiedMessageContentByIds 功能重复。机械合并会让 web 层出现两条平行回水路径。

细微差异(PR 唯一可能增值处):PR 的 getMessagesByIds 显式 ORDER BY seq + 取 max-seq,保证同一 msg_id 多行时返回最新版本;master 的 messageContentByMsgIds 无 ORDER BY,web 层靠 Map.set 后写覆盖(依赖 DB 隐式 seq 升序,偶然等价但非保证)。PR 测试 tests/web-storage-leaf-fallback.test.ts 有一条 "re-sent message with changed bytes serves the NEWEST stored version" 钉住此行为。

选项:

默认(不决策时):A——关闭为冗余(master 已覆盖且更完整)。

当前状态:我停在冲突解决中途,未提交未推送。已解 3/5 文件(src/storage/backend.ts、src/storage/store.ts、src/storage/mysql.ts——含 master 新增的 hydrateMediaAsync(content, conn) 双参签名适配),剩 src/storage/shadow.ts、src/web/sessions-data.ts。等你拍板后再继续或直接关闭。

(注:本环境 tools/add-label.sh 不存在、Gitea labels API 未实现,needs-decision 标签未能程序化打上,靠本评论首行 🔶 需要决策 触发平台标记。)

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

本轮结构化摘要(细节见上一条 🔶 需要决策 评论):

  1. 是什么问题:PR [PR] fix: web blocks-pane message previews survive native client compaction (unified-storage fallback) | web 折叠块消息预览回查统一存储(原生压缩后不再全部 missing) #4761(修 Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738)与 master 冲突;核查后发现 master 已独立实现同一功能,PR 实为冗余的并行实现。
  2. 什么原因导致:master 的 eb2fa5388e(Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738 原始修复,比 PR 主修复早 2.5h 落地)+ 9e3d8e1cf8([Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029,稳定 msg_id 键控 + 快照回退)已覆盖 PR 的全部功能面。
  3. 影响范围与严重程度:仅 web 折叠块预览 / 点击全文回水路径(展示层);master 已正确工作,PR 不合并不影响任何用户。严重度:无外部影响,纯内部冗余。
  4. 怎么修复的:本轮停在决策点。已解 3/5 冲突文件(src/storage/backend.ts、src/storage/store.ts、src/storage/mysql.ts,含 master 新增 hydrateMediaAsync(content, conn) 双参签名适配),剩 src/storage/shadow.ts、src/web/sessions-data.ts。推荐选项 A(关闭为冗余,可选把 PR 的 newest-version 测试 salvage 进 master 的 tests/web-storage-fallback.test.ts)。
  5. 是否有回归问题:选项 A 无回归(master 实现是超集);选项 B 会回退 [Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029 restamp-proof 并丢快照回退(回归);选项 C 产生两条平行回水路径(隐患)。
  6. 是否破坏缓存:不涉及——纯 web 展示层,不触及缓存 / 重试 / 增量游标 / 去重键。
  7. 是否改了设计方向:未定,取决于 owner 选 A/B/C;推荐 A(不改方向)。
  8. 是否引入新配置:不涉及。

判定:停在决策点,等 owner 拍板(默认 A:关闭为冗余)。未提交、未推送。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态更新(daemon 重启后重触发,仍为 attempt 1):

我上一条 🔶 需要决策(floor 11)+ 八问摘要(floor 12)已发,owner 尚未拍板——floor 11 之后全是系统通知(徽标复位 / 重启 / 自动重试),无 owner 回复。

本轮实时核查:

处置:继续停在决策点。PR 冗余是设计取舍(关 / 留 / 双留),不是纯机械冲突,我不擅自机械合并——那会让 web 层出现两条平行回水路径(floor 11 选项 C 的隐患)。不重发完整决策评论(已在 floor 11)。

推荐仍是 A:关闭 PR #4761 为冗余(master 实现是已评审合入的超集)。可选把 PR 的 newest-version 测试 salvage 进 master 的 tests/web-storage-fallback.test.ts,把 master 偶然实现的行为显式钉住。

等你拍板 A/B/C 后再动手。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

进度汇报(响应系统重启提示)—— 不是卡住/挂起,是在等 owner 拍板:

已完成(阶段性成果):

  • floor 11:完整 🔶 需要决策(选项 A/B/C + 推荐 + 默认)
  • floor 12:八问摘要
  • floor 18:状态更新(实时核查:PR 仍 dirty、master 前进、冗余结论不变)

为什么推进不了(设计取舍,非死循环/网络挂起):
PR #4761 与 master 冗余——master 已独立实现 #4738(eb2fa5388e)+ #6029(9e3d8e1cf8,稳定 msg_id 键控 + snapshotRefIndex 快照回退)。当前 merge 进行到 3/5 文件(backend.ts/store.ts/mysql.ts 已解,含 master 新增 hydrateMediaAsync(content, conn) 双参适配),剩 shadow.ts + sessions-data.ts 这 2 个文件不是纯文本冲突:

  • shadow.ts:PR 的 messageRowsByIds vs master 的 messageContentByMsgIds(同功能不同实现,留哪个是取舍)
  • sessions-data.ts:web 层两条竞争回水路径(master 的 hydratedByMsgId+snapshotRefIndex vs PR 的 storageLeavesFor),机械「都留」会产出无法正确工作的代码

所以必须 owner 先拍板,我才能解完并推送。

请求指导(floor 11 的选项):

拍板后我立即继续。拍板前我不重复机械合并(避免制造两条竞争回水路径),也不重发完整决策评论(已在 floor 11)。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

(响应 nudge:我上一条已发完整进度汇报——非卡住/挂起,是在等 owner 拍板,不重复长文。)

一句话状态:PR #4761 与 master 冗余(master 已独立实现 #4738 eb2fa5388e + #6029 9e3d8e1cf8)。merge 进行到 3/5 文件,剩 shadow.ts + sessions-data.ts 是设计取舍(非纯文本冲突:PR messageRowsByIds vs master messageContentByMsgIds;web 层两条竞争回水路径),需 owner 先拍板(floor 11 的 A/B/C):

请拍板,拍板后我立即继续。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):PR 仍 open + mergeable_state: dirty(head 65f2b996a7 未变),master 仍保有 #4738 eb2fa5388e + #6029 9e3d8e1cf8 超集实现,floor 11 的 A/B/C 决策仍无 owner 拍板(floor 18–27 全是系统通知/自动重试)。

我不再重复完整决策评论(floor 11 已发,内容不变)。请 owner 拍板:

拍板前我停在决策点,不机械合并(避免制造两条竞争回水路径),也不重复贴长文。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):PR 仍 open + mergeable_state: dirty(head 65f2b996a7),floor 11 的 A/B/C 决策仍无 owner 拍板(floor 27–30 全是系统通知/自动重试)。

已多次报告(floor 11/18/23/25/28),不重复长文。请 owner 拍板 A/B/C(推荐 A:关闭 PR 为冗余,master 是已评审合入的超集)。拍板前我停在决策点,不机械合并。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):

已多次报告(floor 11/18/23/25/28/31),不重复长文。请 owner 拍板 A/B/C(推荐 A:关闭 PR 为冗余,master 是已评审合入的超集)。拍板前我停在决策点,不机械合并。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态更新(本轮实时核查):

已多次报告(floor 11/18/23/25/28/31/34),不重复长文。请 owner 拍板 A/B/C(推荐 A:关闭 PR 为冗余,master 是已评审合入的超集)。拍板后我立即继续(重新 clone 仓库 + 按拍板结果解冲突/abort)。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):PR 仍 open + mergeable_state: dirty(head 65f2b996a7),floor 11 的 A/B/C 决策仍无 owner 拍板(floor 36–39 全是系统通知/自动重试)。本地工作树仍被 daemon 重置(空 master 分支)。

已多次报告(floor 11/18/23/25/28/31/34/37),不重复长文。请 owner 拍板 A/B/C(推荐 A:关闭 PR 为冗余,master 是已评审合入的超集)。拍板前我停在决策点,不机械合并。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮全部实时核查,非引用旧结论):

本轮动作:

  1. 重新拉取仓库,实测冲突面(5 个文件:backend/mysql/shadow/store/sessions-data)——与 floor 11 判断一致:shadow.ts + sessions-data.ts 是两套同功能竞争实现(PR messageRowsByIds/storageLeaves vs master messageContentByMsgIds/hydratedByMsgId+快照回退),不是纯文本冲突,机械「都留」= 选项 C 的两条平行回水路径。
  2. 打上 needs-decision 标签(floor 11 时镜像 labels API 未实现打不上;本轮实测 API 已可用,经 API 打上;现标签集:bug, needs-decision, p2, pi, pr, scope/ui)。
  3. 本地合并实验已 abort,工作树干净,未推送任何内容,PR head 不变。

请 owner 拍板 A/B/C(floor 11 已列全,推荐 A:关闭 PR 为冗余——master 是已评审合入的超集,restamp-proof + 快照回退都在;可选把 PR 的 newest-version 测试 salvage 进 master 的 tests/web-storage-fallback.test.ts 把偶然行为显式钉住)。拍板前我停在决策点,不做机械合并。

七问:

  1. 是什么问题:PR [PR] fix: web blocks-pane message previews survive native client compaction (unified-storage fallback) | web 折叠块消息预览回查统一存储(原生压缩后不再全部 missing) #4761(修 Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738:web 折叠块消息预览在客户端原生压缩后回查统一存储)与 master 冲突合不进去;且 master 已独立实现同一功能,本 PR 是冗余的并行实现。
  2. 什么原因导致的:master 的 eb2fa5388e(Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738 原始修复,比 PR 主修复早 2.5 小时落地)+ 9e3d8e1cf8([Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029,稳定 msg_id 键控 + snapshotRefIndex 快照回退)已覆盖 PR 全部功能面;PR 分支切自不含这两个提交的旧 base,于是两边各写了一套。
  3. 影响范围与严重程度:仅 web 折叠块预览/点击全文回水路径(展示层);master 已正确工作,PR 不合并不影响任何用户。无外部影响,属内部冗余 + 合并调度阻塞。
  4. 怎么修复的:本轮不改代码——修法本身就是待拍板项。A=关闭为冗余(推荐、默认);B=留 PR 删 master 的(回退 [Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029,不推荐);C=双留(两条平行路径,不推荐)。无新配置面。
  5. 是否有回归问题:本轮零代码改动,无回归;选 A 无回归(master 是超集);选 B 会回退 [Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029 的 restamp-proof 并丢快照回退(回归)。
  6. 是否破坏缓存:不涉及——纯 web 展示层读路径,不触及前缀缓存/重试/增量游标/去重键。
  7. 是否改了设计方向:未定,取决于 owner 拍板;推荐 A(不改方向)。

判定:停在决策点等 owner 拍板(默认 A)。本轮未推送,PR head 不变。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):PR 仍 open + dirty(head 65f2b996a7 未变);master 前进 a259e0b8 → 548e3960(2026-11-18 14:08,#7083),#4738/#6029 超集实现仍在(hydratedByMsgId/snapshotRefIndex/messageContentByMsgIds 逐项验证);floor 11 的 A/B/C 决策仍无 owner 拍板(floor 42–44 为本轮之前的状态回复与系统通知,无 owner 回复)。

本轮零动作(needs-decision 标签已在位,工作树干净,未推送)。请 owner 拍板 A/B/C——推荐 A:关闭 PR 为冗余(master 是已评审合入的超集)。拍板前停在决策点,不做机械合并(会制造选项 C 的两条平行回水路径)。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

本轮七问(本轮动作 = 实时核查 + 状态回复,零代码改动、零推送):

  1. 是什么问题:PR 与 master 冲突(dirty)合不进去;根因是 master 已独立实现同一功能,PR 为冗余并行实现,冲突解法(留哪套)是待拍板的设计取舍。
  2. 什么原因导致的:master 的 eb2fa5388e(Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | Web blocks pane: message previews disappear along with native client compaction because they only read the client's late | web 页面里折… #4738)+ 9e3d8e1cf8([Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029)先于/独立于 PR 落地且更完整;PR 切自旧 base,两边各写一套。
  3. 影响范围与严重程度:无外部影响(master 展示层回水正常工作);仅合并调度阻塞 + 重复唤醒消耗。
  4. 怎么修复的:未修——修法即拍板项(A 关闭为冗余【推荐/默认】/ B 留 PR 删 master 的 / C 双留);本轮仅核查 + 状态回复,needs-decision 标签在位。
  5. 是否有回归问题:零代码改动,无回归;各选项回归分析见 floor 11(仅 B 有回归:回退 [Bug][ui] restampRefsToLedger rewrites messageRefs but leaves the unified store's ref_num stale — ~37% of block detail rows permanently render "no longer in the session" | [Bug][ui] restampRefsToLedger 重排 ref 后未同步统一存储的 ref_num:块明细里约 37% 的直属消息永久显示「该消息已不在会… #6029 + 丢快照回退)。
  6. 是否破坏缓存:不涉及(纯 web 展示层读路径)。
  7. 是否改了设计方向:未定,等 owner 拍板;推荐 A(不改方向)。

判定:停在决策点,等 owner 拍板 A/B/C(默认 A)。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):PR 仍 open + dirty(head 65f2b996a7 未变);master 前进 548e3960 → a446a3c1c8(2026-11-18 23:47,#6827),#4738/#6029 超集实现仍在;floor 11 的 A/B/C 决策仍无 owner 拍板(floor 45–48 均为 agent 状态回复与系统通知)。

本轮零动作(needs-decision 标签在位,工作树干净,未推送)。等 owner 拍板 A/B/C——推荐 A:关闭 PR 为冗余(master 是已评审合入的超集)。拍板前停在决策点。

七问:①问题=PR 与 master 冲突,根因是 master 已独立实现同一功能(eb2fa5388e #4738 + 9e3d8e1cf8 #6029 超集),PR 为冗余并行实现,留哪套是设计取舍;②原因=PR 切自旧 base,两边各写一套;③影响=无外部影响(master 回水正常),仅合并调度阻塞;④修法=未修,修法即拍板项(A 关闭【推荐/默认】/ B 留 PR 删 master 的 / C 双留),本轮仅核查+状态回复;⑤回归=零代码改动无回归(仅 B 有回归:回退 #6029 + 丢快照回退,见 floor 11);⑥缓存=不涉及(纯 web 展示层读路径);⑦设计方向=未定,等拍板,推荐 A 不改方向。判定:停在决策点等 owner 拍板。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

状态未变(本轮实时核查):PR 仍 open + dirty(head 65f2b996a7 未变);master 前进 a446a3c1c8 → 6453feeeb(2026-11-19 19:12,#7151),#4738/#6029 超集实现仍在;floor 11 的 A/B/C 决策仍无 owner 拍板(floor 49–51 均为 agent 状态回复与系统通知)。

本轮零动作(needs-decision 标签在位,工作树干净,未推送)。等 owner 拍板 A/B/C——推荐 A:关闭 PR 为冗余(master 是已评审合入的超集)。拍板前停在决策点。

七问:①问题=PR 与 master 冲突,根因是 master 已独立实现同一功能(eb2fa5388e #4738 + 9e3d8e1cf8 #6029 超集),PR 为冗余并行实现,留哪套是设计取舍;②原因=PR 切自旧 base,两边各写一套;③影响=无外部影响(master 回水正常),仅合并调度阻塞;④修法=未修,修法即拍板项(A 关闭【推荐/默认】/ B 留 PR 删 master 的 / C 双留),本轮仅核查+状态回复;⑤回归=零代码改动无回归(仅 B 有回归:回退 #6029 + 丢快照回退,见 floor 11);⑥缓存=不涉及(纯 web 展示层读路径);⑦设计方向=未定,等拍板,推荐 A 不改方向。判定:停在决策点等 owner 拍板。

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

1 participant