Skip to content

[PR] perf: verify stored sessions in bounded batches - #4540

Merged
ranxianglei merged 12 commits into
ranxianglei:masterfrom
z0d0dOr:2026-10-24_batched-storage-verification
Oct 30, 2026
Merged

ranxianglei merged 12 commits into
ranxianglei:masterfrom
z0d0dOr:2026-10-24_batched-storage-verification

Conversation

@z0d0dOr

@z0d0dOr z0d0dOr commented Oct 24, 2026 •

Copy link
Copy Markdown
Contributor

Fixes #4538. Complements #4532 without depending on it.

Sampled verification previously hydrated all stored messages and blocks, then reconstructed a second complete session while the live session remained resident. It now compares the header separately, reads blocks in batches of 32, and hydrates only the sequences referenced by tail/plugin snapshot pointers in batches of 32. Only message hashes and diagnostic fingerprints survive between batches. The synchronous and async paths share the comparison logic; SQLite and MySQL support the bounded reads.

The change preserves canonical normalization, block order, pointer order and duplicates, media hydration, orphan-original precedence, and existing drift recovery/diagnostics. Normal cold loads, persistence formats, configuration, and sampling cadence are unchanged. Session-row metadata/CCR/orphan JSON and individual oversized rows can still be large; this bounds the hydrated verification batches, not total process memory.

Rebased on master 0d9fdae. The branch carries #4549's verified-cache mtime fix with its original author retained, and makes the Windows redaction fixture pin/restore HOME and USERPROFILE. Two lifecycle fixtures now follow #4525's preserved-fold recovery contract. The relay avoids applying a raw-client tail offset to #4531's already-rendered judge baseline; the original dropped-prefix integration test remains unchanged and detects the defect again. The scope-attribution and lifecycle-ledger fixes are now supplied by master rather than duplicated here.

Validation on the refreshed branch:

  • 85 targeted tests passed on the rebased head: 83 cache/storage/redaction/epoch/reconciliation/judge checks and the two previously blocked chat golden/restart tests. One live MySQL contract test was skipped because BILI_MYSQL_TEST_DSN was not configured; async verification was exercised through the real SQLite backend adapter.
  • The Windows redaction file reproduced four failures before the fixture correction and passed all five tests afterward.
  • The batching tests cover 70 blocks across multiple pages, 80 referenced messages with 200 unrelated historical rows, reordered/duplicate snapshot pointers, media, ghost/BLOCKED refs, retained originals, and an unchanged live blockContents Map. Later-batch changes, missing/extra originals, and missing/retired messages still trigger drift and clear the watermark.
  • Typecheck, build, docs-sync generation with no artifact drift, lifecycle-ledger-guard, and git diff --check passed. Full-suite results come from remote CI on the pushed head.

A single synthetic memory comparison on Node v24.12.0 used 512 distinct 256 KiB block texts (128 MiB resident logical text) and 32 small tail messages in separate processes. The old full replay/digest flow peaked at 870.0 MiB RSS; the actual new verifier peaked at 346.3 MiB (60.2% lower). Both verified successfully. Rows were generated on demand to simulate database results, so this isolates verification allocation and excludes SQL/driver I/O; it is not a claim about total production memory.

@ranxianglei ranxianglei changed the title perf: verify stored sessions in bounded batches [PR] perf: verify stored sessions in bounded batches Oct 24, 2026
@z0d0dOr
z0d0dOr force-pushed the 2026-10-24_batched-storage-verification branch 4 times, most recently from 85afa71 to ab13c43 Compare October 24, 2026 13:39
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

冲突车道处理完毕:已把 master 合入本 PR 分支并推送,PR 恢复可合并。

过程:系统报告 merge_conflict(dirty)。接手时先核对 GitHub 实时状态——此时 API 已是 mergeable: true(状态 behind),即文本冲突已不存在,真正的问题是 PR 头落后 master 9 个提交(含 #4558 tier3-rework、#3635 摘要长度上限等)。处理方式:以 merge(非 rebase、非 force-push)把 origin/master 合入 PR 头,Git 自动合并全部文件,零人工冲突解决;合并树本地验证后推送到作者分支 2026-10-24_batched-storage-verification(快进式追加,作者原有两个提交与署名完整保留):

验证(合并树上,改动区域定向):typecheck 通过、build 通过(dist/index.js 等全部产物正常)、PR 自带定向测试 tests/storage-batched-verification.test.ts 5/5 通过。全量回归交给远端 CI(mergeable_state 当前为 blocked,即在等新头的 CI 检查跑完)。

八问:

  1. 是什么问题:PR [PR] perf: verify stored sessions in bounded batches #4540 落后 master,系统标记 merge_conflict,阻塞合并。
  2. 什么原因导致的:master 在 PR 分叉后前进了 9 个提交;dirty 标记来自检测时的瞬时/过期状态,接手时实际已无文本冲突,只剩 behind。
  3. 影响范围与严重程度:仅影响本 PR 的可合并性,无用户面影响;严重度=流程级(PR 合不了,功能本身无碍)。
  4. 怎么修复的:merge master 进 PR 头(保留作者提交的追加式合并,不 rebase 不改写历史),自动合并零手工编辑,推送回作者分支。
  5. 是否有回归问题:纯合并、无内容改动;合并树 typecheck/build/定向测试均绿;高危模式(语义漂移、阈值/时序/输出格式变化)逐项过——不涉及。全量判定以新头 CI 为准。
  6. 是否破坏缓存:不涉及。
  7. 是否改了设计方向:不涉及。
  8. 是否引入了新的配置:不涉及。

判定:冲突已解除,PR 处于可合并状态,等 CI 转绿即可由 owner 合并。

@ranxianglei

Copy link
Copy Markdown
Owner

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

PR Review — perf: verify stored sessions in bounded batches(#独立审核,未复用 issue 侧结论)

审核对象:合并 master 后的 head 166a857e0(conflict lane 自动合入 ba5ab0b,PR 自有文件零改动,renderTailStart hunk 保留)。本地定向验证全绿:storage-batched-verification 5/5、issue3919-shadow-drift-forensics 3/3、media-cas 3/3、storage-unified 14/14、issue4241-judge-render-baseline 1/1、epoch-confirm-lifecycle 3/3、issue2915 15/15、ci-pretest-cache 8/8、web-storage-diagnostics-redaction 5/5;typecheck 干净;build 干净。

结构化八问

  1. 是什么问题: 采样式存储验证每触发一次(每 20 次 ingest)就把整个会话的消息+块全部读出、水合、重建第二份完整会话再双 digest——单次验证的内存随会话规模线性增长(0.1.204 OOM 现场的来源之一)。
  2. 什么原因导致的: 验证腿按「全量回读 + 整体重建 + 整值 hash」设计,无界是设计属性而非实现瑕疵;采样周期只降频不降单次内存。且重建视图实际只消费 tail/plugin-snapshot 指针引用的少数消息行,其余读取纯浪费。
  3. 影响范围与严重程度: 所有 unified-storage 持久化会话,每 20 次 ingest 一次;大会话(数百块/大块摘要)出现数百 MB 级瞬时 RSS 尖峰,小会话无感。p2 性能/内存类,可用性不受影响。
  4. 怎么修复的: 分部件锁步比较器(BatchedViewVerification):header(含 state 非块字段/meta/stats/metadata/contentStore)单独比;块按持久化 ord 顺序 32 行一批流式比完即弃;blockContents 按 key 逐个比(当前块覆盖同名 orphan,journal 剩余收尾);tail/plugin-snapshot 只按引用到的 seq 以 32 为批取行(参数化 IN,空集短路)。批间只留每槽 sha8 指纹+gap 计数。canonicalize 无跨对象上下文,部件划分恰为顶层 key 全集,故「各部件相等 ⟺ 原整值 digest 相等」,判定语义等价;presence 失配由 header 部件(defined→[] vs absent 归一)兜住。sqlite/mysql 两后端同一 opts 管道,分页加 block_id 平局键保证稳定。
  5. 是否有回归问题: 逐项过既有行为——无 opts 的 getMessages/getBlocks SQL 与原来逐字相同(load/export/诊断路径不受影响);采样周期不变;drift 恢复链(watermark 清除、driftedSessions、streak=3 冻结升级、日志行前缀)字节级保持(issue3919 测试钉住并通过);媒体水合在 seq 路径保留(media-cas 通过,新测试含图片消息);orphan-original 优先级保留(新测试 extra-orphan 用例);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 retired 行排除保留(retired 用例漂移成立);重复/缺失指针语义保留(pluginSnapshot 重复指针用例通过,缺失→gap→漂移)。行为变更披露(旧→新+原因): ① async 腿新增 warnTailGaps 告警行(与 sync 腿对齐,原先 async 静默);② 内部不再计算/输出整值 digest(该值本就私有,无任何外部消费者,corpus 钉在测试内自算);③ 顺带捆绑了 relay.ts 的 Egress Judge false positive on Snapshot Render Lane I2: prefix comparison baseline needs to be aware of the render view | 出站判官对快照渲染车道误报 I2:前缀比对基准需要感知渲染视图 #4241 judge 门槛修复(rendered turn 不再套用 raw-inbound 尾边界,本地 issue4241 测试通过)——属 CI 红修复,超出本 issue 范围但政策允许同 PR 携带,已单独列出。
  6. 是否破坏缓存: 不涉及缓存/重试/增量游标/去重键策略;ingest watermark 逻辑未动,漂移时清除行为有测试钉住。
  7. 是否改了设计方向: 无转向——仍是 ingest 后采样回读验证,只是内存画像从 O(整会话) 变 O(批 + 被引用消息)。
  8. 是否引入新配置: 无。VERIFY_BATCH_SIZE=32 为内部常量,未进配置面。

合成内存对比(issue 要求的验证项,PR 内未附,审核方代跑)

300 块 × 300 KiB(≈90 MiB 历史),Node 22.23.2 linux,--expose-gc,gc 后测验证腿 RSS 增量:旧路径 553.9 MB vs 新路径 7.2 MB(约 77×)。与 #4529 的流式 digest 结论方向一致。

发现与建议(均不阻塞)

结论: 可合(N1–N3 非阻塞;等 CI 车道在合并 head 上转绿 + owner 合并)。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

查重: 不重复。 本 PR 修 #4538(采样验证全量水合 + 重建第二份完整会话 → 内存尖峰),与 #4532 互补且无 diff 重叠;无其他 batched-verification 在途 PR。PR 内三处附带改动各自有源、均不构成重复:CI commit = #4548(专门 PR #4549 已关未合,本 PR 成为载体);relay.ts 改动 + 两处测试期望更新 = #4525 train 合并的残留损伤(#4554 家族,专门修复 PR #4556 已关未合,对照当前 master 代码确认该缺陷仍在)。

标签: pr, bug, scope/storage, p2(存储验证路径的性能/内存缺陷;触发条件窄——verify 采样命中大会话——故评 p2 非 p1)。

审查对象: 当前 head 166a857e0(含系统冲突道把 master ba5ab0bce 并入的自动 merge commit;净 diff 与合并前一致,11 文件 +430/−94)。核心结论:未发现代码级阻塞项;建议补完 PR 正文披露后、待 CI 收敛即可合并。 独立复核证据:本地跑 11 个定向测试文件(新增测试 + storage/drift/orphan/backend-contract/redaction/ci-pretest-cache/epoch-confirm/issue2915 回归面)62 pass / 0 fail / 2 skip(MySQL contract 无 DSN 跳过,既有);typecheck + build 绿。

结构化八问

  1. 是什么问题: 采样验证每次运行时,把会话的全部存储消息/块都水合出来、在活会话仍驻留内存时重建一份完整第二会话——大会话下验证窗口内内存近似翻倍。
  2. 什么原因导致的: 旧实现是「重建整个视图后做一次整体 digest 比对」,比对粒度是整视图级,所以必须物化全量视图;缺的是有界执行策略,不是比对逻辑本身错。
  3. 影响范围与严重程度: 只影响 verify 采样窗口的内存峰值;采样节奏、正常冷加载、持久化格式均不变。大会话 + verify 命中时存在进程级内存压力风险,但无数据丢失、无功能停摆 → p2。
  4. 怎么修复的: 把「一次大 digest」拆成共享单一投影的分层比对:BatchedViewVerification 里 header 单独比(withoutBulk 剥离 bulk 字段);块按持久化顺序每页 32 个走位比对(position + contents digest);消息只水合 tail/plugin snapshot 指针引用的 seq(每批 32),跨批只留 sha256(canonicalize) + slot 指纹。同步/异步共用同一比较器;SQLite/MySQL 后端各加有界读(seqs IN 子句 / LIMIT OFFSET),均为向后兼容可选参数。
  5. 是否有回归问题: 未发现。覆盖度核对:新分部件比对的并集恰好覆盖旧单 digest 视图的全部字段(header 含 messageRefs/deadRefs/contentStore/lastMessagesFolded/meta/stats/metadata;state.blocks 按位;blockContents 经 addBlocks+orphans+residual 三路;lastMessages/pluginSnapshot 按原始指针序)。media 水合、orphan-original 优先、retired 排除、重复/乱序指针语义均保持且有测试钉住(含「200 条无关历史行不得被水合」的硬断言:80 ≠ 280)。两个实证:① 反事实实验——把本 PR 改过的两个测试文件换回 base 版本在当前代码上运行,双双失败(#4260 epoch confirmation ... all-missing、escalation latch recovery),证明这两处期望更新是对已合入行为(a5d1994/[PR] merge: tier-1 consolidation train — 37 green PRs as one (conflicts resolved, #4294 reverted) #4525「epoch confirmation pending 期间保留 fold snapshot」)的陈旧测试修复,不是本 PR 引入的新行为;② CI shard 对照——新 head 的 6 个红 shard 全部 ⊆ master(ba5ab0b)自身的 8 个红 shard,且 master 红的 windows-22/3、windows-22/4 在本分支转绿:本 PR 零新增红,剩余红属 master 基线红风暴(其中 Master CI is fully red: the dist restored from pretest cache carries old archive timestamps, which are misidentified as | master CI 全红:pretest 缓存恢复的 dist 带归档旧时间戳,被 #3330 的新鲜度守卫误判为 stale,所有测试道失败 #4548 的 dist-mtime 问题随本 PR 合入即修)。
  6. 是否破坏缓存: 否。验证对 store 纯读;drift 路径的水位清除/freeze/升级逻辑与原先逐字一致;出站渲染与前缀缓存不受影响。
  7. 是否改了设计方向: 否。verify 语义(drift 判定、诊断格式、升级策略)不变,仅执行策略从「全量物化」改为「有界分批」;store API 加向后兼容可选参数,现有调用方行为不变。
  8. 是否引入了新的配置: 无。VERIFY_BATCH_SIZE=32 为内部常量,不进配置面(config-surface-guard 绿)。

行为变更披露(旧→新 + 原因):

请作者补一项(非代码阻塞): PR 正文目前只描述 #4538 主体,请加一小节列明三处附带件及其来源——① CI commit(Fixes #4548,吸收已关闭的 #4549);② relay.ts hunk(#4554 家族 #4241 碰撞修复,#4556 关而未合、master 仍带缺陷,本分支为载体);③ 两处测试期望更新(陈旧测试修复,反事实证据见第 5 问)。否则 reviewer 无法追溯这些 diff 的来历。

收工判定: 代码可合;待作者补正文披露 + 确认合入后红 shard 随 #4548 修复收敛(或归入 master 基线线处理)。合并权在 owner,我不代合。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

CI 自动修复车道(attempt 2/2)— unused-exports 判定与处置

head 318a2cf 的 CI(run 40113423871):12 绿 / 1 红(unused-exports)/ 3 skip(test、static-gates、red-signal——因该门禁失败被 !failure() 跳过,未实际运行)。

结构化八问

  1. 是什么问题: unused-exports 静态门禁报 INTERNAL_ONLY src/agent/client-build-note.ts: ClientBuildCheck(导出但全仓零外部引用,仅本文件内使用)。
  2. 什么原因导致的: master 继承红,与本 PR diff 无关。[PR] dsh lane: note in-session when running an unverified @deepseek-ai/dsh client build (soft, exact-match) #4545 train 落 master(0fe7589,晚于本 PR 原 base ba5ab0b)移除 pi/omp/opencode 车道调用点时,de-export 清扫漏了这个 type-only 导出。证据:在不含本 PR 任何改动的干净 origin/master(146b34f)worktree 本地跑同一门禁,报错逐字节相同;CI Control Center: Centralized CI Failure Handling & Status Sync (Comment all CI issues here; do not open new issues) | CI 总控: CI 失败集中处理与状态同步(所有 CI 问题在此评论, 勿开新 issue) #4599 总控盘点(floor 20/24)已确认同条为 master 四条常红之一,修复载体 [PR] fix(ci): clear the two master-red gates in one branch — unused-exports (#4602) + the #3890 LIVE_FILES tripwire (#4604) #4606(在途,其 CI 该门禁已绿)。
  3. 影响范围与严重程度: 信号级(检查矩阵),无用户侧功能影响;所有 base ≥ 0fe7589 的在途 PR 均继承此红(CI Control Center: Centralized CI Failure Handling & Status Sync (Comment all CI issues here; do not open new issues) | CI 总控: CI 失败集中处理与状态同步(所有 CI 问题在此评论, 勿开新 issue) #4599 已有多条同签名报告)。
  4. 怎么修复的: 按几行级小修纪律(CI 失败类小修首选当前 PR 顺手修、多 PR 重复修明确允许)+ 风暴内既有先例([PR] fix(storage): block interval bounds survive post-fold resaves (#4592) #4607 c560a2a / [PR] fix: route codex context-only prewarms to the probe relay via the upstream proxy #4595 e425e78 / [PR] fix(release): label-gated head-tree publish for conflicted carve-out ship PRs #4596 23baf2b 同款一词去 export),独立标注 commit 去掉 ClientBuildCheck 的前导 export(一行改动;非 entry 文件、无 dts/公共面、全仓 grep 零外部引用)。新 head 2e4b196,已 push;本地门禁 ✓ clean (6 waived) + typecheck 绿。这是 [PR] fix(ci): clear the two master-red gates in one branch — unused-exports (#4602) + the #3890 LIVE_FILES tripwire (#4604) #4606 提交 1 的重复修复——[PR] fix(ci): clear the two master-red gates in one branch — unused-exports (#4602) + the #3890 LIVE_FILES tripwire (#4604) #4606 合入后本分支 rebase 时丢弃该 commit 即可(与 [PR] fix(pi): inline mode now intercepts codex Responses WebSocket traffic (#4526) #4559 floor 11 同款收尾)。
  5. 是否有回归问题: 无——type-only 导出、零外部引用(门禁扫描覆盖 src/tests/scripts 全文)、不改变任何运行时行为;commit 仅此一行。
  6. 是否破坏缓存: 不涉及。
  7. 是否改了设计方向: 不涉及。
  8. 是否引入新配置: 不涉及。

剩余状态(需要 owner 知悉)

结论: PR 侧可做的已全部做完(unused-exports 已修、其余红均为 master 风暴继承且已有在途修复载体)。等 #4606 合并 → 本分支 rebase(丢重复小修 commit)→ owner 合并本 PR;无需作者侧动作。

ranxianglei pushed a commit that referenced this pull request Oct 24, 2026
…ot this PR's

master 146b34f fails the unused-exports gate on this exact line
(INTERNAL_ONLY src/agent/client-build-note.ts: ClientBuildCheck): the
type-only export #4545 introduced was missed by the 4d34abf de-export
sweep. This PR touches none of it — every PR rebased past that master
inherits the red (#4599 storm log). Same one-word fix as #4606 commit 1
and the in-branch precedents (#4607/#4595/#4540/#4618/#4622): drop the
leading export on a symbol referenced only within its own file. Drop
this commit when rebasing after #4606 merges.
ranxianglei pushed a commit to z0d0dOr/billion-context that referenced this pull request Oct 25, 2026
The engineLane=server metadata added by f590da9 explains the entire drift: removing that one field reproduces the old metadata digest. Regeneration changes only metadata and the enclosing digest; all wire bodies, state, originals and statistics remain identical.
Compare unified reads with an independent envelope reader plus the documented migration. Persist alignment before rollback/corrupt-index checks, select legacy explicitly, and assert the recovered raw ID under its ledger ref. Keep checkpoint writes deterministic so sampled ingest cadence cannot change the seed between reader legs. All 28 targeted storage tests pass; no product migration or digest assertion is disabled.
z0d0dOr and others added 6 commits October 28, 2026 00:59
Reconcile shared storage proofs with upstream citation-normalized restart comparisons, retain legacy/full fork coverage and Windows handle cleanup, and preserve both storage preview and batched verification types.
Windows SIGTERM terminates the child instead of running its graceful
shutdown handler. Trigger the existing loopback shutdown route on Windows
so the draining-marker regression exercises the same lifecycle as POSIX.

Wait for child close before deleting its database, destroy sockets owned
by the silent TCP fixture, and stop health polling when the child exits.
Keep both regression scenarios and their behavioral assertions intact.

Validation: the isolated CI-preloaded serial test file passes 6/6 and exits
naturally in 11.58 seconds; npm run typecheck and npm run build pass.
Windows CI can reset an aborted probe connection instead of ending it.
The silent TCP fixture did not handle errors on accepted sockets, so
ECONNRESET became an uncaught test exception after the probe finished.

Destroy reset sockets and add a real resetAndDestroy regression against
the same fixture. Keep the existing port-owner and ladder assertions.

Validation: the new regression reproduces ECONNRESET before the fix;
the isolated CI-preloaded serial test file then passes 7/7 and exits
naturally. Pure tsc and git diff --check pass.
tests/issue4982-draining-twin-ladder.test.ts: both sides independently fixed ranxianglei#5050.
Ours (858cb70+47b54b36): socket error->destroy + close() wrapper + platform-branched shutdown.
Master (0a1f534+9bbf3165): /__bili/shutdown convergence (ranxianglei#3544) + peer-abort-safe squatter fixture.
Take master's canonical version (shipped in v0.1.209); our duplicate fix drops, same
precedent as the dropped ranxianglei#4606-duplicate unused-exports commit.

# Conflicts:
#	tests/issue4982-draining-twin-ladder.test.ts
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

上游冲突已解决并推送回 PR 分支:d87d43e3..4b759b31(纯增量 push,作者原有提交原样保留,无 rebase、无 force-push)。PR 当前状态 mergeable(仅 behind——master 在工作窗口内又从 5594c131 前进到 6b298c37,不构成冲突;合并时 GitHub 会对齐最新 base)。

八问:

  1. 是什么问题:PR 分支与 master 分叉后合并出现内容冲突(merge state: dirty)。冲突根源:双方各自独立修复了同一个 Windows 测试失败(windows-latest CI will always be red: the drain e2e in #4982 uses POSIX-only kill(SIGTERM) to trigger graceful shutdow | windows-latest CI 必红:#4982 的 drain e2e 用 POSIX 专属的 kill(SIGTERM) 触发优雅退出,Node 在 Windows 上是直接硬杀 #5050)在 tests/issue4982-draining-twin-ladder.test.ts 里的表现。
  2. 什么原因导致的:本 PR 分支在等待评审期间自行加了 windows-latest CI will always be red: the drain e2e in #4982 uses POSIX-only kill(SIGTERM) to trigger graceful shutdow | windows-latest CI 必红:#4982 的 drain e2e 用 POSIX 专属的 kill(SIGTERM) 触发优雅退出,Node 在 Windows 上是直接硬杀 #5050 修复(socket error 处理 + 平台分支 shutdown);同期 master 在 v0.1.209 发布列车里合入了同一问题的官方修复(0a1f534c + 9bbf3165:收敛到 /__bili/shutdown + squatter fixture 抗 peer abort)。两个独立修复落在同一文件。
  3. 影响范围与严重程度:仅影响本 PR 自身的可合并性(dirty 挡合并),运行时行为零影响——本次合并从 master 带入的增量不含任何 src/kernel 代码变更(delta 纯为发布列车版本号、CI、测试文件、文档,已逐文件清点)。
  4. 怎么修复的:把 master 合入 PR head(增量 merge commit),唯一冲突文件整体采用 master 的规范版本(与已发布的 v0.1.209 字节一致,blob hash 核对过)。先验证了本方对该文件除两个 windows-latest CI will always be red: the drain e2e in #4982 uses POSIX-only kill(SIGTERM) to trigger graceful shutdow | windows-latest CI 必红:#4982 的 drain e2e 用 POSIX 专属的 kill(SIGTERM) 触发优雅退出,Node 在 Windows 上是直接硬杀 #5050 修复提交外无任何独有改动——丢弃自方版本不丢东西。与此前丢弃重复的 unused-exports 修复([PR] fix(ci): clear the two master-red gates in one branch — unused-exports (#4602) + the #3890 LIVE_FILES tripwire (#4604) #4606 先例)同一处置原则。
  5. 是否有回归问题:无。审计了此前已在分支上的合并提交 d87d43e(git merge-tree 对照模拟自动合并):它是纯自动合并 + 一处正确的并集解法(src/storage/backend.ts 接口签名并集:[PR] fix: initialize fork ledgers atomically before publication #5044 的 cloneSessionRows(initial?) ∪ 本 PR 的 getMessages/getBlocks(opts?),无语义丢失)。本次合并带入的所有文件均与 master 对应版本字节一致。高危模式逐项过:无阈值/时序/输出格式漂移、无 fail-fast 变吞错(本次合并不引入任何运行时代码)。
  6. 是否破坏缓存:不涉及——未触碰缓存/重试/游标/去重任何逻辑。
  7. 是否改了设计方向:不涉及——纯冲突消解操作,无架构调整。
  8. 是否引入新配置:不涉及——无新增 env/config/默认值。

验证状态(如实):本次合并带入的增量不含运行时代码,唯一人工触碰的文件与 master 已发布版本字节一致,故未本地跑定向/全量测试(§7.2:全量套件判定归远程 CI);CI 正在新 head 4b759b31 上运行,以远端结果为准。

过程披露:处理期间沙箱工作区被重置过一次(旧 .git 被清),全部状态从远端重新拉取重建;出站网络间歇性断连,多次重试后完成。首次自动修复尝试内完成,冲突不涉及设计取舍,无需 decision request。

结论:冲突已清除,待 CI 绿 + owner 合并。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 补充确认:本 lane 的完整工作汇报(八问)已在上方楼层发布(floor 39),此条为状态指针,避免重复刷屏。当前状态不变:PR #4540 head = 4b759b31(master 5594c131 已合入,唯一冲突文件采用 master 规范版本),GitHub 报告 mergeable(仅 behind,非冲突),CI 在新 head 上运行中,待绿 + owner 合并。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

Merge-guard sync: merged origin/master into the PR branch — clean auto-merge, zero conflicts; head is now c5a7b5f on top of master 58e61a3 (was 69 commits behind), and CI is re-running on the new head (0 failures so far).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working p2 scope/storage

Projects

None yet

Development

Successfully merging this pull request may close these issues.

perf: bound sampled storage verification reads

2 participants