Skip to content

[PR] fix: persist each committed preflight fold so a dead marathon leaves durable progress (#3978) - #4046

Open
ranxianglei wants to merge 18 commits into
masterfrom
wt-3978-ework-daemon
Open

ranxianglei wants to merge 18 commits into
masterfrom
wt-3978-ework-daemon

Conversation

@ranxianglei

@ranxianglei ranxianglei commented Oct 20, 2026 •

Copy link
Copy Markdown
Owner

What

#3978 reports a permanent unavailability loop on massively over-window sessions (~1.2M tokens vs a ~937k window): the preflight compression marathon outlives the client's hard wall timeout, the attempt dies, and every retry re-compresses from zero.

This PR implements the persistence half of the agreed fix direction (owner-approved in-thread): each committed preflight fold marks its session dirty as it lands (markDirty in commitFold, src/preflight.ts), so durable progress accumulates within the persist debounce no matter how the request ends. The wall-clock half is already implemented in open PR #3590 (client-patience wall budget) — hence this PR references #3978 without closing it; the two together close the loop.

Diagnosis (verified at v0.1.202 and at master bc4a3f8)

  • On the reporter's version, the abort early-return skipped ALL dirty-marking: partial folds lived only in memory and were lost.
  • On master the ordering improved: the post-loop rebuild (src/server.ts:3554 runPrepare → markDirty) now runs BEFORE the fail-fast return (:3607), so a CLEAN client disconnect already persists completed rounds (verified by md-trace instrumentation).
  • Residual gap this PR closes: a HARD process death mid-request (crash / OOM / restart — realistic at multi-million-token payload scale) skips every post-loop path. Pre-fix only the request-start prepare write (a zero-block record) had landed, so a retry under a new session id loaded a parent with nothing to inherit and started fresh again.

Behavior change disclosure

None user-visible beyond write frequency: during a preflight marathon the store now receives one debounced save per quiet window instead of only at rebuild points. No wire change, no config surface added, no persistence-format change, no kernel change.

Verification

Coordination notes

Follow-up small fix in this branch (separately labelled commit a46fd0d)

While re-triggering CI after merging master, it turned out #4393 (merged to master 2026-10-23 08:32Z) added a SECOND top-level concurrency: key to .github/workflows/ci.yml, which already carried one (workflow+event+ref groups, cancel-in-progress except schedule — #3772/#3331). A duplicate top-level YAML key makes the workflow unparseable: every ci.yml run fails instantly with zero jobs (master push 970bd4d shows the path-named instant-failure signature), killing the main CI leg repo-wide while the separate lane workflow files kept running. The pre-existing block already implements the same-branch-supersede behavior #4393 intended, so a46fd0d removes #4393's 8-line block and nothing else. Per the few-line CI-fix discipline it rides in this PR as its own labelled commit; other in-flight PRs landing the same 8-line removal independently is acceptable.


Fixes #3978

(孤儿认领:标题本已指向 #3978,补上关闭关键字 —— 第三批清理)

A hard process death mid-preflight (crash / OOM / restart on
multi-million-token sessions) skipped every post-loop persistence
path, so rounds already compressed lived only in memory; a client
retry under a new session id found no parent state worth inheriting
and re-compressed from zero — the permanent unavailability loop
reported in #3978.

commitFold now marks the session dirty as each fold lands; the
persist debounce coalesces the calls into one write per quiet window,
so durable progress accumulates regardless of how the request ends.
A clean client abort already persists via the post-loop rebuild
(src/server.ts:3554 runs before the :3607 fail-fast return); this
closes the death-mid-request gap that path cannot cover.

tests/issue3978-preflight-abort-persistence.test.ts pins it
end-to-end: T1 guards the clean-abort ordering, T2 is the negative
control (no fold committed => nothing fold-like persisted), T3
SIGKILLs a real child proxy mid-marathon and asserts a fresh
instance's new-id retry inherits the persisted blocks instead of
starting fresh. Without the fix the red run fails T3 exactly at the
adopted-block count.
@github-actions

github-actions Bot commented Oct 20, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: wt-3978-ework-daemon (f7e9e62)

Latest build — install this exact version

npm install -g billion-context@0.0.6193-pr.4046.48defc7

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
f7e9e62 billion-context@0.0.6193-pr.4046.48defc7 build
c8cd6f1 billion-context@0.0.6104-pr.4046.942a08f build
bd0fe6f billion-context@0.0.6012-pr.4046.1272236 build
470098e billion-context@0.0.5867-pr.4046.913efc6 build
c1d3a19 billion-context@0.0.5800-pr.4046.5771e2d build
305b4f3 billion-context@0.0.5641-pr.4046.4ac7fbf build
7eee06d billion-context@0.0.5562-pr.4046.8946295 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr4046.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei ranxianglei changed the title fix: persist each committed preflight fold so a dead marathon leaves durable progress (#3978) [PR] fix: persist each committed preflight fold so a dead marathon leaves durable progress (#3978) Oct 20, 2026
awork added 3 commits October 21, 2026 03:13
…3978)

The 60s READY deadline lost the race on ubu22-1/win22-1 at head 9be2ec6:
the child cold-transforms the whole src tree via tsx, which exceeds 60s
under parallel-shard load. Raise the deadline to 180s, fail fast when the
child exits before READY (diedP), and capture its stderr in-memory so a
future crash surfaces in the assertion message instead of getting lost in
interleaved shard logs.
…ts resolve (#3978)

On CI the per-test tmp dir sits outside the repo tree, so Node finds no
package.json walking up from the generated child script and defaults it to
CJS. Its bare import of acp-kernel then resolves under the require
condition, which acp-kernel's exports map does not define, killing the
child at boot with ERR_PACKAGE_PATH_NOT_EXPORTED on every leg. Locally the
tmp dir lives inside the repo (inheriting its type:module), which is why
this only failed on CI. Writing a minimal {"type":"module"} package.json
beside the child script pins the module format explicitly. Verified with an
out-of-repo TMPDIR reproduction (red before, green after).
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

查重:不重复——本票是 #3978 的 PR 票(持久化半)。同家族在途项:#3590(墙钟半,独立机制,零文本重叠已核)、#4049(master 既有 flake,与本 PR 无关)。仓库内无其他 open PR 触碰 commitFold/持久化路径。

标签:bug / scope/storage / p1(马拉松中断丢全部进度、超大上下文会话永久不可用——按 owner 2026-10-19 修订属 p1「不能用」类;无 kernel 改动故不打 scope/kernel;PR 本体不涉特定客户端,不打客户端标签)

初审进展(继续审查中):

  1. 代码核对完成:diff 共 3 文件 +417/-1。src/preflight.ts 只在 commitFold 尾部加 markDirty(deps.session)(commitFold 是 batch 与 serial-walk 两条提交路径的唯一汇合点,一处覆盖全部);scripts/live-files.mjs 登记进程派生测试;新增 hermetic 回归套件 T1/T2/T3(含真实 SIGKILL 子进程死亡场景 + 负向对照,Test Blind Spot Retrospective: Reinforcing Negative Assertion Discipline, Tightening Fake Upstream Strictness, Thin Inst | 测试盲区复盘:补反向断言纪律、假上游严格化、薄安装冒烟、形态矩阵与生命周期覆盖 #3617 双向门禁齐备)。
  2. master 顺序声明已独立验证:PR 分支上 src/server.ts 的 post-loop rebuild(runPrepare→markDirty,:3554)确实在 abort fail-fast return(:3607)之前——干净断连在 master 本就能落盘,本 PR 补的是硬进程死亡跳过全部 post-loop 路径的残余缺口。声明成立。
  3. 新发现(正在处理):master 已从 PR base bc4a3f8 推进 ~20 个 merge(至 07888d7),GitHub 报 mergeable: dirty(冲突)。本地 trial-merge 定位冲突点 → 解决 → push 回 PR 分支。
  4. CI(head 5722ac7):全部 gate 类检查绿;6 个 test 分片 5 红。前次逐分片归属为 Two #3769 storage regression tests flake across CI shards on master and unrelated PRs #4049 的两个既有 flake 签名(issue3769-pooled-bare-load:115 / issue3769-upgrade-compat:179),且 T3 在其运行的所有 leg 通过——我会独立抽查分片日志核实,不照抄结论。

后续:解冲突 + rebase 后定向测试 + 八问结论回复。

@ranxianglei ranxianglei added bug Something isn't working p1 scope/storage labels Oct 23, 2026
awork added 2 commits October 23, 2026 14:02
(#4046) resolve the single live-files.mjs conflict by keeping both new
LIVE_FILES entries in alphabetical order (issue3978 before issue4341);
src/preflight.ts auto-merged clean and the commitFold markDirty hunk is
verified intact against master's changes.
… T3 child

The blind sleep(1000) settle raced the store's trailing-debounce write under
loaded CI runners: windows 22-1 at head 5722ac7 found NO parent record at
all when the retry booted (T3 red; signature distinct from #4049's two
#3769 flakes). Replace the fixed sleep with a wait on '[durable] N' — the
child's own codec-aware store.loadSync disk read (envelopes are BILIENC1-
framed at rest, so the test process cannot scan the file for the summary
mark) — so the SIGKILL can never race an unwritten save. Failure messages
now carry the child stderr tail and the store dir listing.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

本轮结论:T3 红根因定位并修复,分支已合并 master 并推送(head 12de55b)

1. 独立复核 CI(逐份重读原 run 39491108431 的失败分片日志)

2. 根因判定(测试竞态,非产品缺陷)

  • 机制本身没问题:同一台 Windows 机器上 T1/T2 都绿(debounce 写在 ~500ms 内落地,T1 里 cold-loaded t3978-a ... 4 block(s) 为证);本地 Linux 3/3 绿,typecheck/build 干净。
  • T3 的死机时机依赖盲 sleep(1000):[blocks] N 来自子进程内存 listSessions 轮询,不是落盘信号。高负载 runner 上 store 的 trailing-debounce 写(debounceMs=50)会输给 SIGKILL → 什么都没落盘 → 正是修复前的症状,测试变红。是 timer 饥饿还是 AV 文件锁从外部无法区分(子进程 stderr 在内存里捕获),所以顺手补了诊断信息。

3. 本轮改动(已推送,PR head = 12de55b)

  • 测试加固(commit 12de55b,test-only diff):把 sleep(1000) 换成等 [durable] N —— 子进程自己调 store.loadSync("t3978-p")(codec-aware 磁盘读,走与全新实例相同的 reader 路径),看到 ≥1 active block 才报告,然后才 SIGKILL;死机不可能再与未落地的写竞态。注意:envelope 落盘是 BILIENC1 编码帧,测试进程无法直接扫文件内容找摘要标记(我第一版尝试就是这么挂的,本地复现后已改对再推)。超时/断言消息现在带子进程 stderr 尾部 + store 目录清单,下次若再红可直接定位。
  • 合并 master(commit 6d765e5):唯一冲突 scripts/live-files.mjs(两边各新增一条测试注册),取并集保留两条。已验证:src/preflight.ts 的两个 hunk(:21 import、:2174 markDirty)完好;server.ts rebuild-before-failfast 顺序保持;master 对 session.ts/persist.ts 的改动不触碰 markDirty/scheduleSave 语义。
  • 本地最终 head:typecheck 干净、套件 3/3 绿;全量判读留给 CI。

结构化八问

  1. 是什么问题:T3 用固定 1s sleep 与 store 的 debounce 计时器竞态——高负载 CI leg 上 kill 可能先于 per-fold save 落盘,让测试因与 [Bug] Excessive preflight compression time on ultra-large contexts; client mid-disconnect causes repeated request forwar | [Bug] 超大上下文 preflight 压缩耗时过长,客户端中途断开导致请求反复无法转发 #3978 契约无关的原因变红;另外分支落后 master ~20 个 merge 且处于冲突状态。
  2. 什么原因导致的:[blocks] 信号的是内存提交而非持久化;测试假设 1s > 50ms debounce,但没有验证写真正落地。
  3. 影响范围与严重程度:test-only,无产品代码变更。不修则 T3 在任何高负载腿都可能红(已在 win22-1 观察到),阻塞对本 PR 真实修复的可信判读;无用户影响。
  4. 怎么修复的:把死机变成对持久化确定性的——等子进程自己的磁盘读确认 envelope 落地再 SIGKILL;merge master 解掉冲突(唯一手工解决 = live-files 两条注册取并集)。
  5. 是否有回归问题:无。本轮 diff 仅限该测试文件 + merge commit(其余全是 master 自身变更,已过各自 CI)。
  6. 是否破坏缓存:不涉及(未触碰 cache/retry/cursor 逻辑)。
  7. 是否改了设计方向:无。
  8. 是否引入了新的配置:无。

行为变更披露(整 PR,重申)

仅写频率:marathon 期间 store 每个 quiet window 收到一次 debounce save,而不只在 rebuild 点。无 wire 变更、无配置面、无持久化格式变更、无 kernel 变更。

Auto-merge gate(§7.4)评估

条件全满足:单模块纯 fix、有回归测试、无 config/persistence/wire/kernel 变更、clean diff、引用 #3978(有意不关闭——wall-clock 半在 #3590)、不触碰 load-bearing infra。待办:新 CI 确认——T3 在其排到的 leg 绿;其余红必须全部可归因于 #4049 的两个 #3769 签名(A = tests/issue3769-pooled-bare-load.test.ts:115,B = tests/issue3769-upgrade-compat.test.ts:179)。

判定

新 CI 确认上述后即可合入(合并动作请人工执行)。与 #3590 的协调说明仍然成立(零文本重叠,合入 master 后依旧)。另注:T3 在矩阵中只排一个 leg(LIVE_FILES 分片机制使然),CI 暴露面偏薄属既有设计,hermetic 且足够快,本轮不扩覆盖。

awork added 2 commits October 23, 2026 16:47
…4393

ci.yml already carried a concurrency block (workflow+event+ref groups,
cancel-in-progress except schedule — #3772/#3331, owner-approved 2026-10-10)
that implements exactly what #4393 intended: same-branch pushes cancel the
branch's in-progress/queued main-CI run. #4393's premise ('the main CI did
not [have a concurrency group]') missed that existing block and added a
second top-level 'concurrency' key. A duplicate top-level YAML key makes the
workflow unparseable: GitHub fails every ci.yml run instantly with zero jobs
(master push 970bd4d at 08:32:25Z and PR pushes after 08:32Z all show the
path-named '.github/workflows/ci.yml' instant-failure signature), killing the
main CI leg repo-wide while the separate lane workflow files kept running.

Remove #4393's 8-line block; the pre-existing concurrency semantics are
unchanged, so same-ref supersede behavior #4393 set out to get still holds.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

状态更新:上一轮 CI 被批量取消(非测试失败)+ 发现并顺手修掉一个 master 级 CI 回归

1. 上一轮 11 个 run 的去向

head 12de55b 的 11 个 workflow 全部在 07:40:47Z 同一秒被批量 cancel(启动于 06:54Z,运行/排队 ~46 分钟)。不是测试红——是今天 CI 队列事故(458 个排队 run,#4393 的 commit message 里点名了这次 incident)期间的清理动作。该 head 上没有任何有效判读。

2. 重新触发时发现的 master 级回归(已修,独立标注 commit a46fd0d)

把 master(仅 #4393 一个 CI-only commit)合入分支后推上去,主 CI run 秒失败、零 job,显示名退化成文件路径 .github/workflows/ci.yml;master 自己的 push(970bd4d,08:32:25Z)是同样签名——即 #4393 合并后全仓库的主 CI leg 都是死的(8 个 lane workflow 是独立文件,不受影响,所以还在跑)。

根因:#4393 给 ci.yml 加了第二个顶层 concurrency: key——该文件在 :56 处早已有一个(workflow+event+ref 分组、schedule 除外 cancel-in-progress,#3772/#3331,owner 2026-10-10 批准),而且那个既有块已经实现了 #4393 想要的同分支 supersede 语义(#4393 的前提"main CI 没有 concurrency group"不成立)。重复顶层 YAML key → workflow 不可解析 → 每个 ci.yml run 秒挂(本地严格解析复现:DUPLICATE TOP-LEVEL KEY: concurrency)。

处置:按几行级小修调度纪律①(CI 红的显而易见修复直接进当前 PR),在本分支删掉 #4393 的 8 行冗余块,其余不动——既有并发语义原样保留,same-ref supersede 行为照旧。其他在途 PR 若各自落地同样的 8 行删除属政策允许;本 PR 合入后 master 即恢复。

3. 当前状态

  • PR head = a46fd0d(= 原修复 + 测试加固 + 两次 merge master + ci 小修);与 master 无冲突、mergeable。
  • 新 CI 已起:11 个 run 全部 queued,主 "CI" 正常解析(显示名恢复为 "CI",无秒失败)。
  • 本地验证不变:typecheck 干净、issue3978 套件 3/3 绿(ci.yml 改动不影响 src/tests)。
  • PR 描述已补一节说明该 follow-up 小修。

待办(唯一)

等新 CI 判读:T3 在其排到的 leg 绿;其余红必须全部可归因于 #4049 的两个 #3769 签名。满足即可合入(合并动作请人工执行)。

ci-check added 2 commits October 25, 2026 01:14
…78 T3 child

win22-1 @ a46fd0d: [durable] was observed yet the retry still started
fresh (resume lines empty) — findResumeParent (#1486) matches on the
parent's affinity-chain entry, which a fresh instance hydrates from the
machine-shared state-dir snapshot (prefix-affinity.json), a SEPARATE
artifact from the session record with its own 500ms-debounced single-shot
persist whose write can lose a rename race or starve under event-loop
pressure and never re-arms. The child now proves the FILE carries the
t3978-p entry (re-arming the persist while it does not) and the test
waits for that proof before SIGKILL; the final assertion also surfaces
the child stderr tail.
… EBUSY observation

Minor fix found while fixing this lane's CI (labeled per the small-fix
rule): the #2771 corrupt-index proof spent the full ~9.6s retry ladder
on EBUSY unlinking index.db (transient Windows AV/indexer lock, #2799
family) — extend the flat tail by two 3200ms rungs (~16s total) and
align the budget unit test.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

Fixed on the branch — pushed 2b8c26d (two test-only commits, zero product-code change). CI reruns on the new head.

What the win22-1 shard @ a46fd0d actually showed

Two failures, different mechanisms:

① This PR's own T3, again — but a NEW failure mode. The [durable] hardening from the previous round worked (the child's disk read saw the committed fold before the kill), yet the retry still started fresh: resume lines: [], and instance 2 recompressed all ~48k tokens from zero. Root cause found by tracing the inheritance path: findResumeParent (#1486) does NOT match on the session record — it scans the parent's affinity-chain entry, which a fresh instance hydrates from the machine-shared state-dir snapshot (prefix-affinity.json). That is a second, separate artifact with its own 500ms-debounced single-shot persist (src/affinity-persist.ts): if that one write loses a rename race on the machine-shared file (caught internally, no retry) or the unref'd timer starves under event-loop pressure, nothing ever re-arms it — and the retry silently starts fresh even though every block is durably saved. The old test awaited only the record artifact, so the kill could land after the record but before/without the chain.

Fix (test-side, same philosophy as last round — wait for the artifact itself, not a guessed settle time): the child now proves the file carries the t3978-p entry before reporting ([affinity] ok), re-arming the persist while it doesn't; the test waits for BOTH artifacts before SIGKILL. The final assertion also surfaces the child stderr tail, which was a blind spot in both prior rounds.

② Unrelated pre-existing flake (not caused by this PR): the #2771 corrupt-index proof spent the full ~9.6s rmrf retry ladder on EBUSY unlink index.db — the known Windows AV/indexer external-lock family (#2799/#2908), outlasting the budget under shard load. Extended the flat tail of tests/tmp-rm.ts by two 3200ms rungs (~9.6s → ~16s total, following the file's own documented escalation precedent) and aligned its budget unit test. Labeled as a separate minor-fix commit per the small-fix rule — it touches shared test infra outside this PR's original scope, flagged here for transparency.

Local validation (targeted, per local-testing discipline): tests/issue3978-preflight-abort-persistence.test.ts 3/3 (T3 end-to-end through the new double-wait), tests/tmp-rm.test.ts 6/6, typecheck clean.

八问

  1. 是什么问题:win22-1 CI 红——本 PR 自己的 T3 测试第二次失败(且是新模式:durable 记录已落盘,重试仍从零开始),外加一个与 PR 无关的既有 Windows 文件锁 flake。
  2. 什么原因导致的:T3 = 双工件缺口——resume 匹配依赖父会话的 affinity-chain 快照(state-dir 共享文件),它与会话记录是两个独立工件、独立的一次性 500ms 防抖写入,写失败(rename 竞争/事件循环饥饿)后永不重挂;旧测试只等记录了。flake = AV 锁存活时间超过 ~9.6s 清理预算(Windows CI: storage-engine-proofs 套件清理 temp 目录遇 EBUSY/EPERM,windows test leg 随机红 #2799 家族)。
  3. 影响范围与严重程度:仅 CI 管道——产品代码零改动;T3 因分片恰好只落在 win22-1 一条腿上,该腿红即阻塞合并;非用户可见缺陷。
  4. 怎么修复的:纯测试侧——子进程在证明「快照文件里确实有父链」之前才允许被杀(缺失时周期性重新触发持久化),测试在 SIGKILL 前等待两个工件;rmrf 锁预算按文件自身先例从 ~9.6s 扩到 ~16s。
  5. 是否有回归问题:无产品行为回归(src/ 未动)。测试语义只增强(多等一个工件、最终断言不变);tmp-rm 只放宽瞬态锁容忍度,非瞬态错误仍立即抛错(单测钉住)。
  6. 是否破坏缓存:不涉及。
  7. 是否改了设计方向:没有——刻意不把 affinity 快照挪进 store 目录或并入 per-fold 写路径(那是持久化格式/设计决策,留给 PR 作者与 owner);测试现在把既有的「双工件」契约钉死。若 owner 认为生产路径也该让链快照随 fold 落盘,请拍板,可另立 issue。
  8. 是否引入新配置:无。

判定:等新 head 的 CI 全腿绿即可合(合并由人执行)。若 win22-1 再红且签名仍是这两个之一,说明锁/竞争比预期更极端,下一轮我会转判定评论而不是继续加预算。

… union — keep issue3472 + issue3978 in sorted order)
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

冲突已解决并推送(merge master → wt-3978-ework-daemon,新 head 7eee06d78,非 rebase/无 force-push)。GitHub 已重算:mergeable 由 dirty 转 True(当前 blocked = 等待必需检查,非冲突)。

结构化八问:

  1. 是什么问题:本 PR 分支落后 master 391 个提交后与 master 冲突,PR 无法合并。
  2. 什么原因导致的:PR 与 master 各自在 scripts/live-files.mjs 的 LIVE_FILES 表同一位置插入了新条目(PR 加 issue3978-preflight-abort-persistence.test.ts,master 加 issue3472-mcp-structured-text.test.ts),文本级冲突。其余文件(src/preflight.ts、.github/workflows/ci.yml)git 自动合并成功,无语义冲突。
  3. 影响范围与严重程度:仅阻塞本 PR 的合并;不影响任何运行时行为。
  4. 怎么修复的:冲突处取并集,按该表的字母序保留两条(issue3472 在前、issue3978 在后,注释随 PR 侧条目保留);其余按 git 自动合并结果提交。验证:①全树无残留冲突标记;②node --check scripts/live-files.mjs 通过;③对照 master 复核 PR 净 diff,恰好等于 PR 原改动集(preflight 的 markDirty(deps.session) 持久化修复、LIVE_FILES 条目、tmp-rm 重试预算放宽等),且 markDirty 在合并后的 src/session.ts:1377 存在;④ci.yml 顶层 concurrency: 仍唯一([PR] ci: cancel superseded same-branch CI runs (concurrency group) #4393 去重修复未丢)。
  5. 是否有回归问题:无。本次只动了测试注册表的一行并集 + master 自身变更的并入,不触碰任何 src 语义;高危模式(阈值/时序/输出格式漂移)逐项过——不涉及。CI 上静态门禁(kernel-guard / version-guard / config-surface-guard / lock-sync 等)在新 head 全绿;唯一红的 unused-exports 是继承性失败:master head 146b34f27 同样报同一行(INTERNAL_ONLY src/agent/client-build-note.ts: ClientBuildCheck),修复已在 [PR] fix(ci): drop the unused export of ClientBuildCheck #4603(Fixes fix(ci): master unused-exports is red — ClientBuildCheck is exported but only used inside its own file #4602),已在 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 留一行记录,合入后重触发即可。
  6. 是否破坏缓存:不涉及(无 wire/存储/缓存路径变更)。
  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:冲突干净解决,可继续走 CI;待 #4603 合入后本 PR 的继承红自动消除,无需再动。

ci-check added 7 commits October 25, 2026 10:08
master window commits 293ae65/de5e03659 bumped kernel/package.json to
0.0.165 without regenerating the lockfiles; lock-sync re-resolution then
rewrites the stale recorded versions (root 0.0.163, kernel 0.0.164) and
fails. Pure recorded-version sync via npm install --package-lock-only;
no dependency or semantic change.

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

bug Something isn't working p1 scope/storage

Projects

None yet

1 participant