Skip to content

[PR] release v0.1.209 - #4982

Merged
ranxianglei merged 9 commits into
masterfrom
2026-10-27_release-v0.1.209
Oct 28, 2026
Merged

ranxianglei merged 9 commits into
masterfrom
2026-10-27_release-v0.1.209

Conversation

@github-actions

Copy link
Copy Markdown
Contributor

Auto-fallback from the one-click release workflow: direct push to master was blocked by branch protection. Merging this PR publishes via the standard release.yml flow.

Changes since v0.1.206

@ranxianglei ranxianglei changed the title release v0.1.209 [PR] release v0.1.209 Oct 27, 2026
@ranxianglei ranxianglei added the ci:full Run the FULL CI matrix (all e2e lanes + 12-leg test matrix) on this PR label Oct 27, 2026
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

查重: 本 PR 不是重复修复,但发现一个需要你知晓的情况 —— v0.1.209 有两个并行的 release PR:

#4982(本,bot 自动 fallback) #4838(10-26 18:32 从 master@0.1.207 cut)
head 状态 最新 master + 1 个纯版本号 commit(package.json/lock + 4 个 plugin manifests) fork 点 + 4 commits;版本行直接 0.1.207→0.1.209(跳过已发的 0.1.208)
与当前 master 落后 0,无冲突 落后 2147 commits;package.json/lock/manifests 与 #4980(sync master to 0.1.208)文本冲突,现状不可直接合并
独有内容 无 ① b251699fe test(#3472/#4836):把 acp_retrieve 钉进 structuredContent wire suite(export pointer 是同一 bare-{outcome} 丢失的最后一个未列载体,带真实断言);② a8efcbe6c fix(release-notes):把 0.1.207(recommended)/0.1.208(critical,#4677 大存储打不开)两条 advisory 条目带进 release notes

已核实这两块内容 master 上都不存在(master 的 release-notes 仍停在 0.1.203 critical;该测试文件还没有 retrieve case)。只合 #4982 会丢掉它们 —— ≤0.1.204 用户升级后看不到「大存储被 0.1.207 弄坏」的 critical advisory。

建议(等你拍板):先把 #4838 那两个内容 commit port 进 #4982(或先落 master),再合 #4982,#4838 之后可关(我不能关 PR)。或者 rebase #4838 解冲突后合它——都行,推荐前者(冲突面小)。

标签: pr + scope/ci + p3(例行发版审核单,构建/发版域,非缺陷)。

审计已开始:窗口 = v0.1.206..本 PR head,477 commits / 466 files(v0.1.207/.208 是 lane/应急车道发布,其修复已分别落 master,本窗覆盖全部新增)。按政策分段执行:diff --stat 总览 → 高危域(自更新/dsh/lockfile/systemd/网络路径)逐行精读 → 其余文件按目录分批通读 → 逐项回归检查 → 行为变更清单。完成后按八问回复。

@github-actions

github-actions Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor Author

📦 Built Package Artifact

Branch: 2026-10-27_release-v0.1.209 (9bbf316)

Latest build — install this exact version

npm install -g billion-context@0.0.6658-pr.4982.e45b04d

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
9bbf316 billion-context@0.0.6658-pr.4982.e45b04d build
0a1f534 billion-context@0.0.6655-pr.4982.609ae67 build
4003e50 billion-context@0.0.6651-pr.4982.8350044 build
2f1828a billion-context@0.0.6614-pr.4982.0937d39 build
f9de49e billion-context@0.0.6604-pr.4982.6470e7b build
9a8fc4d billion-context@0.0.6595-pr.4982.0890753 build
8c06724 billion-context@0.0.6592-pr.4982.d104da0 build
1e642e1 billion-context@0.0.6555-pr.4982.efb0636 build
a7a0af8 billion-context@0.0.6503-pr.4982.4ee3889-a2 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr4982.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei

Copy link
Copy Markdown
Owner

回归 Review 结论(v0.1.208 → v0.1.209,458 files / +51122 −5491)

总体:1 个 MAJOR(fork clone 顺序,建议修复后再发版);其余无阻断性回归。 方法:5 路并行深审(kernel/压缩核心、dsh/客户端 lanes、存储、server/wire/缓存、web/CLI/MCP)覆盖全 diff(非只看 release 条目)+ bili-gate 真实模型回归 + 关键链路人工逐点复核。


🔴 MAJOR:fork clone 顺序使 #4316 PR3/3「number-preserving clone」在生产中完全失效

链路已逐环验证(每行都在本分支确认过):

  1. src/plugin.ts:1558-1559:设置 child.metadata.publicForkReceipt 后立即 publishForkSession(child) —— 在 clone 块(:1561-1579)之前。
  2. src/session.ts:1207-1214:publishForkSession 同步调 getStore().flushSync(child)。
  3. src/persist.ts:2697-2727 → kernel StateStore.flushSync 同步执行 build();guardedBuild 闭包(src/persist.ts:2384-2410)→ this.shadow?.record(record, …) 同步 ingest 进 index.db;因 receipt 已置位,record 携带 pluginSnapshot(gating 在 src/persist.ts:2897)。
  4. src/storage/ingest.ts:178:snapshot 先 intern → child 行被 pack 成 seq 0..M-1;:242 planIngestDelta 见 snapshot 直接返回 null → full ingest。
  5. src/storage/store.ts:2383-2389:cloneSessionRows 幂等 guard if (existing !== null && COUNT(*) > 0) return; —— 此时 child 已有自己 ingest 的 M 行 → clone 对非空 fork 永远 no-op。

后果:

  • 「fork copies every number」(owner decision 2026-10-22)这一主打特性除空 snapshot 边缘情形外从不在生产生效。
  • child 的 kernel 状态继承的是 parent 的 ledger 位置(child.state.messageRefs = structuredClone(parent.state.messageRefs) :1466;blocks structuredClone(b) 含 parent 的 startSeq/endSeq),而 child 的 unified store 是 packed 0..M-1。当 parent 存在 retired 行(holes —— 恰是本 release tombstone 特性制造出来的)时,位置型消费者(seq-interval coverage suppression:prune.ts:525 union disjunct / sync.ts:13-18)会在 packed 行上套用 parent 位置的区间 → 可能误归属,包括把某个 packed seq 恰好落进别的块区间内的存活消息压掉。
  • 补充细节:refLedgerAligned latch 没有拷给 child(fresh state,:1447-1530 逐字段拷贝不含它)→ child 首次冷加载会跑 restampRefsToLedger,把 messageRefs 重编号到 child 自己的 packed seq+1;但 block startSeq/endSeq 从不 restamp(refs.ts:217-287 只动 messageRefs)→ restamp 之后 refs 与 block 区间彼此也不一致。
  • tests/seq-ledger-pr3.test.ts S12a 恰好固化了生产中不会发生的顺序(先 clone、后首次 ingest),所以测试全绿。

修复方向(供参考):把 clone 移到 publishForkSession 的 flush 之前(S12a 契约本来就是 clone-first;注意 ingest 与 clone 的交互 —— UNIQUE(session_id,msg_id,content_hash) + delta refusal),或把 guard 改成与「期望的 parent 行数」比较而非仅判存在。另注意:naive 改顺序会激活下面的 latent minor(lineage 无条件覆写)。


🟡 Minor(建议本 release 或紧随其后处理)

  1. package.json:108 bundle 了 website/articles.json,该文件在本分支不存在(只在 origin/2026-10-24_website-sync-articles,非祖先分支;也没有 build step 生成它)→ 发布包的 Blog 页静默缺失 manifest 列出的长文(含 paper zh/en)。降级优雅(blogfeed.ts:296 stat-check fallback,tests/web-blogfeed.test.ts:46,168 钉住),但发版前应补文件或删除 entry。(已人工验证)
  2. 默认开启的 invariant judge 缺 egress-shed 图片占位符豁免(invariant-judge.ts:185-192 + relay.ts:568,579,712,769):pre-send shedding(No per-prompt image count cap: upstream "Too many images in request: 165 > 30" is forwarded as-is | 缺少按 prompt 的图片张数上限:上游 "Too many images in request: 165 > 30" 被原样转发 #4124 image-count convergence 是本 release 新增)在 sentBody 捕获前改写 wireBody;shed 后图片文本 [image: … omitted for size limit] vs 入站 [image] marker → 图片密集会话出现假 mid-history-rewrite I1 判定。log-only(defaultSink,invariant-judge.ts:413-423)但污染 [design] Pin cache stability with two outbound invariants: no-fold ⇒ exit only appends a tail; fold ⇒ bytes before the f | [design] 用两条出站不变量定死缓存稳定:无折叠=出口只多尾部,有折叠=折叠点前字节不动(证明式断言 + 属性/SMT 搜索分层) #3842 信号。修法:stableEqual 加方向性 shed-placeholder proof,或把 shedStats 传入 judge。
  3. snapshot-render.ts:247 O(n²) gap check(两位 reviewer 独立发现):[...coveredSeqUnion].filter((seq) => rows.some(…)) 每次 render 都跑,月级会话 = 10⁸–10⁹ 次比较/render。一次建 Set 即可。受 snapshotRender rollout 门控(默认 0%),但开启者正是大会话用户。
  4. restamp 碰撞带(stream.ts:861-866):重启后客户端重放 in-flight compress 调用、且所引编号已被 restamp 重新分配时,live-first 解析会 fold 新 owner 的消息。已文档化、限重启后秒级重放窗口;替代方案(span-scan 拒绝)会死锁会话(After the agent restarts, this session can no longer be compressed: the ref suggested by acp_status is flagged as "renum | 代理重启后该会话再也压不了缩:acp_status 建议的 ref 被 compress 判为「已重编号」,两面对 ref 的认知不一致 #4797)。建议 release note 提一句。
  5. decompress 缺 superseded 翻译(decompress-shared.ts:424):引用重启前编号的折叠消息会落到 [decompress FAILED] 而不是翻译(与 compress 侧 supersededRefRejection 不对称)。方向安全(绝不会返回别的消息的字节),可经 acp_status 恢复。
  6. compress 广告 schema 收窄接受集(kernel/src/compress-tools.ts:52-90):array-form content 不再通过 advertised schema 校验(parser 仍接受)→ strict/guided-decoding provider 会在 wire 层拒绝此前合法的调用形态。同时造成一次性全量 prompt-cache 失效(已文档化)。
  7. summary 长度上限开始生效(compress-settings.ts:81-82,342-345):maxSummaryLength/minSummaryLength 此前未映射(配置无效,External summaries over compress.maxSummaryLength are silently discarded after paying full latency — and the cap is not configurable #3635),现在到达 kernel → 一直跑超配置上限的会话升级后会开始被拒。是正确修复,但是真实行为变化。
  8. clone lineage 无条件覆写(latent)(store.ts:2432-2441 / mysql.ts:1232-1236):upsert 无条件更新 parent_session_id/fork_point_seq;当前被上面 MAJOR 的 early-return 掩盖,一旦 naive 修顺序即变 LIVE → 应 COALESCE 保留既有值。
  9. second-process 检测 TOCTOU(store.ts:422-448):.lock holder 先读后无条件覆写 → 并发 boot 互不可见 + pid 复用误报(by design warn-only)。
  10. quick_check 缓存键 {size,mtimeMs}(integrity-check.ts:81):粗粒度 mtime 文件系统上可碰撞 → stale cached-ok。
  11. CI:ci.yml EXCLUDES tests/storage-backend-mysql-e2e.test.ts(引 pre-existing real-MySQL failure CI mysql-e2e hot-path test fails deterministically on real MySQL: async ingest rows never land, failure swallowed below the 3-strike log threshold #4926)——请确认 tracking issue 存在;backport.yml 删除、release-bugfix.yml 重写为 carve-out channel(语义变化已核对,无问题)。

🟠 默认行为变化(均已文档化、可回退;请确认意图)

  • 429 governor 默认 ON(knobs.ts:117-145):连续上游 429(默认 max 5,backoff 2s×2 jitter cap 30s)后,同 host+credential 的后续发送零出站本地应答(429+Retry-After,终态 424);作用于所有非 GET forward 含 passthrough lanes。opt-out BILI_THROTTLE_429=0。本 release 唯一的 default-ON 收紧。
  • HTTP lane egress byte shed 默认 10 MiB(knobs.ts:278-287):此前原样发送(上游 ~48MiB 拒绝,如 Cloudflare 503);现在超预算从出站副本 shed 最旧图片(原件归档、decompress 可恢复)。WS lane 不变。0 恢复原行为。
  • 新 acp_status 报告包 [acp-status]/[/acp-status] 包裹;handoff export 增加 ### assistant 段 + reasoning 终止符(均为 additive,严格解析器需知)。

✅ 回归核查要点(未发现问题的部分)


🧪 bili-gate 真实模型回归

结论:零产品回归证据。 所有红项均机械归因(上游饱和 ×2、窗口/模型标定 ×2),且 Lane A 的 9 项断言在多次运行中全部拿到过绿证据(明细见下)。

🔧 review 期间发现并已修复

本地测试证据(pr4982 分支,均绿)

served-markers 11/11(+split-brain-surface 5/5)、pi-msgids 12/12、opencode-msgids 5/5、omp-msgids 6/6、msgid-lane-flow 6/6、dsh-msgids 6/6、native-intercept 66/66、issue4516 13/13、web-blogfeed / web-blog-route / web-i18n / web-blog-refresh / visibility-markers 39/39、issue4460、issue4666-migration-dedup、storage-state-fields parity、seq-ledger-pr3、ccr-restamp、seq-interval-* 等。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

重跑结果:绿。run 40527486714 conclusion: success(run_attempt 2)——两个红分片(windows-latest node24 s2/s4)在新 runner 上通过,其余 28 项保持绿,30/30 全过。零代码变更,flake 判定成立(同签名未复发)。

最终判定:CI 全绿,可人工合并。 合并后最终树 = master + 版本 bump(#4904 已在 master,GitHub 合并自动带上)。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

Merge guard sync done: merged origin/master (9b6ec2345, #4904 direct-window-guard) into 2026-10-27_release-v0.1.209 cleanly — no conflicts; pushed 1e642e114. PR is no longer behind, mergeable=true; only remaining blocked state is the fresh CI lanes queued/in-progress on the new head.

ranxianglei pushed a commit that referenced this pull request Oct 27, 2026
Unblocks PR #5004's test (windows-latest, 22, 2) red on head da25207:
three abort cells (proxy-openai-sse / proxy-responses-sse / proxy-json)
asserted closed=0 after exhausting the flat 3s (30x100ms) poll on a
loaded Windows runner -- the same signature already recorded under #4599
for #4982/#4774/#4683 (rerun green / in-branch budget bump). Sibling
cell abort x proxy-anthropic-sse passed in 176ms on the same shard and
every other windows/ubuntu shard ran this file green on the same head,
so this is runner load, not a code regression; the PR's diff touches
none of these paths.

Labeled CI-unblock commit; duplicates #4683's pending identical change --
drop this commit when rebasing if #4683 lands first.
ranxianglei added a commit that referenced this pull request Oct 28, 2026
fix: draining proxy no longer claims its freed port — same-lane newcomer ladders instead of dying twin-collision (#4982)
@ranxianglei

Copy link
Copy Markdown
Owner

review代码

ranxianglei added a commit to z0d0dOr/billion-context that referenced this pull request Oct 28, 2026
…n misclassification no longer kills same-lane newcomers (ranxianglei#4982)

A proxy entering graceful shutdown releases its listening socket immediately but kept its registry marker until the drain finished (seconds-to-minutes later). During that window the marker still claimed the port while a different instance could bind it; a same-lane newcomer hitting EADDRINUSE then misclassified the draining predecessor as its twin and died with the adopt signal (exit 3), stranding its launcher behind a cross-lane stranger it is forbidden to attach to (ranxianglei#1225). CI hit this in the ranxianglei#3203 concurrent-sessions e2e (release v0.1.209, run 40574417718).

Two complementary fixes:
- shutdown() stamps portReleasedAt on its own marker the moment the listener closes; classifySameLanePeer skips stamped entries, so a draining instance can never be mistaken for the port holder again (predecessor-wait consumers keep counting them, preserving ranxianglei#3044/ranxianglei#1723).
- the late EADDRINUSE path now verifies the claimed twin's socket ownership via /__bili/health (pid + instance identity, the ranxianglei#4341 criterion) before declaring twin-collision; an unverified claim falls through to the plain foreign-holder policy and ladders.

tests/issue4982-draining-twin-ladder.test.ts pins both halves end to end: a real draining predecessor + silent squatter (newcomer must ladder, not die) and an unstamped marker variant (verification demotes the claim, logged as 'no longer answers as itself').
ranxianglei pushed a commit that referenced this pull request Oct 28, 2026
@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[bot] 进度一行(自动修复车道 attempt 2/2):ci-fixes(哨兵 LIVE_FILES 登记修复 02ef0d1)已 merge 进 release 分支并推送 sha=f9de49e47,新 run 进行中(15 项 pending)。目前唯一红点是 e2e 车道 real opencode native lane vs deterministic fake upstream(非 test 分片、与本次 CI 修复零交集,#1248 家族真实客户端 e2e 在负载下偶发),正在观察;全部 settle 后补终局判定。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

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

查重: n/a(owner 直接指派审核本 PR)。以下为 release 版本窗回归审计结论。

审计窗口与读法:窗口 = 上一发布 tag v0.1.208 → PR head 树(478 文件,+54020/−5642)。注意 v0.1.208 走 bugfix carve-out 车道、不是 master 祖先,实际通读范围是 fork 点 f161c51(=v0.1.206 合入)→ master b08fbf3 → 本 PR 版本提交。读法:--stat 全窗总览 → 高危域文件全文精读(update/dsh 双文件/kernel refs+unit-projection+compress-tools/persist/storage 逐提交)→ 其余约 400 文件按目录+提交主题分批通读。

八问

  1. 是什么问题:发版类 PR(纯版本号 0.1.208→0.1.209 + 插件 manifest 同步),审的不是这个 diff,而是上一版本以来全部变更有没有引入回归。
  2. 什么原因导致的:无单一原因——本窗是约 30 项 fix/feat 的批量发布(seq-ledger 三层对齐 After model switch / client-side history rewrite, all message identities (content hashes) are replaced: ACP numbers re-m | 切模型/客户端重写历史后,消息身份(内容哈希)全部更换:acp 编号重铸、折叠覆盖大规模重挂、死 ref、churn 区临时膨胀重折 #4316、429 本地节流 Provider 429 (concurrency-style limit): turn re-driven every few seconds for minutes, keeps the account pinned at the limit #4204、出站大图瘦身 Image session permanent 503: re-sending 48.6 MiB outbound body every round (fold blocked by cadence, archived images sti | 图片会话永久 503:每轮重发 48.6 MiB 出站体(fold 被 cadence 挡住,已归档图片仍在 wire 上) #3909、MySQL 后端家族、计费 Add model billing feature | 增加模型计费功能 #4899/[PR] feat: per-model USD billing prices + panel money views (#4899) #4916、subagent 分裂 Subagent requests share the parent session id and trip fold-reconcile #4352、codex 共享状态库、D 车道快照渲染修复 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/[Bug] Codex Responses WebSocket blocks long sessions at 100 MiB before ACP compression, including small incremental continuations #4804 等),逐项核对既有行为。
  3. 影响范围与严重程度:影响所有升级到 0.1.209 的用户;其中 4 处默认行为变化(见第 8 问清单),全部是有意功能且带逃生开关;未发现无意回归。
  4. 怎么修复的/怎么审的:三层核对——(a) v0.1.208 carve-out 车道的全部关键修复(v8 migration reads the entire storage content into memory at once: large storage (several GB or more) triggers OOM after | v8 迁移把整个存储内容一次性读进内存:大存储(几GB以上)升级 0.1.207 后 OOM,之后每次启动重炸,老会话永久打不开 #4677 v8 迁移流式化+进度日志、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、SQLite unified storage degraded to legacy-only (bili v0.1.204, node v21.6.1, win32-x64) #4444、CI fully red (on master itself): #3250 WAL cap assertion failing across all platform shards — suspected that this mornin | CI 全红(master 自身):#3250 WAL cap 断言在所有平台分片失败,疑似今晨合入的存储层改动破坏 autocheckpoint 截断行为 #4506、A resume-inheritance miss is a blind spot: when the gate passes but no byte-level chain match exists, zero logs are emit | resume-inheritance 未命中是盲区:门控通过但无字节级链匹配时零日志,克隆继承失败无法归因 #4264/[PR] No longer silent on resume-inheritance miss: add a line of bounded near-miss diagnostics (#4264) | resume-inheritance 未命中不再静默:补一行有界近失诊断(#4264) #4265/[Bug][DSH] Fork loses fold inheritance without adoption diagnostics; interrupted preflight leaves child in repeated local 413 #4215)逐一在新树中确认在场,零丢失;(b) 高危域先读改动前代码再读 diff;(c) golden pin 核对:wire-contract 收缩仅来自 2 次已声明再生成([PR] The compress tool only teaches a single invocation format, to prevent the model from mixing up multiple formats. | compress 工具只教一种调用格式,防止模型把几种格式混着用 #4823 compress 参数 schema 收敛为字符串形态、[hermes] After tool_search defers plugin tools, the entire plugin identity/budget package is lost: /acp always reports " | [hermes] tool_search 延迟插件工具后,插件身份/预算整包丢失:/acp 永远报 "No model request yet",窗口回落 200K #4855 headers manifest),无工具丢失。
  5. 是否有回归问题:逐项过过的既有行为——自动更新路径(src/update.ts 全窗零 diff)、systemd/env 模板(无改动)、lockfile 钉死(kernel 0.0.161→0.0.169 与锁一致,无依赖增删)、nudge 常量与 prompt golden(无漂移)、thinking 通道(tag-echo-filter 只动文本通道,KDD#10 合规)、msgid 不复用承诺(kernel restampRefsToLedger 全程 scratch 交换 + superseded map 进程期保留,被挤占者换新号不重发)。发现并当场修复 1 处真回归:0.1.207/0.1.208 两条 advisory 条目(含 critical「大存储用户会被 0.1.207 迁移砖掉」的实时告警)在 carve-out 车道分叉时遗留在废弃的首个 release 分支(PR [PR] release v0.1.209 #4838 系),若照原样发布,≤0.1.206 用户升级后将失去这条 live 告警——已 cherry-pick 回 PR 分支(commit 2f1828a,恢复后共 17 条 < 20 上限)。
  6. 是否破坏缓存:无。seq-ledger 工作本身就是 ref/ledger 对齐,显式保持前缀缓存字节稳定(scratch 交换、fail-soft、任何失配整轮 no-op);出站瘦身只改出站副本,原图经归档占位可恢复。
  7. 是否改了设计方向:无架构转向。窗内的大功能线(MySQL 存储后端、USD 计费账本、pi/opencode inline 车道)全部默认关/可选,不动主路径。
  8. 是否引入新配置(旧值→新值):

高危域逐项点名:自动更新/自更新 = 零 diff ✓;dsh 相关 = dsh-native.ts/native-intercept.ts 全文精读,只有改善(bootstrap 自修复定时器 #4755/#4811——此前放弃的车道要等 /acp/restart 才恢复、现在自动重试;显式 attach 优先于路由 origin;不可解码 body 保守处理)✓;lockfile/版本钉死 = 仅 kernel 版本号推进且一致 ✓;systemd/env 模板 = 无改动 ✓;网络/代理 = port-owner 绑定校验(#2941)+ upstream-throttle/upstream-fail 两个新模块通读,均保守(仅正面证据才判定)✓。

CI 状态:上一 head f9de49e 有 1 红——「real opencode native lane vs deterministic fake upstream」e2e。该 head 树 = master + 惰性添加(版本号/CI 登记/release-notes JSON),均不可能触到这套真实客户端套件,同期兄弟 PR 该 lane 为 skipped(无法横向对比),判断为疑似继承自 master 或真实客户端 e2e 抖动。advisory 修复已推送(新 head 2f1828a),已触发全量 CI 重跑;若该 lane 在最终 head 仍红,我会拉日志定位,不放行。

判定:可合(条件)——代码面窗口内零回归,唯一发现的回归(advisory 条目丢失)已当场修复推上分支;待最终 head 2f1828a 的 CI 全绿(opencode-native lane 重跑裁决)后即可由你合并。

@ranxianglei

Copy link
Copy Markdown
Owner

🤖 Powered by ework · qwen3.8-27b

[bot] 进度一行:新 run(f9de49e47)健康推进中——最近确认点 04:19Z 为 30 项里 0 红、18 pending(test 分片正常跑),此前那个 e2e 红点未复发。监控进程被看门狗重启了一次,已换成后台守护方式继续盯;settle 后立刻补终局判定(预期全绿)。

This was referenced Oct 28, 2026
ranxianglei pushed a commit to Zhishulo/billion-context that referenced this pull request Oct 31, 2026
…anxianglei#4626)

Resolution notes (3 conflicted files):
* src/served-markers.ts — UNION of the two independent fixes for the same root
  cause (ranxianglei#4626 false split-brain on graceful handover): master's ranxianglei#4982 line
  stamps stoppedAt on dispose(), ranxianglei#4630 stamps disposedAt. dispose() now sets
  BOTH (stoppedAt=Date.now() then writeMarker(true)); reader-side contestedServedAt
  treats either stamp as known-dead, so markers written by old builds
  (stoppedAt only) and new builds (both) are mutually recognizable across
  versions. d-4630-only capability kept in full: peerStillContests
  fail-closed three-state check + runtime-status self-clearing of fired
  split-brain verdicts when the contesting peer provably stopped.
* src/server.ts — import-line unions (peerStillContests, registerSplitBrainEvidenceCheck added on top of master's existing list); registration call already landed conflict-free at startServer.
* tests/served-markers.test.ts — ALL FOUR graceful-stop tests kept (2x stoppedAt family asserting body.stoppedAt / alive-pid skip, 2x disposedAt family asserting tombstone shape + e2e rotation silence) plus display-fix tests in admin/web; each branch's assertions pass against the union implementation because both stamps are now always co-written.
ranxianglei pushed a commit that referenced this pull request Oct 31, 2026
…rsistence adopt-exit (option A)

Ruling (owner-approved 2026-10-31): for an unverified unstamped same-lane
claim the early pre-persistence gate (#4977) wins over #5016's late ladder.
The second row of tests/issue4982-draining-twin-ladder.test.ts now asserts
the code-3 twin-collision adopt-exit BEFORE index.db is opened instead of
ladder-off; the draining (portReleasedAt-stamped) row keeps asserting the
ladder. Fixture cleanup hygiene was already finally-wrapped upstream, nothing
to add.

Refs #5018 #5004 #4982 #5016
@github-actions github-actions Bot mentioned this pull request Oct 31, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment