Skip to content

[PR] fix: 429-hold answered locally before ingesting large request bodies (#5795) - #5801

Merged
ranxianglei merged 5 commits into
masterfrom
2026-11-06_429-hold-early-gate
Nov 7, 2026
Merged

ranxianglei merged 5 commits into
masterfrom
2026-11-06_429-hold-early-gate

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

Closes #5795

What changed

The #4204 egress hold refuses held sends locally (429 backoff window / 424 terminal) without egress — but the refusal was produced only AFTER full ingest, so every host retry of a held send still paid the proxy's entire ingest cost (JSON.parse + openaiToCore + snapshot pinning + processTurn). On the reporter's 51.2 MiB session that is ~5.5–7.2 s of proxy CPU plus a hundreds-of-MB allocation spike PER ATTEMPT, multiplied by the host's retry ladder.

This adds a pre-ingest fast path: a read-only peek at the governor (upstreamThrottlePeekReject) answers ≥1 MiB non-GET/HEAD requests up front in the dispatch path with the exact same local refusal; everything else keeps its existing path, and the post-ingest gate stays unchanged as backstop.

Behavior change (disclosed per policy)

  • Old: a held request was fully ingested first (parse/convert/snapshot-pin/processTurn — server-side session state advanced) and only then refused locally with 429/424.
  • New: requests ≥1 MiB against an active hold are refused locally before prepare — zero ingest cost, zero session-state advance, instant response. Nothing egressed either way; the next admitted turn reconciles off the client's full resend exactly as before.
  • Wire shape UNCHANGED: identical status codes, headers (Retry-After on 429 only) and JSON error body (code: bili_throttle_hold|bili_throttle_terminal, retryable, stage, optional scope/model). Both sites now render through one shared helper (answerLocalThrottleReject), so the post-ingest path's bytes are byte-identical to before.
  • Unchanged by construction: sub-1 MiB traffic; forge/classifier local intercepts (codexForge/autoModeForge, which are decided during prepare); passthrough-mark lanes (「BUG」native-intercept 进程级 globalThis.fetch 补丁按 URL 形状接管所有模型流量,无关调用者(如 MemOS)被卷入宿主压缩流程 #1117); chained-bili hop markers (fix: stitched-stream usage double-count after compress re-request (adapter-anthropic >0 guard) #300, which now get the identical local refusal one stage earlier instead of after their own ingest); GET/HEAD; disabled governor; model-scoped holds whose requested model does not match (peek conservatively skips them → normal path → post-ingest gate decides exactly as today); credential-overridden lanes (providers 统一字典后续:lane 凭据注入 + agent 注册表兜底(#2163 推迟项) #2336 fingerprints invisible to raw headers → conservative fall-through).
  • Rationale for every conservative edge: an early false-negative costs one ingest (= today's behavior); an early false-positive would refuse a request that would have been intercepted locally or belonged to another scope — strictly worse.

Why a separate peek function instead of calling the gate twice

upstreamThrottleGate MINTS a probe token when it admits an expired-window hold. Calling it pre-prepare would consume the probe slot and make the post-prepare call REJECT the very probe request it is supposed to let through. The peek duplicates the reject loop read-only; the gate loop and its mutation stay atomic and single-call-site for admission.

Tests

New tests/issue5795-early-hold-fastpath.test.ts (16 rows):

  • unit: peek/gate parity (active 429, terminal 424 incl. retryAfterSec: null), peek non-mutation (probe slot survives a peek; mid-window both reject without minting), model-scoped conservatism (exact match rejects / other model passes / absent model passes), credential isolation, provisional groups, disabled governor inert, threshold boundary, exact 429/424 response shape + [pre-ingest] log label;
  • hermetic E2E through the real proxy: held big turn = zero egress (upstream capture count unchanged) + local 429 with Retry-After; clean big turn egresses normally (negative row); small turns during a hold keep the post-ingest refusal shape; keyring model-scoped hold does not block other models' big turns.

Adjacent suites green locally: tests/issue4204-throttle-governor.test.ts + tests/issue4776-keyring-model-hold.test.ts (29/29). Full-suite verdict = CI on this head.

Config surface

None. No new knobs/env/config fields — reuses the existing #4204 governor knobs (throttle429*).

Out of scope (tracked in-thread)

The dominant cost — host-side rebuild + re-upload of the full 51 MiB per retry, dsh ignoring Retry-After and classifying the local 424 as retryable SERVER — lives in the dsh repo. That part is a pending owner decision in the issue thread (🔶 需要决策 comment): recommended package = honor Retry-After, reuse the last serialized body on retry, treat local-hold 424/bili_throttle_terminal as non-retryable.

ci-check added 2 commits November 6, 2026 19:03
…5795)

The #4204 egress hold refused requests only AFTER the full ingest
(JSON.parse + openaiToCore + snapshot pinning + processTurn), so every
host retry of a held send still paid ~5.5-7.2s of proxy CPU and a
hundreds-of-MB allocation spike per attempt on a 51 MiB session.

A read-only peek of the governor (upstreamThrottlePeekReject — rejects
only, never mints probe tokens) now answers >=1 MiB POSTs up front in
the dispatch path; sub-MiB traffic, forge/classifier local intercepts,
passthrough-mark lanes, chained-bili hop markers and model-mismatched
model-scoped holds all keep their exact pre-change paths, with the
post-ingest gate unchanged as backstop. The single reject-response
renderer (answerLocalThrottleReject) keeps both sites byte-identical.
Unit: peek/gate parity incl. terminal 424, peek read-only probe-slot
guarantee, model-scoped conservatism, credential isolation, provisional
groups, disabled governor, threshold boundary, exact 429/424 response
shape. E2E hermetic proxy: held big turn = zero egress + local 429 with
Retry-After, clean big turn egresses, small turns during a hold keep the
post-ingest refusal shape, keyring model-scoped hold does not block other
models' big turns.
@ranxianglei ranxianglei changed the title fix: 429-hold answered locally before ingesting large request bodies (#5795) [PR] fix: 429-hold answered locally before ingesting large request bodies (#5795) Nov 6, 2026
@github-actions

github-actions Bot commented Nov 6, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-11-06_429-hold-early-gate (feb8be9)

Latest build — install this exact version

npm install -g billion-context@0.0.8369-pr.5801.92efb5b

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
feb8be9 billion-context@0.0.8369-pr.5801.92efb5b build
c26dbea billion-context@0.0.8359-pr.5801.4be3d85 build
3bb2d87 billion-context@0.0.8349-pr.5801.6d4bb13 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr5801.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)

unused-exports 已修复并推到 PR 分支:commit c26dbea19(branch 2026-11-06_429-hold-early-gate),该检查现已转绿。以下为本次运行总结。

结构化八问

  1. 是什么问题:PR 新提取的共享 helper answerLocalThrottleReject(src/server/relay.ts:257)带着 export,但全仓库没有任何外部消费者——handle.ts:92 只 import tryAnswerEarlyHold,测试文件只 import EARLY_HOLD_GATE_MIN_BYTES / tryAnswerEarlyHold;该函数实际只在 relay.ts 内部两处被调(:309 的 fastpath、:919 的 post-ingest gate)。unused-exports 门禁按 INTERNAL_ONLY 判红。
  2. 什么原因导致的:[bug][dsh] 429-hold + host retry ladder re-upload 51.2 MiB per retry (36/37 attempts failed) | 429-hold 与宿主重试叠加:每次重试重传 51.2 MiB,单步放大 2-5 倍 #5795 重构把 post-ingest 拒绝分支抽成共享函数时留了 export(最终实现改由 tryAnswerEarlyHold 封装,handle.ts 不再直接调用它,export 成了残留)。纯声明可见性问题,零运行时影响。
  3. 影响范围与严重程度:仅 CI 门禁红;无运行时、无用户面影响。
  4. 怎么修复的:按门禁自身的 INTERNAL_ONLY 建议去掉行首 export,函数保持模块私有;两个调用点与原注释("shared by the post-ingest gate below and the pre-ingest fast path")都仍然成立——共享经由模块内部完成。单行 diff,不动任何行为。
  5. 是否有回归问题:无。改动是纯模块可见性标记,运行时字节级不变。证据链:本地 npm run typecheck 绿;CI test-import-graph 门禁绿(机器级确认无外部导入者);CI 上全部 12 项 gate(plan/version-guard/kernel-guard/pi-subagents-guard/lane-guard/lock-sync/config-surface-guard/docs-coverage/lifecycle-ledger-guard/test-import-graph/mysql-contract/unused-exports)+ static-gates×2 全 success;含本 PR 测试文件的三条 leg(ubuntu-22 shard 3/3、ubuntu-24 shard 3/3、windows-22 shard 3/5)全部 success。
  6. 是否破坏缓存:不涉及(不触碰任何存储/缓存/重试路径)。
  7. 是否改了设计方向:不涉及。
  8. 是否引入新配置:不涉及。

验证过程备注

  • 推送前本地跑了门禁原命令 node scripts/check-unused-exports.mjs(✓ clean, 8 waived)+ 全量 typecheck。
  • 本地跑该 PR 回归文件时,第 12 个用例(≥1MiB 无 hold 放行 e2e)在本机环境挂死(proxy 事件循环 watchdog 30s/300s 相继触发)。用同一最小复现脚本在 master(0aace2f20)上得到完全相同的挂死 → 判定为本机环境特有现象,非 PR 引入;随后 CI 上包含该测试的三条 leg 均 ~9 分钟正常 success → 按「本地失败以 CI 为准」关闭,不作缺陷处理。
  • 期间清理了本地产生的 3 个挂死测试孤儿进程(各占 ~100% CPU,共享资源卫生)。
  • 小备注:镜像 check 状态缓存中途停止刷新,后半段改用单次 API 拉取确认(未轮询)。

当前状态

  • 指派检查 unused-exports:已绿。
  • 截至本回复,另有 3 条不含本 PR 变更面的 test shard 仍在跑,目前 0 失败;若后续有红,会按自动修复车道继续跟进。
  • PR 相对 master 处于 behind(推送前即如此,非冲突,不阻塞合并)。

从本车道角度收工:唯一指派的失败检查已修复且全链路验证通过;合并由人执行。

ci-check added 2 commits November 6, 2026 21:18
With a first-response fixture set, n++ sat only on the 200 path, so the fixture replayed its first response forever: the model-scope row's egressing turn received another 429 and its 200 assertion could never pass (deterministic fail, reproduced locally). Counter now advances on every request; fixture behavior for the hold/clean rows is unchanged.
…askUrlsInText)

These e2e requests mint fresh ANONYMOUS sessions, whose provenance path masks the ENTIRE first message through maskUrlsInText before truncating to the preview cap. The #4191 any-scheme regex ([a-z][a-z0-9+.-]*://) backtracks O(n^2) over long scheme-char runs (measured: 7.4s @ 100k chars, exact 4x per doubling), so a megabyte of 'x' froze the event loop for minutes and hung this suite in CI. Whitespace-separated words keep the byte size identical while keeping the scan linear. The underlying masker defect is pre-existing and tracked separately.
@ranxianglei ranxianglei added bug Something isn't working dsh p2 scope/wiring labels Nov 6, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

查重: 不重复。按失败签名(429-hold + pre-ingest fastpath)检索 open+closed 的 issue/PR,仅本 PR 针对 #5795 的 bili 侧修复;dsh 侧根因仍在 #5795 的 🔶 待决评论下等 owner 拍板(跨仓库,不在本 PR 范围)。

标签: bug、dsh、scope/wiring、p2(与源 issue #5795 对齐;不打 needs-decision——该决策项属 dsh 侧,已在 #5795 跟踪)。

结论:可合并(等推送后的新 CI 转绿)——机制本身正确且保守;但审核开始时 head 上 CI 是红的,且新增测试文件有两个确定性缺陷(其一会让任何 runner 挂死),均已修复并推送到 PR 分支。

审核过什么(方向关 / 回归关)

  • peek 对等性:upstreamThrottlePeekReject 与 gate 的 reject 分支逐字节一致(同一 accountKey 匹配、同一 model-scope 跳过条件、同一 slotBlocked+strongerReject,holds 与 provisionalGroups 两表都扫);验证只读——不消费 single-flight probe 槽(单测双向钉住)。
  • 键一致性:快路径 forwardUpstreamUrl(req,opts,route) 与 gate 侧 buildForwardTarget 内部调用同一函数 → URL 维度不可能漂移;头维度用原始入站凭据头,providers 统一字典后续:lane 凭据注入 + agent 注册表兜底(#2163 推迟项) #2336 override 车道保守降级(最坏=多付一次 ingest,永不假拒)。
  • 调度顺序:passthrough-mark 车道(「BUG」native-intercept 进程级 globalThis.fetch 补丁按 URL 形状接管所有模型流量,无关调用者(如 MemOS)被卷入宿主压缩流程 #1117)在快路径之前;count-tokens 请求同样出网(拒绝结果不变);增量车道走独立 /__bili/plugin/turn 路由,不受影响;pf 水位快路径对被标记请求跳过水位判定,不受影响。
  • wire 形状:两处共用一个渲染函数,status/Retry-After/JSON body 与旧 post-ingest 内联块逐字节一致(本地 diff 比对过)。
  • 本地定向验证:新套件 13/13、governor 套件(issue4204,含 keyring model-hold 行)29/29、typecheck、unused-exports gate 全绿;全量以 CI 为准。

审核发现的三个问题(均已修复并推送)

  1. CI 红(阻断级):answerLocalThrottleReject 被 export 但只有 relay.ts 自己调用 → unused-exports gate 红,并经 needs: 连带阻塞整个下游测试矩阵(原 head 的全量套件从未跑过)。修法:去掉 export。注:审核期间 ci-check bot 已并发修了同一处(c26dbea),我的重复 commit 已丢弃,分支保留其版本。
  2. 测试 fixture 计数器 bug(确定性失败):脚本化上游的 n++ 只放在 200 分支——带 first-response fixture 时第一响应被永远重放,model-scope 行「其他模型出网」那一轮永远看不到 200,断言不可能通过。本地复现证实产品代码行为是对的(model-b 确实出网了,拿回的是 fixture 重放的第二个 429)——坏的只是测试本身。修法:计数器每请求递增。
  3. 测试内容触发既有 O(n²) 缺陷(确定性挂死):e2e 请求不带 session 头 → 每次铸造新的匿名会话 → provenance 路径在截断到预览上限之前用 maskUrlsInText 掩整条首消息;Log redaction misses userinfo in non-HTTP protocol URLs (e.g., socks5): socks5://user:pass@host is logged verbatim. | 日志脱敏漏 socks5 等非 http 协议 URL 的 userinfo:socks5://user:pass@host 原样进日志 #4191 引入的 any-scheme 正则 [a-z][a-z0-9+.-]*:// 在长 scheme 字符串上二次方回溯(实测 7.4s @ 100k 字符,每翻倍精确 4×),原 "x".repeat(1.1MB) 填充把事件循环冻住数分钟,该套件在任何 runner 上都跑不完。修法:测试填充改为空格分隔的词(字节尺寸不变、扫描线性),并加注释说明为何不能改回同质字符。

⚠️ 顺带发现的既有产品缺陷(不在本 PR 修,建议单独处理)

上述 O(n²) 是产品代码缺陷,与本 PR 改动无关(log-mask 最后由 #5540 触碰,本 PR 未动它),但与 #5795 共享触发面(大请求):任何匿名请求的首消息含长 [a-z0-9+.-] 连续串(hex dump、数字串、minified blob、小写 base64 类长串)都会让掩码器对全文二次方空转——按实测外推,51 MiB 同类字符 ≈ 数十小时事件循环冻结 = 整个代理不可用(生产若触发属 P1 可用性级)。推荐修法(二选一或都做):① 先把原文截断到预览上限再掩码(输入有界);② 限制 scheme 量词为 [a-z][a-z0-9+.-]{0,31}://(RFC scheme 很短,直接消除二次回溯)。涉及安全相关的掩码逻辑,按 §7.4 留 owner 定夺是否单开处理。

行为变更披露(旧→新 + 原因)

结构化八问

  1. 是什么问题:egress governor 处于 429-hold 时,宿主重试每次都重传多 MiB 历史,bili 先付全量解析/转换/钉存成本(~5.5–7.2s + 数百 MB 内存尖峰)才本地拒绝——纯浪费。
  2. 什么原因导致的:governor 判定点在完整 ingest 之后(forwardCaptured 内),拒绝决策与 ingest 成本串行耦合。
  3. 影响范围与严重程度:需同时满足活跃 hold + 非 GET/HEAD + body ≥1MiB;严重度=成本/性能(CPU/RSS 尖峰、事件循环延迟),无数据丢失、无可用性中断(这些请求本来就会被拒)。
  4. 怎么修复的:prepare 之前加只读 peek(与 gate reject 分支字节对等、不消费 probe 槽),命中即拒;原 post-ingest gate 原样保留作兜底;两处共用一个渲染函数防漂移。
  5. 是否有回归问题:逐项审过既有行为——亚阈值路径、model-scope 双向、凭据隔离、provisional 组、governor 关闭、count-tokens、增量车道、水位快路径——均保持;语义漂移仅上面披露的两处(拦截形状响应形态、链式拒绝时机)。本地定向全绿,全量等 CI。
  6. 是否破坏缓存:不涉及——不触碰前缀缓存/会话状态/消息身份;拒绝路径不写会话任何东西。
  7. 是否改了设计方向:无——在既有 governor 框架内加一条只读快路径;两种压缩模式都推理过(plugin 车道大 body 请求同样受益;anthropic-wire 增量车道走独立路由不受影响)。
  8. 是否引入新配置:无——复用 throttle429* 旋钮;1MiB 阈值是代码常量(EARLY_HOLD_GATE_MIN_BYTES),未暴露为配置。

合并判定:可合并——feb8be9e1 的新 CI 转绿后即可由 owner 合并。本轮推送 commit:bf482e9eb(fixture 计数器)、feb8be9e1(测试内容加固);unused-export 红由 ci-check 并发修复(c26dbea)。既有的 O(n²) 掩码缺陷建议单独处理,不阻塞本 PR。

@ranxianglei ranxianglei left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

评审结论:建议可合并(需人审) — 独立复核通过,无阻塞项。以下为本地复核证据与两条非阻塞备注。

逐点复核(在 PR head feb8be9e1 上,基于 master c45ac21)

  1. peek 与 gate 的 REJECT 分支逐行一致:两个循环(holds + provisionalGroups)、跳过条件(!state.provisional && state.model !== undefined && state.model !== requestModel、slotBlocked(state, now)、strongerReject)完全相同;唯一刻意的差异是 peek 不铸造 probe token、不记 hasApplicable —— 即文档承诺的"只读对照"。新增套件里有专门的非变异钉(peek 后 probe slot 仍在、streak 不漂移)。
  2. 线上形状逐字节不变:post-ingest 原内联渲染被替换为共享 answerLocalThrottleReject,字段、状态码、retry-after(仅 429 带)一字未动;日志标签沿用 session.id / "passthrough",新路径加 "pre-ingest"。
  3. 保守边界都对:GET/HEAD、protocol=null、<1MiB、模型不匹配的 scoped hold(#4776 keyring 行覆盖)、governor 关闭时 inert、凭据覆写车道(#2336,raw header 指纹不可见→落旧路径)——逐一核对代码位置成立。CREDENTIAL_HEADER_NAMES 提为共享常量是对的:hold key 与 peek 读同一组头名,不会再漂。
  4. requestModelFromBody 不构成隐性 ingest:它是顶层结构性扫描(skipJsonValue 不物化值),O(n) 无分配,相对省下的 parse+convert+pin(~5.5–7.2s)可忽略;"零 ingest"合同成立(请求体本就在内存缓冲里)。
  5. 新测试文件对 master 自动失效(import 的 tryAnswerEarlyHold/upstreamThrottlePeekReject 在 master 不存在),回归钉天然有效。

本地验证

  • tests/issue5795-early-hold-fastpath.test.ts + tests/issue4204-throttle-governor.test.ts + tests/issue4776-keyring-model-hold.test.ts:42/42 绿;另跑 tests/issue4725-loop-429-observe.test.ts + tests/issue4743-degenerate-refetch-429.test.ts:10/10 绿。
  • npm run typecheck(tsc --noEmit)干净。
  • CI(plan/gate/build-artifact/static-gates 矩阵)该 head 全 SUCCESS。

两条非阻塞备注(都不需改动,记录在案)

① forge 车道的角落行为变化:fast path 位于 prepare 之前(handle.ts:512),而 codexForge/autoModeForge 判定在 prepare 内(~2100 行起)。若某次 forge/classifier 请求体真 ≥1 MiB 且当时 origin+credential 正被 hold,宿主将收到本地 429/424 而非锻造响应。PR 已论证 forge 形状为 KB 级,且方向安全(零出口、机器可读的退避信号、窗口过后重试即可拿到锻造结果);要在 forge 之前排除它们就必须先解析体,等于自废"免 ingest"目的——维持现状是正确的取舍,留档即可。

② 阈值硬编码 1 MiB(EARLY_HOLD_GATE_MIN_BYTES):当前合理(<1MiB 走旧路径保 shape),若将来想按会话大小调参再提 knob,现在不加是对的(无配置面)。

宿主侧(dsh)的重建+重传大头仍按 issue 线程里的 🔶 待决事项处理,与本 PR 正交。

@ranxianglei ranxianglei left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

补充更正(对上述备注①):我重新核对了代码,之前"forge 形状为 KB 级"的表述不准确——那是转述 PR 描述,与代码事实不符,三条本地应答车道的请求体都随会话历史增长:

  1. dsh compaction 应答/拒绝(handle.ts:2113 isDshCompactionCall,openai/anthropic/responses 三协议):注释明说"the compaction call is a full-budget request",同文件注释记录了 7.5MB 量级的单消息包(#2490);且 #2490/#3013 记录了该调用在压力下会高频重试(现场出现过 5051 次拒绝的风暴)。→ 该组合不罕见。
  2. auto-mode classifier 锻造(handle.ts:2205):isAutoModeClassifier 要求全转录 span(maxTranscriptSpan >= MIN_TRANSCRIPT_SPAN)——它本来就是 "O(n²) full-transcript-resend lane"(原注释),大会话下轻松超 1 MiB。
  3. codex forge(prepare-responses.ts 构建,本地应答 trigger/endpoint/prompt 三种)。

即:在这些车道被 hold 命中的窗口内,宿主会先收到本地 429/424(Retry-After)而非原本的本地锻造/拒绝应答。复核方向性后维持"不改"的判断,理由:(a) 三条车道本来就零出口、零计费,变成带退避语义的 429 不引入新的出口/数据风险;(b) hold 窗口是瞬态的,过期后即恢复原本行为;(c) 要在 fast path 之前识别这些形状就必须解析请求体——精确等于本 PR 要省掉的 O(history) ingest 成本,改动等于自废目标;(d) 若 owner 认为 dsh compaction 风暴在过载窗口内不应被 hold,正确的修法是把这三个本地应答块上提到 tryAnswerEarlyHold 之前(它们需要 parsed,代价已被 prepare 摊销),那是一次有意的区域重排,值得单独讨论而不是顺手改。请人审时知悉此窗口的存在。

@ranxianglei
ranxianglei merged commit 21c950e into master Nov 7, 2026
94 of 135 checks passed
@github-actions

github-actions Bot commented Nov 7, 2026

Copy link
Copy Markdown
Contributor

📦 Merged Build Published

This merge to master was published as 0.0.1065-master.21c950e (rolling master tag).

npm install -g billion-context@0.0.1065-master.21c950e  # this exact build
npm install -g billion-context@master  # always the newest merged build

npm view billion-context@master prints the current version at any time.

ranxianglei added a commit that referenced this pull request Nov 7, 2026
…ge gap)

Picked up from PR #4428 for the CI/test-signal small-fixes batch.
Resolved against current master (post-#5801/#4828): guards keep their new timeout-minutes; kernel-guard keeps the #4828 lockfile-exemption and now also skips only the PR-body declaration check on owner-fired workflow_dispatch.
ranxianglei added a commit that referenced this pull request Nov 7, 2026
…ge gap)

Picked up from PR #4428 for the CI/test-signal small-fixes batch.
Resolved against current master (post-#5801/#4828): guards keep their new timeout-minutes; kernel-guard keeps the #4828 lockfile-exemption and now also skips only the PR-body declaration check on owner-fired workflow_dispatch.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working dsh p2 scope/wiring

Projects

None yet

1 participant