Skip to content

[PR] fix: stop naming sub-gate residue as foldable work in compress receipts (#5872) - #5888

Open
ranxianglei wants to merge 3 commits into
masterfrom
2026-11-07_subgate-success-tail
Open

ranxianglei wants to merge 3 commits into
masterfrom
2026-11-07_subgate-success-tail

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

Closes #5872

What

After a successful fold whose whole offer got exhausted, postCompressTail told the model: "Uncovered history remains — … m00XXX–m00YYY … The context is not at its minimum … inspect with acp_status(scope:"uncompressed")" — even when every listed run sat below compress.minCompressRange by construction, i.e. the kernel would refuse every one of those ranges on submission. Field data from the issue: 187 single-message compress retries after such receipts (83% targeting the model's own pending-reply notes), 18/19 failure streaks starting within seconds of a footer-carrying success, breaker advisories fired 13x and ignored.

How (two host-side fixes, receipt wording only)

1. Success tail (src/stream.ts postCompressTail + src/uncovered.ts). Each audit run now carries chars = Σ msg.text.length over its kept entries — the exact quantity the kernel's too-small gate sums against minCompressRange. When EVERY reported run is below the gate, the receipt keeps the honest mass report but drops the refs, the "not at its minimum" claim, and the acp_status pointer, ending on the contract stop phrase:

Uncovered history remains — all offered ranges are folded, but N msg(s) across K run(s), ~XK tok, sit outside every active block; every run is below the fold-size gate (min 5000 chars), so there is no range you can fold right now. Continue the task without compressing — these runs become foldable on their own as more history accumulates.

Any foldable run present → legacy wording byte-for-byte (#4331's honesty pin stands). Below-floor / drained sessions → verbatim #1387/pi#521 string, byte-stable.

2. Too-small refusal (failure path). While the circuit breaker is unarmed and no conclusive noViableAnywhere verdict applies, the receipt now surfaces the gap the kernel states but never computes: "You picked 2735 char(s) — 2265 short of the 5000-char gate; a single message almost never clears it, widen the range to cover more messages." Once armed, the hard-stop paragraph owns the receipt (#2360 §2 — one instruction per receipt).

Behavior changes (disclosed)

Surface Old New
Success-tail receipt, all runs sub-gate "…not at its minimum… inspect with acp_status" + named refs mass report + gate statement + "Continue the task without compressing"; no refs/claim/pointer
Success-tail receipt, any run foldable legacy wording byte-identical
Success-tail, below floor / drained verbatim minimum claim byte-identical
Too-small refusal, unarmed kernel line + generic hints same + explicit picked-vs-min gap (+ widen advice for 1-span picks)
Too-small refusal, armed breaker paragraph unchanged

No kernel change, no config/env surface added, no wire change (receipts are model-visible text only).

Coexistence

Open PR #5311 (from #5186) adds a [not foldable] prefix to described runs via describeUncoveredRuns; this change keeps that signature untouched (additive chars field only) and the two merge cleanly in either order.

Tests

New tests/issue5872-subgate-receipt.test.ts: P1 all-sub-gate wording (no refs/claim/pointer in tail), P1b audit chars mirrors Σ msg.text.length exactly, P2 super-gate hole keeps legacy wording byte-stable, N1/N2 drained + sub-floor keep the verbatim contract string, P3 too-small gap note appears while unarmed and yields to the breaker paragraph at failure 3. tests/issue4331-uncovered-history.test.ts gains the new required chars field on its formatter fixture (compile fix). Targeted local runs green: both files 16/16 + collateral sweep (fix-1124-compress-receipt, issue2146-compress-loop-breaker, issue2360-compress-receipt, issue2432-breaker-awareness, issue4812-armed-breaker-short-receipt, issue5220-failed-compress-receipt, loop-compress) 51/51; typecheck clean. Full suite: CI.

@github-actions

github-actions Bot commented Nov 7, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-11-07_subgate-success-tail (041c305)

Latest build — install this exact version

npm install -g billion-context@0.0.9449-pr.5888.c874b80

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
041c305 billion-context@0.0.9449-pr.5888.c874b80 build
0572da5 billion-context@0.0.9272-pr.5888.cfca442 build
eed71d3 billion-context@0.0.8543-pr.5888.0ea90a9 build
69a99bb billion-context@0.0.8515-pr.5888.029024b build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr5888.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei ranxianglei changed the title fix: stop naming sub-gate residue as foldable work in compress receipts (#5872) [PR] fix: stop naming sub-gate residue as foldable work in compress receipts (#5872) Nov 7, 2026
@ranxianglei ranxianglei added bug Something isn't working p2 scope/wiring labels Nov 7, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

查重: 不重复。 本 PR(GitHub #5888)是针对 #5872 的唯一在途修复;镜像 open [PR] 行里无其他条目指向同一线程/主题。相关但不同的既有工作(方向核对用):

标签: bug / p2 / scope/wiring(缺陷类;receipt 措辞导致模型空转 187 次 compress 重试、熔断误报——能用但有明显噪音,有绕行"继续任务即可",故 p2;改动面在服务端 receipt 管道 → wiring)。客户端标签不打:机制 lane-agnostic,源 issue 亦未定客户端。

当前状态(刚核实): head 69a99bb1 直接坐在 master tip 5af5f9fe 上,无需 rebase。CI(static-gates ubuntu/windows 已绿;test 矩阵 12 lane 运行中,fetched 2026-11-07 13:15Z);mergeable_state=blocked 仅因 checks pending。代码 review 进行中,结论按八问体例随后给出。

@ranxianglei

Copy link
Copy Markdown
Owner Author

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

Release-cut note (from issue #5872 reporter, verified against npm today): latest is still 0.1.212, and the retry-storm symptom in the field started on builds at/after that version — affected users currently believe "latest" already contains the fix. When this ships, a release-notes entry (#1870 severe-release form) should explicitly state: post-compress receipts no longer name sub-gate residue as foldable work, and too-small refusals now state the char shortfall. Flagging here so it is in context when the release is cut.

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

收到,已记下:发版时 release-notes/package.json(#1870 severe 形式)的条目需明确写两点——①压缩成功回执不再把低于折叠门槛的残留命名为可折叠工作;②too-small 拒绝回执现在会给出 picked-vs-min 的字符缺口。该备注属于发版 PR 的素材,不进本 PR。

继续审 #5888 本体(内核门槛语义已核对完毕,正在过测试质量与 #5311 共存性),结论按八问体例回复本线程。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

审核结论:可合并(已推 1 个标注修复 commit,CI 新 head 运行中)

1. 是什么问题:压缩成功后,若所有未覆盖残留片段都低于 5000 字符折叠门槛,回执仍把它们的 ref 编号、"context is not at its minimum" 断言和 acp_status 指针一起端给模型——等于在模型刚成功的时刻递上一张它必然做不完的工单(kernel 对每个被点名的 range 都会拒)。现场后果:187 次单消息 compress 重试、83% 打在模型自己的待回复笔记上、breaker 触发 13 次被无视。第二处:too-small 拒绝回执只报两个数字不报缺口,模型读不出算术,逐 ref 往前滑。

2. 什么原因导致的:回执措辞层与 kernel 的折叠资格判定不同源——kernel 门槛(kernel/src/compress.ts:531-550)对一次调用的全部 range 求和 raw msg.text.length 再比 minCompressRange,而宿主侧审计只报 token 质量、不报 chars,措辞层无从知道被点名的 run 提交后会不会被拒,只能默认乐观说法。失败路径同理:kernel 报了 picked 和 min,但没人算差值。

3. 影响范围与严重程度:所有车道上「成功压缩 + 残留全低于门槛」的会话(p2):功能不坏,但直接驱动重试风暴与 token 浪费;无该形态的会话零影响。

4. 怎么修复的:给审计 run 加 chars(Σ msg.text.length,与 kernel 门槛求和单元逐字一致,types.ts:492 有契约注释);全低于门槛 → 保留诚实质量报告、去掉 refs/断言/指针、以契约 stop phrase 收尾;存在任一可折叠 run → legacy 措辞逐字节不动(#4331 诚实 pin);below-floor/drain → #1387/pi#521 逐字串不变。失败路径 disarm 期追加 picked-vs-min 缺口注记(spanned=1 时附 widen 建议),armed 后由 hard-stop 段接管(#2360 §2),noViableAnywhere 终局判定时抑制。

5. 是否有回归问题:逐项过——①有可折叠 run 的路径:P2 钉死字节稳定(m00154–m00252 具名 span + claim + pointer 全保留);②drained/sub-floor:N1/N2 逐字串不变;③breaker armed:gapNote 让位(P3 r3 断言 "short of the" 消失);④noViableAnywhere/beyond/clamp/repeatGuard 等既有 hint 拼接顺序未动,gapNote 仅追加在末尾且仅 disarm+too-small+无终局判定三条件同时满足时出现;⑤无 kernel 改动、无 wire 改动、无 config/env 面新增;⑥describeUncoveredRuns 签名未动。审核中发现一处真实缺陷并已直接修掉(见下节)。

6. 是否破坏缓存:不涉及——回执是 tool result 里的模型可见文本,不进前缀缓存字节,不影响重试/增量游标/去重键。

7. 是否改了设计方向:不涉及。纯回执措辞层修复;audit/offer/floor/breaker 机制原样。

8. 是否引入新配置:无。


审核中直接推上的修复 commit(eed71d3a40,独立标注,已在分支上)

新措辞有两处相对 kernel 门槛真实语义过度承诺:

  • "so there is no range you can fold right now" —— kernel 门槛是单次调用跨 range 求和(compress.ts:532-550:totalRangeChars 累加全部非 skip、非 block-boundary 的 range 后再比一次阈值),所以两个 sub-gate run 合一批提交可以过门槛。改为 "no single run can be folded right now"(单 run 单独提交必被拒,此断言严格成立)。
  • "these runs become foldable on their own as more history accumulates" —— 对 covered blocks 之间的内部空洞为假:这类片段永不增长、永不折叠(TECHNICAL-NOTES 自己写的 permanent-hole class,[Bug] compress reports 'no compressible ranges remain' while mid-history spans sit outside every block #4331/出现持续压缩失败,无法自动恢复 #847)。删去该从句,stop phrase 承载指令即可。

TECHNICAL-NOTES.md 同步到同等精度。行为变更披露:all-sub-gate success-tail 句 old→new 即上述两处,其余路径逐字节不变;P1 测试 pin(marker + stop phrase + 三个 absence)无需改动仍成立。本地 16/16(issue5872+issue4331)+ typecheck clean。

跨 PR 发现(属 #5311,未在此修)

#5311 通过 describeUncoveredRuns 给列出的每个 run 加 [not foldable] 前缀,但该 formatter 有两个消费者:legacy success-tail 路径(src/stream.ts:701,恰在"存在可折叠 run"时才走)和 acp_status UNCOVERED HISTORY 段(src/acp-status.ts:329)。两条路径都可能列出超门槛 run——P2 几何(55K 字符中段空洞)下,标签会把一个完全可折叠的 run 标成 "[not foldable]",与同句 "not at its minimum … inspect with acp_status" 自相矛盾,并把 #5872 要消灭的工单式误读重新引回。文本共存已验证干净(两 PR hunk 零重叠、merge-tree 无冲突),语义冲突如上。建议 #5311 把标签谓词改为 r.chars < minCompressRange(本 PR 恰好提供了这个字段)或在 legacy 路径去掉标签;合入顺序无所谓,但该修正应在 #5311 合并前完成。

状态(写稿时 live 核验)

判定:可合并——待新 head CI 绿确认;红了我定向排查。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

收工状态(本线程工作已全部完成,上方楼层为完整记录):

合并权在你:CI 绿后可合 #5888 。

@ranxianglei
ranxianglei force-pushed the 2026-11-07_subgate-success-tail branch from eed71d3 to 0572da5 Compare November 10, 2026 15:23
@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 force-push 完成(rebase 到当前 master 58a8796224,新 head 0572da54ab)。GitHub 现报 mergeable: True(此前 dirty),mergeable_state: blocked 仅因新 head 的 required checks 还在跑——不是冲突。

本次是 CI 冲突修复车道(1/3),只做了 rebase + 解冲突 + 验证,没有改任何 PR 逻辑。

八问

  1. 是什么问题:PR [PR] fix: stop naming sub-gate residue as foldable work in compress receipts (#5872) #5888(sub-gate 残量不再在压缩回执里被称作可折叠工作)与 master 产生合并冲突(merge state dirty),分支无法合并。

  2. 什么原因导致:master 期间把根目录 TECHNICAL-NOTES.md 整体搬到 docs/technical-notes.md(根目录只留一行重定向 stub),而本 PR 恰好编辑了旧文件里的两行内容 → 该文件冲突。src/stream.ts 因 master 在同区有改动(applyRanges 相关)被 git 自动合并为 union,无文本冲突。

  3. 影响范围与严重程度:仅本 PR 分支不可合并;不影响任何已发布行为、不影响其他 PR/用户。

  4. 怎么修复的:① 把 PR 的两个 commit(69a99bb 主修 + eed71d3 措辞)rebase 到当前 master;② TECHNICAL-NOTES.md 保留 master 的重定向 stub,把 PR 那两行 [Bug] post-compress "not at its minimum" footer drives single-message compress retries (18/19 streaks; 83% of retried ranges are the model's own pending-reply notes) #5872 措辞逐字搬进 docs/technical-notes.md(postCompressTail 段的 UNLESS-every-run-below-minCompressRange 子句 + Behavior change 段第二句),另加一个标注清楚的 docs commit;③ src/stream.ts 自动合并后经测试确认 union 正确(PR 的 sub-gate 逻辑 + master 的 applyRanges 改动都在)。

  5. 是否有回归:本地 npm run typecheck clean;定向测试 26/26 通过(issue5872-subgate-receipt / issue4331-uncovered-history / compress-success-tail)+ 邻接 receipt/breaker 套件 51/51 通过;字节稳定 pin([Bug] compress reports 'no compressible ranges remain' while mid-history spans sit outside every block #4331/compress 成功输出缺剩余范围快照与停止提示(pi 侧 #420/#521 对齐) #1387/compress-success-tail)全部保持 → rebase 未引入回归。CI 新 head 静态 gate 已绿(13 success),0 failed,5 个 test-matrix shard 排队中,12 条 gated e2e lane 按环境跳过。本次运行未引入任何新的行为变更(PR 自己披露的行为变更保持不变)。

  6. 是否破坏缓存:不涉及。回执是模型可见文本,不进前缀缓存;本次仅解冲突 + 搬文档,未触碰任何缓存/游标/去重键逻辑。

  7. 是否改了设计方向:不涉及。

  8. 是否引入了新的配置:无。


后续:我继续持有这个 PR 的红项处理权。当前 5 个 test-matrix shard 仍在跑(runner 池排队拥塞,非失败);若其中任何一个转红,我会先查 #4599(CI 总控)判断是否继承性红,再按纪律处置。目前 0 failed,无需介入。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

CI 已全部跑完、全绿(接上一条八问的状态更新):新 head 0572da54ab 的 30 项 check 现读数为 0 failed / 17 success / 13 skipped / pending 0。上一条里"仍在排队的 5 个 test-matrix shard"现已全部完成并转绿;那 13 个 skipped 是环境内零凭据的 gated e2e lane,按设计跳过。

GitHub live 核验:mergeable: true(冲突已消除)、mergeable_state: blocked。这个 blocked 是 branch-protection 的人工合并门禁(合并权在 owner,agent 不合并),不是冲突、也没有失败的 check。

结论:本车道的两项交付——①合并冲突(rebase 到 master 58a8796224,新 head 0572da54ab)、②红项处理——均已完成,无待办;完整审核见上一条八问(floor 13)。CI 绿,可合 #5888 。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

本轮为验证性空跑——live 三源复核后确认无冲突、无需任何改动,未做 rebase/push。

本轮收到"merge state: dirty、需 rebase"的转发,但按状态新鲜度纪律先核 live,结果与前提不符:

  • git fetch 后 origin/master = 58a8796224,与我上一轮 rebase 所用的 base 完全相同,master 自该 rebase 起零新提交;
  • GitHub live:mergeable: True(非 dirty)、base_sha = 58a8796224、head_sha = 0572da54ab(我上轮 force-push 的 head);远端 PR 分支 == 本地 HEAD == 我的 head,无他人另推;
  • CI(镜像缓存):head 0572da54ab 上 0 failed / 17 success / 13 skipped / pending 0,全绿、无运行中项。

八问(本轮)

  1. 是什么问题:转发称本 PR 与 base 冲突、需 rebase 解冲突。
  2. 什么原因导致:实为陈旧/重复触发——我上一轮已把两个 commit rebase 到当时最新的 master(58a8796224)并 force-push,此后 master 未再前进,冲突并未复现,故本轮无冲突可解。
  3. 影响范围与严重程度:无影响;PR 的可合并状态保持不变。
  4. 怎么修复的:无需修复——仅做 live 核验(git fetch + GitHub REST + 镜像缓存三源一致),确认无冲突后不触碰分支(避免无谓 force-push)。
  5. 是否有回归问题:不涉及(未改任何代码)。
  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:无。

结论:PR #5888 当前 head 0572da54ab 已正好坐在最新 master tip(58a8796224)上、CI 全绿、mergeable: True;mergeable_state: blocked 只是人工合并门禁(agent 不执行合并)。无待办,可合 #5888 。

@ranxianglei

Copy link
Copy Markdown
Owner Author

该PR已按根因归类纳入 [session-resolve] 根因汇总 #6298,由汇总单统一跟踪处理;本单不再单独跟进,故关闭。如认为归类有误或问题仍未解决,请直接回复,会重新打开。

@ranxianglei

Copy link
Copy Markdown
Owner Author

状态更正(自动化脚本误操作): 本单此前被「纳入根因聚合」流程误关闭, 现重新打开。机器人工作已停, 本单由根因聚合线程统一跟踪, 根因修复落地后请手动关闭或在本单回复「close」。
[automated: re-open after bot-stop-only policy]

@ranxianglei ranxianglei reopened this Nov 12, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

重新打开:此前被归类聚合脚本误关。按 owner 口径,纳入根因聚合的 issue 只停机器人、不关闭,由根因聚合线程统一跟踪,根因修复落地后统一关闭。
(reopened: absorbed issues must not be closed, only bot work stopped — owner policy 2026-11-12)

@ranxianglei

Copy link
Copy Markdown
Owner Author

(更正)此前按"清理"口径误关闭;按 owner 口径纠正:被根因聚合纳入的 issue 只停机器人、不关闭。本 issue 重新打开,机器人工作已停。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

状态确认(live 三源核验:GitHub API + 镜像 CI 缓存 + 本线程楼层)——本轮无新工作,不推任何改动。

结论:PR 本身已处于「CI 绿、可合」状态(11-10 全绿记录见 floor 17/20),当前仅在 behind 状态等聚合线程 #6298 的根因修复落地后统一处置。无需 owner 决策,无待办。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

收到,确认按 owner 口径(2026-11-12)执行:

  • 本单保持 open,我不关闭——纳入根因聚合的 issue 只停机器人、不关闭,由根因聚合线程统一跟踪,根因修复落地后统一关闭。
  • 机器人工作已停,我 stand down:不再对本 PR 做 rebase / 红项处理。说明一点:master 已前进到 259cd87674,本 PR head 0572da54ab 现为 mergeable_state: behind(落后 base),但无冲突(mergeable: True)。既已停机器人,我不主动 rebase,落后状态留给根因聚合线程统一处理。
  • 当前 live 状态(本轮刚核,git + GitHub + 镜像缓存三源一致):PR open、无冲突、CI 0 failed / 17 success / 13 skipped / pending 0 全绿(head 0572da54ab)。

后续根因修复落地后,由根因聚合线程统一关闭本单即可;在此之前我不做任何改动。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

已确认(与上一条一致,不重复展开):本单我保持不关闭;机器人工作已停,我 stand down,不做 rebase / 红项处理;根因修复落地后由聚合线程统一关闭。无新增待办——后续除非被明确重新触发,我不动本单。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

确认(本条为收尾,不再重复):本单已重新打开,按 owner 口径(2026-11-12)——纳入根因聚合的 issue 只停机器人、不关闭。我(CI 工程师)已 stand down:不关闭本单、不 rebase PR、不做红项处理;根因修复落地后由聚合线程统一关闭。此后除非有新的明确指令,我不再对本单重复回帖。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

提示:同一条 owner 更正(误关→重开;只停机器人不关闭;聚合线程跟踪;根因修复后统一关闭)正被 upstream-sync 重复投递,我已在上方确认多次。为避免刷屏,这是本 bot 的最后一次确认,此后不再逐条回帖。

最终立场(不变):本单我不关闭;PR #5888 我不 rebase、不做红项处理;根因修复落地后由聚合线程统一关闭。若重复投递持续,建议同步层对该 owner 评论做去重。除非出现新的明确指令,我停止对本单的任何操作。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

PR #5888 最终状态快照(本轮 live 三源核验):

  • 合并状态:open、无冲突(mergeable: True)、mergeable_state: behind——master 已前进到 259cd87674,PR head 仍为 0572da54ab(落后 base,但可干净合并)。
  • CI:head 0572da54ab 上 0 failed / 17 success / 13 skipped / pending 0,全绿。
  • 处置:本单已重开、机器人工作已停。我(CI 工程师)不关闭本单、不 rebase、不做红项处理;根因修复落地后由根因聚合线程统一关闭。

风险 / 剩余:无阻塞项。唯一的"落后"是 master 前进所致;既已停机器人,留待聚合线程统一处理,不影响本 PR 正确性(CI 绿、无冲突)。

同一条 owner 更正已被 upstream-sync 重复投递多次,本条为完整状态收尾。此后对相同的重复投递我不再逐条回帖,除非出现新的明确指令。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 (ai: vllm-qwen/qwen3.8-27b)

进度汇报 + 一个需要 ops 处理的阻塞:

PR #5888 状态(本轮 live 核验):open、无冲突(mergeable: True)、behind master(259cd87674)、PR head 0572da54ab、CI 0 failed / 17 success / 13 skipped / pending 0(全绿)。

处置不变:本单已重开、机器人工作已停;我不关闭、不 rebase、不处理红项,根因修复落地后由聚合线程统一关闭。owner 的更正(被聚合脚本误关 → 重开、只停机器人不关单)我已多次确认。

问题:同一条 owner 更正被 upstream-sync 反复重复投递(现在一次带 3 份),我已逐条回帖 5 次。我这边没有工具能停掉同步层的重复投递,结果是我的进程因长时间无新输出被强制重启(32 分钟)。请求 owner/ops 对这条 owner 评论做去重(同一 upstream-sync 内容只投递一次),避免重复唤醒 CI 工程师。

在同步层去重之前,我对相同的重复内容不会再逐条回帖;若出现新的明确指令或新信息,我立即处理。

ci-check added 3 commits November 13, 2026 11:17
…ts (#5872)

After a successful fold whose offer is exhausted, postCompressTail named sub-gate runs (below compress.minCompressRange by construction) with copyable refs, claimed 'the context is not at its minimum', and pointed at acp_status — a work order the kernel refuses on every attempt. Field data: 187 single-message compress retries after such receipts, 18/19 failure streaks starting within seconds of one.

Two host-side fixes, both receipt-wording only (no kernel, no config, no wire change):

1. postCompressTail: each audit run now carries chars (Σ msg.text.length, mirroring the kernel gate sum exactly). When EVERY reported run sits below minCompressRange the receipt keeps the honest mass report but drops the refs, the minimum claim, and the inspection pointer, ending on the contract stop phrase. Any foldable run present -> legacy wording byte-stable (#4331's honesty pin stands); below-floor / drained sessions unchanged (#1387/pi#521 verbatim string).

2. Too-small refusal: while the circuit breaker is unarmed and no conclusive noViableAnywhere verdict applies, the receipt now states the gap ('You picked N char(s) — K short of the M-char gate') plus widen-the-range advice for single-span picks. Once armed, the hard-stop paragraph owns the receipt (#2360 §2, one instruction per receipt).

Tests: tests/issue5872-subgate-receipt.test.ts (positive gates P1/P1b/P2/P3 + negative gates N1/N2); tests/issue4331-uncovered-history.test.ts gains the new required chars field on its formatter fixture.
…rces (#5872)

Review follow-up to the sub-gate success-tail wording: two clauses over-promised relative to the kernel's actual gate semantics.

- "so there is no range you can fold right now" — the minCompressRange gate sums chars across ALL ranges of one call (kernel/src/compress.ts), so a multi-range batch of sub-gate runs can clear it. Say what is true: no single run can be folded right now.
- "these runs become foldable on their own as more history accumulates" — false for interior holes between covered blocks, which never grow and never fold (the permanent-hole class of #4331 / #847). Drop the clause; the contract stop phrase carries the instruction.

TECHNICAL-NOTES.md synced to the same precision. P1 test pins (marker + stop phrase + absences) hold unchanged.
Conflict-resolution follow-up for the rebase onto current master. Master
moved the root TECHNICAL-NOTES.md to docs/technical-notes.md (redirect stub
left behind), so the two #5872 wording lines this PR added to the old root
file have no home there anymore. Port them verbatim (final precision from
the review-fix commit) into docs/technical-notes.md; the root stub stays
untouched. No behavior change — documentation only.

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