Skip to content

[PR] loop: retry degenerate turns that end with whitespace-only text (#732 follow-up) - #5986

Merged
ranxianglei merged 6 commits into
masterfrom
2026-11-09_degenerate-turn-whitespace-gate
Nov 9, 2026
Merged

ranxianglei merged 6 commits into
masterfrom
2026-11-09_degenerate-turn-whitespace-gate

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

问题(plain talk)

思考模型(qwen3.8-27b 等)偶发"停笔":一整轮只输出 reasoning,最后附一两个换行,没有工具调用、没有正文,finish_reason=stop。客户端拿到这样一个空回合,任务就卡死不再推进。

已有的一次性重试机制(#732,针对"纯 reasoning 空回合")会覆盖这种情况,但对"reasoning + 空白文本(如 \n\n)"的变体失效:门控把 2 字节的空白文本读成了"已有可见输出",重试被跳过,空回合被静默放行。实盘表现为:worker 停笔后等宿主 nudge 才恢复,期间无任何重试日志。

改动

验证

  • 定向:tests/loop-degenerate-retry.test.ts 9/9 通过(含 W1/W2);compress 相关 3 文件 35/35
  • tsc --noEmit 干净
  • 字节级回放:从实盘抓包的真实请求(379KB,72 条消息)在两台 GPU 部署上各重放 10 次,20/20 正常产出工具调用,两部署行为一致 → 停笔非部署差异,是模型偶发 + 门控盲点
  • 私有热修版(含抓包脚手架)已在 e-work 虚拟机实盘运行 6 小时:自动重试触发 6 次、全部恢复成功、0 失败;2 次 double-degenerate 均被宿主 nudge 接住,无任务卡死
  • 全量回归套件 A/B 对照跑完后在评论区补结论

说明

本 PR 只含修复本体;用于定位问题的抓包脚手架(出站 body 落盘钩子 + 私有版本号)按惯例不进入公开代码,如需要可单独提议为正式诊断功能。

…follow-up)

A thinking model can end a turn with reasoning + whitespace-only text
(e.g. "\n\n") and finish_reason=stop. No tool call, no visible answer,
client stalls. The #732 one-shot retry gate read this as 'visible output
already reached' (length-2 text) and skipped the retry, passing the
empty turn through silently.

- track non-whitespace visible text (forwardedVisibleText) separately;
  the #732 gate now uses trimmed text + that flag, so whitespace-only
  terminals take the retry path exactly like pure-reasoning terminals
- the #1988 truncation gates keep the original forwardedVisible
  semantics (forwarded, not parsed) — untouched
- observability: warn when the one-shot budget was consumed and the
  retry output was empty again (double-degenerate), so the fleet can
  count the residual case the host nudge still covers
- compress receipt hardening: the FAILED receipt now ends with an
  explicit return-to-task instruction, reducing the observed pattern of
  a model derailing into the compression bookkeeping after a refused
  call and then going quiet
@ranxianglei ranxianglei added bug Something isn't working p2 labels Nov 9, 2026
@github-actions

github-actions Bot commented Nov 9, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-11-09_degenerate-turn-whitespace-gate (8c3a3f1)

Latest build — install this exact version

npm install -g billion-context@0.0.8714-pr.5986.54a1865

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
8c3a3f1 billion-context@0.0.8714-pr.5986.54a1865 build
4c15903 billion-context@0.0.8707-pr.5986.5a40798 build
c5fed0e billion-context@0.0.8705-pr.5986.4b39171 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr5986.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei ranxianglei changed the title loop: retry degenerate turns that end with whitespace-only text (#732 follow-up) [PR] loop: retry degenerate turns that end with whitespace-only text (#732 follow-up) Nov 9, 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。本 PR 属 degenerate-terminal-turn 自动重试家族,open 里有三个同家族兄弟 PR,各自切面不同:本 PR(#5986)改触发判定(whitespace-only 文本也算退化)+ double-degenerate 可观测性;#5897 改开关与措辞(默认关 + nudge 改中性);#4935 加另一种触发形状(bare-announcement);#5965 修重试耗尽后的序列化形态。详见下方「协同风险」。

标签: bug / scope/kernel / p2(缺陷在 loop core 上下文管理;停笔可被宿主 nudge 接住、能用但有明显问题 → p2;修复 client-无关,不打客户端标签)

Review 结论

判定:可合并(待剩余 CI shard 全绿)。方向正确,diff 干净(+77/-3,单 commit),测试有牙;两处非阻塞发现,其中一处我会在 CI 绿后直接补 commit 修掉(见发现 2)。

八问

  1. 是什么问题:思考模型偶发一整轮只输出 reasoning + "\n\n" 这类空白文本,无工具调用、无正文、finish_reason=stop;客户端拿到空回合,任务卡死直到宿主 nudge 兜底。
  2. 什么原因导致的:[BUG] In some cases, with Codex's auto model + DeepSeek, it will suddenly terminate automatically. #732 重试门控把「字节已转发」当成「可见输出已交付」——原门控 assistantText.length === 0 把任何非空字符串(含 2 字节的 "\n\n")读成「已有可见输出」而跳过重试;但宿主回合语义里 whitespace-only 文本和 reasoning 前缀一样属于不可见输出。表面现象 = 停笔;机制 = 门控条件对「可见性」的定义错了一档。
  3. 影响范围与严重程度:所有思考模型 × 所有 wire(openai/anthropic/responses)、两种压缩模式都可能命中该形状;修复前 = 任务卡到宿主 nudge(分钟级等待且无任何重试日志),修复后 = 秒级自动重试一次。p2:能用但有明显问题,有绕行(nudge)。
  4. 怎么修复的:单独追踪「非空白可见文本」(forwardedVisibleText,parse 时置位);degenerate 重试门控改用 trim 判空 + 该标志;Provide a true native, non-server implementation for the pi version. | 为pi版本提供非server的真native实现 #1988 截断门控保持原 forward 语义不动——截断判定要「是否真转发过」、退化判定要「是否可见过」,两种语义刻意分离,这个边界划得对;另加 double-degenerate(预算用完且重试输出仍空)warn 让舰队能量化残留场景;compress parse-failure 收据尾部追加「回到原任务」一句,把收据从「报错」变成「报错+下一步」(实盘观察到模型被拒后纠缠压缩簿记再停笔)。
  5. 是否有回归问题:逐项清点既有行为——
  6. 是否破坏缓存:否。重试机制沿用 [BUG] In some cases, with Codex's auto model + DeepSeek, it will suddenly terminate automatically. #732 既有实现,nudge 只活在重试请求体里;收据变更仅发生在 compress parse-failure 错误路径,同版本内字节确定,不影响后续请求的前缀缓存。
  7. 是否改了设计方向:否。扩展现有门控,无架构/边界/数据流转向。
  8. 是否引入新配置:否,无 env/config/flag。(对照:[PR] Empty-turn continuation retry: off by default (opt-in switch) + neutral nudge wording #5897 提出默认关开关,那属于 config surface、需 owner 批准;本 PR 未碰。)

验证证据(本审核独立执行,非转述)

  • base 新鲜:分支落后 master 0 commit,单 commit c5fed0e,diff 3 文件 +77/-3
  • 本地定向跑改动相关路径 10 个套件 90/90 全过:loop-degenerate-retry(含新增 W1/W2)、issue1029-acp-marker-stale-refs、compress-reject-reasons、issue5220-failed-compress-receipt、issue2303-draft-tail、plugin-empty-turn-retry、issue2756-degenerate-gate、issue-null-toolcalls-degenerate-retry、issue4743-degenerate-refetch-429、issue2612-compaction-draft-tail(全量回归交给 CI,本地不重复跑)
  • tsc --noEmit 干净
  • 测试有效性:W1 断言重试触发(fetchCalls=1)+ nudge 进 body + thinking/空白/重试内容都到达客户端;W2 断言真实正文不重试——双向有牙,门控往松/紧任一方向回归都会被抓
  • CI 实况(镜像缓存,刚查):30 checks,0 failed,static-gates(win/ubuntu)+ 2 个 test shard 已绿,9 个 test shard 仍在跑
  • real-client e2e lane 显示 skipped 是设计使然:ci-e2e.yml 对 PR 事件要求 ci:full 标签才真跑(三层 CI 门控,workflow gate job)。作者以字节级回放补位(实盘抓包在两台 GPU 部署各重放 10 次,20/20 正常)+ 私有热修版实盘 6 小时(自动重试 6 次全部恢复)。若 owner 想合前过一遍真客户端 e2e,给 PR 打 ci:full 标签即可触发(花真实上游 token,我不擅自打)

非阻塞发现

  1. 冗余条件(说明,不改):门控里 !forwardedVisibleText 与 assistantText.trim().length === 0 逻辑等价(标志置位 ⟺ 某 delta 含非空白 ⟺ 拼接文本 trim 非空;两者都在每次 attempt 开头重置,core.ts:846/860)。留作 defense-in-depth + 意图文档没问题,但不要误以为它携带独立信息;注释里「NEVER reuse it in the truncation gates」的护栏写得好,保持。
  2. warn 注释措辞不准(CI 绿后我直接补一个标注 commit):core.ts:1246 注释说 "consumed this round",但 degenerateRetried 是 request 作用域(core.ts:503,从不按 round 重置)——在「重试成功产出工具调用 → 后续 round 又退化 → 放行」这条稀有链上,warn 会触发而 "the retry output was empty again" 归因不准(空的是那个新回合,不是重试输出)。会把注释改成 request 作用域表述;warn 本身数的仍是该数的残留场景,逻辑不动。
  3. double-degenerate warn 行没有直接测试(loggerLog 未被测试捕获):纯观测性,可接受,记录在案。
  4. 「回到原任务」句只加在 "no valid ranges parsed" 这一种收据上:empty-call / arg-corruption / breaker 等其他拒绝收据没加。观察到的失败模式确实在这种收据上,窄范围合理;要不要推广到全部拒绝收据属产品决策,留给 owner。

协同风险(合并顺序建议,比代码细节更重要)

#5897 和 #4935 与本 PR 同时改 src/loop/core.ts 的 degenerate-retry 区域(#5965 不碰 core.ts,但与 #5897 共享两个测试文件)。建议先合本 PR:base 最新、改动最小、触发正交、不依赖任何决策;然后 #5897/#4935 rebase 吸收本 PR——尤其 #5897 的默认关开关会包裹本扩展,其 nudge 改词会影响 W1/D1/O1 的 includes("ended without a tool call") 断言(该 PR 已在小改同一测试文件,rebase 时需对齐)。反过来若 #5897 先合,本 PR 必须 rebase 且 W1/W2 要显式开 opt-in 才能验证。

收工判定

可合并:剩余 9 个 CI shard 全绿后 owner 可直接合;想先过 real-client e2e 就打 ci:full。CI 最终状态我在下一条评论跟进。

@ranxianglei

Copy link
Copy Markdown
Owner Author

非阻塞发现
这里都当场修复

ci-check added 4 commits November 9, 2026 09:01
…te-retry gate (#5986 review)

Review finding 1 of PR #5986: the flag was provably equivalent to the gate's
assistantText.trim().length === 0 check — both reset per attempt at the top of
the retry loop, and assistantText is mutated only by delta accumulation, so the
flag could never disagree with the trim check. Removing it keeps one source of
truth; behavior is unchanged. The truncation gates keep forwardedVisible, which
needs forwarded-not-parsed semantics; that boundary is now documented on the
trim condition itself.
…scoped budget (#5986 review)

Review finding 2 of PR #5986: degenerateRetried is declared once per request
(core.ts top of runCompressLoop) and shared with the #2303 draft-tail retry
path — it is never reset per round. The warn's comment and log line both
claimed 'consumed this round' / 'the retry output was empty again', which
misattributes the pass-through in the chain where an earlier round (or a
draft-tail retry) spent the budget and a LATER round degenerates fresh. Reword
comment and log message to the accurate request-scoped attribution; keep the
stable 'DOUBLE-degenerate' token for fleet grep continuity. No logic change.
…iew)

Review finding 3 of PR #5986: the double-degenerate pass-through (per-request
retry budget already spent, round still ended degenerate) was observable only
in fleet logs — no test captured loggerLog. Add R1/R2 via the setLogCapture
seam: R1 drives the degenerate-re-request + degenerate-retry chain and asserts
exactly one DOUBLE-degenerate warn with the request-scoped attribution; R2
drives a successful retry and asserts NO double-degenerate warn while proving
the capture is live via the auto-retry warn. Gate discipline holds both ways.
…re receipts (#5986 review)

Review finding 4 of PR #5986: the 'return to the original task' sentence was
appended only to the 'no valid ranges parsed' receipt, though the observed
failure mode (model entangled in compression bookkeeping after a rejection,
then stopping out) applies to every parse-failure receipt. Extract it into a
named constant RETURN_TO_TASK_SUFFIX and append it to the empty-call and
argument-corruption receipts as well; the armed-breaker one-liner already
carries its own 'Continue your task without compressing.' and is left
untouched. The policy-refusal receipts (superseded-ref, out-of-recommendation)
are deliberately unchanged: their repair instructions ('re-issue with updated
labels', 'stay within the recommendation set') take precedence over 'move on'.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

收到——四条非阻塞发现已全部当场修复,push 到 PR 分支(4 个独立标注 commit,tree 干净):

commit 对应发现 内容
65dd4b39 refactor(loop) #1 冗余条件 删掉 forwardedVisibleText(声明/重置/置位/门控共 4 处)——已证明与 assistantText.trim().length === 0 恒等价(两者同处 attempt 循环顶重置,assistantText 仅由 delta 累加变更)。行为零变化;截断门控继续用 forwardedVisible,「两种语义不可混用」的护栏注释挪到了 trim 条件上
51825597 fix(loop) #2 warn 措辞 注释与日志行改为 request 作用域归因(degenerateRetried 声明于 core.ts:503、与 #2303 draft-tail 共享预算、从不按 round 重置);保留 DOUBLE-degenerate grep token 不变,舰队日志连续性不受影响。纯注释+日志文案,无逻辑改动
7a513466 test(loop) #3 warn 无测试 新增 R1/R2(走 logger.ts 的 setLogCapture 测试缝):R1 = 退化重请求+退化重试链 → 恰好一条 DOUBLE-degenerate warn 且归因正确;R2 = 重试成功 → 零条 double-degenerate warn,同时断言 auto-retry warn 存在以证明捕获生效(负向断言有牙)
4c159034 fix(stream) #4 收据推广 「回到原任务」句抽成具名常量 RETURN_TO_TASK_SUFFIX,追加到 empty-call 与 arg-corruption 两种 parse-failure 收据尾部

#4 的范围判定(需要你知悉,可推翻):

行为变更披露(旧→新):empty-call / arg-corruption 两种 compress 失败收据的尾部,旧 = guard/loopGuard 文本后直接收尾,新 = 追加一句「This failure only concerns the compress call. Return to the original task now and continue it (visible output or a tool call).」——仅错误路径、同版本内字节确定,不影响正常回合与前缀缓存。其余三条均无对外行为变化(#1 恒等价删除、#2 仅日志文案、#3 纯测试)。

本地验证:改动相关路径 14 个套件 133/133 全过(含新增 R1/R2、原 W1/W2/D1/D2/O1/O2/A1/B1/B2,以及 pin 收据字符串的 issue1502/compress-observability/breaker 三件套);tsc --noEmit 干净。全量回归交给 CI。

CI 状态:push 已触发新一轮 CI(此前在跑的 shard 被 cancel-in-progress 取消,属预期——旧 head 的结果本来就会被新 head 取代)。最终全绿/红结论我在下一条评论跟进。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

CI 最终状态跟进(head 4c15903,4 个 review-fix commit 之上):

先纠正我上一条的一个误判:此前说「real-client e2e lane 因缺 ci:full 标签被 skip」——错了。这个 head 上实际跑了 9 条独立 workflow 且全部 success:E2E (real codex)、e2e-native-pi、e2e-native-opencode、e2e-pi-codex-ws、e2e-fake-codex、e2e-local-registry、e2e-codex-persona、Theorem Lane、CI Windows codex overlay。真客户端覆盖不仅存在,而且是绿的。(之前读的是主 CI workflow 检查矩阵里的 e2e job,与这些独立 workflow 不是一回事。)

主 CI run(42543334847):52 项 check = 37 success + 14 skipped + 1 failure。唯一红:test (windows-latest, 22, 2) —— 该 shard 2005 个用例只挂 1 个:

not ok - #411 plugin stream: client abort keeps sniffed usage, logs info (not error), clears the upstream timer
error: 'condition not met within timeout'   (waitFor, issue411-abort-usage.test.ts:139, 用时 3.5s)

判定:偶发误报(flake),非本 PR 回归,四重证据:

  1. master 绿(已查 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 总控最新评论:master tip 848f59b 17/17 success)→ 排除继承;
  2. diff 不触及故障路径:该用例测的是 client-abort 后的 usage 留存 / upstream timer 清理 / info 日志时序;本 PR 四个 commit 只动 degenerate-retry 门控(core.ts)、compress 收据文案(stream.ts)、degenerate 测试——零交集;
  3. 同分支旧 head c5fed0e(无 review-fix)同一 shard 全绿;
  4. 本地连跑 3 次全绿(每次 5/5)。

而且这个测试文件自己就带着 #2305 的注释:「the info abort log is a SEPARATE async side effect that can land after the timer clears — waiting avoids the synchronous-assert flake seen on a loaded windows-latest runner」——挂掉的正是第三个 waitFor(3s 预算等那条异步 info log),高负载 Windows runner 上的已知 flake 家族(#4099/#4100/#5625)。

处置:按 #4599 既有先例,分支零改动,已对 run 42543334847 触发定向 rerun-failed-jobs(HTTP 201),只重跑那一个失败 shard。已按协议在 #4599 留一行结论。若重跑再红,归属翻转为真回归,我会升级深挖而不是再重跑第二次。

重跑结果出来后我发最后一条确认(预期:52/52 无 failure → 可合并)。

#5986 red CI)

windows-latest/22 shard failed '#411 plugin stream' on a 3s budget for the
async abort info-log; this file was the only suite outside the repo-wide
'timeout waiting for …' idiom, so the #4161 classifier saw a concrete
failure instead of a recoverable timing-window flake. Label every waitFor
call site, throw the idiom, and widen the log-delivery budget to 8s
(belt and suspenders per #4161).
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

CI 最终确认(刚查镜像缓存,快照时间 09:53,head 8c3a3f1):30/30 全绿(16 success + 14 skipped,0 failed,0 pending)。上一条承诺的最后一条确认到齐。

自上一条以来分支上多了一个非我提交的 commit,已审过:8c3a3f149b(ci-check worker,09:31)test(ci): align issue411 waitFor with the #4161 timing-window idiom——它比我触发的「纯重跑」走得更深:把那个红 shard 的根因(测试文件用非惯用语超时字面量,导致分类器无法识别为可恢复时序 flake)在源头修掉。我的审核结论:通过——

  • 声明属实:timeout waiting for … 惯用语在全仓 22 个测试文件 29 处使用,issue411 确为唯一离群者;
  • waitFor(cond, what, ms=3000) 签名变更让编译器强制全部 7 个调用点带具体标签(无默认值 = 漏改即编译错),错误信息从笼统的 "condition not met within timeout" 变为带上下文的 timeout waiting for ${what};
  • 断言零改动;唯一放宽是实际挂掉的异步日志投递等待 3s→8s,注释引本事故;
  • 本地验证:issue411 套件连跑 3 次全绿(每次 5/5)+ tsc --noEmit 干净(新 head)。
    我先前触发的重跑针对旧 head,不再具有判定意义,以此 run 为准。

八问(最终版,覆盖分支全部内容)

  1. 是什么问题:思考模型偶发「reasoning + 空白文本」空回合静默放行,任务卡死等宿主 nudge。
  2. 什么原因导致的:[BUG] In some cases, with Codex's auto model + DeepSeek, it will suddenly terminate automatically. #732 重试门控把「字节已转发」误当「可见输出已交付」,whitespace-only 文本绕过了 trim 判空之外的原条件。
  3. 影响范围与严重程度:所有思考模型 × 所有 wire × 两种压缩模式;修复前卡到宿主 nudge(分钟级、无日志),修复后秒级自动重试。p2。
  4. 怎么修复的:degenerate 重试门控改用 trim 判空(冗余标志位经 owner 拍板后删除,语义并入注释);double-degenerate warn 归因修正为 request 作用域并补双向测试;compress 全部 parse-failure 收据统一追加「回到原任务」指令(breaker 与策略拒绝收据刻意不动);ci-check 的 issue411 测试对齐 CI: windows-latest node-22 shard 2 intermittently red on timing-sensitive e2e tests under load (rotating flakes: issue760, #4079, storage-full-flip #3818) #4161 flake 惯用语(仅测试)。
  5. 是否有回归问题:逐项清点结论不变且新增 commit 为 test-only——Provide a true native, non-server implementation for the pi version. | 为pi版本提供非server的真native实现 #1988 截断门控 / [BUG] In some cases, with Codex's auto model + DeepSeek, it will suddenly terminate automatically. #732 / workbuddy报错Empty stream: upstream gateway sent only placeholder chunks without any model output (chunks=0, bytes=0) #821 / DSH native mode: assistant text ending with </summary> is treated as a completed step, so the turn silently stops (149 samples from session transcripts) #2303 / bug: Responses degenerate retry overlooks custom tool calls and drops completed-attempt usage #1862 既有行为全部保持(定向套件 133/133 + 本次 issue411 3×5/5);行为变更披露维持上一条:①whitespace-only 空回合现在会重试一次(预期效果,代价=每受影响回合多一次上游调用,nudge ephemeral);②empty-call/arg-corruption 收据尾部多一句指令(仅错误路径、同版本内字节确定);③issue411 测试的超时字面量与一处等待预算变化(测试内部,不影响产品行为)。无 fail-fast→吞错、无阈值/时序漂移。
  6. 是否破坏缓存:否(nudge 不落盘不进客户端历史;收据变更仅错误路径且版本内确定;新 commit 不触及任何缓存路径)。
  7. 是否改了设计方向:否。
  8. 是否引入新配置:否。

状态备注:PR 当前 mergeable_state=behind(master 在其后有新提交)——按仓库惯例 BEHIND 不构成合并阻塞,GitHub 可正常合入;若你想先 rebase 最新 master 再合也可以,说一声我来做。

收工判定

可合并:CI 30/30 绿(base 为 8c3a3f1,含原作者 commit + 4 个 review-fix commit + 1 个 ci-check flake 惯用语 commit)、本地定向验证全过、八问无拦项。owner 可直接合并。

@ranxianglei
ranxianglei merged commit 1f311dd into master Nov 9, 2026
52 checks passed
@github-actions

github-actions Bot commented Nov 9, 2026

Copy link
Copy Markdown
Contributor

📦 Merged Build Published

This merge to master was published as 0.0.1108-master.1f311dd (rolling master tag).

npm install -g billion-context@0.0.1108-master.1f311dd  # 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.

Louie0806 added a commit to Louie0806/billion-context that referenced this pull request Nov 9, 2026
…itespace gate, ranxianglei#5849 held-tag flush on inert frames

Advances the base from 7d18b7a to 1f311dd (8 files). Two files overlap with
this lane: src/plugin.ts and src/loop/core.ts. Upstream added a whitespace-only
degenerate-retry gate (ranxianglei#732 follow-up) around the request-scoped budget warn in
runChatLoop, and stopped draining held tags on inert mid-stream frames (ranxianglei#5848).
Both hunks sit clear of the lane's strip gates, so the auto-merge needed no
manual resolution.

Notable upstream commits adopted:
 - c5fed0e loop: retry degenerate turns that end with whitespace-only text (ranxianglei#732 follow-up)
 - 65dd4b3 refactor(loop): drop redundant forwardedVisibleText from the degenerate-retry gate
 - 5182559 fix(loop): correct double-degenerate warn attribution to the request-scoped budget
 - 7a51346 test(loop): pin the double-degenerate warn both directions
 - 4c15903 fix(stream): extend the return-to-task instruction to all parse-failure receipts
 - 48076d2 fix(plugin): don't drain held tags on inert mid-stream frames (ranxianglei#5848)
 - 8c3a3f1 test(ci): align issue411 waitFor with the ranxianglei#4161 timing-window idiom
 - 016d933 test(launcher): cover the ranxianglei#5955 wedged same-lane takeover ladder
ranxianglei pushed a commit that referenced this pull request Nov 10, 2026
…nudge (#5897 rebase fallout)

Master's #5986 added the W1 subtest (reasoning + whitespace-only text)
after this PR branched, asserting the pre-PR nudge wording 'ended without
a tool call'. The PR rewords the nudge to 'produced no visible output'
and updated its own copies of that assertion (D1/O1 in this same file);
the W1 copy is updated here to match — mechanical rebase fallout, no
behavior change.
@github-actions github-actions Bot mentioned this pull request Nov 13, 2026
ranxianglei pushed a commit that referenced this pull request Nov 15, 2026
…nudge (#5897 rebase fallout)

Master's #5986 added the W1 subtest (reasoning + whitespace-only text)
after this PR branched, asserting the pre-PR nudge wording 'ended without
a tool call'. The PR rewords the nudge to 'produced no visible output'
and updated its own copies of that assertion (D1/O1 in this same file);
the W1 copy is updated here to match — mechanical rebase fallout, no
behavior change.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant