Skip to content

[PR] fix: calibrate the dsh ledger floor against the output clamp (#5596) - #6551

Open
ranxianglei wants to merge 6 commits into
masterfrom
2026-11-13_5596-ledger-floor-calibration
Open

ranxianglei wants to merge 6 commits into
masterfrom
2026-11-13_5596-ledger-floor-calibration

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

现场(field shape)

dsh 原生压缩回放被拒(~866KB 信封)→ metadata dshCompactionRefusedBytes → 规划 floor = 866760/4 = 216690 tokens,而 dsh 自家计量 ~184K(bytes/4 高估 ~18%)。请求的 max_tokens 128384(dsh profile 50% 窗口默认)≥ #3384 giant 线(25% 窗口)→ floor 每轮驱动规划;在已偏高的字节代理上再跑 #5278 ×1.12 安全膨胀 + 膨胀值上的 5% margin(合计 ≈1.36×)→ 钳制砸进 4096 能力下限平台区(cap 1172 → 4096)。该泳道此后每个回答只有 4096 预算,两次真实回答被截断。

泳道 修复前 修复后
子泳道(floor 216690 / 256K 窗口 / 请求 128384) 4096(现场平台区) 28475(raw 公式)
本会话(floor 164287) 62797 83498

修复(两个独立失真,两处各修)

A1 — src/server/budget.ts(钳制校准):clampOutputBudget 加 skipSafetyInflation 参数;clampOutgoingOutput 仅在 floor 本身是规划基数(ledgerFloor > inputEstimate,严格大于)时传 true——floor 是 bytes/4 字节代理,本身已偏高,#5278 的膨胀是对低估器的矫正,套在字节代理上就是双重计数。est 驱动时传 false(#5278 逐字节不变);非 dsh 泳道 floor 恒 0,不受影响。

A2 — src/server/handle.ts(过期 floor 清除):authored checkpoint 落地会重置 dsh 的 raw 账本(#3723),旧高水位不再代表任何真实规模(现场 866760 就是这种「已接受的保守残留」)。authored 分支改为 delete 该键;高水位写入移入 refusal-only 分支,拒绝泳道保持 #2490 钉死的粘性行为。

没有采用 issue 字面建议的「拒发不记账」:那会重开 #2490 宿主侧死链(超大输出落进 dsh raw ledger → raw+输出 > dsh 自家窗口 → 死亡循环;tests/issue2490-ledger-floor-doomed-probe.test.ts e2e 钉死),且 floor 防的威胁来自宿主侧、与回放是否发给上游无关。修法是校准 + 落地时失效,不是拆除。

Pre-flight

@github-actions

github-actions Bot commented Nov 13, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-11-13_5596-ledger-floor-calibration (17eed58)

Latest build — install this exact version

npm install -g billion-context@0.0.9693-pr.6551.3dc0578

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
17eed58 billion-context@0.0.9693-pr.6551.3dc0578 build
331eb80 billion-context@0.0.9645-pr.6551.b53ca77 build
2c0ad30 billion-context@0.0.9627-pr.6551.f193510 build
e96d514 billion-context@0.0.9593-pr.6551.bec27a3 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr6551.tgz
npm install -g package

Install instructions are refreshed on each push.

ranxianglei and others added 3 commits November 14, 2026 00:27
…nds (#5596)

An authored compaction answer lands a checkpoint that resets dsh's raw
ledger to the folded snapshot (#3723), so the refused-replay high-water
no longer bounds any future raw scale. The guard recorded the envelope
size on the authored path too (before the branch), leaving a stale
sticky floor that kept planning the output clamp against bytes dsh will
never count again — the field lane sat clamped at 4096 behind an 866KB
envelope whose ledger shrank the moment the answer landed.

The authored path now deletes the key; the high-water write moves to
the refusal-only branch, where it stays sticky for the #2490 kill chain
(refused checkpoints never land, the ledger only grows).
…#5596)

The ledger floor (dshCompactionRefusedBytes / 4) is an over-counting
high-water BYTES proxy — it read ~18% above dsh's own tokenizer in the
field (216690 vs ~184K) — not a low-biased local token estimate. Running
it through the #5278 x1.12 safety inflation double-counts it, and with
the 5% margin on the inflated value (~1.36x total) the clamp slammed
into the 4096 capability-floor plateau near the window: the field sub-
lane left every answer with a 4096-token budget (cap 1172 floored at
4096) where the raw formula gives 28475; the session lane clamped to
62797 instead of 83498.

clampOutputBudget takes a skipSafetyInflation flag that clampOutgoing
Output sets exactly when the floor IS the planning basis (strictly
above the folded estimate). Est-driven calls pass false, so #5278 and
every non-dsh lane stay byte-identical; the pinned #2490/#3384 shapes
(5909, 55000), the #3071 capability floor, and the doomed-probe
surrender all hold on the flag path.

Pinned: tests/issue5596-ledger-floor-calibration.test.ts (field values
unit + two e2e lanes: the refused lane plans raw out of the plateau,
the authored lane goes out unclamped after the floor clears).
…ondition (#5596)

Review finding on the A1 commit: skipSafetyInflation was set on the
bare 'ledgerFloor > inputEstimate', but the floor is the planning basis
only when the #3384 gate ALSO fired (folded view at the enforced
headroom target, or a giant >=25%-of-window request). With the gate
off — the normal post-refusal continuation state, sticky floor above a
healthy folded est — the flag still took the raw path and stripped the
#5278 x1.12 safe pass from the est-driven clamp, contradicting the
fix's own 'est-driven stays byte-identical' claim and the flag's own
doc (the planning basis there IS the token estimate). In the firing
band (est near the headroom target, request below the giant line) the
clamp then went ~12% of est looser or no-op'd entirely, re-exposing
the #5278 under-estimator hole.

floorDrivesPlanning = (viewAtEnforcedTarget || giantRequestedBudget)
&& ledgerFloor > inputEstimate now drives both the flag and the
'planning floor' log clause. Pinned: new unit test s-5596-g (gate off,
floor 82.5K above est 70K -> the safe pass still clamps 20000 ->
17678; the bare-flag code sent 20000 unclamped). Field shapes
(28475/83498), the #3384 pin (5909), the #3071 plateau, the
doomed-probe surrender and both e2e lanes are unchanged.
@ranxianglei
ranxianglei force-pushed the 2026-11-13_5596-ledger-floor-calibration branch from e96d514 to 2c0ad30 Compare November 13, 2026 17:06
@ranxianglei ranxianglei added bug Something isn't working dsh p2 scope/wiring labels Nov 13, 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 —— 本线程只有一个在途 PR(#6551),无同主题并行 PR;本次审核的修复以增量 commit 加进该 PR 分支,未另开。

审核结论:发现并修复 1 处回归,其余通过

我独立审出的问题(A1 flag 条件过宽,已修):src/server/budget.ts:551 的 skipSafetyInflation 传的是裸 ledgerFloor > inputEstimate,但 floor 成为规划基数还需要 #3384 门同时触发(viewAtEnforcedTarget || giantRequestedBudget,见 :546)。门关 + 粘性 floor 高于 est——正是拒绝后的正常续跑态——时,flag 仍会把 est 驱动路径的 #5278 ×1.12 安全 pass 剥掉,与 PR 自己声称的「est-driven 逐字节不变」和 flag 文档注释(「planning basis 不是 est 时才置位」)都矛盾。后果:在 firing band(est 靠近 headroom 目标、请求 <25% 窗口)内,钳制比原来松约 12% est 或干脆不触发,重新暴露 #5278 要防的低估器漏洞(est 低估超过 5% margin 时 input+output 可超窗口)。

修法:floorDrivesPlanning = (viewAtEnforcedTarget || giantRequestedBudget) && ledgerFloor > inputEstimate,flag 与 :559 日志的「planning floor」子句(同一裸条件、同样会误报)一并改用。新增单测 s-5596-g 钉死该 case:门关 + floor 82.5K > est 70K → 安全 pass 仍生效钳到 17678(旧代码此 case 不钳、放行 20000)。现场数值(28475/83498)、#3384 钉值 5909、#3071 平台、#2490 投降语义、两条 e2e 全部不变。

其他审核动作:

  • 分支原落后 master 5 提交(仅涉 src/persist.ts Service crashed, sessions exploded instantly. | 服务崩溃,会话瞬间爆炸 #5219,无文件交集),已 rebase 到当前 master;
  • PR 原无标签,已按其自身 diff 打标:bug / dsh / scope/wiring / p2;
  • 本地定向验证:新测试文件 10/10 + PR 正文所列 9 个回归套件全绿 + typecheck/build 干净;
  • A2 侧我核过:delete 有 markDirty 持久化;refusal 分支高水位写入与 dsh压缩与bili压缩冲突 #2490 粘性行为原样保留;authored 分支的响应写入是条件式(既有防御模式,该守卫点在任何响应字节发出之前执行,headersSent 实际不可达,且与改动前同形,非本 PR 引入)。

八问

  1. 是什么问题:dsh 泳道被拒压缩回放的输出预算规划 floor(bytes/4 高水位,高估 ~18%)上再叠 [Bug] clampOutputBudget's output safety buffer exactly cancels the estimator error: after clamping, input + output is al | [Bug] clampOutputBudget 的输出安全垫与估算器误差精确抵消:钳制后 input+output 恒为 window+1,上游 502 ContextWindowExceededError #5278 ×1.12 + 5% margin(合计 ≈1.36×),把钳制砸进 4096 能力下限平台区,现场泳道每个回答只剩 4096 预算、两次真实回答被截断。
  2. 什么原因导致的:两个独立失真——A1:floor 是已高估的字节代理,却被套上针对低估器的 [Bug] clampOutputBudget's output safety buffer exactly cancels the estimator error: after clamping, input + output is al | [Bug] clampOutputBudget 的输出安全垫与估算器误差精确抵消:钳制后 input+output 恒为 window+1,上游 502 ContextWindowExceededError #5278 膨胀(双重计数);A2:authored checkpoint 落地会重置 dsh 原始账本([Bug][dsh] Unbounded growth of raw history → 39.7MB session file overwhelms dsh desktop renderer process, entire UI beco | [Bug][dsh] 原始历史无上限增长 → 39.7MB 会话文件压死 dsh 桌面端渲染进程,UI 全部无响应 #3723),旧高水位却仍粘性,继续拿 dsh 永远不会再计的字节规划。
  3. 影响范围与严重程度:仅 dsh 泳道(其他泳道 floor 恒 0);触发条件 = 存在拒绝高水位 + floor 驱动规划(giant 请求或 view 达 headroom 目标)。p2:能用但回答被截断,非会话不可恢复/数据丢失/缓存泄露类。
  4. 怎么修复的:A1 在 floor 是规划基数时跳过 [Bug] clampOutputBudget's output safety buffer exactly cancels the estimator error: after clamping, input + output is al | [Bug] clampOutputBudget 的输出安全垫与估算器误差精确抵消:钳制后 input+output 恒为 window+1,上游 502 ContextWindowExceededError #5278 膨胀、按 raw 公式判;A2 authored 路径删除高水位键(落地即失效)、高水位写入移入 refusal-only 分支。未采用 issue 字面建议的「拒发不记账」——那会重开 dsh压缩与bili压缩冲突 #2490 宿主侧死链。
  5. 是否有回归问题:有 1 处,审核发现已修(即上文 flag 过宽,修后 est 驱动路径恢复逐字节不变);已核查的既有行为:dsh native compaction guard refuses 100% → client ledger never shrinks → every step fails from turn 8, last 4 turns output empty | dsh 原生压缩守卫 100% 拒绝 → 客户端本地账本只增不减 → 会话第 8 轮起每步失败、后 4 轮输出全空 #3384 钉值 5909 ✓、dsh压缩与bili压缩冲突 #2490 投降语义与拒绝侧高水位 ✓、[bug] 0.1.195 输出预算三路饿死仍复现:无 output 上限模型 restore 无下限(1→252、fork 原样转发 1)、#2490 planning floor 钳到 1479–2623 → 3 会话 9 次 finish=length 空轮不重试(#546/#1665/#1840/#2904/#2756 残留) #3071 能力下限 ✓、[Bug] clampOutputBudget's output safety buffer exactly cancels the estimator error: after clamping, input + output is al | [Bug] clampOutputBudget 的输出安全垫与估算器误差精确抵消:钳制后 input+output 恒为 window+1,上游 502 ContextWindowExceededError #5278 est 路径 ✓(9 个回归套件全绿)。
  6. 是否破坏缓存:不涉及——改动在出站输出预算规划路径,不触碰前缀缓存/缓存键/重试/增量游标。
  7. 是否改了设计方向:不涉及——floor 仍粘性、dsh压缩与bili压缩冲突 #2490 死链防护原样;仅校准判定公式与落地失效时机。有意的行为变更一处(已披露):authored 路径从「记录高水位」改为「删除高水位」,即 A2 修复本体。
  8. 是否引入了新的配置:不涉及。

状态(实时核过):head 2c0ad301c8(rebase + flag 修复),CI 运行中 0 失败 / 21 待跑(列表里的 cancelled 是旧 head 被新 push 取代的正常取消);mergeable_state = blocked(等 check)。CI 全绿即可合并。

ranxianglei and others added 3 commits November 14, 2026 02:08
…e default (#6480)

The suite-wide baseline (tests/preload-storage-baseline.mjs) pins
BILI_STORAGE_UNIFIED ??= "legacy" for the legacy-written corpus, but this
suite asserts the fresh-install default posture — the exact #6480 pure-sqlite
flip the storage-sqlite-default notice advertises. Under the CI preload chain
the pin hid the notice and two subtests failed deterministically (red in
every shard leg containing the file). Drop the inherited pin so the
production default applies, per the preload's own pin-yourself contract.
The suite baseline preload (preload-storage-baseline.mjs) pins
BILI_STORAGE_UNIFIED="legacy" for the broad corpus. After #6480 flipped the
storage default to pure-sqlite, the notice-registry suite's visibleNotices
assertions expect the pure-sqlite default notice, which is gated on
unifiedSelectionIsSqlite (unified "sqlite" or absent) — the preload's
"legacy" made the notice silent, so the suite was born-red on master. Pin
the selection explicitly (#4011 lesson: a test that depends on the production
default passes locally but fails in CI).
…er-r1

# Conflicts:
#	tests/notice-registry.test.ts
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

CI fix run — PR #6551 (calibrate the dsh ledger floor against the output clamp, #5596), auto-fix lane attempt 1/3

Re-verified the live check state first: the mirror cache was stale (fetched ~10h ago, not refreshing), so I confirmed the two failing jobs directly via a one-off API read. Both test (windows-latest, 22, 1) and test (windows-latest, 22, 3) failed at run-ci-tests.mjs — not green, so I proceeded.

Diagnosis — all 13 failures are ONE signature, and none are this PR's code:

What I pushed (new head 17eed585a9):

  • Merged origin/ci-fixes into the PR branch. ci-fixes had advanced with c6369fb6a1 (pin BILI_STORAGE_UNIFIED="sqlite" in tests/notice-registry.test.ts) — the sentinel's converged fix for the sibling ubuntu visibleNotices born-red from the same [PR] storage: pure-SQLite default — JSON envelopes stop being written unless dual-write is requested (#6479) #6480 storage-default flip.
  • The merge conflicted with this branch's own earlier per-PR fix 331eb8071c (same file, opposite mechanism: delete the pin vs pin "sqlite"). Resolved by taking the ci-fixes version — per the CI-coordination policy the master-level fix converges to ci-fixes, so the redundant per-PR fix is dropped, not kept.
  • The push triggers a full re-run. The ubuntu facet is now on the sentinel's converged fix; the Windows EBUSY facet is a transient flake the re-run may clear.

Honest boundary: the ci-fixes merge does not contain a fix for the Windows EBUSY file-lock flake — that facet is still pending on the sentinel (not yet landed on ci-fixes). If the re-run stays red on the same 13 EBUSY subtests, it is the same inherited flake, not a regression from this PR; the durable fix (retry-ladder budget / storage-mode pinning for the unified-mode suites) belongs on ci-fixes, not this branch.

八问(结构化结论)

  1. 是什么问题:PR 的两个 windows 测试分片(22/1、22/3)红,13 个失败子测试全部是同一签名——teardown 时 EBUSY unlink index.db(Windows 外部 AV/索引器锁住 SQLite 文件,超出 tests/tmp-rm.ts 的 ~9.6s 重试窗口)。
  2. 什么原因导致的:根因是 [PR] storage: pure-SQLite default — JSON envelopes stop being written unless dual-write is requested (#6479) #6480 把存储默认翻成 pure-SQLite 后,更多测试跑在 unified 模式、创建 index.db,放大了既有 windows-22-1 lane randomly red: during test cleanup, unlink index.db hit an EBUSY external lock, couldn't survive tmp-rm | windows-22-1 车道随机红:测试清理期 unlink index.db 遇 EBUSY 外部锁,熬不过 tmp-rm 的 9.6 秒重试梯子(#2799/#2908 族复发) #4347/Windows CI: storage-engine-proofs 套件清理 temp 目录遇 EBUSY/EPERM,windows test leg 随机红 #2799/test(windows): driveWire golden-harness 清理期 index.db EBUSY 复发(#2799 族,#2808 后残留) #2908 家族 Windows 文件锁 flake 的暴露面;锁是外部的(AV/索引器),非进程内泄漏(tmp-rm.ts 注释已探针确认 rmrf 时无自持句柄)。本 PR 的改动(budget.ts/handle.ts)完全不碰这些路径。
  3. 影响范围与严重程度:继承性 master 红——master e1f31ad9d5 自身 run 43469650189 同样失败([PR] fix: normalize inline requests with Fetch semantics #6563/[PR] fix: reject proxy CONNECT handshakes on premature close #6577 已判定),非本 PR 代码问题;属 CI flake,不影响已发布产物,但阻塞本 PR 合并。
  4. 怎么修复的:按 CI 协调政策把 origin/ci-fixes merge 进 PR 分支(新 head 17eed585a9),收敛哨兵对同族 ubuntu visibleNotices born-red 的修复(c6369fb6a1),并丢弃本分支重复的 per-PR 修复 331eb8071c(冲突取 ci-fixes 版);push 触发全量重跑,transient flake 有概率自行转绿。
  5. 是否有回归问题:无。merge 仅带入 ci-fixes 的一个测试文件改动(notice-registry.test.ts 钉 BILI_STORAGE_UNIFIED="sqlite"),与哨兵 CI 验证版逐字节一致;本 PR 的 max tokens 4096 dsh 4096 — from the "emergency output budget clamp" of the billion-context proxy, not from the DSH clien | max token 4096 dsh 4096 来自 billion-context 代理的「紧急输出预算钳制」,不是 DSH 客户端,也不是模型 #5596 修复面(budget/handle)未被触碰。
  6. 是否破坏缓存:不涉及(纯测试/CI 协调,无缓存/重试/游标逻辑变更)。
  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:本 PR 的红是继承性 flake,已按政策 merge ci-fixes 并 push(17eed585a9);Windows EBUSY 面仍待哨兵在 ci-fixes 收敛,重跑若仍红属同一继承 flake 而非本 PR 回归。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

结论 / Outcome (conclusion of this run)

判定:PR #6551 两个红的 windows 分片(22/1、22/3)是继承性 master 红,不是本 PR 代码问题——13 个失败子测试全部是同一签名 EBUSY unlink index.db(teardown 时 Windows 外部 AV/索引器锁住 SQLite 文件,超出 tests/tmp-rm.ts 的 ~9.6s 重试窗口),属 #4347/#2799/#2908 家族 flake;master e1f31ad9d5 自身 run 43469650189 同签名失败(#4599 的 #6563/#6577 已判定)。本 PR 的改动(src/server/budget.ts、src/server/handle.ts、tests/issue5596-ledger-floor-calibration.test.ts)完全不碰这些路径。

本轮已做:按 CI 协调政策把 origin/ci-fixes merge 进 PR 分支(新 head 17eed585a9)——收敛哨兵对同族 ubuntu visibleNotices born-red 的修复(c6369fb6a1),并丢弃本分支重复的 per-PR 修复 331eb8071c(冲突取 ci-fixes 版,逐字节一致)。已 push,触发全量重跑。

已验证:live check 状态(镜像缓存 stale ~10h,改用一次性 API 读确认两 job 均红、非绿);两 job 日志里 13 个失败的 EBUSY 签名;ci-fixes 祖先关系;冲突解决与 ci-fixes 版逐字节一致。

剩余 / 风险:Windows EBUSY 面尚未在 ci-fixes 收敛(哨兵待补)——重跑有概率把 transient flake 转绿;若仍红在同一 13 个 EBUSY 子测试上,属同一继承 flake 而非本 PR 回归,根治(重试窗口预算 / unified 模式套件的存储模式钉定)应落在 ci-fixes 而非本分支。详细八问见上一条评论。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

重跑仍红(继承性 Windows EBUSY flake,非本 PR 代码)

重跑 43512360188(head 17eed58,上轮 merge ci-fixes 后触发)在同样的两个 windows 分片上仍红:

新证据(诊断升级):CI runner 内置的 flake 确认逻辑触发了——同 runner 重试一次仍复现,且本次重跑在新 runner 上也复现。按 CI 自己的判定逻辑(#4347),这从「transient AV/indexer 锁」转向「确定性 handle 泄漏」假设为主(测试/存储代码持有 index.db 句柄未释放,unlink 时文件被锁)。

处置:本 PR 的 diff(src/server/budget.ts、src/server/handle.ts、tests/issue5596-ledger-floor-calibration.test.ts)不触碰任何 persist-zstd/storage/tmp-rm 路径,非本 PR 回归。已按 CI 协调政策 merge ci-fixes(上轮,sha 17eed58);Windows EBUSY 面尚未在 ci-fixes 落地(哨兵待补),已在 #4599 报告新证据(handle 泄漏假设),请哨兵优先排查。本分支不再自行修 master 级 flake。

八问(继承性 flake,非本 PR 代码):

  1. 是什么问题:两个 windows 测试分片 teardown 时 EBUSY unlink index.db 失败——非本 PR 代码,是 master 级继承 flake(windows-22-1 lane randomly red: during test cleanup, unlink index.db hit an EBUSY external lock, couldn't survive tmp-rm | windows-22-1 车道随机红:测试清理期 unlink index.db 遇 EBUSY 外部锁,熬不过 tmp-rm 的 9.6 秒重试梯子(#2799/#2908 族复发) #4347 家族)。
  2. 什么原因:重跑在新 runner + 同 runner 重试均复现,CI 判定从「transient AV 锁」转向「确定性 handle 泄漏」假设为主(测试/存储代码持有 index.db 句柄未释放,unlink 时文件被锁)。根因待哨兵在 ci-fixes 收敛。
  3. 影响范围与严重程度:所有在 windows-latest 上跑 persist-zstd / metadata-fold 套件的 PR 都继承此红(非本 PR 独有,[PR] fix: make codex-responses-ws pre-first-event failures diagnosable; add configurable WS idle watchdog (#6171) #6301/[PR] fix: normalize inline requests with Fetch semantics #6563/[PR] fix: preserve Request fields during inline fallback #6566 同签名);本 PR 无法合并直到 ci-fixes 收敛。
  4. 怎么修复的:本轮未修(非本 PR 代码,禁止在 PR 分支修 master 级 flake);上轮已 merge ci-fixes(收敛 ubuntu visibleNotices 面),Windows EBUSY 面待哨兵在 ci-fixes 落地。
  5. 是否有回归问题:无——本 PR diff 不触碰 persist-zstd/storage/tmp-rm 任何路径。
  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:不涉及。

判定:本轮无代码改动(非本 PR 代码,继承性 flake);新证据已报 #4599 请哨兵排查 handle 泄漏;等 ci-fixes 收敛后本 PR 重跑即绿。

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 dsh p2 scope/wiring

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant