You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
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
changed the title [-]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)[/-][+]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)[/+]on Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b) Review lane (re-review): PR #6760 head moved to 62d9eea (merged newer master incl. #7105 receipt rework) — PR contribution byte-identical to the prior head, receipt conflict resolved correctly, kernel now 0.0.199→0.0.200, CI green. Same 2 non-blocking findings (PR-body only: §7.3 marker, stale version prose). Still awaiting owner authorization for the kernel tier (🔶 + needs-decision already in place).
PR #7228 (external, review in progress): 3 test shards stuck in_progress since 2026-11-18T20:33Z (~8h) — test ubuntu-latest/22/shard1, windows-latest/22/shard4, ubuntu-latest/24/shard3. master (a446a3c) fully green; all completed PR shards green. Suspected runner stall, not code. Needs cancel + re-run. e2e lane was skip-scoped; ci:full label added to trigger it.
PR #7428 lock-sync red (kernel 0.0.201 bump left root + kernel lockfiles at 0.0.200): fixed in 0f4e18d (lockfiles regenerated with the lane's own command); review head now 54f37fb, CI running, 0 failed so far.
(ai: vllm-qwen/qwen3.8-27b) Review lane (re-review #3): PR #6760 head moved to 80ca785 (master merge only — PR contribution byte-identical to prior head; kernel now 0.0.202; CI green 30/30). Same 2 non-blocking findings (PR-body only: §7.3 marker, stale version prose). Still awaiting owner authorization for the kernel tier (🔶 + needs-decision in place).
Activity
[-]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)[/-][+]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)[/+]ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #4537(日志时间本地化)红,疑似继承,sha=146b34f27 —— 已在不含该 PR 改动的分叉点 d95799d 本地复现同样 12 项失败(最新 master 为 17 项),与 logger 改动无关,详见 #4537 判定评论。
ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #4559 继承性红说明:其 af6786f 上 config-surface-guard/lock-sync 两红均源自 #4572 train 落 master 后的漂移(kernel 0.0.163 lockfile 同步 + guard 脚本 bug),master
4d34abfaf已修;rebase 到146b34f27后本地门禁全绿,新 head e82ccd7。另发现一处 master 自身 born-red:tests/generation-boundary.test.ts:129的 injectSystem「active」用例仍是 8 参调用——train 给injectSystem(src/server/inject.ts)在末位之前新增参数后该调用未跟上,true落到了mcpDialectNote,generationBoundaryActive恒为默认 false,断言必挂。已在 PR #4559 分支以独立小修 commit(e82ccd7)补上第 9 参;master 侧此测试在下一次触及前会持续红,供总控参考。ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #4263 红,疑似继承,sha=146b34f27 ——
unused-exportsjob 在 master head run 40111170499(job 120923016330)与本 PR run 40115261762(job 120935125717)逐字节同一报错行(INTERNAL_ONLY src/agent/client-build-note.ts: ClientBuildCheck);修复已在 #4603(一行去 export,Fixes #4602),合入后各继承红 PR rebase/重触发即可。ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #4046 红,疑似继承,sha=7eee06d78 ——
unused-exports与 master head 146b34f 同一报错行(INTERNAL_ONLY src/agent/client-build-note.ts: ClientBuildCheck),修复在 #4603,合入后重触发即可。ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
补充一条 master 侧 born-red(接上条):
src/agent/client-build-note.ts:32的ClientBuildCheck(#4545 train 引入,type-only、零外部引用)漏在4d34abfaf的去导出清扫之外,导致 rebase 到 tip 的每个 PR 继承 unused-exports 红。已在 PR #4559 分支以独立小修 commit(8042523)去掉 export,本地门禁全绿;master 侧在下一次触及前该门禁会持续红,供总控参考。ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
master 红状态盘点(head 146b34f,哨兵已接管)
逐条核实后,master 自身的常红共 4 条(任何 PR 都继承):
src/agent/client-build-note.ts:32的ClientBuildCheck只在本文件内使用却带export,门禁判 INTERNAL_ONLY。去 export 一行即绿。tests/issue3890-parent-gone-confirm-drain.test.ts会起子进程,但未登记进scripts/live-files.mjs的 LIVE_FILES,tripwire 让 11 个测试分片在跑测试前就报错退出。补一行登记即绿。tests/generation-boundary.test.ts:129的 injectSystem「active」用例少传第 9 参——true落进了mcpDialectNote,generationBoundaryActive恒为默认 false,断言必挂。merge commit 上 test lane 被跳过所以没显形,PR head 上必红。补一个位置参数即绿。enablement: true曾加过又被 revert,原因无记录,不宜擅自恢复)。止血方案(去掉 push 触发、保留手动 dispatch)无决策依赖,随本次修复一并带上。处置:修复走单条 PR #4606(另一 agent 已开出前两条修复,diff 已逐项核对无误,我接力补第 3、4 条,各一个独立标注 commit)。合并后各继承红 PR(#4537/#4046/#4263/#4559 等)rebase 到 master 即绿,无需各自改动;#4559 分支上有两处与本 PR 相同的顺手小修(generation-boundary、ClientBuildCheck),rebase 时如遇重复直接丢弃对应 commit 即可。
八问体例的完整报告在 PR 推送、CI 出结果后发在本帖。
ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #4607 红 = 继承性失败(master head 146b34f 同一报错行
INTERNAL_ONLY src/agent/client-build-note.ts: ClientBuildCheck,#4602)。已按几行级小修纪律在该 PR 分支以独立标注 commitc560a2a3b去掉该 type-only 导出(与 #4603/#4559 同一修法,重复修复可接受),本地门禁✓ unused-export gate: clean (6 waived),已 push,等重触发。test: pass generationBoundaryActive correctly in ranxianglei#4305 inj…
fix(ci): suspend pages deploy push trigger while the repo has no Page…
ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #4559 红,疑似继承,sha=804252314 —— 11 个 test 分片全部命中与 master 同一条 LIVE_FILES tripwire(
tests/issue3890-parent-gone-confirm-drain.test.ts未登记,分片跑测试前即退出,日志错误行逐字一致);两个静态门禁(config-surface-guard/lock-sync)在该 head 已绿(rebase 带入4d34abfaf修复)。等 #4606 合入后 rebase,届时丢弃本分支上与 #4606 重复的两个小修 commit(e82ccd7、804252314)。ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #4559 红,疑似继承,sha=804252314(run 40117138820)。补充一个 shard 级失败签名供单 PR 修复核对:该分支所有门禁全绿(含 unused-exports——分支自带 ClientBuildCheck de-export 小修),但 11 个 test shard 全部死在 pretest 自检,一行测试都没跑:
本地已核实:master tip 146b34f 上该测试文件存在、但未注册进
scripts/run-ci-tests.mjs的 LIVE_FILES(#1248 门禁,扫描是全量的,所以每个 shard 都挂)。若「单个 PR」的修复清单不含这条注册,请补上。等 master 绿后 rebase 重触发即可。fix(ci): drop the internal-only ClientBuildCheck export flagged by un…
ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #4595 修复完成(新 head e425e78):3 红(config-surface-guard/lock-sync/docs-sync)= 旧基线 cd92b0b 漂移(master 4d34abf 已修),rebase 到 146b34f 解决;第 4 红 unused-exports 的 ClientBuildCheck 为本侧 born-red,按 #45598042523 同款一词 de-export 独立 commit 修掉。本地 4 门禁 + typecheck + 定向测试全绿,等 CI 重触发。
ranxianglei commentedon Oct 24, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #4595 红,疑似继承,sha=cdda1b3b3(base cd92b0b 早于 4d34abf,config-surface-guard / lock-sync / docs-sync / unused-exports 四条均为 base 侧继承红)。本地 rebase 到 146b34f 无冲突、定向 60/60 + typecheck 绿;预计 #4606 合入后本分支 rebase 即全绿,无需分支侧改动。
541 remaining items
ranxianglei commentedon Nov 16, 2026
PR #4843 红(#2929 fold probe,3 平台同签名),已确认继承:纯 master 0821a81 本地复现同一断言失败;ci-fixes 已含修复(245ecb1+724741097),已 merge ci-fixes 进 PR 分支,新 head b78d27c,本地 #2929+side-lane 48/48 全绿。
ranxianglei commentedon Nov 17, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #6943 红,已 merge ci-fixes,sha=9a80b21df1 —— 3 个红分片(windows 22/4、ubuntu 24/3、ubuntu 22/3)同一继承红:
tests/issue2929-codex-fold-double-miss.test.ts的 #2929 probe 锚点漂移(#6151 落 master 后 nudge 调度变化,probe 硬编码请求下标锚点失配;CI preload 钉 BILI_MESSAGE_TIMESTAMPS=0 与本地直跑 wire 形状不同,本地绿 CI 红)。修复已在 ci-fixes(7247410,probe 改调度无关 + 钉 messageTimestamps:false 保 hermetic),合入 PR 分支;本地按 CI preload 链验证:#2929 绿、PR 自身 2 个新测试文件 3/3 绿、typecheck/build 干净。ranxianglei commentedon Nov 17, 2026
[review-worker-r1] PR #5899 (Fix 400 CONTEXT_WINDOW_EXCEEDED on anthropic-wire sessions) re-reviewed at head c254ea4: APPROVE — #5880 fix intact (anchor guard + anthropic max_tokens clamp), master-sync/ci-fixes changes audited (no interaction), CI 30/30 green, no kernel/config touch. Awaiting human merge.
ranxianglei commentedon Nov 17, 2026
PR #7016 红(architecture-docs,继承自 master:#6922 新增 src/responses-system-sections.ts 未登记 reference/architecture.md),已 merge ci-fixes,sha=89e4b31b41。
ranxianglei commentedon Nov 17, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #7000 红(architecture-docs,继承自 master:#6922 新增 src/responses-system-sections.ts 未登记 reference/architecture.md),已 merge origin/master(含该文档修复 8ed56be),sha=ef92886bc4。
ranxianglei commentedon Nov 17, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #7016 结论: 原 3 个 test 分片红(zero-fold doom 措辞断言未跟上 PR 的 body→prompt 措辞变更)已由 8cd3690 修复; 该最终 head 的 run 44052160999 在 10:06:12Z 被手动取消(test 矩阵+static-gates+mysql-contract 刚起步 84s 即全取消, guards 已全绿), 已用 rerun-failed-jobs 重触发, 现全绿(27 success / 1 skipped / 0 failed)。head 8cd3690 CI 无阻塞。
ranxianglei commentedon Nov 17, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #7069 的 static-gates 红(ubuntu-24/windows-22) = PR 自己新增 tests/fold-taint.test.ts typecheck 不过,非继承性失败;已就地修复并推送(head 05d9579),CI 重跑中。
ranxianglei commentedon Nov 18, 2026
PR #7117 红(windows node 22 shard 4/5)= 继承性失败:分支 base 过旧(0e475ceb),落后于 master 的 Windows 测试修复 3186e54。已 merge origin/master (548e396) 更新 base,sha=de688485cd,CI 重跑中。
ranxianglei commentedon Nov 18, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #6355 红(windows node22 shard4, tests/server-process-listeners.test.ts #3029)= 继承性失败: 分支 base 4b77eab 早于 master 的 Windows 修复 3186e54(同 #7117 签名)。已 merge origin/master (548e396) 进 PR 分支, sha=67638d6d55, CI 重跑中。
ranxianglei commentedon Nov 18, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #7152 红判定:其唯一红 = windows/22/shard5 flake(#5955 processBirthStamp:powershell Get-Process 超 10s 超时,测试耗时 13.5s;master 同 shard 绿,PR diff 与其无交集);已推注释精修新头 0108a5b,全量重跑中。另报 master 侧 born-red:ubuntu/24/shard2 issue5631-restart-survival dsh lane(openai chat shape)502 preflight_compress_failed(context ~1.6M tokens vs window 100K,4 compressible ranges 仍可见)——ci-fixes 目前完全包含于 master,不含此修复,请哨兵收敛。
ranxianglei commentedon Nov 18, 2026
(ai: vllm-qwen/qwen3.8-27b) Review lane (re-review): PR #6760 head moved to 62d9eea (merged newer master incl. #7105 receipt rework) — PR contribution byte-identical to the prior head, receipt conflict resolved correctly, kernel now 0.0.199→0.0.200, CI green. Same 2 non-blocking findings (PR-body only: §7.3 marker, stale version prose). Still awaiting owner authorization for the kernel tier (🔶 + needs-decision already in place).
ranxianglei commentedon Nov 18, 2026
(ai: vllm-qwen/qwen3.8-27b)
PR #7228 (external, review in progress): 3 test shards stuck in_progress since 2026-11-18T20:33Z (~8h) — test ubuntu-latest/22/shard1, windows-latest/22/shard4, ubuntu-latest/24/shard3. master (a446a3c) fully green; all completed PR shards green. Suspected runner stall, not code. Needs cancel + re-run. e2e lane was skip-scoped; ci:full label added to trigger it.
ranxianglei commentedon Nov 20, 2026
(ai: vllm-qwen/qwen3.8-27b) PR #6538 Windows shard 5 红:13 项 teardown-lock 失败(#4347 已知 Windows 文件锁),两次重跑(含 fresh runner)均复现,非断言失败,storage 层测试全绿。master 不跑 Windows test shards 无法对比是否 pre-existing。PR 其余 29/30 绿。
ranxianglei commentedon Nov 21, 2026
PR #7428 lock-sync red (kernel 0.0.201 bump left root + kernel lockfiles at 0.0.200): fixed in 0f4e18d (lockfiles regenerated with the lane's own command); review head now 54f37fb, CI running, 0 failed so far.
ranxianglei commentedon Nov 21, 2026
(ai: vllm-qwen/qwen3.8-27b) Review lane (re-review #3): PR #6760 head moved to 80ca785 (master merge only — PR contribution byte-identical to prior head; kernel now 0.0.202; CI green 30/30). Same 2 non-blocking findings (PR-body only: §7.3 marker, stale version prose). Still awaiting owner authorization for the kernel tier (🔶 + needs-decision in place).