Skip to content

[PR] kernel: preserve existing block coverage when merging blocks (#4484) - #4485

Merged
ranxianglei merged 4 commits into
masterfrom
2026-10-24_fix-merge-coverage-regression
Oct 28, 2026
Merged

ranxianglei merged 4 commits into
masterfrom
2026-10-24_fix-merge-coverage-regression

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

kernel-change: merging existing blocks retracted their own active coverage. The turn-integrity gate judged the fold over this unit's ids only and withdrew ids carried solely by the consumed blocks; those blocks deactivate on apply and the new block does not inherit them, so the folded messages lose all coverage and the host pre-commit guard (src/stream.ts, #3353/#3322) refuses every such merge (gate:coverage-regression), producing repeated rejections of legitimate block merges. The regressive candidate is PRODUCED inside the kernel's projectUnit judgment, so the host side cannot prevent it — the guard can only refuse the symptom. Owner directed the root-cause fix in #3322 ("直接修复根本原因"); this implements it for #4484.

What changed

projectUnit's integrity step (kernel/src/unit-projection.ts), scoped to merges (consumedBlockIds.length > 0); fresh folds keep the legacy path verbatim:

  1. Judge post-fold coverage = this unit's ids ∪ every still-active sibling block's coverage, so a tool half already covered by an unconsumed sibling reads as folded rather than as a spurious visible half (the [bug] Collapsing the "block range" compress actually doubled the live view size (253→519 msgs) → local estimate blew past the window → preflight permanently fail-fast (while the same payload would have passed upstream at 566K/750K) | [bug] 折叠「块范围」的 compr… #3322 facet).
  2. Unfold only ids whose coverage does not depend on a consumed block, re-checking integrity to a fixed point. If consistency would force unfolding such an id, return an honest split fatal instead of emitting a doomed regressive candidate.

Behavior change (flagged)

  • OLD: a block-boundary merge whose range cut through a reasoning/tool group could drop the existing coverage of the consumed blocks → the host refused the fold (state untouched) → the client/model retried and was refused again (repeated coverage-regression rejections).
  • NEW: such merges either (a) succeed while preserving all pre-existing active coverage (the common case — interior hard-protected tool, tail cut mid-tool-group), or (b) fail up-front with a clear "align the range to whole turns / merge fewer blocks" error when no consistent non-regressive candidate exists. Legitimate merges stop being rejected; impossible ones fail fast at the decision point rather than being refused downstream.
  • Host coverage guard ([PR] fix: refuse coverage-regressing folds before commit (#3322) #3353) is unchanged — kept as defense-in-depth.

Tests

  • New: kernel/tests/issue4484-merge-coverage.test.ts — protected-rule interior, parallel-tail cut, clean-merge control; asserts no coverage loss AND final-state integrity validity.
  • Targeted green locally (rebased onto current master): 147 kernel tests (turn/pair integrity, projection, recommend, apply-lane compress/refold/promote) + host guard tests/issue3322-coverage-regression-guard.test.ts. Full-suite verdict via CI.

Bumps acp-kernel 0.0.161 → 0.0.162.

Closes #4484. Related: #3322.

@github-actions

github-actions Bot commented Oct 24, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-10-24_fix-merge-coverage-regression (6178e5a)

Latest build — install this exact version

npm install -g billion-context@0.0.6530-pr.4485.703a216

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
6178e5a billion-context@0.0.6530-pr.4485.703a216 build
31a18db billion-context@0.0.6518-pr.4485.657fc34 build
c4d1b22 billion-context@0.0.5405-pr.4485.d59dbad build
558e14c billion-context@0.0.5401-pr.4485.28c0991 build
3c9749a billion-context@0.0.5400-pr.4485.2413914 build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr4485.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei ranxianglei changed the title kernel: preserve existing block coverage when merging blocks (#4484) [PR] kernel: preserve existing block coverage when merging blocks (#4484) Oct 24, 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 即 #4484 的根因修复(承接 #3322 中 owner「直接修复根本原因」的指示)。相关但机制不同:#3353(host 侧 coverage-regression guard,本 PR 不动它、保留为纵深防御)、#4050/#4085(id churn 下的 fold 覆盖恢复,另一条产线)。

标签: bug + scope/kernel + p2(内核侧根因修复;跨车道不打客户端标签;被拒时旧块仍保持覆盖、无数据丢失 → 功能性缺陷而非可用性缺陷)。

开始审核:已拉取分支(1 commit,3 文件:kernel/src/unit-projection.ts +121/-28、新测试 kernel/tests/issue4484-merge-coverage.test.ts、kernel 版本 0.0.161→0.0.162)。CI 现状:lock-sync 红(kernel 版本 bump 后 package-lock.json 未同步),其余 gate 绿或按文件清单跳过。先在 PR 分支修掉 lock-sync(几行级小修,直接补 commit),随后跑定向测试并给出结构化八问结论。

@ranxianglei ranxianglei added bug Something isn't working p2 scope/kernel labels Oct 24, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

lock-sync 已修复并推送(commit 558e14ccd → 分支 2026-10-24_fix-merge-coverage-regression)。

结构化八问:

  1. 是什么问题:PR 的 CI lock-sync job(package-lock.json version metadata lags package.json (bili 0.1.190 vs 0.1.191, kernel 0.0.112 vs 0.0.114) #2704 锁文件漂移守卫)红了——CI 日志显示根 package-lock.json 里 file:./kernel 链接条目记录的版本与 kernel/package.json 不一致(0.0.161 vs 0.0.162)。
  2. 什么原因导致的:PR 的 kernel 版本 bump(0.0.161 → 0.0.162)只改了 kernel/package.json,没有同步重新生成两处锁文件——根 package-lock.json(file: 链接条目)+ kernel/package-lock.json(自身 top-level 版本)。bump 不触发锁文件再生成,只有 npm install 才会,跳过再生成就留下漂移。
  3. 影响范围与严重程度:仅该 PR 的 CI 门禁,不影响任何用户、运行时行为或产物内容;纯机械性漏步。
  4. 怎么修复的:在 PR 分支上用 lock-sync 完全相同的配方(npm install --package-lock-only --ignore-scripts --no-fund --no-audit,根目录 + kernel/ 各一次)重新生成两个锁文件,独立 chore: commit 推送。diff 只有 3 行版本号(0.0.161 → 0.0.162),无其他 churn;本地复跑命令幂等(二次运行零 diff)。
  5. 是否有回归问题:无——纯锁文件元数据同步,未触碰任何代码、行为、阈值或输出格式;高危模式逐项过:语义漂移 / fail-fast 变吞错 / 默认值·时序·格式变化均不涉及。
  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:CI 红因已消除,等新一轮 CI 跑完 lock-sync 转绿即可;合并侧无需其他动作。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

Review 完成:代码方向正确、回归测试有牙、本地定向全绿;CI 仅 5 个 shard 红,已证明是既有 flake(与 master 同签名红过),rerun 已触发并立项 #4489。判定:可合(kernel 变更属 human-only review tier,最终合并由你执行)。

审核证据(判断层,非流水账)

结构化八问

  1. 是什么问题:合并已有块时,合法合并被 host 覆盖守卫反复拒绝(gate:coverage-regression)——合并后已折叠消息丢失全部覆盖。
  2. 什么原因导致的:projectUnit 的 turn-integrity 只按本单元 ids 判折叠资格,把 consumed 块独占的覆盖当作 apply 后仍在(实际那些块 deactivate、新块不继承它们);且兄弟块已覆盖的半截被误判为可见半截,触发多余回撤。因果链:判定不看「合并后覆盖」→ 产出 regressive candidate → host 只能拒症状。
  3. 影响范围与严重程度:任何 block-boundary 合并、range 切穿 reasoning/tool 组的会话(长 thinking 会话常见)——合并功能不可用,但会话本身可用、fresh fold 不受影响;p2 级功能缺陷,有绕行(range 对齐整轮)。
  4. 怎么修复的:按「合并后覆盖」= 本单元 ∪ 存活兄弟块覆盖来判;定点迭代只回撤不依赖 consumed 块的 id;一致性若强制回撤 consumed-owned id,直接返回诚实的 split fatal(决策点 fail-fast,错误信息给出对齐整轮/少合并块的出路),不再产出注定被拒的 candidate。
  5. 是否有回归问题:审过的既有行为均保持——fresh-fold 路径逐字节未变;[Bug] recommend/apply 对同一区间的 turn-integrity 判定永久不一致:acp_status 反复推荐的区间被 apply 侧整段拒绝(0.1.196,Range would split N turn(s) at the protected-zone boundary) #3236 判定一致性保持;host guard 语义未变;无阈值/时序/输出格式漂移。一处保守行为变化已注明:兄弟块覆盖的 reasoning 与可见 burst 之间的预存分裂(birth-time gate 通常已防住)现在会显式拒绝合并而非产出 regressive candidate——诚实失败 + 可操作错误,判定可接受。PR 正文的 Behavior-change 披露(OLD→NEW)完整、准确。
  6. 是否破坏缓存:不涉及——前缀渲染/摘要锚点位置/缓存失效策略均未触碰;被拒候选发生在决策点,state 不动。
  7. 是否改了设计方向:不涉及——留在 [Bug] recommend/apply 对同一区间的 turn-integrity 判定永久不一致:acp_status 反复推荐的区间被 apply 侧整段拒绝(0.1.196,Range would split N turn(s) at the protected-zone boundary) #3236 shared-unit-projection 教义内,无新配置面、无新模块边界。
  8. 是否引入新配置:无。

CI 状态(c4d1b22,run 40069753327)

32 success / 14 skipped / 5 failure——5 个失败 shard 全部死在同一处:既有 flake tests/issue3250-wal-checkpoint.test.ts:291(WAL custom-cap 断言,left 1095952 bytes, cap 1048576,超出容差 ~11.6 页)。证据链:今天早些 master run 40068050438(sha 02f544d)同样 5 个 shard、逐字节相同的错误红过一次,随后同 sha 的 master runs 全部转绿 → 环境依赖型 flake,与本 PR 的 kernel 变更无关(存储/WAL 定时测试)。已 rerun-failed-jobs;该 flake 已立项 #4489(bug/p3/scope/ci,来源标记指向本 PR 分析)。

收工判定:代码与测试层面可合;等 rerun 转绿后 CI 即全绿。kernel 变更按 Kernel Boundary 属人工合并范畴——rerun 绿后请你自行合并:#4485

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

CI 更新(修正上一条的「rerun 已触发」为完整结果): rerun(attempt 2)同样 5 个 shard、同一测试、逐字节相同错误(left 1095952 bytes)再次失败。追加取证后结论不变且更硬:

判定维持:代码与测试层面可合;CI 的红全部归因于既有 flake(#4489 跟踪)。你随时可以合并(红的归属已有完整证据链);若想要一个全绿的 check 记录,等第三次 attempt。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

CI 取证收尾(修正上一条「单一测试」的框定): 对全部 5 个红 shard 的日志逐槽位核对后发现,红的不是 1 个测试而是 2 个(都在 unified-storage 回收断言、node:sqlite/builtin 引擎):

shard 槽位 失败测试 master run(无本 PR 变更) PR head 三次 attempt
ubuntu 24/1、22/1 issue3250-wal-checkpoint.test.ts:291 WAL custom-cap(left 1095952 bytes) ✅ 同签名红 ×3 同签名
ubuntu 24/3、22/3 media-cas.test.ts stale-row sweep(swept message released its media ref, 1!==0) ✅ 同签名红 ×3 同签名
windows 22/2 两者 ✅ 同签名红 ×3 同签名

测试序号跨树一致(1275/1243/635/835)⇒ 每 shard 文件组成相同(host CI 只分片顶层 tests/*.test.ts,本 PR 新测试在 kernel/tests/ 不入清单)+ 错误字节逐字相同 ⇒ 与本 PR diff 无关是证明而非推测;同 sha 的后续 master runs 全绿 ⇒ runner/池依赖的既有 flake 家族。

处置:

最终判定:代码与测试层面可合(kernel 边界合规、回归钉真实、本地定向全绿);CI 的红全部归因于既有 flake 家族(#4489 跟踪,证据链完整)。按 Kernel Boundary,kernel 变更属人工合并范畴——你可以直接合并(https://github.com/ranxianglei/billion-context/pull/4485),合并时的 CI 即下一次自然抽签。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

自动修复通道(attempt 2/2)判定:5 个红 test shard 不是本 PR 分支的缺陷——是继承自红 master。不在本分支修,到此为止。

证据链:

  • head c4d1b22 的 5 个红 shard(run 40069753327)全部挂在同两个 host 侧存储测试上,kernel 相关测试无一涉及:
    1. media CAS: stale-row sweep releases media refs and mappings — tests/media-cas.test.ts:134,断言 swept message released its media ref expected 0 actual 1;
    2. #3250: a configured cap flows open() → journal_size_limit — tests/issue3250-wal-checkpoint.test.ts:291,WAL must be truncated to the configured cap at the autocheckpoint (left 1095952 bytes, cap 1048576)。
  • master 自己的定时全量矩阵(run 40073163544,head 02f544d = 本 PR 的 base commit)出现完全相同的两个签名,各自在 6 shard 中 4 个上出现,覆盖 ubuntu/windows × node22/24——确定性失败,不是 flake。该红灯已被自动跟踪在 Scheduled full test failed: 02f544da | 定时全量测试失败: 02f544da #4493。
  • 本 PR 只改 kernel/src/unit-projection.ts、其回归测试与 kernel 版本 bump,与上述两个 host 侧存储测试无路径交集。
  • 本 PR 其余门禁全绿:lock-sync(第 1 轮已修漂移,commit 558e14c)、kernel-guard、static-gates(ubuntu+windows)、mysql-contract、docs-sync、version-guard 等;windows 22/1、22/3–5 与 ubuntu 22/2、24/2 shard 也绿。

结构化八问:

  1. 是什么问题:PR head c4d1b22 的 5 个 test shard 全红,但成因是红 master,不是本 PR 的改动。
  2. 什么原因导致的:master 自身(02f544d,即本 PR base)的定时全量矩阵就确定性挂上述两个存储测试(Scheduled full test failed: 02f544da | 定时全量测试失败: 02f544da #4493 跟踪中);基于该 master 的任何 PR 都会继承这些红腿。
  3. 影响范围与严重程度:波及所有基于当前 master 的 PR(CI 无法全绿,直到 Scheduled full test failed: 02f544da | 定时全量测试失败: 02f544da #4493 解决);对用户运行时行为无直接影响(那两个测试是否对应生产回归待 Scheduled full test failed: 02f544da | 定时全量测试失败: 02f544da #4493 根因确认)。
  4. 怎么修复的:未修——按自动修复通道规则,失败不属于本分支的小修时只发判定不硬修;把无关的存储修复捆进这个 kernel PR 违反一 issue 一 scope。第 1 轮的 lock-sync 漂移已在本分支修复(558e14c,现绿)。
  5. 是否有回归问题:本 PR 未引入;除继承自 master 的 test 腿外,所有门禁绿。
  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。

结论:本 PR 代码面自身各门禁通过,待 #4493(红 master)修复转绿后重跑 checks 即可最终确认;review 本身可并行推进。

ranxianglei pushed a commit that referenced this pull request Oct 24, 2026
…fixed ingest count

The #3250 custom-cap case asserted the WAL sat under cap + one page after a
fixed number of ingests. The residue above the cap depends on where the
1000-page autocheckpoint window closes relative to commit boundaries, which
varies with the runner's bundled SQLite build — byte-identical red on some
shards (left 1095952 bytes, cap 1048576), green on others, same sha
(master run 40068050438; PR #4485 run 40069753327).

Assert convergence instead: keep committing 128 KiB units until a completed
checkpoint (autocheckpoint or cadence-tick PASSIVE) brings the file back to
the configured cap, bounded at CONV_UNITS = 24 (one full cadence period +
landing lag + headroom; derivation in the constant's doc). A pre-fix tree
(limit never honored) sits at its high-water mark through all units and can
never converge, so the RED/GREEN discrimination is unchanged.
ranxianglei pushed a commit that referenced this pull request Oct 24, 2026
…fixed ingest count

The #3250 custom-cap case asserted the WAL sat under cap + one page after a
fixed number of ingests. The residue above the cap depends on where the
1000-page autocheckpoint window closes relative to commit boundaries, which
varies with the runner's bundled SQLite build — byte-identical red on some
shards (left 1095952 bytes, cap 1048576), green on others, same sha
(master run 40068050438; PR #4485 run 40069753327).

Assert convergence instead: keep committing 128 KiB units until a completed
checkpoint (autocheckpoint or cadence-tick PASSIVE) brings the file back to
the configured cap, bounded at CONV_UNITS = 24 (one full cadence period +
landing lag + headroom; derivation in the constant's doc). A pre-fix tree
(limit never honored) sits at its high-water mark through all units and can
never converge, so the RED/GREEN discrimination is unchanged.
@ranxianglei

Copy link
Copy Markdown
Owner Author

解决冲突ci

ci-check added 3 commits October 27, 2026 23:50
kernel-change: merging existing blocks retracted their own coverage. The
turn-integrity gate judged the fold over this unit's ids only and withdrew ids
carried solely by the consumed blocks; those blocks deactivate on apply and the
new block does not inherit them, so the folded messages lose all coverage and
the host pre-commit guard refuses the fold (gate:coverage-regression), causing
repeated rejections of legitimate merges.

Fix lives in projectUnit's integrity step, scoped to merges (consumedBlockIds >
0); fresh folds keep the legacy behavior verbatim:
- judge post-fold coverage as this unit's ids PLUS every still-active sibling
  block's coverage, so a tool half already covered by an unconsumed sibling reads
  as folded rather than as a spurious visible half (#3322 facet);
- unfold only ids whose coverage does not depend on a consumed block, re-checking
  to a fixed point; if consistency would force unfolding such an id, fail honestly
  (split fatal) instead of emitting a doomed regressive candidate.

Host coverage guard (src/stream.ts, #3353/#3322) left intact as defense-in-depth.
Regression tests: kernel/tests/issue4484-merge-coverage.test.ts (interior
hard-protected tool; tail cut mid-tool-group; clean-merge control). Bumps
acp-kernel 0.0.161 -> 0.0.162.
The kernel package.json bump skipped lockfile regeneration, so the root
package-lock.json (file:./kernel link entry) and kernel/package-lock.json
still recorded 0.0.161 and the lock-sync drift guard failed. Regenerated
both with 'npm install --package-lock-only' (the exact lock-sync recipe);
diff is only the two version lines.
Gate discipline (#3617): every gate ships with both directions asserted. The new merge stuck-fatal (cannot satisfy turn integrity without unfolding consumed-owned ids) had only success-side coverage. Adds an impossible-merge fixture (b2 tail cut mid-tool-group): asserts the honest split error fires up-front, blocksCreated stays 0, and existing coverage/state are untouched.
@ranxianglei
ranxianglei force-pushed the 2026-10-24_fix-merge-coverage-regression branch from c4d1b22 to 31a18db Compare October 27, 2026 15:51
@ranxianglei

Copy link
Copy Markdown
Owner Author

(ai: google/gemini-2.5-pro)
Rebased onto current master (7f2e119):

  • kernel/package.json conflict resolved as 0.0.168 → 0.0.169 (master's independent 0.0.166–0.0.168 bumps preserved, this change rides on top), both lockfiles regenerated to match.
  • kernel/src/unit-projection.ts applied cleanly — master hadn't touched it since the original base (02f544d).

Validation on the rebased head:

  • kernel suite: 1484 pass / 0 fail (includes kernel/tests/issue4484-merge-coverage.test.ts)
  • bili targeted: issue3322-coverage-regression-guard + issue2627-turn-sep-coverage + issue2202-coverage-accrual — 18 pass / 0 fail
  • npm run typecheck clean, npm run build success

CI lanes on the pushed head will confirm the full suite. Still needs human review (kernel change).

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

冲突处置核验完成:rebase 后的 head(31a18db6)对当前最新 master(9b6ec2345)复核无冲突,GitHub PR 状态已由 dirty 转为 mergeable(state=behind,不阻塞合并)。rebase + push 由上一轮完成(见上方评论),本轮是对 rebase 点之后 master 又前进的 5 个提交做的独立复核。

八问:

  1. 是什么问题:这个 kernel PR 在途期间 master 前进了新提交,分支与 base 产生文本冲突(merge state dirty),PR 无法合并。
  2. 什么原因导致的:纯版本号撞线——本分支把 kernel 版本 bump 到 0.0.162,期间 master 经其他已合入的 kernel 变更把版本推进到 0.0.168,双方改了 kernel/package.json 与两个 lockfile 的同一条 version 行。代码层无语义冲突:master 自原 base(02f544da)起没碰过 kernel/src/unit-projection.ts。
  3. 影响范围与严重程度:只影响本 PR 的合并流程,无任何用户侧行为影响;流程级阻塞,非功能缺陷。
  4. 怎么修复的:rebase 到当时最新 master(7f2e1190),版本冲突解为 0.0.169(master 的 0.0.168 之上 +1,两侧各自的独立 bump 历史都保留),两个 lockfile 重新生成对齐;修复本体(unit-projection.ts + issue4484 回归测试)干净应用、零改动。
  5. 是否有回归问题:无。逐项核过:rebase 点之后 master 新增 5 个提交均不触碰 kernel/(只有 bili 0.1.208 版本同步 + dsh [Bug][dsh] 0.1.208: attributed-but-not-ready direct send has no window guard, and takeover is declared before readiness — long sessions 400 with prompt_too_long | 归因正常但代理未就绪时直发无窗口护栏 + 就绪前就宣告接管 → 长会话必 400 #4894 guard),故 0.0.169 无需再动;对当前 master tip 跑 merge-tree 为干净合并;修复文件 rebase 前后逐字节一致;三处版本号引用(kernel/package.json、kernel/package-lock.json、root package-lock.json 的 kernel 条目)一致为 0.0.169——kernel-guard 的版本纪律(源变更必须伴随 bump)与 lock-sync 门禁都满足,且 CI 上 lock-sync 已实际通过。
  6. 是否破坏缓存:不涉及(仅版本号与 lockfile,未触碰任何缓存/重试/游标逻辑)。
  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。

CI 现状(pushed head):30 项 check,0 红,8 项在跑;静态门禁已过(lock-sync / version-guard / docs-sync / config-surface-guard 等均绿),kernel-guard 排队中。

判定:冲突已解除,PR 恢复可合;kernel 变更仍需人工 review 后合并。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

冲突已解决,PR 现在可合并。 过程说明:我接单后按 merge master 方式解冲突(kernel 版本并集到 0.0.169、lockfile 按 CI recipe 重新生成),推送前发现分支已被 rebase 到新 master(提交作者 ci-check,即原 PR 作者 agent)——两边解法一致,我弃掉自己的未推 merge commit,改为独立复核其 rebase 的正确性,不做双重推送。

复核与验证结果(在等价合并树上本地定向跑):

结构化八问:

  1. 是什么问题:PR 分支与 master 冲突(mergeable=False/dirty),无法合并。
  2. 什么原因导致的:PR 在途期间 master 合入了多个 kernel 变更(acp-kernel 0.0.161→0.0.168)与版本同步,与本 PR 的 kernel 版本 bump 及 lockfile 记录撞在同一行。
  3. 影响范围与严重程度:仅本 PR 自身的可合并性;冲突存在期间无法合入,无用户运行时影响。
  4. 怎么修复的:rebase 三个提交到新 master,冲突按「版本并集」解(0.0.168+本变更→0.0.169),lockfile 用 CI 同款 recipe 重新生成;源码零改动。
  5. 是否有回归问题:未发现——等价树上 typecheck/定向测试全绿,且此前红 master 的两个存储测试已在上游修复并通过。
  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(0.0.169 是 Kernel Boundary CI 强制的版本 bump,非配置项)。

当前状态:mergeable=True(state=behind,非阻塞),CI 重跑中——11 绿 / 8 排队 / 11 skipped(e2e 门)/ 0 失败。CI 全绿后即可合并。

@ranxianglei
ranxianglei merged commit 9563c1b into master Oct 28, 2026
51 checks passed
@github-actions

Copy link
Copy Markdown
Contributor

📦 Merged Build Published

This merge to master was published as 0.0.894-master.9563c1b (rolling master tag).

npm install -g billion-context@0.0.894-master.9563c1b  # 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.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

bug Something isn't working p2 scope/kernel

Projects

None yet

1 participant