fix(autofix): resolve owning package for nested paths; report verify-failed handoffs as not pushed - #7330
Merged
Merged
Conversation
…failed handoffs as not pushed The verify gate mapped each changed file to a flat `packages/<dir>` and read `<dir>/package.json`, which ENOENT-crashed on nested packages such as packages/channels/base — the container packages/channels has no package.json. Walk each changed file up to its nearest package.json in both the issue-fix and review-address verify steps, and skip any candidate that still has none. When such a verify failure follows an agent commit, the review-address handoff rendered the agent's optimistic address-summary.md (which can cite a commit SHA) under a neutral "what I found" heading, so a maintainer chased a commit that was discarded with the runner workspace. An EXIT trap now records any post-commit non-zero exit as outcome=failed, and the handoff states plainly that the change did NOT pass the gate and was NOT pushed. Tests: walk-up detection over a nested package tree, the outcome=failed trap, and the not-pushed handoff wording — each mutation-verified.
Collaborator
|
Thanks for the PR! Template looks good ✓ Problem: observed bug with clear evidence — the verify gate ENOENT-crashed on nested packages ( Direction: CI infrastructure fix for the autofix verification gate — squarely within scope. No auth/sandbox/model/public-contract concerns. Size: not applicable — no core module paths touched. All changes are in Approach: the scope feels right — two focused fixes that both directly address the observed failures. The resolver reads on-disk workspaces globs (not Moving on to code review. 🔍 中文说明感谢贡献！ 模板完整 ✓ 问题：已观测到的 bug，有明确证据——验证门在嵌套包上 ENOENT 崩溃（ 方向：autofix 验证门的 CI 基础设施修复——完全在范围内。无 auth/sandbox/模型/公共契约方面的顾虑。 规模：不适用——未触及核心模块路径。所有改动在 方案：范围合理——两个聚焦修复都直接针对已观测到的故障。解析器读取磁盘上的 workspaces globs（非 进入代码审查 🔍 — Qwen Code · qwen3.7-max Reviewed at |
Collaborator
Author
ReviewBoth defects are real and worth fixing, and the verbatim-extraction test style is the right technique for a workflow-only change. I re-ran Three issues, one of which I think blocks. 1. Walk-up stops at any
|
Collaborator
|
Code review — independent proposal first: I'd replace the flat Reading the diff against that baseline:
No critical blockers. No AGENTS.md violations. Real-scenario testing — this is a CI/workflow-only PR (no user-facing CLI behavior), so the functional test is exercising the resolver against the real repo layout and running the unit tests: Resolver output on the real repo: 中文说明代码审查——先独立提出方案：我会用展开 root 对照该基线阅读 diff：
无关键阻塞项。无 AGENTS.md 违规。 真实场景测试——这是纯 CI/workflow PR（无用户可见 CLI 行为），功能测试为对真实仓库布局运行解析器并执行单元测试。解析器在真实仓库上输出正确：嵌套 workspace、fixture 归属、排除项均符合预期。所有单元测试通过（63/63 + 14/14）。 — Qwen Code · qwen3.7-max Reviewed at |
Collaborator
|
Confidence: 5/5 — clean across every stage; would merge without hesitation. This PR fixes two concrete, observed failures in the autofix verify gate — the nested-package ENOENT crash and the misleading handoff — with a minimal, well-tested solution. The resolver's design choices are all justified: on-disk globs (not node_modules) so branch-added workspaces are visible, longest-prefix matching so fixtures don't steal ownership, shared staging so the two gates can't drift, and a loud failure on an empty workspace set rather than a silent skip. The committed flag's exit-code discrimination (1 vs 128) is the kind of careful detail that prevents the next "chasing a phantom commit" confusion. My independent proposal matched the PR's approach — I didn't find a simpler path. Every change in the diff serves the stated goal; no drive-by refactors or scope creep. The tests are thorough and the resolver works correctly against the real repo layout. If I had to maintain this in six months, the comments and structure would make it straightforward. 中文说明置信度：5/5——每个阶段都干净；毫不犹豫可以合并。 本 PR 用最小化、充分测试的方案修复了 autofix 验证门中两个具体的、已观测到的故障——嵌套包 ENOENT 崩溃和误导性 handoff。解析器的设计选择都有充分理由：磁盘 globs（非 node_modules）使分支新增 workspace 可见，最长前缀匹配使 fixture 不窃取归属，共享暂存使两个门不会漂移，空 workspace 集合大声失败而非静默跳过。committed 标志的退出码判别（1 vs 128）正是防止下一次"追幻影 commit"困惑的细致之处。 我的独立方案与 PR 一致——没找到更简路径。diff 中每个改动都服务于既定目标；无顺手重构或范围蔓延。测试全面，解析器在真实仓库布局上工作正确。 — Qwen Code · qwen3.7-max Reviewed at |
qwen-code-ci-bot
previously approved these changes
Jul 20, 2026
qwen-code-ci-bot
left a comment
qwen-code-ci-bot
left a comment
Collaborator
There was a problem hiding this comment.
LGTM, looks ready to ship. ✅
Contributor
Code Coverage Summary
CLI Package - Full Text ReportCore Package - Full Text ReportFor detailed HTML reports, please see the 'coverage-reports-22.x-ubuntu-latest' artifact from the main CI run. |
wenshao
commented
Jul 20, 2026
wenshao
left a comment
wenshao
left a comment
Collaborator
Author
There was a problem hiding this comment.
Reviewed. Suggestions are inline.
— qwen3.8-max-preview via Qwen Code /review
… script Addresses review on #7330. Extract the changed-file → owning-package walk into .github/scripts/resolve-owning-packages.sh, staged to RUNNER_TEMP from the trusted base alongside check-settings-schema.sh and invoked from both verify gates, so the two gates cannot drift into resolving packages differently (the 8-line walk was otherwise duplicated verbatim in each). Updates the package-scripts test that pinned the old inline grep. Narrow the verify-failed handoff lead-in to "This change was NOT pushed": four paths set outcome=failed BEFORE the deterministic gate runs (agent abort via failure.md, dirty tree, unchanged branch, missing address-summary.md), so the previous "did NOT pass the verification gate" claim was factually wrong for them. The specific reason stays in the headline and the quoted summary.
The repo's shellcheck gate runs --enable=all --severity=style, under which
bare $f/$d references trip SC2250 (prefer ${var}). Brace them to match the
convention already used in check-settings-schema.sh, and update the script
content assertions accordingly. Verified with shellcheck 0.11.0 using the
exact CI flags: clean.
qwen-code-ci-bot
previously approved these changes
Jul 20, 2026
qwen-code-ci-bot
left a comment
qwen-code-ci-bot
left a comment
Collaborator
There was a problem hiding this comment.
No issues found. LGTM! ✅
— qwen3.7-max via Qwen Code /review
…ndoff on commit existence Addresses the deeper review on #7330. Blocking issue: the "nearest package.json" resolver mapped a change under a workspace's fixture/example package (e.g. packages/cli/src/commands/extensions/examples/starter) to that fixture, whose test script is not Vitest — silently SKIPPING packages/cli's own tests, a coverage regression invisible in the log. Resolve against the authoritative `npm query .workspace` set instead and take each file's longest-prefix workspace: nested workspaces (packages/channels/base) match exactly, fixtures and non-workspace paths (packages/sdk-python, packages/README.md, the excluded packages/desktop) drop. Also harden the resolver against a final line with no trailing newline and against an unmatched last line, which under `set -o pipefail` would otherwise abort the script. Handoff wording: keying "was NOT pushed / commit discarded" on outcome=failed was wrong for the abort paths (failure.md, dirty tree, unchanged branch, missing address-summary.md), which set outcome=failed before ever making a commit. Record committed=true right after checkout — before any gate can fail — and key the wording on that; the abort/no-op paths keep the neutral framing. This removes the EXIT trap entirely (its only observable effect was that wording), so it no longer mislabels pre-commit failures either.
Collaborator
Author
|
Thanks — this is an excellent catch on #1; all three are addressed in 3a0cfa8. #1 (blocking) — resolve against the npm workspace set, not "any package.json". Two robustness bugs surfaced while writing that test, both fixed in the script: a final changed-path with no trailing newline was dropped ( #2 + #3 — key the wording on commit-existence, and the trap is gone. The gate now records Smaller notes.
|
Collaborator
Author
Re-review (3a0cfa8)All three previous findings are genuinely fixed — verified, not just read:
One new issue that I think matters, plus smaller ones. A.
|
…tested; harden resolver Addresses the re-review on #7330. The resolver sourced its workspace set from `npm query .workspace`, which reads node_modules — installed from the BASE checkout. A workspace the PR branch ADDS (a new channel adapter, a new sdk — the issue-fix job's whole purpose) was invisible, so its tests were silently skipped, and for a nested new package the ENOENT crash this PR fixes turned into a silent skip. Expand the set from the on-disk root package.json `workspaces` globs instead (shallow `dir/*` + literals, honouring `!` negations, keeping dirs with a package.json): it reflects the branch, matches what `npm run --workspace` accepts downstream, and needs no install. Verified to reproduce `npm query`'s set exactly on the current tree. Also from the review: - Fail the gate loudly on an empty/unreadable workspace set instead of the silent "no package changes" skip, and drop the now-unneeded `|| true` at both resolver call sites (the resolver already exits 0 on legitimate no-match). - Record committed=true at the TOP of the step (ref-only diff), covering an agent that commits then aborts, and count only `git diff --quiet` exit 1 as a commit (128 is a git error, not a discarded commit). - Correct the two call-site comments that still described the superseded nearest-package.json approach. Also hardens the resolver against a final changed-path with no trailing newline and an unmatched last line under `set -o pipefail`.
Collaborator
Author
|
Sharp catch on A — that one would have quietly re-opened the exact hole. All fixed in 1de8989. A — resolve from the on-disk manifest, not B — loud on empty, C — comments corrected. Both Smaller notes:
Also hardened the resolver against a final changed-path with no trailing newline and an unmatched last line under Local re-verification with the exact CI toolchain: shellcheck 0.11.0 ( |
Collaborator
Author
Re-review (1de8989) — LGTMEverything from the last round is addressed, and I verified the substantive parts rather than reading them. A — on-disk workspace expansion. The replacement is exactly equivalent to what it replaced on the current manifest, which is the property that matters for not regressing: …and it now also resolves branch-added workspaces, which was the bug. B — loud failure. C — both call-site comments now describe the actual algorithm. Minor notes — both fixed, and the risk I was implicitly asking you to take on the 81/81 pass locally; Optional polish (non-blocking)
Nothing here blocks. Nice iteration — the resolver ended up more correct than the |
wenshao
commented
Jul 20, 2026
wenshao
left a comment
wenshao
left a comment
Collaborator
Author
There was a problem hiding this comment.
Reviewed. Suggestions are inline.
— qwen3.8-max-preview via Qwen Code /review
Comment on lines
+1039
to
+1040
| # Map each changed file to its OWNING package — the nearest ancestor | ||
| # with a package.json — via the trusted staged resolver, shared with |
Collaborator
Author
There was a problem hiding this comment.
[Suggestion] The comment describes the resolver as "the nearest ancestor with a package.json" — the exact algorithm the resolver deliberately does NOT do. The resolver script's own header (resolve-owning-packages.sh lines 10–14) explicitly states the workspace set is npm query .workspace, NOT "any ancestor directory that has a package.json". — Concrete cost: a future maintainer debugging fixture-package resolution reads this comment, concludes the resolver should do nearest-package.json walking, and "fixes" it — reintroducing the fixture-shadowing bug where packages/cli/src/commands/extensions/examples/starter's package.json shadows packages/cli's tests.
Suggested change
| # Map each changed file to its OWNING package — the nearest ancestor | |
| # with a package.json — via the trusted staged resolver, shared with | |
| # Map each changed file to its OWNING package — the longest-prefix | |
| # npm workspace (from `npm query .workspace`) — via the trusted staged resolver, shared with |
— qwen3.8-max-preview via Qwen Code /review
Comment on lines
+2501
to
+2502
| # Map each changed file to its OWNING package — the nearest ancestor | ||
| # with a package.json — via the trusted staged resolver, shared with |
Collaborator
Author
There was a problem hiding this comment.
[Suggestion] Same misleading comment as the issue-fix verify gate (~line 1039): "the nearest ancestor with a package.json" describes the algorithm the resolver was written to replace. See the resolver script's header for the correct description. — Concrete cost: same misdirection risk, doubled across both call sites.
Suggested change
| # Map each changed file to its OWNING package — the nearest ancestor | |
| # with a package.json — via the trusted staged resolver, shared with | |
| # Map each changed file to its OWNING package — the longest-prefix | |
| # npm workspace (from `npm query .workspace`) — via the trusted staged resolver, shared with |
— qwen3.8-max-preview via Qwen Code /review
qwen-code-ci-bot
left a comment
qwen-code-ci-bot
left a comment
Collaborator
There was a problem hiding this comment.
Reviewed. Not reviewed: six-months-later-maintainer — Agent 6c subagent execution failed. Not reviewed: build-and-test — Agent 7 infrastructure failure (worktree inaccessible during review). Not reviewed: coverage — could not read the agents' transcripts (no subagent transcripts at /home/github-runner/actions-runner-22/_work/_temp/qwen-home/projects/-home-github-runner-actions-runner-22--work-qwen-code-qwen-code/subagents/c32eeb27-1b8d-42e5-8542-f31885cdd009 (ENOENT: no such file or directory, scandir '/home/github-runner/actions-runner-22/_work/_temp/qwen-home/projects/-home-github-runner-actions-runner-22--work-qwen-code-qwen-code/subagents/c32eeb27-1b8d-42e5-8542-f31885cdd009'). The harness writes one per agent; if there are none, either no agents ran or the harness could not write them.), so this run cannot show that any of the diff was read. Not reviewed: verification — could not check that Step 4 and Step 5 ran (no subagent transcripts at /home/github-runner/actions-runner-22/_work/_temp/qwen-home/projects/-home-github-runner-actions-runner-22--work-qwen-code-qwen-code/subagents/c32eeb27-1b8d-42e5-8542-f31885cdd009 (ENOENT: no such file or directory, scandir '/home/github-runner/actions-runner-22/_work/_temp/qwen-home/projects/-home-github-runner-actions-runner-22--work-qwen-code-qwen-code/subagents/c32eeb27-1b8d-42e5-8542-f31885cdd009'). The harness writes one per agent; if there are none, either no agents ran or the harness could not write them.).
— qwen3.7-max via Qwen Code /review
Collaborator
Author
|
@qwen-code /triage |
ytahdn
approved these changes
Jul 20, 2026
ytahdn
left a comment
ytahdn
left a comment
Collaborator
There was a problem hiding this comment.
LGTM. 两个 autofix verification gate 缺陷修复均正确。
嵌套包解析：resolve-owning-packages.sh 从磁盘 workspaces globs 展开真实 workspace 集合（非 npm query），最长前缀匹配 == ""/* 正确处理嵌套包，set -euo pipefail + if 形式避免 pipefail 下 no-match 退出，空集合 exit 1 大声失败。两个 verify gate 均从 trusted base checkout 暂存脚本，防止 PR 分支篡改。
commit 状态记录：|| committed_rc=$? 防止 set -e 触发，仅 exit 1（真实 diff）设置 committed=true，exit 0/128 均不设置。Handoff 措辞基于 COMMITTED（非 OUTCOME）正确区分 committed-but-unpushed vs abort/no-op。
测试覆盖 5 个新用例，边界条件处理到位。
— qwen3.7-plus via Qwen Code /review
qwen-code-ci-bot
approved these changes
Jul 20, 2026
qwen-code-ci-bot
left a comment
qwen-code-ci-bot
left a comment
Collaborator
There was a problem hiding this comment.
LGTM, looks ready to ship. ✅
This was referenced Jul 20, 2026
Merged
RehDen
pushed a commit
to RehDen/qwen-code
that referenced
this pull request
Jul 21, 2026
* feat(autofix): raise the strict round cap from 5 to 10 Measured across the last 40 bot-authored PRs: 17 finished at round 0, 12 at 1, 4 at 2, 2 at 3, 1 at 4, and 3 reached the cap of 5. All three that reached it merged AT it rather than stalling — and one of those spent two of its five rounds on the verify-gate ENOENT that QwenLM#7330 has since fixed. So the ceiling was never the thing that stopped a PR, but it sat close enough to bind on a bad day with no headroom. 10 gives that headroom. The cap exists to stop an unproductive LOOP, not to ration ordinary iteration; a genuinely stuck PR still stops, just later. Deliberately not larger: retries for a transient model or gate failure increment the same counter, so the cap also bounds how much one bad provider window can spend. API_AUTH_MAX_ROUNDS stays at 3 and still short-circuits the errors only a maintainer can fix. Replaces the literal `MAX_ROUNDS: '5'` assertion with the ordering the numbers must satisfy — auth cap < strict cap < takeover cap — so the values stay tunable and a cap that stops binding fails instead. * fix(scripts): anchor round-cap regex with word boundary (QwenLM#7412) --------- Co-authored-by: wenshao <wenshao@example.com> Co-authored-by: qwen-code-ci-bot <qwen-code-ci-bot@users.noreply.github.com>
This was referenced Jul 21, 2026
Collaborator
|
Released in v0.20.1. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this PR does
Fixes two defects in the autofix verification gate — the per-package test re-run and the failure handoff — that together stranded and misreported #7246.
1. Resolve each changed file to its owning npm workspace. The gate mapped every changed file to a flat
packages/<dir>and read<dir>/package.json. That ENOENT-crashed on nested packages (a change underpackages/channels/basemapped to the containerpackages/channels, which has nopackage.json), taking down the whole verify step. The first fix walked up to the nearestpackage.json— but "has apackage.json" ≠ "is a workspace": a change under a fixture package inside a workspace's src tree (e.g.packages/cli/src/commands/extensions/examples/starter) would resolve to the fixture, whose test script is not Vitest, and silently skippackages/cli's own tests. So the gate now resolves against the real workspace set — expanded from the on-disk rootpackage.jsonworkspacesglobs (shallowdir/*+ literals, honouring!negations, keeping dirs with apackage.json) — and takes each file's longest-prefix workspace: nested workspaces (packages/channels/base) match exactly, and fixtures / non-workspace paths (packages/sdk-python, a top-levelpackages/README.md, the excludedpackages/desktop) match nothing and are dropped. Reading the globs on disk rather thannpm query/node_modules matters because node_modules reflects the base checkout — a workspace the branch adds (a new channel adapter, a new sdk, which is the issue-fix job's whole point) would otherwise be invisible and its tests silently skipped. The logic lives in a shared script.github/scripts/resolve-owning-packages.sh, staged toRUNNER_TEMPfrom the trusted base alongsidecheck-settings-schema.shand invoked from both verify gates (with no|| true, so a resolver error fails the gate loudly), so the two cannot drift.2. Only call a change "not pushed" when a commit actually existed. When a verify failure follows an agent commit, the review-address handoff rendered the agent's own
address-summary.md— which reads like a success and can cite a commit SHA — under a neutral "What I found before stopping:" heading, so a maintainer chased a commit that was discarded with the runner (the "no commit record" confusion on #7246). The gate now recordscommitted=trueat the top of the step (a ref-only diff), before the failure.md early-exits, and the handoff keys its "was NOT pushed — any commit referenced below … has been discarded" wording on that, not onoutcome=failed. The abort / no-op paths (failure.md, dirty tree, unchanged branch, missingaddress-summary.md) setoutcome=failedbefore ever committing, so they correctly keep the neutral framing.Why it's needed
The repo's
packages/channels/*layout meant a single nested-package review comment crashed the gate; the "nearest package.json" heuristic would have turned that crash into a silent test-coverage hole around fixture packages; and the resulting handoff pointed the maintainer at a discarded commit. This makes the gate resolve the repo's real workspace set and makes the failure report honest.Reviewer Test Plan
How to verify
npx vitest run scripts/tests/qwen-autofix-workflow.test.js— 67/67. Key coverage:workspacesglobs + members, nonpm install) including a branch-added top-level and nested workspace, a fixturepackage.jsoninside a workspace's src tree, and a!-excluded workspace; asserts the branch-added packages resolve, the fixture change resolves topackages/cli(never the fixture), and the excluded/non-workspace paths drop. A separate test asserts the resolver fails loudly (non-zero + message) on an empty workspace set.git, records the flag only whengit diff --quietexits 1 (a real diff) — not on 0 (no commit) or 128 (a git error).COMMITTED— "was NOT pushed" when a commit existed, neutral "What I found before stopping" otherwise. Neither verify gate carries an EXIT trap.cpand invokeresolve-owning-packages.shfromRUNNER_TEMP(mirrors the schema-gate staging assertions).npx vitest run scripts/tests/package-scripts.test.js— the verify step invokes the resolver.--enable=all --severity=styleclean on the script; actionlint 1.7.12 clean on the workflow;js-yamlparses; bothverifysteps passbash -n.Evidence (Before & After)
Tested on
Environment (optional)
Unit tests only — no
npm install; the resolver reads the on-disk manifest directly.Risk & Scope
package.jsonworkspaces globs (no install needed) — verified to reproducenpm query .workspace's set exactly on the current tree. It is hardened against a final changed-path with no trailing newline and an unmatched last line underset -o pipefail, and fails loudly on an empty set. A workflow-only PR (nopackages/changes) still skips package tests as before.Linked Issues
Fixes the verification-gate crash and misleading handoff observed on #7246. Same "make the managed-PR loop fail honestly" line as #7229 (crash), #7247 (model API retry).
中文说明
本 PR 做了什么
修复 autofix 验证门里的两个缺陷(逐包测试重跑 + 失败 handoff),它们共同导致 #7246 先搁置、又被误报。
1. 把每个改动文件解析到其归属的 npm workspace。 门原先把改动文件映射成扁平的
packages/<dir>再读<dir>/package.json,对嵌套包会 ENOENT 崩溃(packages/channels/base的改动映射到无package.json的容器packages/channels)。第一版改成"向上走到最近的package.json",但"有package.json"≠"是 workspace":改动落在某 workspace 源码树内的 fixture 包(如packages/cli/src/commands/extensions/examples/starter)会解析到该 fixture(其 test 脚本非 Vitest),从而静默跳过packages/cli自己的测试。所以现在门解析真实 workspace 集合——从磁盘上 rootpackage.json的workspacesglobs 展开(浅层dir/*+ 字面量,遵守!排除,保留含package.json的目录)——取每个文件的最长前缀 workspace:嵌套 workspace(packages/channels/base)精确匹配,fixture / 非-workspace 路径(packages/sdk-python、顶层packages/README.md、被排除的packages/desktop)不匹配即丢弃。读磁盘 globs 而非npm query/node_modules 很关键:node_modules 反映的是基座 checkout,分支新增的 workspace(新 channel adapter、新 sdk,正是 issue-fix job 的目的)否则不可见、测试被静默跳过。逻辑在共享脚本.github/scripts/resolve-owning-packages.sh,从可信基座 stage 到RUNNER_TEMP、两个门调用(无|| true,解析器出错即让门大声失败),不会漂移。2. 只有确有 commit 才称"未被推送"。 验证失败发生在 agent 提交之后时,review-address 的 handoff 会把 agent 的
address-summary.md(读起来像成功、可能引用 commit SHA)放在中性的 "What I found before stopping:" 标题下,维护者去追一个随 runner 丢弃的 commit(#7246 的"看不到提交记录")。现在门在 step 顶部(ref-only diff)、failure.md 早退之前记录committed=true,handoff 的"was NOT pushed …"措辞键控它、而非outcome=failed。中止 / no-op 路径(failure.md、脏树、分支无改动、缺address-summary.md)在提交前就置了outcome=failed,因此正确保留中性措辞。为什么需要
仓库的
packages/channels/*结构使一条嵌套包 review 评论就能崩门;"最近 package.json"启发式又会把崩溃变成 fixture 包周围的静默测试覆盖漏洞;随后的 handoff 还把维护者指向被丢弃的 commit。本 PR 让门解析真实 workspace 集合,并让失败报告诚实。评审验证
npx vitest run scripts/tests/qwen-autofix-workflow.test.js—— 67/67:解析器构建 workspace 树(root workspaces globs + 成员,免npm install),含分支新增的顶层与嵌套 workspace、workspace 源码树内的 fixture、!排除的 workspace;断言分支新增包被解析、fixture 改动解析到packages/cli、排除/非-workspace 路径丢弃;另有测试断言空 workspace 集合时大声失败(非零 + 消息);committed片段用 PATH-stubgit驱动 0/1/128,仅退出 1 记为 commit;handoff 键控COMMITTED;两个门都无 EXIT trap。npx vitest run scripts/tests/package-scripts.test.js—— verify 步骤调用解析器。--enable=all --severity=styleclean;actionlint 1.7.12 clean;YAML 可解析;两个verify步骤过bash -n。风险与范围
package.json的 workspaces globs(免安装)——已验证在当前树上逐字复现npm query .workspace的集合。对"末行无换行"与"末行无匹配"(set -o pipefail下会中止脚本)做了加固,空集合时大声失败。仅改 workflow(无packages/改动)的 PR 仍如常跳过包测试。关联 Issue
修复在 #7246 上观察到的验证门崩溃与误导性 handoff。与 #7229(崩溃)、#7247(模型 API 重试)同一"让被托管 PR 循环诚实失败"主线。