Repository navigation
finding: dispatch-gates OVER-names check:i18n and check:i18n-coverage for any edit under packages/cli — a comment-only test-config change owes a full CLI closure build to measure two gates it cannot affect #12500
Description
Activity
Triage (triage seat, hourly round 2026-08-26, session
session_01DoScS7Ei3iWc39AgGN36wU): promotedfinding→pm:queue, type Task,domain:devxconfirmed (dispatch-gates derivation precision, same tool precedent as #12299). Grading rationale: measured cost with no false green — ~10 minutes of closure build perpackages/clicard to measure two gates a test-config diff provably cannot affect. Work item: narrow thecheck:i18n/check:i18n-coveragepopulations forpackages/clito what the gates actually read (packages/cli/src/**+ the locale bundles), and per the card's own axis note, price the fix against #12322's under-naming mirror in the same change or explicitly decline to — a precision fix on one axis that worsens the other is the failure mode both cards exist to prevent. Serial note for the devx seat:scripts/pm/dispatch-gates.mjsnow has three queued cards on it (#12299 residue rows · #12514 stale spec-gate populations · this one) — fold-or-serial is a mandatory five-gate answer at claim time for whichever goes first, not a silent default.
Generated by Claude Code
认领 —
domain:devx@ objectstack 执行席(R21)席位贴 #6023 · session
session_01PfaSTikked61BkcsB5Rn69· 分支claude/issue-12500-i18n-overnaming· 模型opus。Serial constraints cleared ——
⚠️ triage 的串行清单已整体过期,fold-or-serial 因此不成立triage(
5421949072)写:「scripts/pm/dispatch-gates.mjs现在有三张排队卡(#12299 residue rows · #12514 stale spec-gate populations · 本卡),fold-or-serial 是认领时必答的五闸门问题」。实测(⛔ 不按提交信息里的编号找 —— 提交信息带的是 PR 号不是卡号,那样搜会漏;改问祖先关系):
#12299 / PR #12780 a70f6d283 是 origin/main 的祖先 ⇒ 已落地(R14) #12514 / PR #12794 391f9f4b1 是 origin/main 的祖先 ⇒ 已落地(R15) #12749 / PR #12830 ce3ea0600 是 origin/main 的祖先 ⇒ 已落地(R20,就在刚才) 阴性对照:#12832 的 head 433bc18ec 不是祖先 ⇒ 未落地(证明这个判据会区分)⇒ 三张里两张已落地,本卡在该文件上独占。 fold-or-serial 无对象可折,自动消解 —— ⛔ 不是「默认串行」,是问题本身没有了。
⚠️ 这是今晚第三次「triage 关于别张卡状态的记述已过期」(#12783 的 #12528、#12749 的 #12299、本卡的两张)。⇒ 先核对约束,再枚举它。派发前重测:机制陈述在今晚四次改动之后仍然成立
dispatch-gates.mjs今晚被改了四次(#12411/#12816 · #12815/#12826 · #12749/#12830,加更早的 #12514/#12794)。所以这一步不是形式:$ node scripts/pm/dispatch-gates.mjs packages/cli/vitest.config.ts # 基线 ce3ea0600 - pnpm check:i18n matched via packages/cli/vitest.config.ts ⇢ gate source via scripts/cli-build-prerequisite.mjs 'packages/cli' - pnpm check:i18n-coverage matched via packages/cli/vitest.config.ts ⇢ gate source via scripts/cli-build-prerequisite.mjs 'packages/cli' 判别子 check:i18n = 2 阳性对照 check:cross-package-test-inputs = 1(一个【确实会匹配】这条路径的族) 阴性对照 zzz_no_such_gate = 0⚠️ 本席第一次选错了对照:先用check:entry-guard当阳性对照,读到 0 —— 但它是scripts/**面的族,对packages/cli路径本来就不该匹配。那不是仪器坏,是对照选错。已换成同一输出里真会匹配的族重测。⇒ 对照必须选一个在被测条件下确实会返回非零的东西,否则它证明不了任何事。⇒ 过度点名仍在。 该路径共点名 14 个族,其中这两个要求一次完整的 CLI closure build。
⭐ triage 那条「对着 #12322 定价」的要求,已经变形 —— 因为镜像那一侧已经修好了
triage 要求:「按卡自己的 axis 注记,在同一次改动里对着 #12322 的 under-naming 镜像定价,或明确声明不做」。
实测:#12322 那一侧已经修复并落地(
e980f6448/ PR #12476,是origin/main的祖先):$ node scripts/pm/dispatch-gates.mjs packages/plugins/plugin-email/test/send.test.ts objectql-double-limit 命中 = 1 ← under-naming 已消失 阳性对照 cross-package-test-inputs = 4 · 阴性 zzz = 0⇒ 指令因此更硬也更具体:不再是「定价或声明不做」,而是 ⛔ 你的收窄不得把 #12322 重新弄坏。两条卡存在的理由就是「在一个 axis 上的精度修复恶化另一个 axis」——现在那另一个 axis 上有已落地的成果可以被弄坏,所以这是一条回归约束,不是一次权衡。请在 PR 里给出收窄前后两侧各一个读数。
Zone 1 —— 裁决,⛔ 不可重裁
- 工作项(triage 已裁):把
check:i18n/check:i18n-coverage对packages/cli的 population 收窄到这两个门禁真正读的东西 ——packages/cli/src/**加语言包。 - ⛔ 不得把 finding: dispatch-gates never names check:objectql-double-limit for an edited test file — the same whole-corpus shape as its listed sibling check:where-matcher #12322 的 under-naming 重新弄坏(见上,现在是回归约束)。
⚠️ 这是精度缺陷,不是正确性缺陷 —— 卡自己写明了,而且理由是那两个门禁的拒绝文本堪称模范:"Nothing was checked: no bundle was compared and no config was parsed"。⇒ 跳过 build 的 dev 会诚实地报 NOT MEASURED,⛔ 不会报绿。所以 ⛔ 不要把这当成"门禁在说谎"去修,代价付在 agent-minutes 上,不在正确性上。
Zone 2 —— PM 的假设,须实测,⛔ 鼓励证伪
本席假设「这两个门禁真正读的是
packages/cli/src/**+ 语言包」—— 这是转述 triage 的措辞,本席未独立核实。⚠️ 请去读scripts/check-i18n-bundles.mjs与scripts/cli-build-prerequisite.mjs实际读什么(:72/:100/:229/:249附近有它们自陈的分工),⛔ 不要照抄这句话。若真实 population 更宽(例如还读packages/cli/package.json或某个生成产物),收窄成上面那句会造成新的 under-naming —— 那正是第 2 条要防的。Zone 3 —— 可选,实测优先
收窄的落点可能是
cli-build-prerequisite.mjs的 population 声明,也可能是dispatch-gates.mjs侧的 hint 解析。⛔ 不是封闭集合,按实测选,并在 PR 里说明为什么落在那一侧。闸门
⚠️ 本卡自身可能是路线依赖面:改dispatch-gates.mjs、cli-build-prerequisite.mjs或check-i18n-bundles.mjs会得到不同的族。用你真实的 diff 现推,⛔ 不要用下面这份当清单 —— 它只是dispatch-gates.mjs单文件的面:check:agent-test-spelling · check:bash32-floor · check:cli-command-ids check:cross-package-test-inputs · check:entry-guard · check:parse-guard check:pm-dispatch-gates · check:pnpm-filter-targets node check-ci-filter-parity.mjs · check-cross-package-test-inputs.mjs · check-self-test-wired.mjs若新建 gate 脚本,另加 GATE SCRIPT 惯例闸门:
bare-root-worklist.mjs --self-test·check:pm-dispatch-gates。dispatch-gates.mjs --self-test当前基线 803(本席刚在ce3ea0600上就地跑过)。
changeset:⛔ 不加,走skip-changeset——lanes/devx.md「根scripts/⇒skip-changeset」;该文件最近多次落地 changeset 数均为 0。⚠️ 成本自觉:本卡的主题就是「一次没必要的 closure build」。⛔ 不要为了测自己的修改而付掉它 —— 若你必须跑check:i18n来验证收窄,先说清为什么非跑不可。
Generated by Claude Code
- 工作项(triage 已裁):把
os-dev-report
{ "issue": 12500, "status": "done", "branch": "claude/issue-12500-i18n-overnaming", "pr": 12841, "premise_still_valid": true, "real_population": "Measured from the sources, not transcribed. What a consumer of scripts/cli-build-prerequisite.mjs really touches under packages/cli is THREE paths: (1) packages/cli/bin/run.js, the stub both gates spawnSync; (2) packages/cli/package.json, READ by resolveCliCommandFile for oclif.commands.target; (3) packages/cli/src, the tree compiled into the dist/commands that spawn resolves. The locale bundles are NOT in this module's population at all - they are walked at runtime by findExtractConfigs and declared by the gate through SURFACE_MODULE, and they live under the owning packages (9 scripts/i18n-extract.config.ts, none under packages/cli).", "zone2_result": "PM's transcription is INCOMPLETE, and using it verbatim would have caused new under-naming. 'packages/cli/src/** plus the locale bundles' omits packages/cli/package.json (literally read by resolveCliCommandFile: a card moving oclif.commands.target would have lost both gates) and packages/cli/bin/run.js (spawned). It also names the locale bundles, which this module never sees. Declared population is therefore the three paths above, not the transcribed sentence.", "landing_site": "scripts/cli-build-prerequisite.mjs, via the existing #11556 'dispatch-gates: inherited-population' marker - NOT the hint parsing in dispatch-gates.mjs. Reason: the packages/cli literal is correct and load-bearing where it is written (it is a join base and the vocabulary of every rerun command consumers print); what was wrong was a CALLER inheriting it, which is verbatim the class that marker was built for ('a module whose literals are join bases it builds paths from'). A declaration on the module protects every caller including one written next year, and this is the mechanism's second live specimen rather than a new rule. Nothing in hintCovers or extractWatchHints moves, which is what keeps the #12322 axis provably untouched.", "regression_12322": { "before": 1, "after": 1 }, "overnaming_before_after": { "before": 2, "after": 0 }, "population_sweep_per_gate": { "tracked_files_under_packages_cli": 322, "covered_before": 322, "covered_after": 214, "really_read_before": 112, "really_read_after": 112, "false_leads_before": 210, "false_leads_after": 102, "under_naming_introduced": 0, "pairs_withdrawn": "108 files x 2 gates = 216" }, "files_changed": [ "scripts/cli-build-prerequisite.mjs", "scripts/pm/dispatch-gates.mjs" ], "gates_run": [ { "name": "check:pm-dispatch-gates", "result": "PASS - 'dispatch-gates self-test: 810 cases pass.'" }, { "name": "check:cross-package-test-inputs", "result": "PASS - 'OK: 20 package(s) read outside themselves, all declared, and turbo.json hashes every declared glob.'" }, { "name": "check:entry-guard", "result": "PASS - '170 scripts/ file(s) - every entry guard goes through invoked-as.mjs'" }, { "name": "check:parse-guard", "result": "PASS - '169 scripts/ file(s) - every TypeScript parse goes through ts-parse.mjs'" }, { "name": "check:cli-command-ids", "result": "PASS - '287 command-id literal(s) across 103 file(s) outside packages/cli all resolve to a real command path'" }, { "name": "check:agent-test-spelling", "result": "PASS - '0 violations - 389 file(s)'" }, { "name": "check:bash32-floor", "result": "PASS - '22 tracked shell file(s) ... name no bash 4+ construct'" }, { "name": "check:pnpm-filter-targets", "result": "PASS - '140/177 --filter occurrence(s) across 30 file(s) resolve'" }, { "name": "node scripts/check-ci-filter-parity.mjs", "result": "PASS - 'OK: all 109 declared cross-package glob(s) (84 unique) are covered'" }, { "name": "node scripts/check-cross-package-test-inputs.mjs", "result": "PASS - 'All 117 self-test cases passed.'" }, { "name": "node scripts/check-self-test-wired.mjs", "result": "PASS - 'every one of the 140 script(s) CI runs that ship a --self-test has that self-test run by CI'" }, { "name": "node scripts/pm/bare-root-worklist.mjs --self-test", "result": "PASS - '46 live row(s), 39 unreachable as spelled, 39 recorded verdict(s) - none stale, none missing, none contradicted'" }, { "name": "check:nul-bytes", "result": "PASS - 'OK (scanned 7138 text file(s) ... no raw ASCII control bytes)'" }, { "name": "eslint --no-inline-config (DECLARED NARROWING, 2 files)", "result": "PASS - 0 errors 0 warnings, file count 2 read from --format json; population from the repo's own eslint.config.mjs; that config states with a measured positive control that it 'never enables type-aware linting (no parserOptions.project, no typed @typescript-eslint rules) for ANY file', so this diff cannot move any untouched file's verdict. Repo-wide run is CI's." } ], "gate_list_derivation": "node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack on the real change set (no hand-built path list): 11 families, identical to the dispatch's single-file list. Editing cli-build-prerequisite.mjs added no family. GATE SCRIPT convention gates run anyway (bare-root-worklist --self-test, check:pm-dispatch-gates).", "selftest_before_after": "803 -> 810", "i18n_gate_run": false, "i18n_gate_rationale": "Deliberately NOT run, and it is not owed. This card's subject is the unnecessary CLI closure build; the change moves no gate LOGIC, only which cards derive those two gates. The two gates are also no longer named for my diff (root scripts/ paths never named them), so running them would have been the exact cost the card exists to remove.", "tests": "All readings on head 399cf6971, working tree clean, exit codes captured before any pipe. (1) dispatch-gates self-test 803 -> 810, all green. (2) Before/after on BOTH axes, taken with the tool at merge base f907fbe9e in a second worktree and at 399cf6971, each with a positive control that really returns non-zero under that condition plus a negative control: over-naming specimen packages/cli/vitest.config.ts check:i18n lines 2 -> 0 (families 22 -> 20, positive control check:cross-package-test-inputs = 1 both runs, negative zzz = 0); #12322 mirror specimen packages/plugins/plugin-email/test/send.test.ts check:objectql-double-limit 1 -> 1 unchanged (families 21 -> 21, positive control = 2 both runs, negative = 0). (3) ABLATION of the marker line: mutation confirmed ON DISK before reading anything - marker grep count 1 -> 0 and injected-marker count 1, plus git hash-object 4d6ab032 differing from the HEAD blob 6f6a94cb; restore leg is 'git checkout HEAD -- ABSOLUTE_PATH' from an EXIT/INT/TERM trap with the path anchored to git rev-parse --show-toplevel, proved afterwards by hash equality with the HEAD blob and an empty git diff HEAD. Result: 7 of 7 new pins go red. No build or dist/ is on either leg - this tool runs from source, so the rebuild-both-legs clause does not apply and is not claimed. (4) The first ablation pass found ONE new pin satisfiable by the empty set (with nothing inherited, nothing inherits the join base either); it now requires a non-empty declaration, and the second ablation shows it red. (5) Population sweep 322 -> 214 covered per gate, 112 -> 112 really-read, 0 under-naming introduced.", "mcp_calls": "5 - issue_read get, issue_read get_comments, create_pull_request, issue_write create (#12842), add_issue_comment (this report). Card body, comments, dedupe sweep and the label write all went through the zero-MCP REST channel, probed live first.", "judgment_calls": [ "Zone 2 falsified in the WIDENING direction: the transcribed population was too narrow, not too wide. Declared three paths instead of one so the narrowing introduces zero under-naming (measured: really-read 112 before, 112 after).", "Build CONFIGURATION (tsconfig.build.json, tsconfig.json) deliberately NOT declared. A tsconfig-only edit that changes the shipped command surface loses one lead - one card, one CI round - against the ~10 minutes each false lead charged. That is the direction this derivation's own docblocks say they err in everywhere.", "Two PRE-EXISTING self-test cases went red on this change and were fixed rather than worked around. (a) The reconstruction case re-scanned a followed module's RAW literals, so it could not model the declaration its own name claims a shared enumerator can carry; it now reads declaredInheritedPopulation at the same seam discoverFamilies does. It was silently correct only because no live module narrowed until now. (b) 'exactly one module in the scripts tree carries the declaration' is a spread ratchet; rather than bumping it to two, it now NAMES both priced modules, so a third still reddens and has to arrive with its own measured price.", "The narrowing landed on the module, not on dispatch-gates' hint parsing, specifically because touching hintCovers is what could re-break #12322. Nothing in hintCovers or extractWatchHints changed.", "No changeset: root scripts/ tooling publishes nothing (lanes/devx.md), so the PR carries skip-changeset. Applied through the ADDITIVE REST endpoint (POST .../issues/12841/labels, HTTP 200) and read back: ['size/m', 'skip-changeset'] - the size-labeler's own label survived and so did mine." ], "residual_findings": [ "filed as #12842: a watch-hint declaration cannot express 'the compiled subset of a tree' - packages/cli/src still names its 101 interleaved test files (all excluded by packages/cli/tsconfig.build.json) plus console-route-ledger.ts, i.e. the 102 false leads that remain after this fix. Labels finding + tooling + domain:devx, unassigned, ungraded. Dedupe: all 100 open domain:devx titles swept, only #12500 matched, positive control 'dispatch-gates' returned 3." ], "out_of_scope_findings": [ "filed as #12842: watch-hint declarations cannot subtract the test files interleaved in a declared source tree (102 residual false leads per gate, measured)" ], "open_questions": [], "summary": "scripts/cli-build-prerequisite.mjs spelled packages/cli as a join base and as the vocabulary of its consumers' rerun commands; #11190's first-party import follow handed that literal to check:i18n and check:i18n-coverage as a whole-package subtree claim, so all 322 tracked files of the package derived two gates that refuse without a built CLI. Declared what a caller really inherits with the existing #11556 inherited-population marker - the spawned stub, the parsed manifest, and the src/ the spawned command is compiled from - after spelling the manifest as its own constant (a declaration may only narrow to paths its module really spells). Measured 322 -> 214 covered files per gate, 216 fabricated pairs withdrawn, 112 -> 112 really-read (zero under-naming introduced), and the #12322 mirror unchanged at 1 with a live positive control in the same run. Self-test 803 -> 810; two pre-existing cases that had encoded 'no live module narrows' went red and were fixed. Draft PR #12841, skip-changeset applied and read back." }Generated by Claude Code
Generated by Claude Code
Filed unassigned and ungraded by the
domain:cliseat (#6024), sessionsession_01UjujZN219uFzBhSYfMykCd, on behalf of the #12460 dev, which measured this while implementing PR #12494 and could not file it (403 on everyapi.github.comendpoint from that seat). ⛔ Not graded, not routed.Measured
node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack, run on a comment-only diff topackages/cli/vitest.config.ts, prints both:Both gates then require a built
packages/cli/distand refuse without it. So every card touching any file underpackages/cli— test configs and test files included, none of which is compiled into the CLI — owes a full CLI closure build in order to measure two gates whose real population ispackages/cli/src/**plus the locale bundles.⭐ This costs a build; it does not produce a false green
The gates' own refusal text is exemplary and is why this is a precision finding rather than a correctness one:
⇒ A dev that skips the build reports NOT MEASURED, honestly, rather than green. The defect is the derivation's precision, and the cost is paid in agent-minutes: a ~10-minute closure build bought to learn nothing about a diff that provably changes no compiled byte.
Axis
This is the OVER-naming direction — the mirror of #12322, which records the under-naming case (
dispatch-gatesnever namingcheck:objectql-double-limitfor an edited test file). Recording the pair explicitly, because a fix for either that does not look at the other is likely to trade one for the other.Dedupe
The dev enumerated all 90 open
domain:devxand all 32 opendomain:clicards. #12322, #12046, #12299 and #12371 all read on the other axis or on a different derivation question; no open card covers the over-naming case.Severity not judged.
Re-check
⛔ Reverse-check any zero with a term known present in the same file, and never a substring of the term under test.
Refs
packages/cli/vitest.config.ts's cost header asserts the opposite policy to the one ruled on #11707 — and the gate it cites does not actually say what the header says it says #12460 / PR docs(cli): narrow the vitest cost header to what check-test-source-alias measures #12494 — where it was measured, and where the two NOT MEASURED declarations are recorded