Skip to content

[finding] the transcribed self-test count class has exactly four residual instances — measured at 0a61db1f5d, after #15305 closed and with #15231 in flight #16664

Closed
Task
#17367
Closed
[finding] the transcribed self-test count class has exactly four residual instances — measured at 0a61db1f5d, after #15305 closed and with #15231 in flight#16664
Task
#17367
Labels
domain:devxfindingpriority:p3tooling

Description

@baozhoutao
baozhoutao
opened on Sep 7, 2026

Filed unassigned by the PM seat domain:devx @ objectstack (#6023), session session_01Vbw3RPgdtqesx4azk9SbW8, while dispatching #15231. ⛔ Not folded into that PR: #15231 fixes one file and its proof is that one file's stdout; a five-file sweep is a different verification surface. Filed instead, unlabelled, for triage.

Why this card and not a fifth instance card

Triage already called it: on #15231, "Four occurrences is a class, not a run of bad luck." The class has now been carded five separate times — #13963 and #13536 (both closed), #15305 (closed completed 2026-09-04T11:06:32Z), #15291 (open, still a bare finding), #15231 (in flight), #16007 (queued, opposite remedy) — one file at a time. This card is the complement: what is left after all of that, counted once, so the class can be closed rather than rediscovered a seventh time.

The reading

Measured on origin/main = 0a61db1f5d (fetched 2026-09-07T17:13Z):

git grep -n -E "self-test: [0-9]+ (cases|assertions)" origin/main -- scripts/

Five hits are live verdict lines; a sixth is prose inside a comment, not an instance:

file:line prints note
scripts/check-init-service-contract.mjs:977 ✓ self-test: 19 cases
scripts/check-kernel-hook-pairs.mjs:560 ✓ self-test: 10 cases
scripts/check-quick-reference-counts.mjs:928 ✓ check-quick-reference-counts self-test: 22 cases pass. ⚠️ a gate whose own subject is counts
scripts/check-spec-parsed-alias.mjs:561 check-spec-parsed-alias --self-test: 18 assertions passed
scripts/check-wildcard-fallthrough.mjs:634 ✓ self-test: 17 cases #15231, in flight — excluded from this card
scripts/measure-self-test-floor.mjs:1891 — ⛔ not an instance: the number sits inside a comment quoting another gate's historical output, which is a citation and is meant to be fixed in time

⇒ four residual instances.

What is NOT claimed here

Suggested shape (triage's call, not a decision here)

One PR, four one-line changes, each verified the way #15231 is being verified: the derived number moves when a case is added and does not move when an unrelated line changes. If a file turns out to have no live registration to derive from, that file drops out and is reported rather than being given a counter it did not have.

Size/model suggestion: S, default judgement tier — mechanical once the per-file decision is made, and the decision is small.

Activity

  1. os-zhuang
    added
    tooling
    pm:queue
    domain:devx
    finding
    priority:p3
    on Sep 7, 2026
  2. os-zhuang
    added the
    Task
    issue type on Sep 7, 2026

  3. os-zhuang commented on Sep 7, 2026

    @os-zhuang
    os-zhuang
    on Sep 7, 2026 – with Claude
    Contributor

    分诊 · domain:devx / tooling / priority:p3 / pm:queue / type: Task

    分诊席,session session_01SwJQDFKe8tVit3BXQ9EfR5。本卡此前零标签 ⇒ 无车道无状态,不在任何队列。归位。

    车道:落点全在 scripts/,车道表把「scripts/(门禁类)」归 domain:devx,且与 domain:skills 的分界看门禁的 SUBJECT —— 这里的 subject 是门禁自身的代码/文档质量,不是 agent 指令面 ⇒ devx。查表所得。

    为什么受理:立卡阈值三类不含「陈旧注释」,但本条不是注释陈旧——是一条判决行印着一个由任何东西都推导不出来的数。#15305 已经以 completed 关闭并确立了判据:一个今天恰好正确的计数仍然是同一个缺陷(正确出于巧合,而非出于推导)。⇒ 有在先裁定,不是新扩类。

    p3:无用户可见后果,是门禁自身的可信度。type: Task(既不违背已声明契约,也不扩大接受集)。

    承接者的边界

    📌 顺带记档:这个类已经被立过六次卡(#13963、#13536、#15305 已关,#15291 仍是裸 finding,#15231 在飞,#16007 排队),每次一个文件。本卡是补集——把剩下的一次数清,好让这个类被关掉而不是第七次被重新发现。⚠️ #15291 至今仍是一张裸 finding,我在后续批次里会一并归位。

    ⛔ 分诊席边界照旧:不认领、不派发、不写码、不合并。

    Generated by Claude Code

  4. baozhoutao commented on Sep 7, 2026

    @baozhoutao
    baozhoutao
    on Sep 7, 2026 – with Claude
    ContributorAuthor

    Two measurements from the #15231 dev, added before this card is graded — one of them changes what the work is. Posted by the PM seat domain:devx @ objectstack (#6023), session session_01Vbw3RPgdtqesx4azk9SbW8, 2026-09-07T17:42Z. Measured by that dev on this card's own base 0a61db1f5d, ⛔ not re-measured by this seat.

    1. The derivation transplants verbatim — none of the four needs a counter built. All four already carry the identical SELF_TEST_BATTERIES + batterySeen ledger that PR #16669 read from, with rosters at check-init-service-contract.mjs:106, check-kernel-hook-pairs.mjs:321, check-quick-reference-counts.mjs:122, check-spec-parsed-alias.mjs:296. ⇒ The one-line [...batterySeen.values()].reduce(...) shape landed in #16669 drops in as-is. That makes the suggested "one PR, four one-line changes" cheaper than this card assumed.

    2. ⭐ grep -c "^\s*assert(" is the WRONG instrument on three of the four, and using it would land a fresh false number in three files. They do not share an assertion helper:

    file helper raw assert( sites printed
    check-init-service-contract.mjs assert(...) 26 19
    check-kernel-hook-pairs.mjs check(...) 2 10
    check-quick-reference-counts.mjs expect(label, got, want) 0 22
    check-spec-parsed-alias.mjs check(...) 0 18

    ⇒ The runtime ledger is the only honest source in all four. ⚠️ This is the trap this whole class is about, re-armed one level up: a repair that reaches for the obvious counting instrument reproduces the defect it is fixing, and in three of the four files the wrong number would not even be close. ⛔ Whoever takes this must derive from batterySeen, not from a grep — and ⛔ must not "correct the literal" to a grep count.

    Precedent for the shape and for what verification is owed: PR #16669 (Fixes #15231) — positive control (a deliberately added case moves the number), negative control (an unrelated edit does not), deletion control (the floor reds), and the pre-change control showing the old literal not moving. ⛔ No assertion total was pinned as a new floor there, and none should be here.

    Generated by Claude Code

  5. baozhoutao
    added a commit that references this issue on Sep 10, 2026

    fix(ci): honor the duration-unit-keys "placed last" masking contract,…

    22e529d
  6. baozhoutao
    added
    pm:dispatched
    and removed
    pm:queue
    on Sep 10, 2026
  7. baozhoutao
    self-assigned this
    on Sep 10, 2026

  8. baozhoutao commented on Sep 10, 2026

    @baozhoutao
    baozhoutao
    on Sep 10, 2026 – with Claude
    ContributorAuthor

    Claim: session_012GKcPZbMoGq7WPzKLfRBTU · claude/issue-16664-derive-self-test-verdict-counts

    派发(本评论来自 domain:devx 执行 PM 席 · 座位贴 #6023)。assignee 与本条 claim 由本席代 dev 落;dev 继承二者,⛔ 不再发第二条 claim,⛔ 不写 assignee。

    复验 —— 四个残留一个不多一个不少

    origin/main 706ad0fcc4,复跑卡面那条 git grep:

    scripts/check-init-service-contract.mjs:977   console.log('✓ self-test: 19 cases');
scripts/check-kernel-hook-pairs.mjs:560         console.log('✓ self-test: 10 cases');
scripts/check-quick-reference-counts.mjs:928  console.log('✓ check-quick-reference-counts self-test: 22 cases pass.');
scripts/check-spec-parsed-alias.mjs:561       console.log('check-spec-parsed-alias --self-test: 18 assertions passed');
scripts/measure-self-test-floor.mjs:1891      //   … `✓ dispatch-gates self-test: 1415 cases pass.` (cf6b67164, …
对照(无意义变体 `bananas`)= 0

    ⭐ 卡面表里那条「#15231 在飞,本卡排除」的 check-wildcard-fallthrough.mjs 今天已经不在命中列表里 —— #15231 落地了。⇒ 卡面的「恰好四个」在今天的树上依然成立,而且这四个就是它列的那四个。

    判据(⛔ 不是建议,是验收线)

    范围:上面头四个文件,一个不多。

    ⛔ scripts/measure-self-test-floor.mjs:1891 不是实例,不许动。 卡面写清了:那个数字在注释里,是在引用另一个门禁的历史输出,是一条引文,引文本来就该定在时间里。改它是把对的改错。

    每文件的决定,卡面已经把规则给了,不是留给你发明:

    验证方式(卡面点名要的,照做):每个改动都要证两件事——

    1. 加一个用例,这个数会动;
    2. 改一行无关的,这个数不动。
      ⭐ 只证第 1 条不够;⛔ 一个前后读数相同的探针不是对照。

    ⛔ 明确切出去

    ⭐ 卡面点名 check-quick-reference-counts 值得最仔细看:一个自己的主题就是「计数」的门禁,却打印一个关于它自己的、没被推导的数——这是表里最锋利的一个。

    交付纪律

    • worktree-first;⛔ 不 git stash;⛔ 已推的分支不 rebase / amend / force-push。
    • changeset:scripts/** 不发包 ⇒ 大概率 skip-changeset。⛔ 先量后断言,读数写进 PR body。
    • ⭐⭐ skip-changeset 是一个 LABEL,不是 PR body 里的一句话。 量完之后给 PR 打上 skip-changeset 标签;打不上就在报告里说一声,由 PM 补。fix(devx): check-comment-mask-corpus no longer drops argv[0] when --masker is absent #17352 就是「body 里写对了、没打标签、Check Changeset 照样红」。
    • ⭐⭐ Clause-② 写进 PR BODY,不只写在评论里。本卡判 Clause-②: no。行首:**Clause-②**: no。发之前拿 scripts/pm/check-clause2-carriers.mjs 的 readClause2Line() 自验读得到。
    • PR 开 draft,Closes #16664。

    Generated by Claude Code

  9. baozhoutao commented on Sep 10, 2026

    @baozhoutao
    baozhoutao
    on Sep 10, 2026 – with Claude
    ContributorAuthor

    os-dev-report

    {
    "issue": 16664,
    "status": "done",
    "branch": "claude/issue-16664-derive-self-test-verdict-counts",
    "pr": "#17367",
    "pr_url": "#17367",
    "premise_still_valid": true,
    "summary": "Derived all four residual self-test verdict counts named on the card from their existing SELF_TEST_BATTERIES + batterySeen ledger, matching the shape PR #16669 landed for check-wildcard-fallthrough.mjs (Fixes #15231). check-init-service-contract.mjs, check-kernel-hook-pairs.mjs and check-quick-reference-counts.mjs derive from batterySeen.size (the printed number has always meant batteries-that-ran, verified by instrumenting each file: size matched the printed literal, sum(batterySeen.values()) did not). check-spec-parsed-alias.mjs derives from sum(batterySeen.values()) since its roster is a single battery and its line says assertions, not cases. All four print byte-identical output to before the change. scripts/measure-self-test-floor.mjs:1891 was left untouched (it is a citation, not a live verdict line).",
    "tests": "Both verification legs run per file by temporarily mutating on disk, observing node FILE --self-test, then reverting exactly (git diff clean before the next leg, confirmed each time): check-init-service-contract.mjs 19->20 (case added) / stayed 19 (unrelated comment); check-kernel-hook-pairs.mjs 10->11 / stayed 10; check-quick-reference-counts.mjs 22->23 / stayed 22; check-spec-parsed-alias.mjs 18->19 (one extra assertion, no roster change needed since floor is a minimum) / stayed 18. pnpm run check:nul-bytes green (run standalone under the verify lock, exit 0, judged from its own printed OK line, plus again inside the batch). Full local gate surface derived via node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack against this diff: 35 commands (17 pnpm check:, 18 direct node scripts/**), run sequentially under scripts/pm/os-verify-lock.sh (held ~15m31s, shared-box seconds per its own warning, mostly pnpm check:pm-dispatch-gates's ~7min self-test), every exit code captured before any pipe. Reconciled with --ran (command :: exit N lines): dispatch-gates --ran: 35 derived famil(ies) accounted for — 35 run, 0 NOT-MEASURED (a DERIVED zero — all 35 recorded an exit code and none of them is 3). Includes the four directly-matched gates for the touched files: pnpm check:init-service-contract, pnpm check:kernel-hook-pairs, pnpm check:quick-reference-counts, pnpm check:spec-parsed-alias — all exit 0. Root scripts/ is outside every turbo-visible package dependency graph, so local-verification-scope legs ① (dependency-closure build) and ② (own-package test/typecheck) are both empty for this diff by measurement, not by assumption.",
    "mcp_calls": "8 — 5 reads (issue get, issue get_comments, PR #16669 get_files, PR #16669 get, PR #16669 get_labels [errored: PR number resolves as not-an-Issue, as documented]), 1 create_pull_request, 1 issue_write label update, 1 PR #17367 get read-back to confirm the label landed",
    "open_questions": [
    {
    "question": "Two report/footer formats conflict for this card: the per-card dispatch prompt's '## Report back' names a report shape {issue, branch, pr_url, files_changed[], per_file:[...], dropped_out:[...], changeset_decision+reading, skip_changeset_label_applied, clause2, blocked_on[], not_done[]}, and its PR footer instruction says bare _Generated by [Claude Code](https://claude.ai/code)_. The standing os-dev instructions define a different terminal-JSON envelope (issue/status/branch/pr/premise_still_valid/summary/tests/mcp_calls/open_questions/out_of_scope_findings) and require the session-URL footer form for PR bodies specifically (matching the actual precedent, PR #16669, which used the session-URL form). Per the standing instructions' own precedence rule, I followed the standing envelope and session-URL footer, and folded the dispatch's requested per-file/changeset/clause2 fields in as additional keys on this same JSON (see files_changed, per_file, dropped_out, changeset_decision, changeset_reading, skip_changeset_label_applied, clause2 below) rather than silently picking one schema and dropping the other's content.",
    "options": ["A: standing envelope + extra keys (what I did)", "B: dispatch's shape verbatim, dropping the standing envelope", "C: two separate messages"],
    "recommendation": "A, because it satisfies the standing instructions' explicit terminal-message contract while losing none of the dispatch-requested content; PM's mechanical parser can read whichever keys it expects and ignore the rest."
    }
    ],
    "out_of_scope_findings": [],
    "files_changed": [
    "scripts/check-init-service-contract.mjs",
    "scripts/check-kernel-hook-pairs.mjs",
    "scripts/check-quick-reference-counts.mjs",
    "scripts/check-spec-parsed-alias.mjs"
    ],
    "per_file": [
    {
    "file": "scripts/check-init-service-contract.mjs",
    "before": "console.log('✓ self-test: 19 cases');",
    "after": "console.log(✓ self-test: ${batterySeen.size} cases);",
    "moves_on_added_case": "19 -> 20 cases (added a 20th battery + assertion, reverted after)",
    "static_on_unrelated_edit": "stayed 19 cases (added/reverted an unrelated comment above const ROOT = ...)"
    },
    {
    "file": "scripts/check-kernel-hook-pairs.mjs",
    "before": "console.log('✓ self-test: 10 cases');",
    "after": "console.log(✓ self-test: ${batterySeen.size} cases);",
    "moves_on_added_case": "10 -> 11 cases (added an 11th battery + assertion, reverted after)",
    "static_on_unrelated_edit": "stayed 10 cases (added/reverted an unrelated comment above function selfTest())"
    },
    {
    "file": "scripts/check-quick-reference-counts.mjs",
    "before": "console.log('✓ check-quick-reference-counts self-test: 22 cases pass.');",
    "after": "console.log(✓ check-quick-reference-counts self-test: ${batterySeen.size} cases pass.);",
    "moves_on_added_case": "22 -> 23 cases (added a 23rd battery + assertion, reverted after)",
    "static_on_unrelated_edit": "stayed 22 cases (added/reverted an unrelated comment above function selfTest())"
    },
    {
    "file": "scripts/check-spec-parsed-alias.mjs",
    "before": "console.log('check-spec-parsed-alias --self-test: 18 assertions passed');",
    "after": "const assertionsRun = [...batterySeen.values()].reduce((total, count) => total + count, 0); console.log(check-spec-parsed-alias --self-test: ${assertionsRun} assertions passed);",
    "moves_on_added_case": "18 -> 19 assertions (added one more check(...) call to the existing single battery, no roster change needed since the roster value is a floor not an equality, reverted after)",
    "static_on_unrelated_edit": "stayed 18 assertions (added/reverted an unrelated comment above function selfTest())"
    }
    ],
    "dropped_out": [],
    "changeset_decision": "skip-changeset",
    "changeset_reading": "All four files are under root scripts/, which is not inside any pnpm-workspace.yaml package glob (packages/    , packages/apps/, ..., apps/, examples/). The only package.json referencing any of the four filenames is the root manifest itself, which is "private": true. A repo-wide grep for symbols unique to these files (SELF_TEST_BATTERIES, check-quick-reference-counts) inside any existing dist/**/.js returned zero hits. Nothing published moves.",
    "skip_changeset_label_applied": true,
    "clause2": "no",
    "blocked_on": [],
    "not_done": []
    }

    Generated by Claude Code

  10. baozhoutao
    closed this as completedin #17367on Sep 10, 2026
  11. baozhoutao
    removed
    pm:dispatched
    on Sep 10, 2026
  12. baozhoutao
    removed their assignment
    on Sep 10, 2026
  13. akarma-synetal
    added a commit that references this issue on Sep 17, 2026

    fix(ci): quote workflow step name YAML truncates at " #", add guard (o…

    c632a8f
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    domain:devxfindingpriority:p3tooling

    Type

    Task

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions