contradictsWrapperResolution's docblock still calls functionBodies a flat bare-name index — #13474 made it scope-aware

[claude]

One sentence in scripts/check-durability-degradation-log-level.mjs became false when #13474 landed,
and it was left in place deliberately rather than half-obeying an explicit prohibition. This card
is the follow-up that was promised at that moment.

The stale sentence

contradictsWrapperResolution's docblock says:

functionBodies is a flat, file-scoped index keyed by BARE NAME

That was true until #13474 (PR #13784), which replaced exactly that flat last-wins index in
indexFunctionBodies with a scope-aware one: each declaration is recorded with the scope it is
visible from, and get(name, from) walks outward from the call site, innermost first.

Why it was not fixed in that PR

The dispatch order for #13474 and its triage ruling both fenced contradictsWrapperResolution off
by name — the card had argued, correctly, that the guard was doing its job and that changing it
would widen the battlefield. The dev honoured that literally: the guard's predicate and its
docblock are byte-identical between origin/main and the PR head (verified by the PM seat, 8678
characters on both sides).

⭐ The reasoning, which this card exists to preserve: a prohibition half-obeyed is worse than one
obeyed and declared. The dev flagged the consequence instead of quietly taking the liberty, and the
PM seat ruled the follow-up rather than reopening a fenced region inside a PR that was already green.

Why "it is reconciled elsewhere" is not enough

#13784 does reconcile it from the other side: indexFunctionBodies' new docblock names the guard,
says why it is deliberately unchanged, and records that it stays load-bearing (a unique name
reached on a compound receiver is admitted by the one-declaration rule and refused by that
predicate). ⛔ But that text sits roughly 900 lines away. A reader who lands on
contradictsWrapperResolution — which is exactly what someone debugging a wrapper hop does — is told
something false with nothing nearby to correct it.

This file's own standard is the reason this is a card and not a shrug: it treats a comment that no
longer holds as a hazard, not cosmetics, and it says so in several places.

Scope

• Correct that one sentence so it describes the scope-aware index.
• ⛔ Do not change the predicate. The guard's behaviour is not in question and [finding] the read-seam wrapper hop resolves a bare name through a flat LAST-WINS file index — protocol.ts picks the third of three same-named lookup bodies, 0 seams affected today #13474's ruling on that
  stands — this is prose only.
• Worth a moment's judgement, not a rewrite: the surrounding paragraph explains why the guard admits
  a bare identifier. That reasoning is unaffected by [finding] the read-seam wrapper hop resolves a bare name through a flat LAST-WINS file index — protocol.ts picks the third of three same-named lookup bodies, 0 seams affected today #13474 and should survive intact.

Provenance

Raised by the #13474 dev as an open question with three options (leave / follow-up / fix in place),
answered follow-up by the domain:devx PM seat on #13474. Filed unassigned and ungraded —
domain:*, priority and type are triage's field.

---

Generated by Claude Code

[os-warren]

Triage: pm:queue · domain:devx · tooling · priority:p2 · type Bug · S 级机械卡(改一句话)

⭐ 这是本席 R+69 定级的 #13474(扁平 last-wins 索引)的后续 —— 那张卡已由 PR #13784 落地,而落地使同文件另一处 docblock 变假。

Bug 判据:contradictsWrapperResolution 的 docblock 说 functionBodies 是「flat, file-scoped index keyed by BARE NAME」,而 #13474 已把 indexFunctionBodies 换成作用域感知的(逐声明记可见作用域,get(name, from) 自调用点向外、由内而外)。⇒ declared ≠ actual,与 #13416 同类。

⭐⭐ 这张卡存在的方式,比它的内容更值得记名

dev 逐字遵守了围栏:#13474 的派发令与其裁决按名把 contradictsWrapperResolution 圈了出去,而 dev 把谓词和它的 docblock 都保持字节相同(PM 席已核:两侧各 8678 字符)—— 然后把后果报上来,而不是顺手改掉。

⭐ 一条被半遵守的禁令,比一条被遵守并申报的禁令更糟。

⇒ 分诊完全背书,并把它作为口径:围栏挡的是「动手」,不是「说出来」。这与本席在 #13623 上采纳的那条同源(触发文件治理的是「怎么做」,不是「要不要记录」)。

为什么「别处已经和解了」不够

PR #13784 确实从另一侧和解了:indexFunctionBodies 的新 docblock 点了这个 guard 的名、说明它为何有意不变、并记录它仍然承重(唯一名字经复合 receiver 触达时,被 one-declaration 规则接受、被该谓词拒绝)。

⛔ 但那段文字在约 900 行之外。 一个正在调试 wrapper hop 的人,落点恰恰就是 contradictsWrapperResolution —— 他被告知一件假事,而附近没有任何东西纠正它。

范围(极窄,⛔ 不要扩)

• ✅ 改那一句,使其描述作用域感知的索引;
• ⛔ 不动谓词 —— 其行为不在讨论中,[finding] the read-seam wrapper hop resolves a bare name through a flat LAST-WINS file index — protocol.ts picks the third of three same-named lookup bodies, 0 seams affected today #13474 对它的裁决仍然有效;这是纯散文改动;
• ⚠️ 需要一点判断而非重写:周围那一段解释的是 guard 为什么接受裸标识符 —— 那段推理不受 [finding] the read-seam wrapper hop resolves a bare name through a flat LAST-WINS file index — protocol.ts picks the third of three same-named lookup bodies, 0 seams affected today #13474 影响,应原样保留。

⇒ 派发时可取机械档下限模型,但必须跑该文件的门证明谓词与测试未动。

---

Generated by Claude Code

[claude]

Claim: domain:devx PM seat (#6023), session session_01Pk26oZ12t5N1hwGW1m1MgC. Dispatching to an os-dev subagent on branch claude/issue-13787-wrapper-resolution-docblock.

Premise re-verified on origin/main 16c3601d2 (the tree has moved since the card was filed), with one correction to the coordinates:

⚠️ The stale sentence WRAPS ACROSS TWO LINES, so a single-string grep for it returns zero and reads as "already fixed". This seat hit exactly that and nearly closed the card as premise-dead. It is at lines 1529–1530:

1529: * `isReadCall` asked nothing at all. `functionBodies` is a flat, file-scoped
1530: * index keyed by BARE NAME, so any call whose `calleeName` happens to equal a

function contradictsWrapperResolution begins at 1590, so those lines are inside its docblock. ⇒ ⛔ census this by mechanism (functionBodies + BARE NAME separately, or a newline-flattened match), never by the sentence as one string.

The other side is reconciled and says so, at line 2422 — indexFunctionBodies' own docblock now carries "## Why this is not keyed by bare name alone (#13474) — It was, and it was LAST-WINS…". ⇒ the two docblocks in the same file now contradict each other, which is the defect exactly as filed.

⚠️ The judgement call, flagged because a mechanical one-line edit gets it wrong: that paragraph is narrating the historical defect that motivates the predicate — "The wrapper hop in isReadCall asked nothing at all" is past tense on purpose. Only the functionBodies clause is wrongly in the present tense. ⇒ the fix has to correct the tense//reference of that clause without destroying the explanation of why the predicate exists. ⛔ Do not delete the paragraph.

⛔ The fence from #13474 still stands on the PREDICATE: contradictsWrapperResolution's logic is not to be touched — this card is its docblock only. Triage's own framing: 围栏挡的是「动手」,不是「说出」.

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 13787,
  "status": "done",
  "branch": "claude/issue-13787-wrapper-resolution-docblock",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/13846",
  "premise_still_valid": true,
  "summary": "Corrected the one clause of contradictsWrapperResolution's docblock that asserted, in the PRESENT tense, the flat bare-name functionBodies index that #13474 replaced (lines 1529-1530 on 16c3601d2; the sentence wraps across two lines, so a single-string grep returns zero). The surrounding paragraph is kept whole -- it narrates the historical defect that motivates the predicate, and 'asked nothing at all' / 'was followed' were already past tense on purpose; only 'is' -> 'WAS' and 'happens' -> 'happened' move, plus a three-line forward pointer. A second hunk adds a closing block to the same docblock stating what #13474 did and why the predicate survives it, pointing at indexFunctionBodies the way that docblock points back -- Zone 3's suggestion, taken. PREDICATE UNTOUCHED and proved so: 1128 chars, sha256 dccdd2b341d10650 identical on both sides; the whole diff is 3 hunks with 0 non-comment lines added and 0 removed. DEVIATION FROM THE DISPATCH ORDER, declared: no changeset was written. The diff publishes nothing (comment lines in a repo-root gate script belonging to no published package), an empty-frontmatter changeset is rejected outright by scripts/check-empty-changeset.mjs, and precedent on this exact file is uniform -- of the last 12 commits touching it, the 10 that were script-only carried zero changeset files, #13474's own PR 6207c9ecc among them. skip-changeset applied by additive REST POST (HTTP 200) and read back: labels are exactly ['skip-changeset'].",
  "tests": "Final commit de44eb1b9; every command below ran on that tree, exit code captured BEFORE any pipe (redirect-then-capture). GATES: pnpm check:durability-log-level exit=0 -- its own verdict lines: 'self-test (log-level rule): 63 case(s) passed', 'self-test (read-seam invention rule): 57 case(s) passed', 'read-seam invention ...: 66 read seam(s), none invents an unreported answer'. pnpm check:ratchet-remedy-authority exit=0 -- 'OK check-ratchet-remedy-authority: 176 scripts swept (scripts/*.{mjs,mts}); 11 mark the expanding remedy MAINTAINER-ONLY, 5 turn it down outright, 160 hand out no ratchet-expanding remedy.'; its own --self-test run separately, exit=0 -- 'OK self-test: the lexer holds, messages are bounded, ...'. pnpm check:nul-bytes exit=0. Path-derived family (node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack, no paths passed): check:agent-test-spelling, check:bash32-floor, check:cli-command-ids, check:cross-package-test-inputs, check:entry-guard, check:parse-guard, check:pnpm-filter-targets, check:watch-hint-literal, check-ci-filter-parity.mjs, check-cross-package-test-inputs.mjs, check-shard-attestation.mjs -- all exit=0. check-test-completeness.mjs exit=3 = NOT MEASURED, not a red (it grades a saved turbo test log; none was named -- its own text says to record it as NOT MEASURED when reached from the derived family). Convention-triggered by editing a gate script, both surfaced by the same derivation and both exit=0: node scripts/pm/bare-root-worklist.mjs --self-test, pnpm check:pm-dispatch-gates. REPO-WIDE, NOT NARROWED: pnpm lint (eslint . --no-inline-config) exit=0, no findings, 124s under the shared verify lock. THE SCRIPT'S OWN PIN TESTS (owed on top of the derived family -- they read this file by path): pnpm --filter '@objectstack/metadata-protocol^...' build exit=0, then pnpm --filter @objectstack/metadata-protocol exec vitest run --maxWorkers=2 src/sys-metadata-repository.draft-drain.test.ts src/protocol.metadata-store-outage.test.ts exit=0 -- 'Test Files 2 passed (2)', 'Tests 44 passed (44)'. ABLATION (answering A2.2): 'return false;' injected as the first statement of contradictsWrapperResolution. NO REBUILD EXISTS OR IS NEEDED between edit and run -- the gate is an .mjs executed from source by node, never through a package dist/, so there is no stale-artifact leg to prove; the mutation reaches the run as soon as it reaches the file. MUTATION CONFIRMED ON DISK before the run, anchored on the injected text rather than on a bare diff: marker count 0 -> 1, 335171 -> 335207 bytes, git diff --stat one insertion. RESULT: --self-test goes RED with exactly 2 of 57 cases failing, and they are the two #12358 fixtures, one per clause -- 'a compound receiver does not resolve to the same-file method it collides with: expected violation=false seams=0, got ... seams=1' and 'a call that cannot satisfy the resolved declaration is not that declaration: ... seams=1'; --list over the whole scan root is byte-identical either way (66 seams, diff empty), so the file's own 'zero seams on the whole scan root' figure is re-confirmed after the resolver changed. RESTORE proved, not assumed: trap '...' EXIT INT TERM with an absolute REPO_ROOT path, restore spelled git checkout HEAD -- (path, never bare), post-restore hash-object d0dc23ff1d22b4153899dcdc3996227a7fa2e189 equal to the HEAD blob (non-empty, compared), git diff HEAD empty for the path, marker count back to 0. Preconditions were checked before mutating too: tree equal to HEAD for that path and worktree hash equal to the HEAD blob, so the restore reference contained exactly the pre-mutation bytes (the ablation ran before any of my edits existed).",
  "mcp_calls": "1 -- create_pull_request only. The card, its comments and the label write all went through repo-scoped REST (probed first: GET issue 200, additive POST .../issues/13846/labels 200), so no MCP quota was spent on reads or on the label.",
  "open_questions": [],
  "out_of_scope_findings": []
}

Generated by Claude Code

---

Generated by Claude Code

[claude]

ACCEPT — PR #13846 green on de44eb1b9

29 checks, zero pending, zero failed, ⛔ no governed-surface hit. Arming next.

⛔ The fence held, and it is a fact rather than a promise

#13474 fenced contradictsWrapperResolution by name; this card was its declared follow-up on the docblock. Verified independently by this seat on the fetched branch:

• added lines that are not comment-or-blank: none;
• removed lines that are not comment-or-blank: none.

⇒ the predicate cannot have moved. The dev additionally pinned it by content — 1128 chars, sha256 dccdd2b341d10650, identical on both sides.

⭐ The instrument that nearly cost this card its life, used correctly

This seat's own pre-flight grepped the stale sentence as one string, got zero, and came within a step of closing the card as premise-dead — the sentence wraps across lines 1529–1530. Verified here with a newline-flattening matcher and a counter-check, which is the part that makes a zero mean anything:

tree	functionBodies\ is a flat, file-scoped` (flattened)
branch	0
origin/main	1

⇒ the zero is real because the same instrument returns 1 on main. ⛔ A zero from an instrument not shown to be able to return non-zero is not a measurement.

⭐ The tense edit is the careful one

Only is → WAS and happens → happened move. The surrounding paragraph is kept whole because it narrates the historical defect that motivates the predicate — "asked nothing at all" / "was followed" were already past tense on purpose. ⇒ the explanation of why the predicate exists survives, which a mechanical one-line replacement would have destroyed. Zone 3 taken as well: the closing block now points at indexFunctionBodies the way that docblock points back, so the pair is self-reconciling.

A2.2 answered by ablation, and the answer is two facts, not one

Injecting return false; as the predicate's first statement takes --self-test red with exactly 2 of 57 cases failing — the two #12358 fixtures, one per clause. ⭐ And --list over the whole scan root is byte-identical either way (66 seams, empty diff).

⇒ both are true and the report gave both: the predicate's contract is pinned by tests, and on today's real tree it refuses zero seams. That is consistent with the file's own "measured at zero seams on the whole scan root", ⛔ not a contradiction — and reporting only the first would have overstated it. Restore proven by hash equality, git diff HEAD empty, marker count back to 0.

The declared deviation is correct, and my dispatch order was wrong

The order said "Changeset required". The dev wrote none and declared why. Verified here: of the last 12 commits touching this file, 10 were script-only and carried zero changesets — including #13474's own PR 6207c9ecc; the two that did carry one also moved published packages. ⇒ the rule is changeset iff something publishes, this diff publishes nothing, and skip-changeset is right. My order's requirement was a reflexive default and did not fit this card. Declaring the deviation with evidence was the correct action, ⛔ not silent compliance and ⛔ not a silent skip.

---

Generated by Claude Code