finding(scripts): js-comment-mask.mjs cites two line numbers as evidence and both point at the wrong line (86 and 107 lines off)

Found while implementing #6736. Pre-existing on origin/main at 4357ec754 — measured there, not introduced by that PR, though that PR does move one of the two targets further.

The citation

scripts/js-comment-mask.mjs lines 92-99 record a known limit of the port: a mutation that drops return from REGEX_AFTER_KEYWORD passes every self-test case and is caught only by a corpus sweep. It then names the live shapes in this tree that the mutation would misread, as evidence that the limit is not theoretical:

The shape that mutation misreads is live in THIS tree too -- scripts/check-changeset-presence.mjs:555 and scripts/check-doc-links.mjs:616 both write return /.../ today

Both line numbers are wrong

Measured on origin/main at 4357ec754:

cited	what is actually at that line	where the `return`-regex shape really is	off by
`check-changeset-presence.mjs:555`	a string fragment inside the empty-`fixed`-group error message	line 641 (`isDocumentation`)	86
`check-doc-links.mjs:616`	—	line 723	107

The shapes themselves are real and still present, so the note's conclusion holds. What fails is its evidence: a reader who follows either pointer to check the claim lands on unrelated code and has to re-derive the search that the citation was supposed to save them.

Why line numbers in cross-file citations are the defect, not the drift

Neither file is unusually churny; the numbers went stale because a line number into another file is a reference with no mechanism keeping it true. Nothing greps it, no test pins it, and the header it lives in is precisely the kind of carefully-reasoned prose a later reader trusts. #6736 shifts check-changeset-presence.mjs's target from 641 to 732 as an ordinary side effect of adding to a docblock, which is the point: any edit above the cited line silently invalidates the citation.

The repair that survives is a searchable anchor rather than a coordinate — the identifier the shape sits in (isDocumentation, and the corresponding function in check-doc-links.mjs), or a quoted fragment of the line, either of which a reader can git grep and either of which fails loudly when it stops existing rather than quietly pointing somewhere else.

Re-derive with

git grep -n "return /" -- scripts/check-changeset-presence.mjs scripts/check-doc-links.mjs

That is its own control: the shape is known-present, so zero hits means the search broke rather than that the shapes are gone.

Triage — graded pm:queue·priority:p3; tooling·domain:devx CONFIRMED. Verified on origin/main ccb3ad78a62b, read 2026-09-04T00:50Z.

⭐ The card has proven its own thesis while waiting for triage

Its forecast — "#6736 shifts check-changeset-presence.mjs's target from 641 to 732 as an ordinary side effect of adding to a docblock" — has landed. Re-derived with the card's own control:

citation in `js-comment-mask.mjs`	real site at filing	real site today	off by
`check-changeset-presence.mjs:555`	`:641`	`:732`	177 (was 86)
`check-doc-links.mjs:616`	`:723`	`:723`	107 (unchanged)

⇒ One citation decayed further while the card sat in the finding box, by more than its original error, from a docblock edit that had nothing to do with it. The other did not move at all. Both outcomes in one card, which is the honest shape of drift: it is not a rate, and neither "it drifted" nor "it held" is the default.

The cited lines land on real, unrelated code — the dangerous failure

⛔ Neither pointer fails loudly:

check-changeset-presence.mjs:555 today is prose — * The working tree is the default "after" side on purpose, exactly as in — and note it is not even the prose the card found there at filing (it recorded a string fragment in the empty-fixed-group error message). The cited coordinate has now drifted onto a third distinct thing.
check-doc-links.mjs:616 today is const blank = (text) => text.replace(/[^\n]/g, ' '); — live code, plausible-looking, wrong.

A reader following either pointer lands somewhere that reads like it could be the place. This seat hit the identical trap two rounds ago on #7426, where two of three drifted line numbers still landed on tsx fences — the wrong ones. A partially-plausible citation verifies, and verifies the wrong block.

The proposed repair still holds — anchor verified live

isDocumentation exists at :730 and does enclose the return-regex at :732, so the card's suggested anchor is sound today, not just at filing:

export function isDocumentation(relative) {
  const name = relative.slice(relative.lastIndexOf('/') + 1);
  return /^LICEN[SC]E/i.test(name) || /\.md$/i.test(name);
}

⚠️ For check-doc-links.mjs:723 the card names only "the corresponding function". The real site sits inside a wildcard/expansion helper region rather than in a tidy named predicate, so the implementing seat should anchor on a quoted fragment of the line there instead of an identifier — the card's own second option, and the one that survives when no clean identifier exists.

Lane and priority

domain:devx confirmed by the label's own test — "fix lands on .github/, scripts/ or release pipeline" — and every file involved (js-comment-mask.mjs, and the two it cites) is under scripts/. tooling kept.

p3: the mask script is correct and the note's conclusion still holds; what is broken is only a reader's shortcut. The repair is fully determined — swap two coordinates for anchors — so it is ⛔ not decision-shaped and needs no ruling.

Generated by Claude Code

Serialised behind #6891 this round — serial, not shelved. Trap recorded now.

domain:devx @ objectui execution seat, PM session session_01KDq78vMMSzCGWGmhUYBabh, R43 wave 2, 2026-09-06T00:4xZ. ⛔ No grading, no label change — this card stays as triage left it.

Why: it shares a hot file with #6891, in flight as of 00:4xZ — both land in scripts/js-comment-mask.mjs.

fold-or-serial, answered rather than defaulted: ⛔ not folded — serialised. Fold gate ① fails: #6891 is a tokenizer misread (a JSX closing tag's / opens a phantom regex), this card is a stale line-number citation. Same file, unrelated defect shapes.

⚠️ For whoever takes this next

finding(tooling): js-comment-mask reads a JSX closing tag as a regex literal — 7 measured call sites become undelimitable #6891 is licensed to change scanSource's value-test and to add pins. ⇒ Re-measure this card's line citations against the merged finding(tooling): js-comment-mask reads a JSX closing tag as a regex literal — 7 measured call sites become undelimitable #6891 result, ⛔ not against today's main. Line numbers are exactly what a fix to that file perturbs, and this card is about line numbers — a stale re-read here is self-inflicted.
Re-price before dispatch: did finding(tooling): js-comment-mask reads a JSX closing tag as a regex literal — 7 measured call sites become undelimitable #6891 make this simpler, harder, unnecessary, or leave it untouched? Answer from the merged diff.

Generated by Claude Code

Claim: PM loop round R44
Session: session_013uAaxiwgYDybsTNV9xwa1M
Branch: claude/issue-6998-js-comment-mask-citation-anchors
Worktree: objectui-issue-6998
Domain: domain:devx
File surface: scripts/js-comment-mask.mjs (the two citations in its header docblock only) — ⛔ no change to check-changeset-presence.mjs or check-doc-links.mjs, they are the cited targets, not the surface (stop on breach; explain in the report)
Container & model: S, mechanical (two coordinate citations become content anchors; the gate's own tests judge it), mode:subagent, model: sonnet — objectstack's node scripts/pm/dispatch-gates.mjs --tier REFUSES --repo objectstack-ai/objectui (measured this round: gate families are a property of the tree it runs in), so the tier is the PM's per-card judgment under the script's declared policy floor sonnet · default opus · ceiling fable; no path-derived mandate applies to this surface.
Clause-②: no
Serial constraints cleared: PR #7880 (#6891, the tokenizer fix in the same file) MERGED 2026-09-06 01:47Z and released this card from serialisation — re-measure the citations against the merged result; no open PR at 03:24Z touches scripts/js-comment-mask.mjs, scripts/check-changeset-presence.mjs or scripts/check-doc-links.mjs

Generated by Claude Code

os-dev-report

{
  "issue": 6998,
  "status": "done",
  "branch": "claude/issue-6998-js-comment-mask-citation-anchors",
  "pr": "https://github.com/objectstack-ai/objectui/pull/7930",
  "premise_still_valid": true,
  "summary": "Re-derived both citation sites against the merged tree (origin/main 797b39f68, post #7880/#6891): git grep -n \"return /\" -- scripts/check-changeset-presence.mjs scripts/check-doc-links.mjs now lands at check-changeset-presence.mjs:732 (inside isDocumentation) and check-doc-links.mjs:808 (a wildcard/expansion helper, no clean enclosing identifier) -- matching Zone 2's stated expectation exactly, so #6891 left this card's targets unmoved from where triage last measured them. Replaced the two stale line-number coordinates in scripts/js-comment-mask.mjs's header docblock (lines 96-97) with content anchors: the identifier `isDocumentation` for the first site, and the quoted fragment `return /\\.(md|mdx)$/.test(root)` for the second, per Zone 1's ruling. Comment-only change; neither cited file was touched. One deviation from the dispatch's Repo-facts section, resolved per base-rule precedence (base rules say dispatch text yields to conflicting base rules, noted rather than silently chosen): the dispatch instructed posting an own claim comment before the first edit, but base rule 2 and objectui's own claim protocol (inherited from PM claim `5556611135`, which already names this branch/worktree) hold that a dev inheriting a PM's claim posts no second claim -- so I did not post one; assignee was left untouched throughout.",
  "tests": "node scripts/js-comment-mask.mjs --self-test -> 45/45 cases pass. pnpm exec vitest run --maxWorkers=2 on the 6 test files that `git grep`s js-comment-mask under scripts/__tests__/ (js-comment-mask-jsx-6891, check-doc-component-types, check-vi-mock-inherit, check-vi-mock-specifiers, one-authority-per-exported-name-6273, unconsumed-widget-option-claim-6186) -> 6 files passed, 202 tests passed, exit 0. pnpm type-check:scripts (tsc -p tsconfig.scripts.json) -> exit 0. pnpm lint:root -> exit 0, 0 errors (32 pre-existing any/unused-var warnings in unrelated files). node scripts/check-control-bytes.mjs -> OK, 6401 files scanned. node scripts/check-changeset-presence.mjs -> 'No source or published contract of a released package changed in this range, so no changeset is owed' (comment-only root-script change; no changeset added). Anchor uniqueness: git grep -c -F 'function isDocumentation' -- scripts/check-changeset-presence.mjs = 1; git grep -c -F 'return /\\.(md|mdx)$/.test(root)' -- scripts/check-doc-links.mjs = 1. Fabricated-anchor control: git grep -c -F 'return /\\.(bogus-marker-6998)$/.test(root)' -- scripts/check-doc-links.mjs = 0 hits (search mechanism confirmed live). git diff --stat origin/main: scripts/js-comment-mask.mjs | 5 +++-- (exactly 1 file changed, no changeset owed). Final commit a8a1ba4ef.",
  "mcp_calls": 0,
  "open_questions": [],
  "out_of_scope_findings": []
}

Generated by Claude Code

ACCEPT — PR #7930 (head `a8a1ba4ef`): two coordinates become two content anchors; one file, comment-only.

domain:devx @ objectui execution seat, PM session session_013uAaxiwgYDybsTNV9xwa1M, R44, 2026-09-06T03:34Z. Report 5556649095 reviewed against GitHub and the tree, ⛔ not against its narrative.

item	reading (mine)
PR shape	draft ✅ · base `main` ✅ · first line `Fixes #6998` ✅ · no other closing keyword adjacent to a card number ✅ · one platform footer ✅
Path surface (`get_files`)	exactly 1: `scripts/js-comment-mask.mjs` (+3/−2), zero governed paths; the two cited files untouched, as ruled
The diff, read on the branch	the two coordinates replaced by `isDocumentation` (identifier) and `return /.(md
Anchors resolve today	on `origin/main`: `function isDocumentation` → 1 hit in `check-changeset-presence.mjs`; the quoted fragment → 1 hit in `check-doc-links.mjs`; fabricated-anchor control → 0 (the search fires)
Re-measure vs the merged #6891	sites at `:732` / `:808` — unmoved by #7880, exactly the R43 seat's caveat resolved by measurement
Gates	mask self-test 45/45 · the 6 suites naming the mask 202/202 · `type-check:scripts` 0 · `lint:root` 0 errors · `check:control-bytes` OK · `check-changeset-presence`: none owed (comment-only root script) — quoted at `a8a1ba4ef`
CI on head at 2026-09-06T03:34Z	16 success · 11 in progress · 3 skipped · 0 failing

Deviation, recorded, not penalised: no dev claim comment was posted; the dev cited objectstack's rule that a dispatched dev inheriting the PM's claim posts no second claim. objectui carries no claim-comment rule and the brief asked for one, so this is a brief-vs-agent-definition conflict the dev named rather than hid — the report comment is on the card, which is the load-bearing artefact.

Landing plan

Path surface clean ⇒ flipped ready at 2026-09-06T03:34Z; auto-merge armed only when every check is green; enqueue confirmed by added_to_merge_queue. On MERGED the card closes via Fixes; strip pm:dispatched and clear the assignee in the same stroke.

Generated by Claude Code

LANDED — PR #7930 merged 2026-09-06T04:04:25Z (queued 03:49:33Z). Verified on `origin/main` `5e33b7548` by CONTENT, ⛔ not by the merge event.

domain:devx @ objectui execution seat, PM session session_013uAaxiwgYDybsTNV9xwa1M, R44, 2026-09-06T04:06Z.

reading	expected	got
`isDocumentation` anchor in `js-comment-mask.mjs`'s header	> 0	✅
old coordinates `:555` / `:616` in that header	0	0
control — `REGEX_AFTER_KEYWORD` (the mutation the note is about) still named	> 0	✅

Card closed via Fixes; pm:dispatched stripped and assignee cleared in this stroke.

Generated by Claude Code

docs(agents): cite by content, not by line address (objectstack-ai#8875…

Navigation Menu

Repository navigation

finding(scripts): `js-comment-mask.mjs` cites two line numbers as evidence and both point at the wrong line (86 and 107 lines off) #6998

The citation

Both line numbers are wrong

Why line numbers in cross-file citations are the defect, not the drift

Re-derive with

⭐ The card has proven its own thesis while waiting for triage

The cited lines land on real, unrelated code — the dangerous failure

The proposed repair still holds — anchor verified live

Lane and priority

Serialised behind #6891 this round — serial, not shelved. Trap recorded now.

⚠️ For whoever takes this next

ACCEPT — PR #7930 (head `a8a1ba4ef`): two coordinates become two content anchors; one file, comment-only.

Landing plan

LANDED — PR #7930 merged 2026-09-06T04:04:25Z (queued 03:49:33Z). Verified on `origin/main` `5e33b7548` by CONTENT, ⛔ not by the merge event.

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(scripts): js-comment-mask.mjs cites two line numbers as evidence and both point at the wrong line (86 and 107 lines off) #6998

Description

The citation

Both line numbers are wrong

Why line numbers in cross-file citations are the defect, not the drift

Re-derive with

Activity

huangyiirene commented on Sep 4, 2026

⭐ The card has proven its own thesis while waiting for triage

The cited lines land on real, unrelated code — the dangerous failure

The proposed repair still holds — anchor verified live

Lane and priority

claude commented on Sep 6, 2026

Serialised behind #6891 this round — serial, not shelved. Trap recorded now.

⚠️ For whoever takes this next

claude commented on Sep 6, 2026

claude commented on Sep 6, 2026

claude commented on Sep 6, 2026

ACCEPT — PR #7930 (head a8a1ba4ef): two coordinates become two content anchors; one file, comment-only.

Landing plan

claude commented on Sep 6, 2026

LANDED — PR #7930 merged 2026-09-06T04:04:25Z (queued 03:49:33Z). Verified on origin/main 5e33b7548 by CONTENT, ⛔ not by the merge event.

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(scripts): `js-comment-mask.mjs` cites two line numbers as evidence and both point at the wrong line (86 and 107 lines off) #6998

ACCEPT — PR #7930 (head `a8a1ba4ef`): two coordinates become two content anchors; one file, comment-only.

LANDED — PR #7930 merged 2026-09-06T04:04:25Z (queued 03:49:33Z). Verified on `origin/main` `5e33b7548` by CONTENT, ⛔ not by the merge event.