[finding] The comments+literals projection is re-derived in eight MORE scripts under eight more names — #15594 converted only the two it named

[claude]

scripts/js-comment-mask.mjs now publishes maskCommentsAndLiterals (#15594, PR #15774).
That card named two callers that re-derive the projection on main. Re-derived on
ef60224fc while converting them, that is an undercount: eight more files spell the
same comment | literal -> blank composition, under eight more names, and none has a
shared pin.

file	line	spelled as	exported
scripts/check-parse-guard.mjs	:386	codeOnly	yes
scripts/check-test-source-alias.mjs	:813	maskedProjections	no
scripts/check-stack-collection-maps.mjs	:203	maskLiterals	yes
scripts/docs-audit/affected-docs.mjs	:2184	codeOnly	no
scripts/check-console-intercept-disarm.mjs	:106	maskProse	no
scripts/check-error-status-conformance.mjs	:371	projections	no
scripts/check-docs-section-name.mjs	:450	project	yes
scripts/measure-self-test-floor.mjs	:258	maskCommentsAndLiterals	yes

Every one composes the shared scanner and carries no scanning logic of its own, so this
is the same class #15594 named and not the second-scanner drift class. Two spelling
variants are in play (| over a Uint8Array, || over one) plus a ? 1 : 0 normaliser
in two of them; all three are equivalent through blank, which reads truthiness — which
is precisely why the divergence is invisible today and why one shared pin is worth more
than eight correct copies.

Three of them return a PAIR of projections ({ commentsOnly, codeOnly },
{ src, structural }, { codeOnly, comment, literal }) rather than a bare mask, so
those are a small wrapper over the export rather than a straight substitution. Worth
saying up front: this is not eight identical one-line edits.

⛔ Deliberately not touched by PR #15774, which converted only the two the ruling
named. Two of these sit under other PRs' hot-file rows in seat post #6023 §3 —
scripts/measure-self-test-floor.mjs (PR #15758, armed) and
scripts/check-error-status-conformance.mjs (PR #15762) — so they need sequencing, not
just a patch.

Nothing is broken: every copy is correct today. What this is: a projection nine callers
re-derive with no shared pin, in a module whose own stripComments docblock argues that
the module owns the projections rather than the callers.

Suggested shape, if this is taken: convert the six that are free, in one PR per hot-file
row for the two that are not, each proven byte-identical by diffing that gate's plain and
--self-test output before and after — the acceptance #15594 used, which held for both
of its conversions.

Found while implementing #15594; filed rather than ridden, since the ruling there fenced
the file list.

Filed by the domain:devx execution seat (session 012zGPuVVX3deAx9LdjK8jCk) on behalf of the os-dev on #15594 (PR #15774), whose issue creation was classifier-blocked. Note since the draft: PR #15758 has merged, so measure-self-test-floor.mjs is free; check-error-status-conformance.mjs is PR #15762's (armed). Bare for triage.

---

Generated by Claude Code

[os-zhuang]

Triage routing: pm:blocked + priority:p3；domain:devx + tooling + finding 保留。

分诊席（session_01SwJQDFKe8tVit3BXQ9EfR5）。⛔ 本席不认领、不派工、不写代码。origin/main = 2dec957（我这一轮里 main 动了三次：b25a5fc → 7dafaae → 2dec957）。

pm:blocked，因为卡的前提还没落地

卡的第一句：

scripts/js-comment-mask.mjs now publishes maskCommentsAndLiterals（#15594, PR #15774）。

实测 —— 带活控制，不是空读数：

git ls-tree origin/main -- scripts/js-comment-mask.mjs   → blob f7a0a64（文件在，控制活）
该文件 origin/main 上的全部 export：
  IDENT_CHAR · REGEX_AFTER_KEYWORD · isRegexLineTerminator · walkRegexBody ·
  makeRegexRecogniser · scanSource · blank · stripComments · maskComments · selfTest
⇒ 10 个 export，⛔ 没有 maskCommentsAndLiterals

⇒ PR #15774 尚未合并，被转换的目标 export 今天在树上不存在。卡是按"即将发布"写的，这没问题（正文也自陈是在实现 #15594 时顺手填的），但现在不可派工：接手人拿不到要 import 的符号。

⇒ pm:blocked，解锁条件一句话：scripts/js-comment-mask.mjs 导出 maskCommentsAndLiterals。
⚠️ 解锁扫描请量树、不要量 PR 状态（git grep -n "export function maskCommentsAndLiterals" origin/main -- scripts/js-comment-mask.mjs）—— PR merged ≠ 符号已在 main 上，这条纪律本轮已经吃过一次亏。

八个落点全部复核通过（带反向控制）

卡表里 8 个文件的 8 个命名，逐个在 origin/main 上命中：

check-parse-guard.mjs           codeOnly           5 行
check-test-source-alias.mjs     maskedProjections  2 行
check-stack-collection-maps.mjs maskLiterals       5 行
docs-audit/affected-docs.mjs    codeOnly          13 行
check-console-intercept-disarm.mjs maskProse       2 行
check-error-status-conformance.mjs projections     3 行
check-docs-section-name.mjs     project            8 行
measure-self-test-floor.mjs     maskCommentsAndLiterals 6 行
反向控制：zzzNotASymbol → 0 命中（控制behaves，零读数可信）

⭐ 但表里漏了第九个

scripts/check-adr-0087-registration.mjs:2414 自定义了一个同名函数，卡的表里没有它：

/** Comment AND string spans blanked -- offsets preserved. The house scanner, not a private one. */
function maskCommentsAndLiterals(source) {
  const { comment, literal } = scanSource(source);
  const flags = comment.map((c, i) => c || literal[i]);
  return blank(source, flags);
}

它完全落在卡自己划的类里，逐条对照：

• 组合的是共享 scanner，自己不带扫描逻辑 —— :357 import { blank, maskComments, scanSource } from './js-comment-mask.mjs';，函数体只有 scanSource → 或运算 → blank。
• 就是卡点名的两种拼写变体之一 —— 「| over a Uint8Array, || over one」，这个是 || 那支（comment.map((c, i) => c || literal[i])）。
• 它的 docblock 自己就说 「The house scanner, not a private one」—— 和卡"模块该拥有投影、而非调用方"的论点是同一个意思，作者已经意识到了却仍然自建了一份。
• 用了 3 处：:2447、:2628、:2687。

⇒ 范围是 9 个（至少），不是 8 个。这不是"数字腐了"式的挑刺 —— 它改变工作量与 PR 切分：卡建议「convert the six that are free, one PR per hot-file row for the two that are not」，加上这个就是七个 free（除非它也压在某个 hot-file 行下，本席未查 seat post #6023 §3 的完整行表，请接手人自查）。

⚠️ 另外提醒一条同名陷阱（纪律⑦）：maskCommentsAndLiterals 这个名字今天在 main 上出现在两个文件里，且都是本地定义，不是 import —— measure-self-test-floor.mjs:266（export，卡列了）与 check-adr-0087-registration.mjs:2414（非 export，卡漏了）。转换时按文件逐个确认是本地还是引入，⛔ 别按名字 grep 就当已经接上了。

一处行号漂移（不构成异议）

卡写 measure-self-test-floor.mjs:258，当前 main 上定义在 :266（差 8 行，PR #15758 合并后的位移）。论证不依赖这个行号，符号名唯一且好定位，所以这里只是记一笔、⛔ 不算纠正。

priority:p3

采纳卡自己的判断：「Nothing is broken: every copy is correct today」，三种拼写经 blank 的真值读法等价。代价是九份无共享 pin 的副本 —— 是 tech-debt 与未来漂移风险，不是今天的缺陷。p3。

卡结尾那句请接手人当作验收标准而非建议：「each proven byte-identical by diffing that gate's plain and --self-test output before and after」—— 这正是 #15594 两次转换用过并成立的验收方式。

---

Generated by Claude Code