chore(devx): the three sequential-double-blank comment-mask callers take maskCommentsAndLiterals (#15776, residual)

[baozhoutao]

Fixes #15776

js-comment-mask.mjs publishes maskCommentsAndLiterals (#15594, PR #15774). PR
#16300 converted the eight callers the card originally named that spelled the
comments+literals projection as comment[i] | literal[i] -> blank. Its Release
comment (5559693595) recorded a third spelling three more files carry — a
sequential double blank: blank(source, comment) then blank(masked, literal) — a shape the card's original grep could not match, measured at 0
disagreements over 6,206 files vs the export (live control 6,176). This PR
converts those three, closing the card's class.

Per-file table

file	line (pre-edit)	spelled as	shape	verdict
scripts/check-keyed-text-bounds.mjs	:402	project	wrapper, keeps pair { masked, struct }	converted
scripts/check-runner-env-posture.mjs	:258 (inline in findRunnerEnvReads)	commentMasked / bothMasked	inline composition, not extracted	converted
scripts/check-widget-option-census.mjs	:183	structureMask (exported, no external importers)	straight substitution	converted

All three now call maskComments / maskCommentsAndLiterals from
js-comment-mask.mjs instead of re-deriving the composition by hand.
js-comment-mask.mjs itself is untouched: no new export, no semantics change.
check-keyed-text-bounds.mjs keeps one unrelated scanSource/blank call site
(:1443, blank(ownSource, scanSource(ownSource).comment) — a single
comments-only mask, not the comments+literals class this card is about) — out
of this card's scope, left alone.

Residual check

A recursive scan of scripts/** for the sequential double-blank spelling
(blank(blank() and for the comment[i] | literal[i] spelling, after this
PR's edits, finds nothing outside js-comment-mask.mjs itself and the
Release's adopted out-of-class list (check-entry-guard.mjs,
pm/dispatch-gates.mjs's maskSelfTests, and the two per-index guards in
check-undeclared-dep-imports.mjs / eslint-fatal-guard.mjs, all of which
materialise no comments+literals mask). The card's class is empty after this PR.

Acceptance — byte-identical before/after

Method: plain + --self-test output captured on origin/main (3e270d4e2)
before the edits and again on the edited tree after; exit code captured by
redirect-then-$?, never across a pipe.

gate	plain (stdout+stderr)	--self-test (stdout+stderr)
check-keyed-text-bounds	EMPTY diff, exit 0→0	EMPTY diff, exit 0→0
check-runner-env-posture	EMPTY diff, exit 0→0	EMPTY diff, exit 0→0
check-widget-option-census	EMPTY diff, exit 0→0	EMPTY diff, exit 0→0

6 runs, 6 empty diffs, exit 0 on every side.

Ablation control (proves the diff could have failed)

Two legs against js-comment-mask.mjs's maskCommentsAndLiterals, each
trap-restored (proven by git diff HEAD empty and the blob hash returning to
7956343dd85bdef8924372cb2cdec45fcd54501d):

• Leg 1 — literal half dropped (degrades to comments-only)
• Leg 2 — identity (nothing blanked at all)

gate	plain / leg1	plain / leg2	self-test / leg1	self-test / leg2
check-keyed-text-bounds	MOVED (red: 3 unclassifiable declarations)	MOVED (same)	unmoved	MOVED
check-runner-env-posture	unmoved	MOVED (red)	MOVED (red, 3/34 self-test cases fail)	MOVED (red, 6/34 fail)
check-widget-option-census	unmoved	unmoved	unmoved	MOVED (red)

Every gate has at least one reading that moves under at least one leg, so the
byte-identical acceptance above is not vacuous for any of the three. Honest
exception: check-widget-option-census's plain run is unmoved by BOTH
legs — the real spec/parser corpus this gate reads today does not happen to
contain comment or string-literal content that a degraded mask would
mis-parse (its self-test's own literal-masking fixture sits at a nested
bracket depth the depth-walker already excludes regardless of masking, so it
does not discriminate this ablation either — a pre-existing property of that
fixture, not introduced here). Not filed separately: unlike #16299 (fully
vacuous under both legs, both plain and self-test), this file's self-test
channel does move under leg 2, so the gate's overall acceptance is not
fully unpinned — only its plain-run channel is silent on this particular
degradation today.

Differential proof (closes the gap regardless)

Because behaviour cannot be told apart from output alone, the deleted
sequential-double-blank spelling was run against its replacement
(maskCommentsAndLiterals) directly, over the whole tree, with a live
control:

files read:                    6,192  (106,979,676 chars)
old spelling vs new export:    0 disagreements
comments-only vs new (control): 6,167 files disagree

Zero disagreements over the full corpus is what actually retires the risk —
identical to the equivalence PR #16300's own report measured for the other
spelling variant.

Gates

node scripts/pm/dispatch-gates.mjs --changed --commands --repo objectstack-ai/objectstack on the final tree (10691a60e) derives 34
commands over the 3 changed paths. --ran reconciliation: 34 derived, 34 run,
0 NOT-MEASURED, 0 UNRUN. All 34 of 34 exit 0 — pnpm check:pm-dispatch-gates
(initially still running at draft-open time) finished green (1,534 self-test
cases pass) shortly after. check-comment-mask-corpus, check-scripts-symbol-anchors,
check-self-test-wired, check-declaration-mirrors (the families that would
redden on a dropped export, an unwired self-test, or a stale mirror) all exit
0. check:nul-bytes exit 0, plus an independent
grep -naP '[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]' over the three files, no hits.

TURBO_SCM_BASE=3e270d4e2 pnpm exec turbo ls --affected → 0 packages: the
diff is scripts/** only, so no package test/typecheck is owed.

Scope note

skip-changeset applied: internal gate scripts, nothing published from any
package. No content/docs/releases/ change.

---

Generated by Claude Code