lib: banner purge + WHY-only comment sweep (rule 8)

Every // ---- section banner dies (132 -> 0): names carry the WHAT.
Narration deleted (filename restatements, run-with lines, what-the-
next-line-does); every ref/hare cite, task cite, divergence, ABI/
layout contract, and ownership qualifier kept (borrowed-view lines
restored where the sweep over-cut). Comment-only proven: all 442
walk-workdir .s and 32 import-probe .s byte-identical before/after;
libbyteid 56-roster all-ID.

lib/strconv/decimal.ww

@@ -1,6 +1,5 @@
-// strconv — arbitrary-precision decimal engine for float↔string
-// conversion. Mirrors ref/hare/strconv/decimal.ha (Hare in turn ports
-// Go's lib/strconv/decimal.go). Pure integer arithmetic; no f32/f64
+// Mirrors ref/hare/strconv/decimal.ha (Hare in turn ports Go's
+// lib/strconv/decimal.go). Pure integer arithmetic; no f32/f64
 // references (#121 residual-guard SAFE).
 //
 // Spelling divergences from Hare (mechanical, ww-side parser shape):

lib/strconv/ftos.ww

@@ -1,4 +1,3 @@
-// strconv — float→string via Ryū (shortest round-trippable decimal).
 // Mirrors ref/hare/strconv/ftos_ryu.ha (the algorithm core) +
 // ref/hare/strconv/ftos.ha:432 (the f64tos driver). Ryū: Ulf Adams,
 // https://doi.org/10.1145/3192366.3192369 — Hare translated it from the
@@ -339,12 +338,12 @@ fn f64todecf64(mantissa: u64, exponent: u32) decf64 = {
 	return decf64 { exponent = (exp: i64), mantissa = output };
 };
 
-// ==== f32 Ryū sub-path (ftos_ryu.ha). The *32 helpers below mirror their
+// f32 Ryū sub-path (ftos_ryu.ha). The *32 helpers below mirror their
 // u64 siblings at 32-bit width; they reuse the SHARED f64computeinvpow5/
 // f64computepow5 (and thus the f64 SPLIT2 tables) per ftos_ryu.ha — there
 // is no separate f32 table. Same scalar-PARAM-mutation → copy-to-local,
 // comma-split, expr-yield → block divergences as the f64 path
-// above. ====
+// above.
 
 // ref/hare/strconv/ftos_ryu.ha:52. Largest p with 5^p | value (32-bit).
 fn pow5fac32(v: u32) u32 = {
@@ -523,13 +522,13 @@ fn f32todecf32(mantissa: u32, exponent: u32) decf32 = {
 	return decf32 { mantissa = output, exponent = (exp: i64) };
 };
 
-// ==== G-format encode layer (ftos.ha) — only the ffmt::G / prec=void /
+// G-format encode layer (ftos.ha) — only the ffmt::G / prec=void /
 // fflags::NONE-REACHABLE logic. The SHOW_POINT/precision/E-vs-uppercase
 // arms (ftos.ha:88-105, 127-145, 170-213's zeros/caps) are UNREACHABLE
 // for G/void/NONE (ffpoint(NONE)=false, prec is never uint, f is always
 // G) and are NOT ported — porting them stubbed would be untested dead
 // code. The parametric ftosf/ffmt/fflags surface is deferred (task #64;
-// needs a parametric consumer + io::handle + #158). ====
+// needs a parametric consumer + io::handle + #158).
 
 // ref/hare/strconv/ftos.ha:49. Decimal digit-count of n (n <= 1e17).
 fn declen(n: u64) uint = {

lib/strconv/ftos_data.ww

@@ -1,4 +1,3 @@
-// strconv — Ryū float→string lookup tables + bit-count constants.
 // Mirrors ref/hare/strconv/ftos_ryu.ha:159-222 byte-exact. Pure data
 // fold (strconv #106 fold-5): no logic, consumed by ftos.ww's
 // f64computeinvpow5 / f64computepow5 (the Ryū power-of-five cores).

lib/strconv/ftos_test.ww

@@ -1,6 +1,3 @@
-// ftostest — exercises lib/strconv/ftos.ww (Hare ftos.ha / ftos_ryu.ha
-// Ryū port).
-//
 // Ports ref/hare/strconv/+test/ftos_test.ha's ffmt::G / prec=void /
 // fflags::NONE rows (the f64tos cases) verbatim — these are exactly
 // strconv.f64tos's output. encode_f_dec is covered by 13.37/1100/0.011/…

lib/strconv/int_test.ww

@@ -1,6 +1,3 @@
-// inttest — exercises lib/strconv integer parse: parseint / stoi64 /
-// stou64 / the iN/uN width wrappers.
-//
 // Verbatim port of ref/hare/strconv/stoi.ha:56-86 (stoi/stoi_bases) and
 // stou.ha:116-138 (stou/stou_bases). Hare's strconv integer tests are
 // flat assert SEQUENCES, not row-array tables — mirrored here as inline
@@ -210,7 +207,6 @@ fn ck_uint_ovf(id: i32, s: str, b: base) void = {
 	ck_uint(62, "110101", base.BIN, 53u64: uint);        // 0b110101
 };
 
-// ---- format side: u64tos / i64tos / machine-word wrappers ------------
 // Verbatim ports of ref/hare/strconv/utos.ha:74-103 (utos/utos_bases) and
 // itos.ha:54-87 (itos/itos_bases) — Hare's format tests are flat assert
 // sequences too. Radix-literal inputs are written in DECIMAL (ww value

lib/strconv/stof.ww

@@ -1,6 +1,6 @@
-// strconv — string-to-float. Mirrors ref/hare/strconv/stof.ha
-// (Hare in turn adapts Go): Eisel-Lemire fast path [1] with the
-// Simple-Decimal-Conversion slow path [2] (decimal.ww) as fallback.
+// Mirrors ref/hare/strconv/stof.ha (Hare in turn adapts Go):
+// Eisel-Lemire fast path [1] with the Simple-Decimal-Conversion slow
+// path [2] (decimal.ww) as fallback.
 //   [1]: https://nigeltao.github.io/blog/2020/eisel-lemire.html
 //   [2]: https://nigeltao.github.io/blog/2020/parse-number-f64-simple.html
 //

lib/strconv/stof_data.ww

@@ -1,5 +1,4 @@
-// strconv — stof/ftos lookup tables. Mirrors ref/hare/strconv/stof_data.ha
-// byte-exact. Pure-data fold (strconv #106 fold-2, was fold-3 before drew
+// Mirrors ref/hare/strconv/stof_data.ha byte-exact. Pure-data fold (strconv #106 fold-2, was fold-3 before drew
 // re-sequenced 2026-05-26): no logic, exercised transitively when fold-3's
 // `leftshift_newdigits` lands (ref/hare/strconv/decimal.ha:35).
 //

lib/strconv/stof_test.ww

@@ -1,5 +1,3 @@
-// stoftest — exercises lib/strconv/stof.ww (Hare stof.ha port).
-//
 // Ports ref/hare/strconv/stof.ha's @test vectors (stof64 / stof32 /
 // stofhex). Comparisons are BIT-level (math.f64bits / f32bits) so a
 // sign flip (-0.0 vs 0.0) or a 1-ulp miss fails the row rather than
@@ -14,8 +12,6 @@ import strconv;
 import math;
 
 
-// ---- unwrap helpers (bit-exact value checks) -------------------------
-
 fn chk64(s: str, b: base, want: f64) bool = {
 	match (stof64(s, b)) {
 	case let v: f64 => { return math.f64bits(v) == math.f64bits(want); };
@@ -89,7 +85,6 @@ fn nan32(s: str) bool = {
 	return false;
 };
 
-// ---- stof64 ----------------------------------------------------------
 // ref/hare/strconv/stof.ha:530.
 
 @test fn stof64_dec() void = {
@@ -141,7 +136,6 @@ fn nan32(s: str) bool = {
 	assert(!(!nan64("naN")));
 };
 
-// ---- stof32 ----------------------------------------------------------
 // ref/hare/strconv/stof.ha:560.
 
 @test fn stof32_dec() void = {
@@ -180,7 +174,6 @@ fn nan32(s: str) bool = {
 	assert(!(!chk32("9.19100241453305036800e+20", base.DEC, 9.19100241453305036800e+20f32)));
 };
 
-// ---- stofhex ---------------------------------------------------------
 // ref/hare/strconv/stof.ha:590. Hex-float surface-form literals (0x1.fp-2,
 // math::F64_MAX_NORMAL, …) are spelled as IEEE-754 bit patterns since ww
 // has no hex-float literal lexer / no F*_MAX_NORMAL math consts.

lib/strconv/strconv.ww

@@ -1,5 +1,3 @@
-// strconv — number↔string conversions.
-//
 // Mirrors Hare's strconv:: surface. The *tos functions return a
 // `const str` view into a module-level buffer that is overwritten on
 // the next call to the same function; callers must copy the bytes if
@@ -71,7 +69,7 @@ let lut_lower: [16]rune = [
 // fidelity; the initial value is irrelevant (only the freshly-written
 // prefix is ever read) but the fill form is exercised (probed: emits
 // byte-identically cross-stage).
-let u64tos_buf: [64]u8 = [0...];	// 64 binary digits
+let u64tos_buf: [64]u8 = [0...];
 
 // u64tos — convert u to a base-b numeric string. Returns a view into
 // `u64tos_buf`, overwritten on the next call; copy via strings.dup to
@@ -114,7 +112,7 @@ export fn u64tos(u: u64, b: base) str = {
 // i64tos_buf — independent from u64tos_buf so i64tos's own u64tos call
 // (the magnitude) doesn't clobber the in-flight result. 65 = 64 digits
 // plus the leading '-'. Hare's `static let buf: [65]u8` (itos.ha:18).
-let i64tos_buf: [65]u8 = [0...];	// 64 binary digits plus '-'
+let i64tos_buf: [65]u8 = [0...];
 
 // i64tos — convert i to a base-b numeric string. Returns a view into
 // `i64tos_buf`. Verbatim port of ref/hare/strconv/itos.ha:10-32.