JS-108: five divergent copies of the u8-slice-to-JS-string helper

Now: DONE

Proposed by the 2026-07-07 jab multi-agent review; confirmed by both skeptic verifiers. The same slice->JSString conversion is reimplemented five ways with different truncation caps and OOM behavior; the divergence already bit as JS-107.

Input

Context

Copies: uri.cpp:12 JABCSliceStr (2048 stack, silent truncate), io.cpp:615 JABCSliceStr (PAGESIZE stack, silent truncate; io.cpp:24 fwd-decl notes it "mirrors uri.cpp's"), pack.hpp:31 JABCStrOf (malloc, undefined on OOM), ulog.hpp:109-137 inline memcpy dance, codec.cpp:104-181 inline memcpy dance.
Every copy double-copies (slice -> NUL scratch -> JSC internal copy) where a length-explicit constructor needs one.
Owner norms flag duplicated helpers; the caps/OOM behavior have already diverged.

Goals

One length-explicit JABCStrOfSlice(ctx, u8cs) in JABC.hpp/utf8.cpp that fails loudly (throw, never truncate); delete the five copies.

Constraints

Fixes JS-107 by construction; behavior change is truncate->throw — audit callers that relied on clamping (none should).

WIP

TODOs

hoist the helper; migrate uri/io/pack/ulog/codec call sites
suite green; long-string cases added where missing

2026-07-10 revision: still OPEN — no JABCStrOfSlice hoisted in JABC.hpp; copies live uri.cpp:12, io.cpp:615, pack.hpp:31; ulog.hpp AND codec.cpp each carry TWO dances (ulog.hpp:117+130, codec.cpp:~104+176) — 6+ copies, not 5