test262 built-ins/String — 32 (self-contained worklist)

[proggeramlug]

Self-contained worklist — 32 failing. Snapshot 2026-07-03 full sweep (perry 0.5.1218, node v26.3.0). Everything you need is in this issue; no history to read, no re-measure to start.

String methods on non-generic receivers, normalize, well-formed, symbol coercion. Crate: perry-runtime (string), perry-codegen.

Setup (any box)

git clone https://github.com/PerryTS/perry && cd perry
cargo build --release -p perry -p perry-runtime-static -p perry-stdlib-static
git clone --depth 1 https://github.com/tc39/test262 vendor/test262   # checkout test-compat/test262/pinned-sha.txt
# node on PATH (differential oracle). See test-compat/test262/README.md.

Measure/verify: scripts/test262_subset.py --root vendor/test262 --dir built-ins/String --jobs 8 (the .failures.txt sidecar lists failures; diff before/after for zero-regression).

Failure reasons

n	reason
4	Uncaught exception: #1.1: Exception === "intostr". Actu
3	Uncaught exception: #1: __instance = new Object(true);
2	Uncaught exception: #1: __instance = new Boolean; __ins
2	Uncaught exception: #1.1: Exception is instance of Type
2	Uncaught exception: Expected a TypeError to be thrown b
1	Uncaught exception: #2: toString=function(){return "__T
1	Uncaught exception: #1: Function.prototype.indicator =
1	TypeError: Cannot assign to read only property 'length'
1	Uncaught exception: #2: var __str__instance = new Strin
1	Uncaught exception: #1: delete String.prototype.toStrin
1	Uncaught exception: #2: delete String.prototype.toStrin
1	Uncaught exception: #1: var x; __instance = {toString:f

All 32 failing cases (path — reason)

built-ins/String/S15.5.1.1_A1_T9.js    Uncaught exception: #2: toString=function(){return "__THIS__";}; __str = String(
built-ins/String/S15.5.3_A2_T2.js    Uncaught exception: #1: Function.prototype.indicator = 1; String.indicator === 1
built-ins/String/S15.5.5.1_A4_T1.js    TypeError: Cannot assign to read only property 'length' of object '#<Object>'
built-ins/String/S15.5.5.1_A5.js    Uncaught exception: #2: var __str__instance = new String("ABC\u0041\u0042\u0043"
built-ins/String/prototype/S15.5.4_A1.js    Uncaught exception: #1: delete String.prototype.toString; String.prototype.toStr
built-ins/String/prototype/S15.5.4_A3.js    Uncaught exception: #2: delete String.prototype.toString; String.prototype.toStr
built-ins/String/prototype/concat/S15.5.4.6_A4_T1.js    Uncaught exception: #1: var x; __instance = {toString:function(){return "one"}};
built-ins/String/prototype/constructor/S15.5.4.1_A1_T2.js    TypeError: String.prototype.valueOf called on incompatible receiver
built-ins/String/prototype/indexOf/S15.5.4.7_A1_T1.js    Uncaught exception: #1: __instance = new Object(true); __instance.indexOf = Stri
built-ins/String/prototype/indexOf/S15.5.4.7_A1_T2.js    Uncaught exception: #1: __instance = new Boolean; __instance.indexOf = String.pr
built-ins/String/prototype/indexOf/S15.5.4.7_A4_T4.js    Uncaught exception: #1.1: Exception === "intostr". Actual: Test262Error: #1: "va
built-ins/String/prototype/indexOf/S15.5.4.7_A4_T5.js    Uncaught exception: #1.1: Exception === "intostr". Actual: Test262Error: #1: "va
built-ins/String/prototype/lastIndexOf/S15.5.4.8_A1_T1.js    Uncaught exception: #1: __instance = new Object(true); __instance.lastIndexOf = 
built-ins/String/prototype/lastIndexOf/S15.5.4.8_A1_T2.js    Uncaught exception: #1: __instance = new Boolean; __instance.lastIndexOf = Strin
built-ins/String/prototype/lastIndexOf/S15.5.4.8_A4_T2.js    Uncaught exception: #1.1: Exception === "intointeger". Actual: Test262Error: #1:
built-ins/String/prototype/lastIndexOf/S15.5.4.8_A4_T4.js    Uncaught exception: #1.1: Exception === "intostr". Actual: Test262Error: #1: var
built-ins/String/prototype/lastIndexOf/S15.5.4.8_A4_T5.js    Uncaught exception: #1.1: Exception === "intostr". Actual: Test262Error: #1: var
built-ins/String/prototype/localeCompare/15.5.4.9_CE.js    Uncaught exception: String.prototype.localeCompare considers ö (\u006f\u0308) ≠
built-ins/String/prototype/padEnd/observable-operations.js    Uncaught exception: Expected SameValue(«"abc"», «"abcdefdefde"») to be true
built-ins/String/prototype/padStart/observable-operations.js    Uncaught exception: Expected SameValue(«"abc"», «"defdefdeabc"») to be true
built-ins/String/prototype/replace/S15.5.4.11_A1_T1.js    Uncaught exception: #1: __instance = new Object(true); __instance.replace = Stri
built-ins/String/prototype/replace/S15.5.4.11_A1_T10.js    Uncaught exception: #1: var x; var __obj = {toString:function(){return "AB";}}; 
built-ins/String/prototype/replace/S15.5.4.11_A1_T12.js    Uncaught exception: #1.1: Exception === "insearchValue". Actual: Test262Error: #
built-ins/String/prototype/replace/S15.5.4.11_A1_T15.js    Uncaught exception: #1.1: Exception is instance of TypeError. Actual: Test262Err
built-ins/String/prototype/replace/S15.5.4.11_A1_T16.js    Uncaught exception: #1.1: Exception is instance of TypeError. Actual: Test262Err
built-ins/String/prototype/replace/S15.5.4.11_A1_T2.js    Uncaught exception: #1: var x; __instance = new Boolean; __instance.replace = St
built-ins/String/prototype/replace/S15.5.4.11_A1_T5.js    Uncaught exception: #1: function(){return "gnulluna"}().replace(null, Function()
built-ins/String/prototype/replace/S15.5.4.11_A1_T9.js    Uncaught exception: #1: __obj = {valueOf:function(){}, toString:void 0}; functio
built-ins/String/prototype/replace/replaceValue-evaluation-order-regexp-object.js    Uncaught exception: Expected SameValue(«0», «1») to be true
built-ins/String/prototype/split/separator-regexp.js    SyntaxError: Invalid regular expression: /\k<x>/: invalid pattern
built-ins/String/prototype/toString/non-generic.js    Uncaught exception: Expected a TypeError to be thrown but no exception was throw
built-ins/String/prototype/valueOf/non-generic.js    Uncaught exception: Expected a TypeError to be thrown but no exception was throw

Workflow

Pick a coherent single-root subcluster, localize (cargo run --release -- f.ts --trace hir --focus <fn> / --trace llvm), fix, verify listed cases pass AND re-run the slice for zero regressions. Code-only PR (no Cargo.toml/Cargo.lock version bump, no CLAUDE.md Current-Version, no CHANGELOG — maintainer folds metadata). cargo fmt --all -- --check + bash scripts/check_file_size.sh must pass (files < 2000 lines). Reference this issue. Never reference internal infrastructure hostnames.

[proggeramlug]

Localized the biggest subcluster (String.prototype methods borrowed onto non-string objects — obj.indexOf = String.prototype.indexOf; obj.indexOf(true, false) → -1 instead of coercing this; also lastIndexOf/concat).

Root is at the HIR lowering layer, not runtime dispatch: x.indexOf(a, b) on an any-typed receiver is eagerly folded to Expr::ArrayIndexOf { array: x, ... } (confirmed via --print-hir), so a borrowed String.prototype.indexOf thunk stored as an own property never runs — the array-like engine searches the receiver's indexed elements and returns -1. String.prototype.indexOf.call(obj, ...) works (the .call fold-exclusion routes it through the generic string thunk); only the direct member form obj.indexOf(...) mis-folds.

The fix belongs in the .indexOf/.lastIndexOf/.concat fold in perry-hir (don't fold to ArrayIndexOf when the receiver has, or may have, an own property of that name). That's a broad optimization change with regression surface across all array indexOf/lastIndexOf/concat call sites, so leaving it for a dedicated maintainer pass rather than rushing it. A runtime-layer classify_own_slot attempt does not help (codegen never reaches the runtime dispatch for these folded calls).

[proggeramlug]

Localized S15.5.5.1_A4_T1 — it is not a String bug, it is sloppy-mode read-modify-write

Filing the diagnosis rather than a fix: the root is a missing field on a HIR variant, which is wider than this worklist entry and deserves its own PR.

The failing line in that test is neither of the plain assignments — it is __str__instance.length++ (and the with(...) length = 0 further down). Plain __str__instance.length = -1 already behaves correctly.

Characterization (node 26.5.1 vs perry @ 0f9bb6cc5, all sloppy mode):

case	node	perry
new String("globglob").length++	8	TypeError
new String("globglob").length += 1	8	TypeError
new String("globglob").length = -1	8	8 ✅
Object.freeze({a:1}).a++	1	TypeError
Object.freeze({a:1}).a = 2	1	1 ✅
Object.defineProperty(o,"x",{value:5,writable:false}); o.x++	5	TypeError
(function fn(a,b){}).length++	2	2 ✅
frozen array a[0]++	1	1 ✅

So this is not String-specific — Object.freeze(cfg); cfg.count++ throws on any frozen plain object, which is ordinary real-world code. Only the named-property read-modify-write path is affected; closures and array indices take other paths and are already correct.

Root cause. The two forms lower to different HIR, and only one of them carries the strict flag:

o.a = 2   →  PutValueSet { target, key: String("a"), value, receiver, strict: false }
o.a++     →  PropertyUpdate { object, property: "a", op: Add, prefix: false }
                                                        ^ no `strict` field at all

Expr::PropertyUpdate (perry-hir/src/ir/expr.rs) has no strict-mode information, so its codegen cannot make the sloppy-vs-strict decision and ends up on the runtime setter that throws unconditionally — e.g. the if is_frozen { throw_immutable_write(0, …) } gates in perry-runtime/src/object/field_set_by_name/tail.rs. Per spec these are [[Set]] failures that return false; only strict mode turns that into a TypeError.

Fix shape. Give PropertyUpdate the same strict flag PutValueSet already carries, populate it in perry-hir/src/lower/expr_misc.rs from the enclosing strictness, and have codegen select the throwing vs. silent setter from it. Mechanically that means the variant plus stable_hash/expr.rs, monomorph/substitute_expr.rs, and the codegen consumer; the ~50 files that mention PropertyUpdate mostly match with .. and are unaffected.

Note IndexUpdate is the sibling variant and is worth checking for the same gap while in there.

The with(__str__instance) length = 0; line in the same test is a separate question (whether with is supported at all) and is not covered by the above.

Found while landing #7470 / #7471.