fix(runtime): dispatch inherited Array statics on a subclass constructor (#7541)

[proggeramlug]

Fixes #7541.

Stacked on #7603 (fix/7574-array-subclass-raw-paths), and not by
convenience — see Why this is stacked below. Review the top commit only.

The bug is not in spread

class MyArr extends Array {}
const sub = MyArr.from([1, 2, 3]);
console.log([...sub]);                // node: [ 1, 2, 3 ]   perry: TypeError: value is not iterable
console.log(Array.isArray([...sub])); // node: true          perry: TypeError

The issue guessed that array_from_spread_value's subclass arm
(is_array_subclass_instance / array_subclass_has_iterator_override) was
answering wrong. It is not — a directly constructed instance spreads
correctly today, and did before this PR:

typeof MyArr.from  = undefined      <-- node: function
const d = new MyArr(); d.push(1,2,3);
[...d]                              -> [ 1, 2, 3 ]   (already correct)

MyArr.from resolves to nothing. Array.from / Array.of /
Array.isArray are folded in the HIR on the literal identifier Array
(perry-hir/src/lower/expr_call/array_only_methods.rs:293), so a subclass
receiver matches no fold and the call falls through to
js_class_static_method_call, whose documented miss-fallback returns the
receiver. MyArr.from([1,2,3]) therefore evaluates to the class ref — which
is genuinely not iterable, so the spread throws exactly where the issue saw it.
This is the CLAUDE.md "native base-class subclassing… keying any of that on a
literal extends name loses it" weak area, on the STATIC side.

Fix

js_class_static_method_call already has arms for inherited builtin statics on
a class X extends Promise (X.all / X.resolve) and on a
class X extends Buffer (nm_static_buffer_proto_chain). Add the Array one
beside them, gated on is_array_subclass_class_id(class_id) — the same bounded
class-chain walk the instance-side dispatch already uses.

Both spec statics are already implemented constructor-aware:
array::array_from_full(c, items, mapfn, thisArg) and
array::array_of_full(c, vals) run Construct(C, …) whenever
IsConstructor(this), and is_constructor_value recognizes an INT32 class ref.
So passing the subclass receiver through as this builds a real subclass
instance — MyArr.from(x) behaves as Array.from.call(MyArr, x) does in
node, rather than degrading to a plain Array. isArray needs no receiver and
routes to js_array_is_array.

Why this is stacked on #7603 (not merely convenient)

array_from_full's element install is CreateDataPropertyOrThrow, whose
implementation branches on jsv_is_array(result) — and js_array_is_array
answers true for an Array-subclass instance. So it calls
js_array_set_f64_extend on what is physically an ObjectHeader. On main
that is precisely the #7574 corruption this PR would newly start triggering:
without #7603 the fix half-works and then writes into the object header —

isArr true        <-- constructed a MyArr
len undefined     <-- elements and `length` never landed
[]

With #7603's funnel underneath, the same code is correct. Shipping this against
main alone would trade one wrong answer for a memory-safety hazard, so it
must land after #7603.

Validation (local; CI has a deep backlog, so local is what this rests on)

• test-files/test_gap_7541_array_subclass_inherited_statics.ts —
  byte-identical to node --experimental-strip-types (v26.5.1), exit 0, 19
  lines. Covers the issue's exact repro plus from with a mapFn, from a Set,
  from an array-like, of, isArray, an indirect subclass
  (class Indirect extends MyArr), and the whole iteration surface on a
  static-produced instance (for…of, spread, Array.from, destructuring,
  map, indexing, length). Controls: the base Array.from/Array.of
  intrinsics and an ordinary user-class inherited static are unchanged.
• Sabotage: with crates/perry-runtime/ reverted in full to this PR's
  base (i.e. fix(runtime): an Array subclass in a base-typed binding was read as a raw header (#7574) #7603 present, this change absent) and the test file untouched,
  the same file exits 1 with TypeError: value is not iterable — the
  reported symptom, verbatim. Restored, it exits 0, byte-identical.
• test_gap_7574_array_subclass_declared_base_type.ts still byte-identical on
  this branch.
• cargo test -p perry-runtime: 1853 passed, 0 failed.
• python3 scripts/raw_handle_debt.py: 998 (baseline 998).
  python3 scripts/addr_class_inventory.py,
  python3 scripts/class_id_collisions.py, ./scripts/check_file_size.sh,
  cargo fmt --all -- --check: all clean.

Scope — deliberately NOT fixed here

• The property-GET form. typeof MyArr.from still reports undefined;
  only the CALL form is dispatched (the GET path is a different resolver, and
  the same is true today for the Promise/Buffer subclass statics that
  already work as calls). Called out inline in the test file.
• sub instanceof MyArr is false for a static-produced instance — the
  pre-existing class-registry parent-edge gap, the Array sibling of instanceof a Map/Set SUBCLASS is false (m instanceof MyMap); only the native base edge survives #7575.
• ArraySpeciesCreate on sub.map(f) still yields a plain Array (node:
  MyArr). Pre-existing, identical on the unannotated path, noted in fix(runtime): an Array subclass in a base-typed binding was read as a raw header (#7574) #7603.

No version bump (maintainer bumps at merge).

Summary by CodeRabbit

• Bug Fixes

  • Fixed inherited Array.from, Array.of, and Array.isArray behavior for Array subclasses.
  • Array.from and Array.of now correctly create instances of the subclass, including indirect subclasses.
  • Improved support for iterable, array-like, mapped, and empty-input scenarios.

• Tests

  • Added regression coverage for subclass inheritance, iteration, destructuring, and related array operations.

• Chores

  • Updated the project version to 0.5.1344.

[coderabbitai]

Caution

Review failed

The pull request is closed.

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 5ca33ad9-09d7-45b7-8317-86503499f64d

📥 Commits

Reviewing files that changed from the base of the PR and between 87b5d39 and 780f5aa.

⛔ Files ignored due to path filters (1)

• Cargo.lock is excluded by !**/*.lock

📒 Files selected for processing (5)

• CLAUDE.md
• Cargo.toml
• changelog.d/7605-array-subclass-inherited-statics.md
• crates/perry-runtime/src/object/class_registry/parent_static.rs
• test-files/test_gap_7541_array_subclass_inherited_statics.ts

---

📝 Walkthrough

Walkthrough

The runtime now dispatches inherited Array.from, Array.of, and Array.isArray calls for direct and indirect Array subclasses. Regression tests cover construction, iteration, mapping, array-like inputs, and unaffected statics. Version metadata and the changelog were updated.

Changes

Array subclass inherited statics

Layer / File(s)	Summary
Runtime dispatch and regression coverage crates/perry-runtime/..., test-files/test_gap_7541_array_subclass_inherited_statics.ts	Array subclass calls now use subclass-aware from and of construction and isArray checks. Tests cover direct and indirect subclasses, multiple input types, iteration, and base or unrelated statics.
Release documentation and version metadata changelog.d/7605-array-subclass-inherited-statics.md, Cargo.toml, CLAUDE.md	The changelog records the behavior and remaining limitations. The workspace and documented current version are updated to 0.5.1344.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• PerryTS/perry#7569: Modifies related Array-subclass runtime dispatch.
• PerryTS/perry#7603: Adds related Array-subclass dispatch and tests.

Suggested labels: bug

Suggested reviewers: thehypnoo

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/7541-array-subclass-inherited-statics

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}