[andrewtdiz]

Summary

Fixes #4033 by filling the ordinary ArrayBuffer/DataView semantics needed by binary-heavy packages:

• add validated ArrayBuffer / SharedArrayBuffer constructor length coercion so huge or negative sizes throw RangeError before truncation/allocation
• install ArrayBuffer.isView as a real callable static and support value extraction (const isView = ArrayBuffer.isView)
• install ArrayBuffer.prototype.byteLength as an accessor with brand checks and descriptor metadata
• keep full-span DataView instances distinct from their backing ArrayBuffer while preserving view registry write propagation
• recognize direct DataView instances, DataView subclasses, and existing typed views in ArrayBuffer.isView / util.types.isArrayBufferView

DeepWiki

Required reference saved at:

• target/deepwiki/arraybuffer-dataview-semantics.md

Queried boa-dev/boa for ArrayBuffer constructor/prototype/static surface, ArrayBuffer.isView, DataView internal slot/subclass recognition, byteLength getter descriptors/brand checks, slice bounds/defaults, and avoiding SharedArrayBuffer/resizable scope creep.

Evidence

Base SHA: 132fb26cf461eced3da6bc86e04f680557c49400
Commit: bbf96e3e3

Issue baseline for the requested subset was:

• pass=40 runtime-fail=40 parity=50.0%

After this change:

scripts/test262_subset.py --root vendor/test262 --dir built-ins/ArrayBuffer built-ins/DataView --max 80 --timeout 8 --sample-cap 20 --report /tmp/perry-c262-arraybuffer-dataview.json

pass           47
diff           0
runtime-fail   33
compile-fail   0
skip           0
judged         80
parity:        58.8%
report:        /tmp/perry-c262-arraybuffer-dataview.json

Focused smoke covered representative #4033 failures:

• huge ArrayBuffer lengths throw RangeError
• typeof ArrayBuffer.isView === "function"
• extracted const isView = ArrayBuffer.isView is callable
• ArrayBuffer.isView(new DataView(ab)) === true
• ArrayBuffer.isView(new (class extends DataView {})(ab)) === true
• ArrayBuffer.isView(ab) === false
• ArrayBuffer.prototype.byteLength descriptor/getter/brand checks work
• ab.slice(1).byteLength === 3

Additional commands run:

cargo fmt --all --check
git diff --check
cargo build --release -p perry
cargo build --release -p perry-runtime -p perry-stdlib
PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry run test-files/test_parity_buffer.ts
PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry run test-files/test_gap_fetch_reqresp_2640_2643.ts
PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry run test-files/test_issue_3988_typedarray_static_accessors.ts
PERRY_NO_AUTO_OPTIMIZE=1 target/release/perry run test-files/test_gap_dataview_2878_2877_2879.ts

Residual risks

• This intentionally avoids SharedArrayBuffer/Atomics, resizable/growable buffers, transfer semantics, and broad TypedArray transform-method work.
• Remaining c262 failures include broader constructor/prototype descriptor completeness, species/slice edge cases, and typed-array subclass view recognition beyond the DataView subclass recognition fixed here.
• DataView subclass recognition is modeled for view/brand predicates; this does not attempt a full JS internal-slot object model for arbitrary subclass byte access beyond Perry's existing DataView paths.