crypto.createECDH returns a native ECDH handle backed by mbedTLS for
prime192v1, secp224r1, prime256v1, secp384r1, secp521r1, secp192k1,
secp224k1, secp256k1, brainpoolP256r1, brainpoolP384r1 and
brainpoolP512r1. generateKeys, computeSecret, getPublicKey,
getPrivateKey, setPrivateKey, setPublicKey and ECDH.convertKey follow
Node 24: compressed, uncompressed and hybrid points, input and output
encodings, the private key range, key pair validation, and Node's error
codes and messages.
Curve names follow OpenSSL's short-name lookup: names of other curves
from crypto.getCurves() are refused at compile time when literal and
throw a coded SC2020 error when computed, other OpenSSL short names
throw ERR_CRYPTO_OPERATION_FAILED, and unknown names throw
ERR_CRYPTO_INVALID_CURVE. Calls whose arguments no @types/node overload
accepts, and computeSecret(view, encoding), are refused at compile time.
The runtime unit scr_ecdh.c links, with the mbedTLS archive, only into
programs that use ECDH. wasm32-wasi builds refuse it.
Construction from checked-dynamic code (`new F(options)` with F an
untyped value holding a ctor-less subclass of a stream class or
EventEmitter) dispatched to the class's constructor and completed the
arguments against the inherited signature, which takes none here, so the
options were dropped. The checked construction dispatch now reports
those arguments at run time with the message typed construction reports
at compile time, located at the class: a fixed argument count refuses at
the call, and a spread call refuses when its argument list is not empty.
Reading Object, Date, RegExp, Map, Set, Promise, WeakMap and WeakSet as
interned function values made every use of those globals a dynamic value,
including the receivers of member reads. The interned function models only
name, length and constructor, so `Map.prototype`, `Date.prototype.getYear`,
`Object.isFrozen` as a value, `Map["prototype"]` and `"prototype" in Map`
stopped reaching their SC2020/SC1090 fences and read undefined through the
dynamic paths instead (769 Test262 variants went from a refusal to a wrong
result).
A builtin global read for its members now keeps its member lowerings and
fences; only name, length and constructor read through the value. A member
read through an alias typed as the constructor's static side (`const D =
Date; D.prototype`) reports SC2020 before the generic field read. Passing
the value to Object or Reflect statics that reflect on or redefine its
members (`Object.hasOwn(Object, ...)`, `Object.defineProperty(Promise,
...)`, `hasOwnProperty.call(Map, ...)`) is refused; identity and the empty
enumerable key lists still compile. The interned functions answer Node's
lengths (Date 7, RegExp 2, Promise 1).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A member export resolves to its binding only when the binding is declared
before the write and never reassigned, so a consumer reads the value the
member stored. A module that re-exports another module's exports through
module.exports = require("./x") now counts its own writes after the
re-exported module's: a member it adds resolves, and a member both modules
write stays refused. The exports rebinding refusal and its test now cover
the form that stays refused, exports = value without module.exports.
A factory's parameter counts as the argument it was called with only
while the factory can't change it. The check saw assignments and updates
of the name, but not a write through a destructuring pattern
(`[body] = ...`, `({ body } = ...)`) or a for-in or for-of target, so a
factory could replace its callback with one that reads `this` and the
call still compiled without the module object.
The parameter now counts as changed by any assignment, update,
destructuring or for-in/for-of target that writes it, by a declaration
of the same name, by a write to an element of `arguments`, and by a call
of `eval`.
The exemption for `f.apply(this, ...)` and `f.call(this, ...)` dropped
SC1090 for every module export that passes `this` on. When the function
it reaches reads `this`, Node gives it the module object, and the direct
module call has none, so the result differed from Node.
A `this` that is passed on now counts as unread only when the function
it goes to provably ignores its `this`: an arrow, a function that uses
`this` only to pass it on to such functions, or what a known function
returns, with its parameters bound to the call's arguments and never
assigned. The analysis follows at most six calls and functions, and
refuses otherwise. Effect 4's provideService still compiles: it passes
`this` to what dual(3, body) returns, which passes it on only to the
arrow it was given.
Passing `this` on to a function that reads it now refuses, directly,
through such a factory, or when the factory reassigns its parameter.
crypto.subtle.importKey('raw', ...) for HMAC keys, HMAC sign and verify,
and the SHA-1/SHA-2 digests compile through the crypto global, with
CryptoKey values as native handles. scr_webcrypto.c is only the WebCrypto
layer (Web IDL conversion, algorithm normalization, the CryptoKey handle,
promise wrapping and Node's TypeErrors and DOMExceptions) over node:crypto's
digest core and deferred promise queue. It links only into programs that
call these methods (the new webcrypto link feature). Other algorithms, the
jwk format and the other SubtleCrypto methods are SC2020 refusals.
The real jose@6.2.12 HS256 path (SignJWT, jwtVerify) now runs natively. It
also needed four compiler fixes: object patterns of positions and length
over arrays, the Uint8Array base64/hex feature probes, a compile-time
String.fromCharCode.apply, and structuredClone of a parameter whose default
is {}.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`m.f()` where `m` is `const m = require("./x")` or `import * as m from
"./x"` lowers to a direct call of the exported function, so a function
that reads `this` saw no receiver (or another frame's) where Node passes
the module object: `exports.hello = function () { return this.name; }`
threw a TypeError natively. Such calls now refuse with SC1090 (deferred
in JavaScript); exports that do not read this compile as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A class constructor's `return expr` was evaluated and dropped, so
`new C()` answered the constructed instance even where Node answers the
returned object (`if (C.instance) return C.instance;` built a second
instance).
A return that yields an instance of the class itself stores it in a
per-class result slot, read and cleared by one helper at every
construction site (direct new, class values, the checked dynamic
construction dispatchers, record-to-class coercion). A return of any
other object refuses with SC1090 (compile time in TypeScript, a
deferred fence in JavaScript); a JavaScript value typed only at run time
is tested when it returns. super() into a class that can return an
existing instance refuses, since Node would rebind the subclass's this.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The npm compatibility probes ranked these CommonJS forms among the top blockers. They now compile with Node's semantics inside the static CommonJS model, where exports stay module globals or alias plumbing to declarations. Every remaining form keeps a refusal that names the construct.
- Inline require() of program modules (cjs-inline-require.ts). `require("./x").f`, `.f(...)` and `new (require("./x").C)()` over a table module run the module's run-once %init at the call's position, then resolve the member like a binding does. `require("./x")` as a value over a single-value module is that value (`require("debug")("ns")`, table entries, construction). Inline requires are preflight module edges. An edge in a function body is lazy: it orders the dependency after the requiring module and does not close initialization cycles. Other require() calls name their construct instead of reporting an unlowerable 'Require' value. A package that nothing installed resolves throws MODULE_NOT_FOUND.
- module.exports = require("./y") or a require binding (re-export), = TypeError (a builtin global), and = any computed value (`a || b`, ternary, `new X()`). Chained `exports.a = exports.b = v` evaluates v once. Export members assigned in value position (`const src = exports.src = []`) or inside functions write the export global. A fresh `exports = module.exports = {}` reset at the top of a module is a no-op. Any other rebinding of `exports` is refused by name.
- A module's own exports object (cjs-exports-object.ts). Identity comparisons, typeof and truthiness of `exports` / `module.exports` use the object that top-level `this` already represents, while the module keeps its original exports. This covers the `root === module.exports` statement that used to vanish from `.call(this)` wrappers. `exports.f(x)` calls the member's export global when no assigned value can read `this`. Other value uses are refused by name, including in CommonJS files where the message used to say "ES module".
- tsc interop (cjs-tsc-interop.ts). `const x_1 = __importDefault(require(...))` and `__importStar(...)` are require bindings, stamped and unstamped. The helper declarations lower to nothing once every call is such a binding.
- Export tables: method entries compile, and `...require("./x")` spread entries now evaluate x at their place. Before, the module was never initialized.
- TypeScript reads of members a JavaScript module attached to an exported function go through the function's own property table.
- Nested require destructuring of program modules, `const { codes: { E } } = require("./errors")`, destructures the named export at the require (SC1012). `const { A } = ns` over a module binding snapshots the export.
The capture-buffer fence tests merged from opt/cjs-exports-value used `root === module.exports`, which now compiles. They use eval() to keep testing the fence.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A refused statement in a JavaScript function body was dropped without a diagnostic or runtime fence when the body lowered inside a capture buffer whose owner kept the result. tryLowerExpression lowers the receiver of `(function () { ... }).call(this)` or `.apply(...)` and a computed class base that way: the refused statement (for example `console.log(root === module.exports)`) was missing from the kept function, its diagnostic was handed back to the enclosing statement, and that statement, having lowered cleanly, filed it as an informational runtime-fence entry. The program printed nothing for that statement.
lowerStmts now emits the statement's runtime fence inside a buffer as well (Lowerer.capturedRuntimeFence, the same message deferToRuntimeFence would give), leaving the diagnostics with the buffer's owner. The three JavaScript class initializer sites that skipped their fence under a buffer (static field, instance field, static initializer block) do the same. An owner that discards the candidate discards the fence with it, so every program that compiled before keeps byte-identical IR unless it contained such a dropped statement: all 3006 corpus programs produce the same IR as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A Hash or Hmac that flows into an unknown or any slot, or into untyped
JavaScript code, is now a native handle (Hash or Hmac) instead of a
compile error. The handle keeps the value's identity, so a cast back to
Hash or Hmac answers the same value. update(), digest() and Hash's
copy() dispatch through it, called directly or read as values: update
is one function for Hash and Hmac and digest is one per class, as in
Node 24.22, and update() answers the receiver. The stream and event
members throw a coded SC2020 error there, as do the methods called with
another receiver; other names read as undefined.
The to-dyn conversion boxes a Hash or Hmac through
scr_crypto_hash_to_dyn, which installs the handle ops on first use, so
programs that never box one link none of the handle code.
A native handle read through `as any` now dispatches its method calls
through the handle. Calling copy() on an Hmac throws Node's TypeError
("mac.copy is not a function") after the arguments are evaluated,
instead of the method-call fence.
Corpus 5768 and 5769 cover the TypeScript and JavaScript forms against
Node 24.22, and errors.test.ts covers the SC2020 cases. corpus-costs.json
records their plain and sanitized costs.
Hash and Hmac values are Transforms, as in Node: they can be written
and ended, piped into, used as pipeline() and stream/promises pipeline
stages, read, and listened to, and setEncoding() applies to the digest
they push. Each value gets one native Transform on its first stream use.
The stream refers to its value without a reference; a pipe destination
or pipeline stage holds the value until the stream flushes or is
destroyed, and a destroyed stream is released so that listeners that
capture the value do not keep it alive.
After the stream flushes, a Hash answers one more digest() with the
same digest and then throws ERR_CRYPTO_HASH_FINALIZED, and update()
throws ERR_CRYPTO_HASH_UPDATE_FAILED; an Hmac answers digest() with an
empty digest. A stream write after digest() reports
ERR_CRYPTO_HASH_UPDATE_FAILED through 'error'. copy() of a Hash whose
stream flushed, and stream use of a value after its stream closed,
throw a coded SC2020 error.
Writing a checked-dynamic value that is not null, a string or a view to
any stream now throws ERR_INVALID_ARG_TYPE; it used to report the
null-values error.
Corpus 5760-5762 cover the piping, pipeline, setEncoding and event
forms, the states after a flush, and the JavaScript forms, and
errors.test.ts covers the two SC2020 errors.
createCipheriv, createDecipheriv, update(), setAAD() and setAuthTag()
now read the value a typed-ref box holds when it is a string, view,
ArrayBuffer or KeyObject. ERR_INVALID_ARG_TYPE names the class of a
class instance that arrives through an untyped value, as Node does,
instead of Object. An unknown output encoding that is not a string is
rendered as Node's %s formats it (util.inspect at depth 0, or the
object's own toString); a function, or an instance of a class that
defines toString, throws a coded SC2020 error. Cipher programs link the
inspect unit for that rendering.
5250 passes each argument through `any`, including class instances,
records and arrays that Node rejects, and 5254 covers output encodings
that are not strings from JavaScript.
- Every mbedTLS call in the cipher runtime is checked. A failed update(),
final() or setAAD() returns no output and retires the context, so later
calls throw and no tag is returned or accepted. An update past the
AES-GCM limit of 2^36 - 32 bytes per message throws Node's "Trying to
add data in unsupported state" error; it used to return zeroed output
and leave the tag covering only the earlier data.
- chacha20-poly1305 decryption without setAuthTag() fails in final()
again. Node 24.22 skips the tag check when authTagLength is not passed
either and returns unverified plaintext; limitations.mdx records the
difference.
- A runtime unit test moves contexts next to the GCM message and AAD
limits and checks that the failure is permanent; errors.test.ts covers
decryption without a tag.
`x as Sub` lowered to a call to an interned %class.assert helper that
unwrapped a union box, retained the payload twice, compared the class-id
range, and threw on mismatch; every use paid a call, three retains, two
releases and a pending check (2,152 sites in tsc-ts). Plain operands
(locals and field reads, which isSafeToRepeat already admits) now lower
to the check itself: per compatible union arm a tag test plus, where the
arm is not already the target or a subclass, an instanceOf range test,
selecting the downcast payload, with the helper's exact TypeError as the
final arm. Other operands keep the helper so they are evaluated once.
The backend borrows through the result: instanceOf reads its operand as
a receiver instead of retaining a temporary, and downcast/upcast are
borrowable, edge-preserving projections, so `(node as Sub).field` and
method receivers on a cast no longer touch reference counts. The mismatch
arm unwinds directly (previous commit) and stays out of the hot path.
Corpus 4432 pins the Node-observable behavior of failing casts (a
TypeError before any later read, unwinding through finally blocks and
callbacks) and of successful casts from bases, subclasses and unions; a
harness test pins the exact messages, which Node cannot oracle.
- Enforce oxlint and oxfmt for owned JavaScript and TypeScript with independent CI checks.
- Preserve compiler import boundaries and protected fixtures while cleaning up lint findings.
- Standardize source formatting and validate import, fixture, and embedded-source protections.
* Expand static compilation for common Effect 4 patterns
* Run every Effect 4 fixture slice in CI
* Preserve existing corpus behavior across Effect 4 lowering
* Preserve Effect platform callbacks and native stream state
* Record new corpus baselines and retain child stdin refusals
* Retire the obsolete stateful regex diagnostic
* Preserve promise reference count layout for cycle collection
* Expand static Effect support across common APIs
* Preserve compiler parity across expanded Effect patterns
* Fix compiler regressions exposed by Effect coverage
* Bound shared type conversion traversal
* Clarify conversion contracts and tidy fixtures
* Run every Effect fixture in CI shards
* Preserve inferred array assignments in concurrent Effect programs
* Record inferred array fixture compiler ordering
- Remove C emission, debugging options, and automatic backend fallback.
- Complete LLVM lowering and native bootstrap coverage across supported targets.
- Link executables and libraries with precompiled C runtime packs, including iOS and Android.
- Run production scope and closure capture logic with native TypeScript 7 symbols.
- Give lowering caches explicit pass ownership and stable symbol identities.
- Support reference collection arrays and optional Set seeds on both native backends.
- Store checked values in native collections while preserving supported object identities and JavaScript key equality.
- Support collection iteration and reject reference conversions that would copy or adapt stored objects.
* Fence unsupported Array.from elements
- Share the backend-supported array-element contract across type mapping, callback producers, and IR validation
- Diagnose or defer unsupported Array.from mapper results before emission
- Cover the JavaScript runtime-fence path that previously crashed the C emitter
Co-authored-by: xXCrash2BomberXx <91439080+xXCrash2BomberXx@users.noreply.github.com>
* Fence unsupported array producer results
---------
Co-authored-by: xXCrash2BomberXx <91439080+xXCrash2BomberXx@users.noreply.github.com>
- The written-binding decline yields when the declared type has no mapping or the write's RHS roots at an ambient-undefined name
- A class extending a collection-poisoned base flushes the base's diagnostics eagerly: the heritage evaluates at module init, so deferring the compiler's own fence manufactured a silent wrong output
- trapDeclRootOf's written-binding decline now requires a mappable declared type: ordinary storage only buys anything when it compiles — an unmappable recursive call shape traded the compiling no-storage trap for a guaranteed SC2009 on a value that never exists
- bindingEverWritten skips statement-position writes whose RHS chain roots at an ambient-undefined name: that statement IS the RHS root's throw (lowerExprStatement's first claim) and never touches the target's storage, so it cannot justify the decline
- A poisoned base a top-level class extends flushes its deferred diagnostics eagerly (collectClassShape): the heritage evaluates at module init and the base's fence is the compiler's own — Node defines the base fine — so the deferred trap compiled a binary that refused at startup where Node runs; leaf and derived-only poisoned classes keep the JS deferral
- Fixtures: corpus 2632 (unmappable written trap binding) and 2633 (ambient-rooted rewrite), diagnostics js-extends-poisoned-base (the eager report names the base's real blocker), errors.test pins for both deferral edges that must NOT report eagerly; order-parity baselines recorded for the new entries
- The ratified rename rider (world-unification §7 lane 5): "DOM" collided with the browser DOM and, post-lane-1, the structure is no longer a document tree. Comments and doc prose now say "the checked-dynamic tree" for the noun and "dyn" adjectivally; runtime symbol names (scr_dyn_*), the error.domClone libCall, and the real web DOMException family are unchanged.
- The few user-facing strings carrying the nickname move with it (the structuredClone cyclic message, the dc subscriber/publish fences, two emitter ICE texts); diagnostics snapshots regenerated (96 pass). One private compiler constant (DOM_PROTO_METHOD_NAMES -> DYN_PROTO_METHOD_NAMES) rides along.
- Mechanical sweep, 142 files, comments/prose only — no behavior change.
- corpus 2578 (TS) and 2579 (JS, the JSDoc @param {object} repro shape) pin typeof/truthiness/String()/===, scalar normalization, representation-free narrowing, and the identity round trip against Node on both backends
- npm case jsval-into-dyn drives the prettier-plugins shape over a real package surface (the new plugstub fixture package: any-declared exports carrying engine functions)
- dyncheck jsval-* tests pin every loud fence (keyed read/write, member calls, 'in', destructuring, JSON.stringify, structuredClone, Object.keys, deepStrictEqual, inspect) plus the RC-clean catch-and-continue unwind under the SAN lane
- the errors.test.ts SC1101 pin flips to the now Node-exact behavior (the retired "one unbridgeable mix": the wrapped argument compiles and the unbound-this read throws Node's own TypeError); ts7 order-parity baselines record the three new fixtures
- Array patterns over checker-typed string sources split once through %str.chars (code points, astral characters whole) and ride the array machinery — declarations, assignments, for-of heads in both declaration and expression position, and parameters; object patterns bind the wrapper's one own data property, length, with renames and dead defaults. Static builds claim exactly what staticStringPattern vets; every other string form keeps the engine gate, and --dynamic keeps the engine route wholesale.
- Object patterns over stdlib globals in JavaScript files bind MEMBER identity tokens (const { subtle } = globalThis.crypto, const { Console } = console — 27 of the node suite's 28 SC1031 first-blockers), so programs run past the destructure and each use meets its own per-site fence; globalThis members are provenance-vetted, bind the member global's own token, and register the stdlibGlobalAliasDecl twin, so surfaced members (Math) keep their receiver-keyed lowerings.
- Members of SURFACED globals (Math.max, process.env, console.log) fence eagerly by member and global name instead of degrading detached uses to quiet token reads, and IR-string sources the checker does not type as strings fence naming the checker's type, never the token's carrier.
- Corpus 2556-2558 pin the string forms and the token identity/truthiness/alias agreements under the Node oracle, byte-identical through both backends; errors.test.ts pins the surfaced-global fence and the lazy per-site token failure.
- order-parity baselines re-recorded additively: the new entries plus the 2510-2555 fixtures that landed without one.
- island ('any') arguments at intrinsic/libCall/callValue slots take the validated exit (dynCheck's stance); non-exit-safe slots fence with the slot named, and the boundary pass backstops every jsOp/dynCall/dynInvoke argument so world-mixing sites can never ICE
- a checker-'any' receiver/callee that LOWERED checked-dynamic dispatches to the DOM machinery: ambient-this reads are dynKeyGet, rest-args map/forEach ride the runtime dispatch, dyn callees take dynCall
- Object.entries/values push unit-armed (null | undefined) fields as the null literal; new Set(seed) exits an island-handle seed as a validated copy
- fence-bodied functions emit even when their ABI names an unregistered class, and a run()-end sweep rewrites leftover unregistered-class type slots to the inert f64 placeholder — the clang invalid-C escape closes with the ICE
- RegExp[] compiles (REF elements, both backends; map callbacks returning other unrepresentable kinds meet a named fence), and the C emitter appends program-sized line arrays by loop, not spread (500k-line TUs overflowed the stack)
Chris Tate