A union, unknown or any value is converted to a dynamic value before it
is thrown, and promise rejections and other runtime throws carry dynamic
values too. A catch binding then holds a dynamic payload, which the
native class test does not recognize: `e instanceof C` was false for
every class of a hierarchy, and a read narrowed by it trapped. A test
through a parameter of type unknown was already correct.
When a class or one of its subclasses crosses into unknown, the class
dynamic-dispatch pass now routes the catch-binding tests of that class
(`e instanceof C`, the reads it narrows and `e as C`) through helpers
that keep the native test and answer a dynamic payload with the dynamic
class test, as standalone classes already do. The new "dynamic" caught
test selects that payload, so native payloads are not converted.
Programs whose classes never cross into unknown compile as before.
fs.promises.rm(path, options) accepts the { recursive, force, maxRetries,
retryDelay } literal options that rmSync accepts. It runs the same removal
and rejects with its errors.
Removing a directory without recursive now throws Node's ERR_FS_EISDIR
SystemError from rmSync and fs.promises.rm. A retryDelay outside int32 >= 0
or a maxRetries outside uint32 throws ERR_OUT_OF_RANGE. rmSync checks these
options before the path, and fs.promises.rm checks the path first, as Node
does.
Fixes#1615
Co-authored-by: jlucaso1 <55464917+jlucaso1@users.noreply.github.com>
The runtime's ToIntegerOrInfinity for string index arguments (slice,
substring, charAt, indexOf, startsWith, padStart, repeat and the others
in scr_string.c) converts through int64 instead of calling trunc, which
is a libm call on baseline x86-64. Each slice made two such calls.
The frontend's syntax trees and program state refer to each other, so
reference counting frees almost none of them when a build disposes the
frontend. Whether a cycle pass reclaimed them before the build ended
depended on where the collector's schedule happened to be. Passes during
the frontend found its data live and widened the growth fractions, so the
next pass waited until the heap had grown well past its size at disposal,
and the dead frontend and the emitter's garbage stayed until the build
ended.
The native compiler now runs scr_collect_cycles_phase right after it
disposes the frontend, through a host function. The runtime entry point
sweeps, then restarts the growth fractions and the death rule's backoff
from the heap the sweep leaves, as after a productive pass. Under Node
the hook is unset.
When a script thread parks its allocator arena, each class slice still
holds the pages past its bump pointer that were faulted in but never
carved: the rest of the huge page the bump pointer is in, and the pages
the pre-fault helper faulted in ahead of it. The same applies to region
chunks when the thread exits. Until another thread claims the arena,
nothing carves there, so a program that continues on threads that never
claim it (parse threads ahead of a check on the main thread) kept them.
- scr_sa_park_arena returns those pages with MADV_DONTNEED. The next
thread to claim the arena faults them in again as it carves, and they
read as zero, as fresh pages do.
- scr_rgn_thread_exit returns each chunk's pages past its bump on worker
threads; nothing carves in an exited thread's regions.
- scr_mem_release_ahead drops pre-fault requests that overlap the range
and waits for one in progress before returning it, so the helper does
not fault the pages in again.
- scr_alloc.test.c checks that a worker that shares and carves past the
huge-page switch leaves no resident pages past its frontier, keeps its
blocks, and that the next thread carves on from the frontier, zeroed.
A generic arrow or method held by an object literal is also lowered with
its type parameters erased, for the literal's own value. In that lowering
the object it returns builds as a checked-dynamic value, so a callback
member with an optional or defaulted parameter, such as
`decode(value, at = [])`, met a slot type that maps to a dynamic value.
The completed-signature check refused it with SC1090, although the same
code in a standalone generic function compiled.
A slot whose type maps to a dynamic value now takes the function the same
way an 'any' slot does: the value boxes into a checked-dynamic function,
which completes omitted arguments with undefined.
Fixes#1373
Co-authored-by: blurname <50226652+blurname@users.noreply.github.com>
TypeScript accepts parentheses and non-null assertions around the
reference an assignment, a `++`/`--` update or a for-in/of head writes
(`(x) = 1`, `h.sums[k]! += 10`), and both are erased at run time. The
compiler now reads its source files with the AstFile writeTargets option:
the writer's `left`, `operand` or `initializer` is the reference itself
and the reference's `parent` is the writer, so lowering and every analysis
of a binding's or member's writes see one shape. The reference keeps its
own source span. Files read without the option, such as type nodes from
the checker and the AST conformance tests, keep TypeScript's exact tree.
Fixes#1625
Co-authored-by: ejc3 <1694850+ejc3@users.noreply.github.com>
ClassDynamicDispatch.process runs once per reachability wave and builds
dispatch arms for every boxed class. Several lookups inside it grew with
the program and repeated for every dispatch or member:
- propertyReceivers was recomputed at every call site, including once
per property dispatch. It now returns its last plans for a mode while
the boxed classes, the registered classes, the subclass links
(Lowerer.subclassLinks) and the constructed classes
(Lowerer.constructedClassCount) are unchanged, and each boxed class
still maps to the same ClassInfo. In canonical mode it also skips the
subtree of a class it has already seen, which adds no plans.
- The property dispatch loop tests the answered-class set before the
member lookups, and the method and protocol loops test candidates in
a Set instead of with Array.includes. A dynamic method call scans the
boxed constructors for a static method only when no instance candidate
answers the method, the only case that reads the result.
- Lowerer.liftedFnNamed indexes liftedFns by name and replaces the
linear liftedFns.find scans in the dispatch, prototype, descriptor and
method-value helpers.
The emitted IR is unchanged.
A warm build that finds the program's LLVM in the cache read the file into
one string before emitting the program object. A program whose LLVM is
past the engine's string limit (three.js nodes, 816 MB) failed there with
`RangeError: Invalid string length`, though its cold build, which keeps
the LLVM in parts, succeeds. The cache hit now reads the file with
readTextParts: parts of at most 64 Mi UTF-16 units, decoded across reads,
none ending inside a surrogate pair, which the emitter takes as a cold
build hands them over.
A WebAssembly module the build finds but does not translate (one that uses
a proposal outside the translated feature set, an invalid one, or one the
translator rejects) now rejects with a CompileError that says the module
was found and names the reason, for example "it uses simd, outside the
translated WebAssembly feature set". Bytes the build did not see keep the
existing message.
The build emits the hash, length and reason of each such module in a
second table, scr_wasm_untranslated, next to scr_wasm_registry; the runtime
reads it when a lookup misses. WebAssembly.validate still answers false for
these modules, and the error stays a catchable CompileError.
Fixes#1617
Co-authored-by: jlucaso1 <55464917+jlucaso1@users.noreply.github.com>
TypedArray.of(...items) on the numeric typed-array constructors creates a
new array of its arguments. Each argument converts with ToNumber and then
to the element type, in argument order after all arguments are
evaluated, as in Node. Arguments that are all numbers copy directly;
other values and spreads collect into an array that goes through
TypedArray.from's conversion.
Spreading a dynamic iterable into a call's items (Array.of, push,
TypedArray.of) now drains it through the iterator protocol with the
spread call's TypeErrors, and a typed iterable spread into dynamic items
boxes and drains the same way. Before, both reached a run-time SC2020.
Fixes#1440
Co-authored-by: saboursepideh <25549001+saboursepideh@users.noreply.github.com>
A plain function's own properties (name, length, prototype, and the ones a
program assigns, as in `hash.outputLen = 32`) live in its property table.
Object.freeze, Object.seal and Object.preventExtensions of a function now
act on that table and answer the function, and Object.isFrozen,
Object.isSealed and Object.isExtensible answer from it, as in Node:
- Object.freeze of a function was refused as a possibly-aliased value,
and Object.isFrozen of one threw SC2020;
- Object.isExtensible answered false and Object.isSealed true for every
function.
Adding a property to a frozen or non-extensible function throws Node's
TypeError; assigning a read-only one throws the runtime's existing
read-only TypeError. Class constructors keep their fences. This lets
@noble/hashes' createHasher return its frozen hasher, so the sha256
program in the issue prints the digest Node prints.
Fixes#1437
Co-authored-by: saboursepideh <25549001+saboursepideh@users.noreply.github.com>
merge-profiles wrote the merged profile to `<output>.tmp` and renamed it
into place. Opening that fixed path followed an existing link there and
truncated its target. The temporary file is now created exclusively with
createUniqueFile next to the output, as emission does, and is removed when
writing fails or the stream reports an error.
everyExprChild, everyStmtChild, mapExprChildren and mapStmtChildren passed
an arrow function to every() or map() for each child list. In the native
compiler that arrow is a closure allocated and released at every visited
node, and each element passes through Array.prototype.every's callback
adapter. The child lists are now walked with index loops in the same
order, so every IR pass that traverses the module visits the same nodes
with less work per node.
A comparator that chains key differences with `||`, such as
`(b.score ?? 0) - (a.score ?? 0) || a.id - b.id`, ran through the
comparator sort: TimSort called the closure once per comparison. Each
term orders one number field, ascending or descending, and a term that is
0 or NaN falls through to the next, so the chain orders elements by their
keys with the first deciding. With no key NaN or -0 that order is
consistent, and a stable sort by the keys gives the comparator sort's
permutation.
The recognizer now accepts such chains of up to eight terms over number
fields of records and class instances, and a key may read an optional
number field with a number default (`x.k ?? 0`), also as a single key. The
runtime reads and checks every key of every element before it moves a
slot, then runs stable keyed passes from the last key to the first. A NaN
or -0 key, or a chain with any other term, keeps the comparator sort.
The frontend, lowering and LLVM emission of an optimized executable build run in a child
process when no earlier build of the same invocation emitted its native program object.
The parent passes the LLVM text to the native helper as the file the child wrote, so the
helper runs after the pages the frontend used are released. Later builds of the
invocation find most native partitions cached and peak in the frontend, so they, along
with development builds and builds with SCRIPTC_FRONTEND_PROCESS=0, prepare their input
in-process. The child runs with the parent's Node options without --eval, --print and
their code, or inspector, watch and test runner options. With SCRIPTC_TIMING=1, it
reports its phases on the parent's clock.
`a instanceof Uint8Array || (... a.BYTES_PER_ELEMENT === 1)`, the
@noble/hashes isBytes guard, narrows `a` to never on the right side when
the guard is specialized for a Uint8Array argument, so the read missed
the typed-array member lowering; it now answers the element size from
the receiver's storage. The checked-dynamic read of BYTES_PER_ELEMENT on
a typed array or Buffer (an untyped argument) answered undefined; it now
answers the element size, as each kind's prototype does in Node.
Refs #1437
Co-authored-by: saboursepideh <25549001+saboursepideh@users.noreply.github.com>
Each class slice a thread opens took one madvise(MADV_DODUMP) call,
which takes the mmap lock for writing, and the threads of a worker
executable open hundreds of slices as they start. The slices are now
made dumpable only when the kernel would write a core of the process:
when kernel.core_pattern starts with '|' or '@', or when RLIMIT_CORE is
above 0. The reservation stays MADV_DONTDUMP in every case.
The evolving-array analysis walked the binding's whole function and
resolved the symbol of every identifier in it, once per empty-array
binding, so a function with many such bindings cost quadratic time. It now
takes the file's identifiers spelled like the binding from the per-file
index (identifiersNamed), keeps those inside the binding's function, and
resolves only them.
A benchmark workload or the self-hosted compiler that fails to build
with both checkouts now fails the gate, and so does a run in which no
program built with both. Before this, a run whose builds all failed the
same way, such as with a helper of another version, reported PASS. A
build error now prints in full under its program; its first line is
often a warning.
An element write whose index the checker types as a number but whose
storage may hold undefined (a for...of variable over an array that may
have holes) converted the index with ToNumber only on plain-array writes.
Typed-array writes, compound and update writes, element assignments used
as values and element targets in destructuring assignment now convert it
the same way, as the element reads already do, instead of refusing it as
a non-number key.
Fixes#1622
Co-authored-by: ejc3 <1694850+ejc3@users.noreply.github.com>
forEach, map, filter, some, every, find, findIndex, findLast,
findLastIndex, reduce and reduceRight compile on every typed-array kind
through one shared helper per method shape. The receiver, the callback and
reduce's initial value evaluate in order before the helper runs; the
helper reads the length once and each element when it visits it, so the
callback's writes show in later elements. Callbacks receive the element,
its index and the typed array (reduce's receive the accumulator first).
map and filter return a new typed array of the receiver's kind, map
storing each result with that kind's element conversion, and a Buffer's
map and filter return a Buffer. reduce without an initial value throws
Node's TypeError on an empty typed array. A thisArg keeps the fence, as
for arrays.
Fixes#1620
Co-authored-by: ejc3 <1694850+ejc3@users.noreply.github.com>
Each call of a function makes a new function object for a generic arrow
or function expression bound inside it, so `make() !== make()` when make
returns the binding, and such values are distinct Map and Set keys. The
compiled instances are shared module functions, which cannot give that
identity, so a value use of such a binding keeps an SC1090 fence that
names the reason. Calls of the binding still compile.
An unannotated `const xs = []` (or `let`) inside a function, which
TypeScript types by the elements written before each read, takes the array
type its settled reads have: records pushed in a loop give a record array,
numbers a number[], strings a string[]. Reads inside closures, which the
checker leaves unsettled, do not count. When settled reads disagree or the
binding is reassigned, the binding is a dynamic array, as the same binding
already is at module scope. Before, such a binding was always number[] and
pushes of other values were refused.
Fixes#1623
Co-authored-by: ejc3 <1694850+ejc3@users.noreply.github.com>
A JavaScript package method that returns one of two literal shapes (a
result reply or a failure reply) through one declared interface with
optional members hands its consumer the members of the shape it built,
including the ones only one shape has. scriptc 0.3.4 dropped them (the
failure reply arrived as { version, id } and read as a result); the
corpus program keeps that case covered: direct and looped calls, an
async method and a sync one, JSON output, Object.keys and `in`.
Fixes#1314
Co-authored-by: devin-ai-integration[bot] <158243242+devin-ai-integration[bot]@users.noreply.github.com>
A typed array or Buffer in a compiled program is always a built-in kind
(programs cannot subclass them, add own properties or change their
prototypes), so Node 24's prototype chains for them are known exactly.
- A read of a member the type does not declare (in JavaScript, or under
@ts-ignore) answers undefined when that chain lacks the name, as Node's
GetV does: the base64 and hex feature probes (`typeof bytes.toHex`) that
libraries such as @noble/hashes make. Names Node has, including the
Buffer prototype's members on a Uint8Array that may hold a Buffer, keep
the fence.
- `"name" in bytes` with a literal member name answers from the same
chains, and `bytes.BYTES_PER_ELEMENT` is the element size.
- Array destructuring assignment into typed-array elements
(`[high[i], low[i]] = [h, l]`) evaluates the receiver, the index and the
value in order and writes the element as `b[i] = v` does.
Refs #1437
Co-authored-by: saboursepideh <25549001+saboursepideh@users.noreply.github.com>
Math.max and Math.min with a spread and other arguments (`Math.max(...xs,
5)`) pack their arguments, in order and with each spread flattened, into
one number[] that the runtime folds, as Math.hypot already does. The fold
keeps NaN poisoning, -0 below +0 and the empty fold's infinities, and the
arguments evaluate left to right before it. The refusal hint now names
the supported forms.
Fixes#1621
Co-authored-by: ejc3 <1694850+ejc3@users.noreply.github.com>
Chris Tate