11 Commits
Author SHA1 Message Date
Chris Tate b49052a389 Recursive record types compile statically as named recursive shapes
- mapType back-references mint placeholder shapes/unions finalized at frame exit; identity is per checker type, and context-sensitive (generic/mixin) knots stay fenced
- every type walk survives cyclic shape graphs: coinductive JSON/dyn-safety predicates, guarded formatIrType/jsvalLiftable, cycle headers via the existing collector fixpoint on both backends
- JSON.stringify throws V8's byte-exact circular TypeError (hop lines, ellipsis rule); cyclic values crossing into checked-dynamic slots trap instead of hanging
- console.log/util.inspect render <ref *N>/[Circular *N] exactly (SEMANTICS 129 resolved; depth:null lowers) and deepStrictEqual terminates through Node's pair memo
- corpus 2480-2488 green on C, LLVM, and the sanitized RC-audit lane, which proves child->parent cycles collect
2026-07-23 13:24:13 -05:00
Chris Tate 0803a0fcd3 Library mode: compile a TypeScript module set to a linkable archive
- scriptc build --lib --profile emits ccc wrappers, init/sink/reset/collect entries, and a static archive
- async_free is structural: the async and event-loop units never link, symbol-audit proven
- Every runtime trap routes through one funnel; the sink contract covers both emissions
- The K-fixture conformance matrix runs a real C host against both backends
2026-07-23 09:42:26 -05:00
Chris Tate b5e3d019fc Rename the capability: library mode, never "core", surface and internals
- CLI: the core verb becomes `scriptc build --lib --profile <p.json>` (no positional; profile names the entry); artifact renames to <name>.lib.a with .lib.ll/.lib.c/.lib.ir.json siblings
- SC4xxx diagnostic text and hints say library, never core; trap-message strings likewise (LIBRARY RC AUDIT, library reset registry, library inbound bytes)
- internals: compileLibrary/compileLibArchive, IrModule.lib + IrLibSection/IrLibExport, moduleLibAsyncSurface, libRoots, loadLibraryProfile/LibraryProfile under src/library/, frontend/lib-exports.ts, emitLibEntries/emitLibDefs
- runtime: scr_core.c -> scr_library.c with scr_library_* symbols and the SCR_LIB macro (scr_lib_* was taken by the stdlib unit); scr_lib_session_cleanup
- tests: tests/library-mode/ fixtures (entries lib.ts), library-*.test.ts harness files
2026-07-23 09:24:44 -05:00
Chris Tate 7bdd40453a node:querystring lowers statically — parse, stringify, escape, unescape
- scr_qs.c ports Node v24's legacy codec quirk-faithfully (the scan state machine with maxKeys' pair budget, unescapeBuffer's code-unit byte truncation with U+FFFD replacement decode, encodeStringified's value rules over the DOM crossing); link-gated by moduleUsesQs, and escape rides the always-linked component encoder
- parse's result is the call site's ParsedUrlQuery dictionary (the networkInterfaces verification stance; the runtime fills the overflow map and groups repeats into string[] buckets), sep/eq/maxKeys complete to Node's defaults, and custom encoder/decoder options fence by name
- decode/encode lower as Node's own aliases; both backends emit the same runtime calls byte-identically
- querystring joins SUPPORTED_BUILTIN_MODULES, the fallback declarations, and the regenerated surface manifest, so npm-static packages requiring it unguarded stop falling back to the island (the shim stays for island code)
- corpus 2460-2464 pin the grammar corners against Node: malformed escapes, '+' vs %2B, repeated keys, maxKeys with empty segments, multi-char/multi-byte separators, astral truncation, and every CJS acquisition spelling
2026-07-23 02:38:44 -05:00
Chris Tate fc2a467408 Funnel home fix + K8 fiber/loop symbol audit
- non-core scr_trap/scr_trap_fmt move from scr_console.c to scr_exception.c: the failure-channel unit is in every link, the runtime's own unit-test TU subsets included (the console unit is not)
- K8 additionally asserts no scr_loop/fiber/timer/child symbol is defined in or referenced by the archive — the design's mechanical backstop for the narrowed core base set
2026-07-23 02:00:43 -05:00
Chris Tate 50676fc9dd K-fixture conformance suite: K1-K10 over both emissions
- five generic fixtures under tests/core-mode (scalars, buffers, reinit, traps, init-trap) with C probe drivers; the harness flips the profile emission and pins identical outputs across the two lanes
- K1 nm symbol exactness both directions; K8 ambient audit (no sigaction/signal/pthread_create/atexit/setvbuf references — scr_init and scr_lib_init compile out under SCR_CORE)
- K3 covers both arena postures over one core; K5-K7 cover sink-exactly-once, init traps, pre-registration abort, Uncaught rendering, poisoned-core aborts; K10 reruns K4/K5/K7 under ASan + the RC audit's re-init zero-heap seam
- K9 drives SC4002/SC4003/SC4004/SC4005/SC4007 end to end with the profile-teaching rider verbatim
2026-07-23 01:37:42 -05:00
Chris Tate b5f6b0dc46 Cite patterns, not target projects, across comments and hints
- Lowering rationale names the idiom; one user-facing hint reworded
- Diagnostics snapshot re-pinned to the neutral hint text
2026-07-23 01:18:50 -05:00
Chris Tate 59274de9d7 Trap funnel: scr_trap/scr_trap_fmt over every core-linkable trap site
- executable expansion in scr_console.c is byte-identical fputs/vfprintf + abort; core expansion in the new scr_core.c poisons, delivers to the registered sink exactly once, aborts pre-registration or if the sink returns
- 89 fputs/fprintf+abort pairs and 10 bare OOM aborts across the 20 core-linkable TUs now funnel; lazy atexit teardowns (json errdyn, regex bytecodes, symbol registry) reroute to the core reset registry via scr_atexit
- scr_core.c owns sink/poison/result-arena/reset-registry/marshalling helpers; scr_exception.c renders escaped exceptions as "Uncaught ..." into the funnel; scr_lib exposes its cleanup to the reset seam
2026-07-23 01:14:32 -05:00
Chris Tate 871284a382 Mutable-default snapshots, npm named re-exports, and perf_hooks performance.now
- export default <mutable let> registers Node's snapshot as const storage keyed by the default alias symbol; resolveValueSymbol's default-hop walk stops importer reads there instead of chasing to the live let (corpus 2426 pins the snapshot/live split against export { x as default })
- named re-exports from installed npm packages lower as import-plus-export plumbing: the island load registers at the statement's position and bindings key the aliased symbols direct imports use (npm case npm-reexport covers named, renamed, CJS, and as-default spellings)
- node:perf_hooks joins the supported builtins: performance.now() lowers to scr_perf_now (the uptime anchor's monotonic clock, fractional ms), and performance.now.bind(performance) lowers as a () => number function value (corpus 2427); both backends
2026-07-22 23:21:59 -05:00
Chris Tate 5d6abeaefe JS-exact StringToNumber in the runtime, fuzz-verified against Node
- scr_string_to_number: ECMA-262 StringNumericLiteral exactly — StrWhiteSpace trim (empty → +0), signed decimals through strtod over the validated span, unsigned 0x/0o/0b through the exact bignum rounding, trailing garbage → NaN
- oracle suite mirrors the formatting precedent: gen-tonumber-cases.mjs (Node is the oracle), 50k committed cases asserting bit-equal doubles under ASan + RC audit
- fuzzed once against Node on 1,000,000 mixed-shape strings: 1000000/1000000 bit-exact
2026-07-22 20:02:10 -05:00
Chris Tate 23d5918a53 init 2026-07-22 18:08:04 -05:00