AGRA RULING (issue 0089, attempt 7): bare reserved-name MEMBER positions are
intentionally exempt from the reserved-type-name rule, and the implementation
already does the right thing — this is a docs + one-example change, no code.
The exempt member positions are struct FIELD names, union TAG names, and protocol
method-SIGNATURE names: they sit in a member slot, are reached via obj.name (or
dispatched by string), and are never type-classified, so they never mis-lower.
The backtick is optional there. The exemption stops at member DEFINITIONS: an
impl method is a real function (reached through the impl_block -> fn_decl arm), so
a reserved-spelled impl method still needs the backtick, exactly like a free
function (cf. examples/1122) — and every bare reserved-name value binding /
declaration name still errors (0076 preserved).
- specs.md / readme.md: replace the "every binding site" / "any binding site"
overclaim with the precise rule — required positions (value bindings +
declaration names + impl method definitions) vs the exempt member-name
positions (field / tag / protocol signature; backtick optional).
- examples/0158-types-reserved-name-member-exempt.sx: pins the exempt behavior —
bare reserved-name struct fields + union tag read & written bare AND via
backtick, and a protocol with a bare reserved-name method dispatched through
the protocol (impl definition takes the backtick).
- issues/0089: document the member-name exemption in the RESOLVED banner + add
0158 to the regression list.
Gate: zig build, zig build test, bash tests/run_examples.sh — all green
(430 passed, 0 failed, 0 timed out).
The codegen-side resolver was already raw-aware for the universal model;
the sema/LSP editor index (the second classifier) only honored the DIRECT
raw type. A COMPOUND raw type (`*`s2`, `?`s2`, `[N]`s2`, `[]`s2`, `[*]`s2`)
stores its inner type-name as a bare string on the Type info struct, and
every resolution site re-read it with skip_builtin=false — so the index
reclassified a user type named `s2` as the builtin int, diverging from
codegen (issue-0083 class, LSP surface only; codegen unchanged).
Structural cure: every compound info struct (Pointer/Optional/Slice/
ManyPointer/Array) carries a REQUIRED is_raw bit (no default — a future
construction site cannot drop it). is_raw is set at every construction
site (resolveTypeNode arms, fieldType arms, variadic slice, .ptr/slice_expr
derivation, for-loop by-ref, substType) and passed as skip_builtin at every
resolution site (elementTypeOf, field-access pointer unwrap, index, deref,
optional unwrap/null-coalesce, if/while optional binding, match subject).
Optional-unwrap + deref sites converted from Type.fromName/pointerPointeeType
(builtin-only, divergent) to resolveTypeNameStr(name, is_raw); the now-dead
pointerPointeeType removed.
Tests: src/sema.test.zig gains pointer/optional/array raw-vs-bare
regressions (raw → user type, bare → builtin control) — each FAILS on
pre-fix sema, PASSES after — plus a parameterized-raw coverage test.
Closes the remaining three F0.6 findings so the universal backtick raw
identifier holds in BOTH classifiers and at EVERY parser construction site.
1. Struct-body constants thread is_raw + name_span. The struct-body const
forms (untyped `` `s2 :: 5 `` and typed `` `s2 : T : v ``) built the
const_decl node without name_span/is_raw, so a backtick const was falsely
rejected and a bare reserved-name const caretted at 1:1. They now capture
both. Structural cure: `ast.ConstDecl`'s name_span + is_raw carry NO
default, so the compiler rejects any construction site that omits them
(mirrors checkBindingName's required `is_raw` arg). FnDecl keeps its
defaults — every parser fn_decl routes through parseFnDecl whose
`name_is_raw` is a required parameter (equivalent guarantee).
2. Raw identifier in TYPE position flows through the normal continuations.
parseTypeExpr no longer returns a terminal type_expr for a raw atom; the
raw flag rides the atom through the qualified-path / Closure / parameterized
continuations, so `` `s2(s64) ``, `` *`s2 ``, `` ?`s2 `` all parse.
ParameterizedTypeExpr carries is_raw; resolveParameterizedWithBindings
skips the `Vector` intrinsic when raw.
3. sema/LSP (the second classifier) honors is_raw. Type.fromTypeExpr returns
null for a raw type_expr; resolveTypeNode skips the builtin classifier when
raw; resolveTypeNameStr takes a skip_builtin arg threaded from te/id.is_raw
(compound inner names pass false). A backtick reserved-name annotation now
resolves to the user type in the editor index, not the builtin.
Tests: examples/0156 (struct-body const), 0157 (parameterized raw type +
wrappers), 1142 (bare struct-body const errors, caret on name); src/sema.test.zig
pins the LSP raw-type resolution (fail-before verified). Gate: 365 unit tests,
429 examples, 0 failed.
AGRA ruling (attempt 4): `` `name `` is THE LITERAL identifier `name`, usable in
EVERY position — the backtick only means "treat this token as a plain identifier,
never the reserved keyword/type", and is never part of the name's text.
- Raw in TYPE position is now VALID (reverses attempt-2 "raw is not a type"):
`parseTypeExpr` emits a raw `type_expr`; `TypeResolver.resolveNamed` gains a
`skip_builtin` flag (threaded from `te.is_raw` via lower.zig + type_bridge) so a
`` `s2 `` reference resolves to a `` `s2 ``-declared type (struct/enum/union/alias),
else a normal "unknown type 's2'" error (reportIfUnknownType skips the builtin
exemption when raw). Bare `s2` in type position stays the builtin int.
- Every declaration-name site is is_raw-exemptible: `is_raw` added to TypeExpr +
StructDecl/EnumDecl/UnionDecl/ErrorSetDecl/ProtocolDecl/ForeignClassDecl/UfcsAlias/
NamespaceDecl/ImportDecl/CImportDecl/LibraryDecl; parser threads name_is_raw to
every decl parse fn; namespace imports carry it through imports.addNamespace.
Typed-const path (`` `s2 : s64 : 5 ``) now threads name_span+is_raw (fixes the
1:1-caret bug).
- Check<->exemption made structurally symmetric: checkBindingName/checkDeclName take
is_raw as a REQUIRED argument and skip inside the check, so no call site can
validate a name without honoring the exemption (the desync cause of prior rounds).
- Bare reserved-name declarations of every kind still error (0076 preserved);
`#import c` foreign names stay auto-raw + bare-callable.
specs.md + readme.md updated to the universal model. issue 0089 RESOLVED banner
rewritten. Examples: replace 1139 (raw-not-a-type) with 0154 (raw type reference);
add 0155 (typed const + union tag) and 1141 (bare type-decl negatives).
Gate: zig build + zig build test + run_examples (426 passed, 0 failed).
A bare reserved-type-name `::` declaration was silently accepted, and the
attempt-2 lowerCall rewrite then made a bare `s2 :: (…) {…}` function callable —
bypassing the backtick rule for handwritten sx. The reserved-name binding check
covered `:=` / typed-local / param / captures but NOT the `::` declaration form.
- ast: `ConstDecl`/`FnDecl` carry `is_raw` + `name_span` threaded from the parser
(parseConstBinding / parseFnDecl, all call sites incl. struct/impl methods).
- semantic_diagnostics: reject a bare reserved spelling at EVERY declaration-name
site — const, function (incl. struct/impl methods), struct/enum/union/error-set,
protocol, foreign-class, ufcs alias, namespaced/library/c-import name. Backtick
(`is_raw`) and the compiler's `#builtin` definition (`string :: []u8 #builtin`)
are the only exemptions; a value whose node is itself a named decl defers to
that node's own check.
- c_import: synthesized foreign fn_decls are `is_raw = true`, so a C function
whose own name collides with a reserved spelling (`int s2(int);`) imports and
bare-calls unedited.
- lower: scope the `.type_expr`→`.identifier` call rewrite to a callee FnDecl of
RAW provenance (`is_raw`) — only a backtick / `#import c` foreign fn can carry a
reserved-name spelling, so a non-raw match never gets rewritten.
- examples: 0153 (positive — backtick `::` const + fn, bare + tick call), 1140
(negative — bare `::` const + fn rejected).
- docs: specs.md + readme.md state the backtick is required at every binding site
including `::` const / function / type declarations; issue 0089 banner updated.
Completes the issue-0089 backtick raw-identifier / `#import c` exemption
across all remaining identifier positions and closes three boundary gaps
the F0.6 review found.
1. Exhaustive raw-binding coverage. The `is_raw` bit now threads through
`ast.Identifier` and EVERY binding/capture form — `IfExpr`/`WhileExpr`
optional bindings, `ForExpr` capture + index, `MatchArm` capture,
`CatchExpr`/`OnFailStmt` tag bindings, `DestructureDecl` per-name, and
the protocol-default-body / foreign-class method param lists — not just
`var_decl`/`param`. `UnknownTypeChecker` skips the reserved-name check at
each arm when raw, so a backtick works in every identifier position while
a bare reserved spelling still errors (issue 0076 preserved).
2. Raw identifier is never a type. `parseTypeExpr`'s atom rejects a raw
identifier in type position (`x : `s2 = 1`, `List(`s2)`) with an accurate
diagnostic instead of silently type-classifying it.
3. Reserved-name function bare-callable. A bare `s2(4)` parses its callee as
a `.type_expr` (reserved spelling); `lowerCall` now rewrites a type_expr
callee to an identifier when a function of that name is in scope, so a
backtick-declared sx fn and a `#import c` foreign fn whose C name collides
with a reserved type spelling both resolve by their bare name.
(`TypeName(val)` is not a cast, so there is no ambiguity.)
Tests: examples/0152 (every control-flow/capture form + bare ref/call/member
access), examples/1054 (catch/onfail tag bindings), examples/1139 (raw in
type position rejected), examples/1220 extended (foreign reserved-name
function bare-call). 0076 negatives 1119/1121/1122/1123/1124/1125 stay green.
Gate: zig build + zig build test + 422 examples pass. specs.md + readme.md
updated; issues/0089 RESOLVED banner refreshed.
Reserved type-name spellings (s1, s2, u8, …) can now be used as value
identifiers two ways, resolving issue 0089:
1. Backtick raw identifier: a leading backtick (`s2) lexes to an
.identifier token carrying a new Token.is_raw flag, with the backtick
excluded from the text. A raw identifier is never type-classified — the
parser skips Type.fromName for it — so it is always a value identifier.
The flag threads to VarDecl.is_raw / Param.is_raw at binding sites, and
the reserved-type-name check (UnknownTypeChecker) skips raw bindings.
Because the token tag stays .identifier, the escape works in every
position (local, global, param, field, fn name, struct member, later
reference) with no per-site parser change.
2. #import c exemption: c_import.zig synthesizes foreign decls with
Param.is_raw = true, so generated C param names that collide with
reserved type names (s1, s2) import unedited.
A bare reserved-name binding in sx still errors (issue 0076 preserved):
the is_raw-gated skip only fires for backtick / foreign names, and a raw
binding's address-of / autoref lowering stays correct because every
occurrence is an .identifier, never a .type_expr.
Tests: examples/0151 (backtick, every position),
examples/1220 (foreign exemption, compiled+run), lexer unit tests.
1119 (bare-binding rejection) stays green. specs.md + readme.md updated.
emitCmpNe lowered float `!=` to `LLVMRealONE` (ordered not-equal), which
is false when either operand is NaN. That made `nan != nan` false in
native code — breaking the canonical `x != x` NaN test, making `!=`
non-complementary with `==` for NaN, and disagreeing with the interpreter.
Change the float predicate to `LLVMRealUNE` (unordered not-equal): true
if either operand is NaN OR they are unequal. For all non-NaN operands
`UNE` ≡ `ONE`, so only NaN-involving comparisons change (toward correct).
The integer predicate (`LLVMIntNE`) and `emitCmpEq` (`OEQ`) are unchanged,
so `nan == nan` stays false and `!=` is now the exact complement of `==`.
- Regression: examples/0150-types-float-ne-unordered-nan.sx (fails before,
passes after; also pins #run/comptime == runtime agreement).
- specs.md: documents float comparison / NaN semantics (Operators).
- Resolves issue 0091 (issues/0091-float-ne-ordered-nan.md).
A field-like access on a builtin INTEGER type name folds to a compile-time
constant of the queried type, driven by (width, signedness) arithmetic:
sN: min=-(2^(N-1)), max=2^(N-1)-1; uN: min=0, max=2^N-1
for every width s1..s64 / u1..u64 (not just power-of-two), plus usize/isize.
- type_resolver.zig: extract the single width parser (parseWidthInt) reused by
resolveNamed AND the new accessors (no second parser — issue-0083 class);
add resolveBuiltinName / integerWidthSign / integerLimitBits / integerLimitFor.
- lower.zig: lowerNumericLimit intercept beside the error.X / Struct.CONST /
pack-arity identifier-receiver intercepts; folds ints via constInt, emits a
clean diagnostic for a non-numeric receiver (bool/string/void/Any/noreturn),
falls through for floats (NL.2).
- expr_typer.zig: mirror the result type so inferExprType reports the queried type.
- program_index.zig: recognize the accessors in the comptime-int / array-dim path
so [u8.max]T (255) / [s16.max]T (32767) work; [u64.max]T is rejected oversized.
- u64.max / usize.max stored as the all-ones bit pattern with TYPE u64 (i64 -1),
asserted via union { u: u64; s: s64 } reinterpret.
Docs: specs.md numeric-limits subsection (formulas + result-type + u64 note);
readme.md language overview. Examples 0148 (positive) / 0149 (negative-receiver).
Unit tests for the value computation in type_resolver.test.zig.
Gate: zig build, zig build test (359/359), tests/run_examples.sh (416 ok, 0 failed).
Per Agra ruling: user-facing sx changes must update all relevant docs
including readme.md. Replaces the prior 'Original syntax sketches. Do
not modify.' rule so the docs-track-changes review criterion is
enforceable from the next step onward.
The count description claimed every count must be "positive integral",
which is wrong: zero is context-dependent. Verified at HEAD — an array
dimension (`[0]s64`) and a generic value-param count (`Box(0)`, $N:u32)
both accept zero as a length-0 instantiation, while a `Vector` lane
count stays strictly positive (`Vector(0,f32)` rejected). Negatives are
rejected for array dims and unsigned value-params, but a signed
value-param accepts a negative; only the integral requirement (folds
4.0, rejects 4.5) is common to all three.
Split the count paragraph into per-consumer bullets stating the exact
range each accepts. Range-bound paragraph unchanged. Pin the zero
contrast with examples 0147 (array-dim + value-param zero accepted) and
1505 (Vector zero-lane rejected). No compiler-code change.
specs.md lumped `inline for` / `for` range bounds in with counts (array
dimension, Vector lane count, generic value-param count) under the
count negative-rejection rule. A range bound is a range ENDPOINT, not a
count: negative endpoints are valid and an empty/inverted range runs zero
iterations. The compiler already implements this correctly (Agra ruling:
spec-text bug, no code change).
- specs.md: counts and range bounds are now described separately. Counts
reject negatives; bounds accept any compile-time integer (negatives
valid, integral floats fold) but still reject a non-integral float
because the loop cursor must be an integer.
- examples/0612-comptime-inline-for-range-bounds.sx: `inline for -2..1`
and `for -2..1` both sum -3; `inline for 0..(-2.0)` runs zero
iterations (empty range). Runtime/comptime parity asserted.
- examples/1138-diagnostics-inline-for-non-integral-bound.sx: a
non-integral float bound `inline for 0..4.5` is a clean diagnostic,
exit 1 (must-be-integer still applies to bounds).
Count consumers (1132/1133/1134/1135) unchanged and green.
A failed value-param bind on a type-returning function (e.g.
`MakeC :: ($K: Count, $T: Type) -> Type { return [K]T; }` with
`a : MakeC(5_000_000_000, s64)`) emitted its correct range diagnostic
but then `instantiateTypeFunction` returned `null`, so
`resolveParameterizedWithBindings` fell through to an empty-struct
placeholder named after the function. The binding `a` got that
placeholder type, so a later `a.len` cascaded a bogus second error
`field 'len' not found on type 'MakeC'`.
The struct binder (`instantiateGenericStruct`) already returns
`.unresolved` here; the type-fn binder now matches it — a failed
value-param bind poisons to `.unresolved` instead of `null`, so the
caller propagates the diagnosed poison and the existing
`emitFieldError` suppression yields one clean diagnostic. Covers
every type-fn value-param failure mode: overflow via an aliased
constraint, a non-const arg, and an unknown type arg.
Regression: examples/1137-diagnostics-value-param-type-fn-no-cascade.sx
Three adjacent cells of the shared count surface still diverged from the
rest; all now route through the same leaf+fold+narrow+diagnose path.
1. Aliased integer constraint bypassed the value-param range gate — only
builtin constraint names matched intTypeRange, so Box(5_000_000_000)
with `$K: Count` (Count :: u32) compiled and bound a truncated value.
resolveValueParamArg (shared by both the struct AND type-fn binder) now
resolves the constraint to its underlying builtin via
canonicalIntConstraintName (Count -> u32, Small -> s8) before
range-checking, so an aliased integer constraint behaves exactly like
the builtin it names.
2. A named const with an expression RHS (M :: 2; N :: M + 1) did not fold
as a count — moduleConstInt read only a literal RHS node. It now folds
every const's RHS through the shared evalConstIntExpr, cycle-guarded
(mutual / self cycles fold to null, not a stack overflow), and pass-0
pre-registers expression-RHS consts. N :: M + 1 == 3 at every consumer:
dim (direct + alias), Vector lane, value-param (struct + type-fn),
inline for.
3. Stateful resolveArrayLen still fabricated length 0 after a failed fold;
it now returns null -> the .unresolved sentinel (no fabrication). The
binding's lowering never reaches sizeOf (alloca defers it; hasErrors
aborts first) and a field access on an already-diagnosed .unresolved
value is poison-suppressed (emitFieldError), so a failed-fold dim emits
ONE clean diagnostic with no panic.
Regressions: examples/0146 (full positive matrix — every consumer x leaf
form), 1135 (aliased u32 + s8 overflow), 1136 (direct non-const dim halts
cleanly). The cascade cleanup also tightened 1502/1503 to one diagnostic.
Unit test added for moduleConstInt expression-folding + cycle detection.
Item 2 (Agra ruling): a compile-time INTEGRAL float (`4.0`, `N : f64 :
4.0`, `N :: 4.0`) used as an array dimension / Vector lane / generic
value-param count / `inline for` bound now folds to its integer at the
shared leaf — `program_index.floatToIntExact`, used by both the
`.float_literal` arm of `evalConstIntExpr` and `moduleConstInt`. All four
consumers route through the one evaluator, so `[4.0]s64` lays out the same
`[4]s64` uniformly; a non-integral (`4.5`) or negative value stays
rejected by the downstream `foldDimU32` gate. Pass-0 now pre-registers
float-valued module consts for forward-alias parity with int consts.
Item 1: a generic value-param bind (`Box($K: u32)`) never range-checked
the folded arg, so `Box(5_000_000_000)` compiled and ran. The bind now
range-checks against the param's declared type — a `u32` count through the
shared `foldDimU32` gate (making program_index's "single u32 gate for
value-param counts" doc true), any other integer type through the new
`program_index.intTypeRange` — and emits a clean "value N does not fit in
u32 parameter K" otherwise. The declared type is threaded via a new
`TemplateParam.value_type`.
Regressions: examples 0145 (integral-float array dim), 1504 (Vector lane),
0611 (inline-for bound), 0209 (value-param integral-float), 1132
(non-integral float dim rejected), 1133 (negative float dim rejected),
1134 (oversized u32 value-param rejected) + program_index float-fold unit
tests. Gate: zig build, zig build test, 406/0 run_examples.
The stateless alias-registration array-dim path collapsed foldDimU32's
distinct .too_large / .below_min outcomes into null, so an oversized type
alias (Big :: [5000000000]s64) emitted the FALSE 'an array dimension is not
a compile-time integer constant' message while the direct form correctly
reported 'array dimension 5000000000 does not fit in u32'.
Add program_index.reportDimError as the single source of dim-error wording
(the stateful path now emits through it too) and type_bridge.foldArrayDim to
surface the DimU32 reason at the alias-registration site. An oversized/negative
alias dim now routes to reportDimError for the same precise message as the
direct form; a genuinely non-const alias dim keeps the alias-specific message.
Regression: examples/1131-diagnostics-array-dim-oversized-u32-alias.sx
Two remaining siblings in F0.4's comptime-int path.
1. Type-returning function with a value param used as a TYPE annotation
(`b : Make(N, s64)` where `Make :: ($K: u32, $T: Type) -> Type`):
- `isValueParamPosition` (semantic_diagnostics) now also skips a value
param of a `fn_ast_map` type-returning function, so `N` is not walked
as the type name "N" ("unknown type 'N'").
- `resolveParameterizedWithBindings` routes a type-returning-function
name to `instantiateTypeFunction` (the `.call` path already did).
- `instantiateTypeFunction` resolves a general return-type expression
(`return [K]T`) with bindings active — not just struct/union returns.
`Make(N, s64)`, `Make(M + 1, s64)`, `Make(3, s64)` all resolve to one
`[3]s64`.
2. Oversized dim/lane fold panicked the compiler (0087): an array dim /
Vector lane folded to a valid i64 (5e9) then narrowed to u32 with an
unchecked `@intCast`. New single gate `program_index.foldDimU32` folds
via `evalConstIntExpr` then range-checks `[min, maxInt(u32)]`; the three
narrowing sites (resolveArrayLen stateful + stateless, resolveVectorLane)
all route through it and emit a clean diagnostic + halt instead of
panicking. Value-param args stay i64 until used as a dim/lane, where the
same gate checks them.
Regressions: examples/0208 (value-param type function), examples/1130
(oversized array dim clean halt), examples/1503 (oversized Vector lane
clean halt). Marks issue 0087 RESOLVED.
Gate: zig build, zig build test, bash tests/run_examples.sh — 398 passed,
0 failed, 0 timed out.
While fixing 0083 (attempt 5) noticed a distinct, pre-existing bug:
writing to a Vector component (`v.x = 1.0`) aborts with "unresolved type
reached LLVM emission" in emitStore. Reading a lane works; a literal lane
count triggers it, so it is NOT the lane-count class. Confirmed
reproducible on the pristine pre-attempt-5 compiler (not introduced by
the lane-count fix). The standard vector idiom (`.[…]` construction +
component reads / arithmetic, examples/1500) is unaffected. Filed for a
separate session; not worked around here.
Attempts 1–4 fixed the array-dimension paths but the same length-0
fabrication class survived on every other site that resolves a
compile-time integer. Unify them all on the single shared
`program_index.evalConstIntExpr` so they cannot diverge:
- All three Vector lane resolvers (resolveTypeCallWithBindings,
resolveParameterizedWithBindings, resolveArrayLiteralType) and both
generic value-param binders (instantiateGenericStruct,
instantiateTypeFunction) hand-rolled an `else => 0` switch. A
module-const lane `Vector(N, f32)` fabricated a 0-lane `<0 x float>`
(LLVM "huge alignment" abort); a value-param `Vec(N, f32)` fabricated
a 0 binding / wrong mangled name. They now fold through the shared
evaluator and emit a clean diagnostic + `.unresolved` on a non-const
operand (resolveVectorLane / resolveValueParamArg) — never 0.
- evalComptimeInt (inline-for bounds) delegated to the shared evaluator,
so `inline for 0..M` / `0..(M+1)` fold like array dims. The `<pack>.len`
leaf moved into the shared folder via a new `ctx.lookupPackLen`.
- The unknown-type semantic checker no longer walks a value-param
position (`Vector(N, …)` / `Vec(N, …)`) as a type name (was reporting
"unknown type 'N'").
- The parameterized-type-arg parser and the function-body lookahead
(hasFnBodyAfterArrow) accept a const-EXPRESSION in a value position, so
`Vector(M + 1, f32)` and `[M + 1]T` parse as a return type too (the
latter a pre-existing array-dim sibling that the same heuristic broke).
Regressions: examples/1501 (named-const + const-expr lane, direct +
alias, 3/4-lane reads), 1502 (runtime lane clean-halts, exit 1, no LLVM
crash), 0207 (Vec(N)/Vec(M+1) == Vec(3) instantiation), 0610 (inline-for
const bounds). Shared-evaluator unit test extended with the pack-len arm.
zig build && zig build test && bash tests/run_examples.sh: 395 passed,
0 failed.
A constant-FOLDABLE expression array dimension (`[M + 1]`, `[M * N]`,
`[N - M]`, nested `[M + N - 1]`, parenthesised `[(M + 1) * 2]`, mixing
untyped and typed module consts) was wrongly rejected as "not a
compile-time integer constant" even though every operand is
compile-time-known. Attempts 1-3 resolved only a bare named-const dim or
a literal; an expression dim must be EVALUATED, not rejected.
Fix: the shared dim resolver now routes the dimension through a single
constant integer-expression evaluator (`program_index.evalConstIntExpr`)
that folds integer `+ - * / %` and unary negate over literals and
named/typed module consts, recursively (parentheses carry no AST node).
The leaf-name lookup is delegated via `ctx.lookupDimName`, so the
stateful body-lowering path (`Lowering`, which also sees comptime
constants and generic `$N` values) and the stateless registration path
(`type_bridge.StatelessInner`, module consts only) share the EXACT SAME
folding logic and cannot diverge — an expression dim via a type alias
resolves identically to the direct form.
No-fabrication discipline unchanged: a genuinely non-comptime dimension
(runtime local, non-comptime call, unbound name) or arithmetic that
overflows / divides by zero still yields null -> `.unresolved` -> the
same clean compile-halting diagnostic, never a fabricated length.
- examples/0144-types-const-expr-array-dim.sx: every expression form,
direct vs alias, scalar / string / struct element types (fails on the
pre-fix compiler, passes after).
- examples/1129 re-pointed at a genuinely non-const dimension
(`[get()]s64`, a runtime call) so it still proves the stateless
clean-halt (a foldable expression is no longer an error).
- program_index.test.zig: unit test for evalConstIntExpr folding and
clean-halt-on-non-const.
A type alias whose dimension is a named const (`Arr :: [N]T`) resolves its
dimension eagerly during scanDecls pass 1, on the stateless registration path,
which can only read `module_const_map`. Typed consts (`N : s64 : 16`) register
only in pass 2 and a forward-declared untyped const had not registered yet, so
the stateless resolver saw an empty table, printed a non-fatal warning,
fabricated length 0, and continued — yielding a 0-byte alloca, garbage reads,
and a segfault for slice/struct elements.
- scanDecls pass 0 pre-registers every integer-valued module const before any
type alias resolves, so typed, untyped, and forward-referenced consts all
resolve identically.
- Both dim resolvers now share `program_index.moduleConstInt`, so the stateful
body-lowering path and the stateless registration path cannot diverge.
- `resolveArrayLen` returns `?u32`; `resolveCompound` yields `.unresolved` on
null instead of a 0-length array. The stateful path emits a diagnostic; the
alias-registration path surfaces an unresolved alias as a clean compile error
that aborts the build. The Vector lane-count `else => 0` is fixed the same way.
Regressions: examples/0143 (typed-const dim direct + via alias for s64/string/
struct, forward-ref alias, nested) and examples/1129 (an unresolvable computed
dim halts with a clean diagnostic + non-zero exit). Both fail on the pre-fix
compiler (garbage/segfault; warning+exit0) and pass after.
Makes the F0.4 fixes exhaustive across every resolution / nesting path.
0083 — named-const array dimension, stateless paths. Attempt 1 fixed the
stateful resolver (direct local decls, struct fields, params, returns) but the
binding-free registration-time resolver (`type_bridge`, used for type aliases
`Arr :: [N]T` and inline union/enum field types) still resolved a named dim with
a silent `else 0`, so `Arr :: [N]s64; a : Arr` and `union { a: [N]s64 }` were
still miscompiled (garbage / bus error). Thread the module-global const table
(`ProgramIndex.module_const_map`) into `type_bridge` alongside the alias map, so
`StatelessInner.resolveArrayLen` resolves a named module-const dim to the same
length everywhere. The remaining unresolvable case (a computed/comptime dim on
the binding-free path, which the stateful path hard-errors) now bails LOUDLY
instead of fabricating a 0 length.
0085 — nested slice-literal elements. `lowerArrayLiteral` lowered each element
with the element type as target but appended the raw value. A nested `.[...]`
element at a slice element type (`[][]s64`) still lowers to an aggregate array
`[N]T`, so the outer aggregate held raw arrays where slice {ptr,len} headers
were expected — indexing the inner slice read a garbage pointer and segfaulted.
After lowering each element, coerce a same-element array to the slice element
type via the existing `array_to_slice` op. The coercion recurses with the
nesting, so `[][]T` and deeper materialize at every level — local-bound AND
direct-call-argument forms.
Regressions (fail-before/pass-after demonstrated on the pre-fix compiler):
examples/0140-types-named-const-array-dim.sx — extended with type-alias,
nested [N][M]T, and union-field named dims (s64 / string / struct elems)
examples/0142-types-nested-slice-literal-elements.sx — [][]s64 + [][]string,
local-bound vs direct-arg
src/ir/type_bridge.test.zig — named-const dim resolves to literal length
Gate: zig build, zig build test, bash tests/run_examples.sh (388 passed).
Issues 0083 and 0085 marked RESOLVED.
Two silent-miscompile codegen fixes:
0083 — named-const array dimension. `TypeResolver.resolveCompound`'s array
arm resolved the dimension with `if int_literal ... else 0`, so a named const
(`N :: 16; [N]T`) hit the silent `else 0`: the array became 0-length / 0-byte
and element access ran out of bounds (garbage for scalars, bus error for
slice/pointer/struct elements). The arm now delegates the dimension to
`inner.resolveArrayLen` (symmetric with `inner.resolveInner` for the element).
The stateful `Lowering.resolveArrayLen` evaluates it as a compile-time integer
across the comptime-constant / generic-value / module-global const tables and
emits a diagnostic — no fabricated length — when it isn't one.
0084 — `.[...]` literal passed directly as a call arg. `lowerArrayLiteral`
always yields an aggregate array value; the array→slice conversion is the
caller's job. The local-bound var-decl path did it, but the call-arg coercion
path had no array→slice arm, so `classify([N]T, []T)` returned `.none` and the
raw array was passed where a slice was expected (callee read its {ptr,len}
header off the wrong bytes → 0 / garbage / segfault). `classify` now returns a
new `.array_to_slice` plan for same-element `[N]T → []T`, and `coerceToType`
emits the existing `array_to_slice` op — identical to the local-bound path.
Regressions (fail-before/pass-after demonstrated on the pre-fix compiler):
examples/0140-types-named-const-array-dim.sx (s64 + string + struct elems)
examples/0141-types-slice-literal-direct-call-arg.sx (string + []s64)
Gate: zig build, zig build test, bash tests/run_examples.sh (387 passed).
Issues 0083 and 0084 marked RESOLVED.
Example 0717 now asserts the (token, index) Diag for ALL SIX raise sites
in cli.sx, closing the two the reviewer found still unasserted:
- zero-arg UnknownCommand: parse([], ...) -> index -1, token ""
(the args.len == 0 sub-branch of cli.sx:237, distinct from the
one-arg too-few form already covered at index 0 / token args[0]).
- TooManyFlags (cli.sx:256): a command declaring 17 flag specs (> the
inline 16 cap) is rejected, not truncated -> index -1, token command.
The three index==-1 cases (zero-arg, too-many, missing-req) seed their
Diag with a sentinel before parse, so each assertion proves parse WROTE
the -1/"" rather than merely matching the `.{}` default. Verified
non-vacuous: flipping any expected value makes that line FAIL.
Test-only: cli.sx logic and src/ are untouched.
Extend example 0717 to pin the offending token VIEW and its args index
for every failure the parser's Diag populates: unknown-command,
unknown-group, too-few-args, missing-value, value-eats-flag, and the
missing-required index. Closes the test-coverage gap flagged in review;
cli.sx parser logic unchanged.
Extend std/cli.sx with a zero-heap argument parser that the caller drives
over a logical argv ([]string), separate from the F3.1 os_args accessor.
Grammar: <group> <command> [--flag VALUE | --bool]... [--json] [-- rest...]
- (group, command) dispatched against a caller-provided Command table;
no match -> error.UnknownCommand.
- value-taking vs boolean flags fixed by each command's FlagSpec list;
--json is a reserved global boolean surfaced as parsed.json.
- `--` or the first bare operand ends flag parsing; the remainder is
parsed.rest (operand views).
Heap discipline (heap-discipline.md): zero heap, zero copy. group/command/
flag values/rest are all VIEWS into args. Parsed is a by-value stack struct;
flag presence/values live in a fixed [16]FlagValue inline array indexed by
spec position (no per-flag allocation, no context.allocator). The flag-spec
list and command table are caller storage passed as views.
Failure surfacing (no silent skip): unknown command, unknown flag, a
value-flag missing its value, and an absent required flag each raise a
specific CliError variant; a caller-owned Diag records the offending token
(index + view) before each raise, since error tags carry no data.
examples/0717 drives the parser over explicit []string vectors: a valid
group/command/--flag/--bool/--json case (asserting parsed values + that
values are views into argv), subcommand dispatch, `--`/bare-operand
separators, and the five failure variants each asserted via destructure +
Diag. zig build && zig build test && run_examples.sh green (385 passed).
The global-init constant serializers in emit_llvm.zig printed a diagnostic
on an unserializable value and then RETURNED an undef/null placeholder and
CONTINUED emitting. For a comptime `#run` global that yields a function
reference (`fp :: #run pick();` where pick returns a function), the build
fell through to the JIT and segfaulted calling through the undef pointer
(exit 134) — a silent miscompile dressed up as a printed error.
Route every genuine bail in the serialization family through a new
`failGlobalInit` helper: it sets `comptime_failed` (so core.generateCode
aborts with a non-zero exit after emit()) and returns an undef placeholder
that never ships, because the halt fires before object emission / JIT. This
covers the comptime func_ref leaf, the require_resolved aggregate func_ref
leaf, the top-level + vtable func_ref globals, the comptime-init catch, and
the remaining heap-walk / aggregate-shape bails. Unresolved-function
diagnostics now name the function instead of its (stdlib-unstable) IR index.
The require_resolved=false Pass-0 placeholder is unchanged (func_map is
empty until Pass 1; the aggregate is re-emitted with require_resolved=true).
Regression: examples/1128-diagnostics-comptime-global-funcref-rejected.sx —
a `#run` global returning a function ref now exits 1 with the diagnostic
(was: exit 134 segfault). Fail-before/pass-after verified.
A module-global initialized with an enum literal silently zero-initialized
to the first tag (`chosen : Color = .green` read back as `.red`), and an
enum tag inside a global array/struct was rejected as non-constant. The
constant serializer had no enum-literal arm.
Add `Lowering.constEnumLiteral`: serialize an enum literal to a
`ConstantValue.int` holding the variant's tag value, resolved against the
destination enum type and respecting explicit variant values; the global's
type drives the backing width at emit time. Wired into `globalInitValue`
(scalar global) and `constExprValue` (array element / struct field / nested
aggregate). A non-enum destination or unknown variant is diagnosed loudly,
never silently zero-initialized. The compiler-injected OS/ARCH globals now
serialize to their real `.unknown` tag (6 / 4); runtime reads are unchanged
(they resolve through comptime_constants), so only the static initializer in
the pinned .ir snapshots changes.
Remove the silent `func_ref => orelse LLVMConstNull` fallbacks in the LLVM
constant emitters: aggregate func_ref leaves carry a `require_resolved` flag
(transient null in Pass 0, loud diagnostic if still unresolved in the
Pass-1.5 re-emit), a top-level func_ref global is resolved in
initVtableGlobals, and the comptime (#run) path bails loudly instead of
emitting a null function pointer.
Regression: examples/0139-types-global-enum-literal-init.sx (scalar, array,
struct field, explicit-value enum u16 stride, struct-array with enum field);
negative: examples/1127-diagnostics-global-enum-literal-bad-variant.sx.
Mark issue 0082 RESOLVED.
A module-global aggregate initializer rejected a `null` literal in a
pointer (or optional-pointer) field as "must be initialized by a
compile-time constant". `Lowering.constExprValue` had no `.null_literal`
arm, so the null leaf returned no constant and the whole aggregate looked
non-constant — even though `null` is the compile-time zero pointer (a
top-level scalar `p : *s64 = null;` already serialized fine).
Add `.null_literal => .null_val` to constExprValue. While here, make the
two LLVM constant emitters exhaustive: emitConstAggregate and the
top-level init_val switch in emit_llvm.zig previously ended in a silent
`else => LLVMConstNull(...)` catch-all (the silent-arm class CLAUDE.md
mandates rooting out). They now handle every ConstantValue tag explicitly
(.null_val/.zeroinit -> all-zero constant, .undef -> LLVMGetUndef,
.func_ref resolved, nested .vtable is a hard @panic tripwire). The
reject-loud path for genuinely non-constant fields is preserved.
Regression: examples/0138 (array-of-struct null ptr fields, array of
all-null pointers, nested struct-in-struct null ptr) and the negative
examples/1126 (null ptr field beside a non-const field still errors).
Fail-before/pass-after verified.
A module-global array of struct literals (`pairs : [2]Pair = .[ .{...}, .{...} ]`)
was emitted as `zeroinitializer`, silently dropping every declared field — reads
returned 0 with no diagnostic. Global struct literals and struct-with-array
already worked; the gap was struct literals used as ARRAY elements.
Root cause: `Lowering.constExprValue` (the const-aggregate serializer for global
initializers) had no `.struct_literal` arm. `constArrayLiteral` serialized each
element through `constExprValue`, so a struct-literal element returned null,
collapsing the whole array initializer to null; `globalInitValue` then emitted no
payload and the LLVM backend zero-initialized the global — the same silent-zero
class as 0071/0072, one level inside an array literal.
Fix: make `constExprValue` type-aware — thread the destination element/field
TypeId so a struct-literal leaf routes through `constStructLiteral` and a nested
array-literal through `constArrayLiteral` with the correct element type.
`constArrayLiteral` derives its element type from the array TypeId;
`constStructLiteral` passes each field's type. A global aggregate initializer that
still does not fully reduce to a compile-time constant is now rejected loudly
(`diagnoseNonConstGlobal`) instead of silently zeroing. `emitConstAggregate`
already recurses over nested aggregates, so `sx run` (JIT) and `sx build` (AOT)
both materialize the declared values.
Regression: examples/0137-types-global-aggregate-literal-init.sx (global
[N]Struct literal, global struct literal, struct-with-array, nested
array-of-struct-with-array; values read back with no prior store, plus a store on
top). Fails on the pre-fix compiler (array-of-struct fields read 0), passes after.
Marks issues 0079 (already resolved) and 0080 RESOLVED.
A store to a module-global array element (`g[i] = v`) was silently dropped:
a subsequent `g[i]` read the array's initializer, not `v`. Constant index,
variable index, and cross-function stores were all affected, in both `sx run`
and `sx build`. Global scalars and local arrays were fine.
Root cause: `Lowering.lowerExprAsPtr` (the lvalue/address path) handled only
local identifiers. A module-global identifier fell through to the value
fallback `lowerExpr`, which emits `global_get` — loading the whole array by
value. The LLVM backend's `emitIndexGep` then allocas a throwaway temp, copies
the value in, and GEPs into the temp, so the store wrote a discarded copy.
Fix: teach `lowerExprAsPtr`'s identifier arm about globals — emit `global_addr`
(a pointer into the global's live storage), or `global_get` for a pointer-typed
global (mirroring the local pointer case). Route the `address_of(index_expr)`
array base through `lowerExprAsPtr` too so `&g[i]` is likewise an lvalue into
the global. `index_gep` now GEPs directly into the global for const and variable
index, across functions. This also fixes global struct field stores, which
shared the same root cause.
Regression: examples/0136-types-global-array-element-store.sx (const-index,
var-index, cross-function store on a scalar global array; struct-element array
for stride; nested-array global for the recursive lvalue). Fails on the pre-fix
compiler, passes after.
Add library/modules/std/cli.sx: a pure-sx command-line argument accessor
backed by the macOS C runtime (_NSGetArgv/_NSGetArgc), no compiler change.
os_argc() -> s64
os_args(buf: []string) -> []string
Zero heap, zero per-arg allocation: os_args fills a caller-provided buffer
(stack array) with string VIEWS over the process's own argv block, which
lives for the whole process. The returned slice header is a by-value stack
return; nothing touches context.allocator.
Documents the `sx run` reality: under `sx run <prog.sx> ...` the process
argv is the interpreter's argv (sx, run, prog.sx, ...), not a program's
logical args. This accessor reports the real process argv truthfully;
mapping to logical args is a later consumer concern (distribution P3.1).
Non-macOS platforms bail loudly (message + _exit) rather than returning a
silent empty.
examples/0716-modules-cli-argv.sx asserts only deterministic structural
invariants (argc >= 1, argv[0] non-empty, os_argc() == filled length).
Add 0715-modules-json-suite as the single comprehensive pinned suite for
std.json (mirrors 0711 for std.hash), alongside the focused 0713/0714 demos:
- ROUND-TRIP build->write->parse->write over a document covering EVERY value
kind (a string with every escape form \" \\ \b \f \n \r \t plus a \u00XX
control, integers 0 / negative / s64 MIN / s64 MAX, bool, null, array,
nested object) with insertion-order assertions, exact writer bytes, and
parse-then-rewrite idempotence.
- DECODE positives: \/, the full named-escape set, \uXXXX (BMP 1- and 2-byte)
plus a surrogate pair, the escaped control forms, and raw multi-byte UTF-8
round-tripping through writer + reader.
- MALFORMED matrix: one assertion per JsonParseError variant and its key
edges (UnexpectedToken, UnexpectedEnd, BadEscape, BadNumber incl. leading
zero / lone '-' / fraction / exponent / overflow, TrailingGarbage,
BadControlChar), each asserted to raise.
Pure test work: src/ and library/ untouched, no json.sx change needed. Every
model is built through an explicit Arena allocator (heap discipline).
parse_string scanned for `"` and `\` but accepted every other byte,
including raw control characters. RFC 8259 §7 requires those bytes to be
escaped inside a string; an unescaped one is invalid JSON and must surface
a parse error, not be silently accepted.
Add `BadControlChar` to JsonParseError and reject any unescaped byte < 0x20
in the string body scan (which gates the decode path too, so escaped forms
like \t/\n/ still decode correctly; 0x20 and 0x7F are not over-rejected).
Regression test in examples/0714: raw 0x09/0x0A/0x00 each raise
BadControlChar via `?`/`!`; a positive case proves the escaped forms still
decode to the right bytes. All prior assertions kept.
Issue 0078 (string == as an and/or operand emitting an invalid PHI) is
resolved on this branch, so the example no longer needs the split that
worked around it. Restore the natural combined assertion
sub.items[0].key == "k" and sub.items[0].val.str == "v"
(one nested-pair report), and the in_range containment helper to
return x >= lo and x < hi;
Drop the now-stale issues/0078 references. Re-captured expected stdout
(nested-key/nested-val -> nested-pair). json.sx and src/ untouched.
A string `==`/`!=` used as an operand of a short-circuit `and`/`or` emitted
invalid LLVM (`PHI node entries do not match predecessors!`). String compares
expand into their own memcmp sub-CFG during LLVM emission, so the operand
finishes in a later basic block (`str.merge`) than the one the IR block
started in. `fixupPhiNodes` wired the short-circuit merge PHI's incoming edge
to `block_map[ir_block]` (the block the IR block started as), recording a
stale predecessor (`%entry`/`%and.rhs.0`).
Fix: record the builder's actual insertion block after emitting each IR
block's instructions (`term_block_map`, via `LLVMGetInsertBlock`) and use it
as the PHI predecessor. General — corrects the incoming block for any operand
that emitted intermediate basic blocks (string `==`, value `match`, …), not
just string `==`.
Regression: examples/0045-basic-string-eq-short-circuit.sx (string `==` on
both sides of `and` and of `or`, plus a match-value + enum-payload `==` shape).
Fails (LLVM abort) pre-fix, passes after.
Add the JSON reader (parser) to library/modules/std/json.sx, the inverse
of the F2.1 writer over the same value model: insertion-ordered objects,
arrays, strings (full unescaping incl. \uXXXX + surrogate pairs), s64
integers, bool, null.
Heap discipline (binding): exactly two allocation kinds, both through the
EXPLICIT `alloc` parameter, never the implicit context allocator —
composite backing stores (Array/Object.items via add/put) and decoded
escaped-string buffers (bounded by the raw span). Un-escaped string
values are zero-copy VIEWS into the input buffer (valid only while it
lives); scalars carry no heap.
Failure surfacing (hard contract): malformed input raises a meaningful
JsonParseError variant (UnexpectedToken / UnexpectedEnd / BadEscape /
BadNumber / TrailingGarbage) on the error channel, never a bogus value.
Trailing non-whitespace is TrailingGarbage; fractions/exponents,
out-of-s64 magnitudes, and leading zeros are BadNumber. Number
accumulation runs in negative space so s64 MIN parses exactly.
examples/0714-modules-json-reader.sx asserts the parsed structure
(insertion order, every kind), proves the view-vs-decoded heap split by
pointer containment, round-trips back through the writer byte-for-byte,
decodes a surrogate-pair into 4 UTF-8 bytes, and checks every malformed
variant.
Filed issues/0078: a string `==` (or any sub-CFG operand) used in a
short-circuit `and`/`or` emits invalid LLVM IR (stale PHI predecessor),
hit while writing the example's assertions and worked around there by not
combining comparisons with `and`/`or`. src/ untouched.
Close the coverage gap from attempt 1: example 0713 now builds integer
fields holding s64 MIN (-9223372036854775808) and s64 MAX
(9223372036854775807) — plus zero, a small negative, and a small positive —
and asserts the EXACT emitted bytes. This permanently pins the edge that
write_int is specifically engineered for (folding positives into negative
space so MIN's non-representable-positive magnitude serializes correctly).
s64 MIN is expressed as (0 - 9223372036854775807 - 1) because its magnitude
is not a representable positive s64 literal.
Test hygiene: stream to a repo-local, gitignored .sx-tmp/ path (created if
missing) instead of a fixed /tmp name, and unlink it right after read-back
so nothing leaks. Writer/model logic and src/ are untouched.
Add library/modules/std/json.sx — the JSON value model and writer
(reader lands in a later step).
Value model: a tagged union over null/bool/integer(s64)/string/array/
object. Objects are an ORDERED list of (key,value) pairs preserving
INSERTION ORDER (no hash map, never sorted/deduped). Integers only — no
fraction/exponent this milestone.
Heap discipline:
- Scalars carry no heap; string values are VIEWS into caller memory
(never copied into the node).
- Composite nodes (Array/Object) own growable child storage, allocated
through an EXPLICIT allocator parameter on the builder methods
(arr.add(v, alloc) / obj.put(key, val, alloc), mirroring List.append)
— never the implicit context allocator.
- The writer adds ZERO output allocations: it emits into a caller-
provided Sink, either a fixed []u8 buffer (overflow raises, never
truncates) or streaming straight to an fs.File through a small caller
staging buffer (no whole-document string; peak memory O(staging)).
Integer digits format in a stack [20]u8; s64 MIN is handled by
formatting in negative space. Sink/IO/overflow surface on the !
error channel.
examples/0713-modules-json-writer.sx builds a nested object + array +
string with every escape kind + negative int + bool + null, then asserts
the EXACT bytes (insertion order, escaping) from both the buffer sink and
the file-streaming sink, plus the overflow-raises path.
