Commit Graph

2 Commits

Author SHA1 Message Date
agra
12552e125d fix(ir): resolve named-const array dims (0083) + materialize literal slice args (0084)
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.
2026-06-04 08:22:45 +03:00
agra
f3bda369f6 refactor(ir): extract CoercionResolver (conversions.zig) for coercion planning (A4.3 step 2)
Coercion classification now lives in src/ir/conversions.zig behind a *Lowering
facade (CoercionResolver), mirroring CallResolver / GenericResolver /
ProtocolResolver. Two pure classifiers:
- classify(src, dst) -> CoercionPlan (15 kinds: no_op / unbox_any / box_any /
  closure_to_fn_reject / tuple_elementwise / optional_unwrap / void_to_optional /
  optional_wrap / erase_protocol / int_to_float / float_to_int / ptr_int_bitcast /
  widen / narrow / none) — the built-in coercion ladder.
- classifyXX(src, dst) -> XXPlan (unbox_any / no_op / erase_protocol /
  protocol_to_pointer / coerce) — the xx-operator head.

coerceToType and lowerXX now `switch (classify…)` then emit; branch order
mirrors the originals exactly and every arm reproduces the prior lowering — the
f32/f64 Any match dispatch, buildProtocolErasure (lowerXX) vs buildProtocolValue
(coerceToType), tuple/optional recursion, and the user-Into fallback + pointer
materialization + recursion-guard/diagnostics (which stay in lowerXX /
tryUserConversion). IR emission stays entirely in Lowering; the classifiers are
pure. lowerXX keeps the operand's lowered Ref type as src_ty. `.none` means no
built-in applies (pass through; the Into fallback runs) — no silent default.

New pub: isFloat / isIntEx / typeBitsEx / resolveConcreteTypeName (the classifier
reads them); coercionResolver() accessor. lower.zig net -54 lines.

conversions.test.zig drives CoercionResolver directly: the full classify ladder
(no-op, Any box/unbox, widen/narrow, int<->float, ptr<->int, optional
wrap/unwrap, void->optional, tuple, closure-reject, .none for two unrelated
structs), erase_protocol for a concrete source, and classifyXX (all 5 kinds incl.
protocol-to-pointer vs coerce and pointer-materialization -> coerce).

zig build, zig build test, tests/run_examples.sh (357/0) all green — no .ir churn.
2026-06-02 22:45:56 +03:00