Docs reference bool types (bool64, bool128, bool256, bool512) that the compiler does not implement

[davidfrogley]

Summary

Web/src/docs/types/bool.md introduces a "family of explicitly sized boolean types — bool8 through bool512" and lists seven distinct widths. The compiler implements only three: bool8, bool16, bool32, plus bool as an alias for bool8. Code copied from the docs that uses bool64/bool128/bool256/bool512 fails at name resolution.

Compiler reality (source of truth)

Rux/Include/Rux/Type.h:17-50 — TypeRef::Kind defines exactly:

Bool8, Bool16, Bool32,
...
Bool = Bool8,   // alias

No Bool64 / Bool128 / Bool256 / Bool512 exist anywhere in the compiler.

Type-name resolution accepts only "bool", "bool8", "bool16", "bool32":

• Rux/Source/Hir.cpp:220-222, 687-689
• Rux/Source/Sema.cpp:460-462, 1008-1010

For independent corroboration, the codebase already gets this right elsewhere: Web/src/blog/language-without-llvm.md:100 correctly lists "bool8, bool16, bool32" — only this docs page disagrees.

Repro

let a: bool64  = true;   // unknown type
let b: bool128 = true;   // unknown type
let c: bool256 = true;   // unknown type
let d: bool512 = true;   // unknown type

All four declarations should compile per the docs; none do.

Every doc/asset that needs editing

Web/src/docs/types/bool.md

• L5 — overview: "Rux provides a family of explicitly sized boolean types — bool8 through bool512".
• L13-21 — types table; rows for bool64, bool128, bool256, bool512.
• L28 — example let c: bool64 = true;.
• L41 — recommendation aside referencing wider types ("FFI requirements or SIMD vectorization") — wording can stay if bool32 covers it; remove the implication that wider widths exist.
• L57 — example let z: bool128 = true;.
• L129 — "Wider boolean types (bool32, bool64, etc.) also support bitwise operations." bool64 doesn't exist; fix the wording (e.g., "bool16 and bool32" or just "wider boolean types").
• L166 — recommendation: "Reserve wider types for SIMD, FFI, and memory-layout-sensitive contexts." Same issue as L41; either point at bool16/bool32 specifically or scope the claim down.
• L222 — note: "Wider boolean types (bool256, bool512) are primarily intended for SIMD vector masks and low-level bit manipulation. They may require aligned memory on certain platforms." Whole sentence is fictitious.

Web/src/examples/Bool.rux

• L5-8 — bool64, bool128, bool256, bool512 declarations.

(L1-4 and L9 onward are correct and should stay.)

Web/.vitepress/grammars/rux.tmLanguage.json

• L562 — bool type-name regex: \b(bool|bool8|bool16|bool32|bool64|bool128|bool256|bool512)\b — should be \b(bool|bool8|bool16|bool32)\b.

Doc that's already correct (don't touch)

• Web/src/api/index.md:415-419 — bool8 Display extension entry is real and matches the stdlib.
• Web/src/blog/language-without-llvm.md:100, 128 — already lists the right three widths and uses bool32 for FFI.
• Web/src/docs/types/slices.md:63 — bool8[] example is correct.

Suggested fix

Trim docs/example/grammar to match the compiler — bool, bool8, bool16, bool32 only. If wider boolean types are on the roadmap, gate the relevant sections behind a "Planned" / "Not yet implemented" callout instead of presenting them as available.

The "Wider boolean types are primarily intended for SIMD vector masks" sentence at L222 is worth a closer look from a language-design perspective, not just a docs cleanup: the existing bool32 already covers Win32 BOOL interop and 4-byte SIMD lanes; if there's no concrete use case for bool128+, the cleanest fix is to drop those plans rather than promise them.