fibers B1.0c: support params in abi(.pure) (read from registers)

Adversarial review of B1.0b found a param-bearing abi(.pure) function
emitted invalid LLVM ("cannot use argument of naked function" — loud
verifier error, not silent) because the param-alloca loop spilled the
args to stack slots, which a naked function cannot have.

Fixed forward — this ENABLES the B1.3 context-switch use case rather
than rejecting it: gate the param-alloca loop on fd.abi != .pure in
decl.zig (both body-lowering paths) and generic.zig. A naked function's
args stay in their ABI registers and are read directly by the asm body
(e.g. swap_context reads from/to from x0/x1); the LLVM args are
declared-but-unused, which the verifier allows.

examples/1803-concurrency-pure-asm-param.sx: naked add(a, b) reads x0/x1
(add x0, x0, x1; ret) -> 40 + 2 = 42. aarch64-pinned.

Pack abi(.pure) (variadic + naked — nonsensical, can't read a runtime
pack from registers) left unsupported: pack.zig's param loop is
intertwined with comptime-param/#insert handling, so that case still
hits the loud verifier error. Documented in the checkpoint.

Also updates PLAN-FIBERS / CHECKPOINT-FIBERS for B1.0 completion.
B1.0 complete. Suite green (725/0).

current/CHECKPOINT-FIBERS.md

@@ -4,55 +4,67 @@ Companion to [PLAN-FIBERS.md](PLAN-FIBERS.md). Update after every step (one step
 per the cadence rule). New corpus category: `18xx` concurrency.
 
 ## Last completed step
-**B1.0a (`abi(.pure)` lock commit) — DONE.** Plumbed the `is_pure` flag end-to-end and made
-emit bail loudly:
-- IR `Function.is_pure: bool` ([inst.zig](../src/ir/inst.zig)) — set from `fd.abi == .pure`
-  at both `declareFunction` decl sites ([decl.zig](../src/ir/lower/decl.zig)).
-- `funcWantsImplicitCtx` returns false for `.pure` (mirrors the `.c` skip, decl.zig:515) —
-  a `.pure` fn gets no synthetic `__sx_ctx`.
-- Both body-lowering paths bypass `lowerValueBody` for `.pure`: lower the asm body as
-  statements + cap with `unreachable` (a `.pure` body has no sx return — the asm rets
-  itself; this avoids the implicit-return diagnostic).
-- `emit_llvm` Pass 2 (~line 402) **bails loudly** when `func.is_pure`
-  ("`abi(.pure)` function '…' LLVM emission not yet implemented") via `comptime_failed`
-  (driver aborts nonzero) — NOT a framed body (whose epilogue would corrupt a context
-  switch's SP-in ≠ SP-out).
-- `examples/1800-concurrency-pure-asm.sx` — one host example (no `.build` pin; the bail is
-  host-independent, fires before any asm/instruction selection), locked to the bail snapshot
-  (exit 1, empty stdout, the loud diagnostic on stderr).
-- **Adversarial review (closed in-step):** the review caught that `is_pure` was set ONLY at
-  the two `declareFunction` decl sites — generic monomorphization
-  ([generic.zig](../src/ir/lower/generic.zig)) and pack expansion
-  ([pack.zig](../src/ir/lower/pack.zig)) create the `Function` via a different path and left
-  `is_pure` false, so a generic `.pure` instance silently shipped a framed body (returned 42
-  but leaked the prologue's stack adjustment — the exact corruption the lock prevents). Both
-  paths now set `is_pure` + route `.pure` bodies through the asm-only + `unreachable` cap.
-  Locked by `examples/1801-concurrency-pure-generic-bail.sx`. (The review's other CRITICAL —
-  a `.pure` *lambda* — is a **false positive**: `isLambda`'s return-type scan
-  (parser.zig:3652) breaks on the `abi` keyword, so a `.pure` lambda is unparseable and
-  `parseLambda`'s abi-handling is never reached. Latent `isLambda`/`parseLambda`
-  inconsistency, not a B1 concern.)
-- **Naming:** the sx-facing name is **`pure`** throughout (field, diagnostic); LLVM's
-  `naked` attribute is only the B1.0b lowering mechanism (per user direction — don't call
-  the function "naked").
-- `zig build && zig build test` green: **723 ran, 0 failed**.
+**B1.0b (`abi(.pure)` real emission) — DONE. B1.0 complete.** Replaced the emit bail with
+real LLVM `naked` emission:
+- `emit_llvm` declaration pass: for `func.is_pure`, add the LLVM `naked` + `noinline` +
+  `nounwind` attributes and **skip** the `frame-pointer=all` attribute (incompatible with a
+  frameless function). Pass 2 now emits the `.pure` body normally — `naked` makes the
+  backend emit it verbatim (the inline asm + its own `ret`) with no prologue/epilogue.
+- IR shape (verified): `; Function Attrs: naked noinline nounwind` / `define internal i64
+  @answer() #0 { entry: call void asm sideeffect "…ret…", ""() unreachable }` /
+  `attributes #0 = { naked noinline nounwind }`. The caller invokes it as an ordinary
+  `() -> i64` call (`.pure` is `call_conv == .default`).
+- `examples/1800-concurrency-pure-asm.sx` — now GREEN, aarch64-pinned (`.build {"target":
+  "macos"}`): runs end-to-end → **exit 42** on this host, ir-only on a mismatch; `.ir`
+  snapshot captured.
+- `examples/1801-concurrency-pure-generic.sx` (renamed from `-bail`) — the generic `.pure`
+  now emits a correct naked `answer__i64` (exit 42), proving generic.zig produces a naked
+  body, not a framed one. aarch64-pinned.
+- `examples/1802-concurrency-pure-asm-x86.sx` — x86_64 cross sibling (`.build {"target":
+  "x86_64-linux"}`, ir-only here): `.ir` locks `naked` + `movl $42, %eax` / `ret`.
+- Unit test `emit: abi(.pure) function gets the naked attribute (no frame-pointer)` in
+  `emit_llvm.test.zig` (asserts `naked` present, `frame-pointer` absent).
+- **B1.0c (review-hardening):** a param-bearing `.pure` fn emitted invalid LLVM (loud
+  verifier error "cannot use argument of naked function") because the param-alloca loop
+  wasn't gated. Fixed forward (this *enables* the B1.3 context-switch use case rather than
+  rejecting it): gated the param-alloca loop on `fd.abi != .pure` in decl.zig (both paths) +
+  generic.zig; a naked fn's args stay in registers (read by asm), declared-but-unused in
+  LLVM. Locked by `examples/1803-concurrency-pure-asm-param.sx` (`add(a,b)` → x0+x1 → 42).
+- `zig build && zig build test` green: **725 ran, 0 failed** + unit tests.
+
+### Earlier — B1.0a (lock + review hardening)
+Plumbed `Function.is_pure` (set from `fd.abi == .pure` at both decl sites + generic.zig +
+pack.zig); `funcWantsImplicitCtx` skips `.pure` (no synthetic ctx, like `.c`); all
+body-lowering paths bypass `lowerValueBody` for `.pure` (asm body + `unreachable` cap — no sx
+return); `emit_llvm` Pass 2 bailed loudly (since flipped to real emission). Adversarial
+review caught the generic/pack `is_pure` gap (a generic `.pure` silently shipped a framed
+body); closed + locked. The review's `.pure`-lambda CRITICAL was a false positive
+(unparseable — `isLambda` breaks on the `abi` keyword).
 
 ## Current state
-Stream A (atomics) is feature-complete (✅) and unblocks B2-channels. Stream B1: **B1.0a
-landed**; the `abi(.pure)` ABI is plumbed but emit deliberately bails (B1.0b flips it to
-real LLVM `naked` emission). No fibers/Io/scheduler code yet. Grounded floor facts:
+Stream A (atomics) is feature-complete (✅). Stream B1: **B1.0 complete** — `abi(.pure)`
+emits a real LLVM `naked` function end-to-end (decl, generic, pack paths), the substrate for
+the fiber context-switch. No fibers/Io/scheduler code yet. Grounded floor facts:
 - `context` is already an implicit `*Context` param (slot 0) + `push Context` is a stack
   `alloca` ⇒ **fiber-local for free**. Only shared root = `__sx_default_context` global
   (entry-point bind). B1.1 expected to be a **library convention** (spawn trampoline
   snapshots the spawner's ctx into slot 0), **likely zero compiler change** — probe first.
 - Inline asm works end-to-end (lower→emit→JIT, aarch64 + x86_64) — the `.pure` body reuses it.
+- **`.pure` with PARAMS works** (B1.0c, the B1.3 substrate): the param-alloca loop is gated
+  on `fd.abi != .pure` in decl.zig (both paths) + generic.zig — a naked fn's args stay in
+  ABI registers (read by the asm body), declared-but-unused in LLVM (verifier-legal).
+  Example `1803-concurrency-pure-asm-param.sx` (`add(a,b)` reads x0/x1). **Unsupported (loud,
+  not silent):** a `.pure` *variadic-pack* fn (pack.zig's param loop is intertwined with
+  comptime-param/`#insert` handling, and a naked fn can't read a runtime-sized pack from
+  registers anyway) → loud LLVM-verifier error for that nonsensical construct. Acceptable
+  boundary; a sharper sx diagnostic for it is a candidate polish, not a blocker.
 
 ## Next step
-**B1.0b (`abi(.pure)` real emission)** — per PLAN-FIBERS.md "Phases → B1.0 → B1.0b" and the
-kickoff prompt at the bottom of that file. Replace the emit bail with LLVM's `naked`
-attribute + asm-only body; pin `1800` aarch64 (run end-to-end → exit 42, capture `.ir`); add
-x86_64 cross sibling `1802` (ir-only); add an `emit_llvm.test.zig` unit test asserting the
-`naked` attr. Separate commit (cadence rule — B1.0a locked, B1.0b greens).
+**B1.1 (per-fiber `context` root) — probe-first.** Per PLAN-FIBERS.md "Phases → B1.1". Write
+a probe confirming a spawn trampoline can pass a snapshotted `Context` as slot 0 with no
+compiler change (grounded as likely zero-change); lock the behavior with an `18xx` example +
+a checkpoint note on the convention. Only if the probe surfaces a real gap (a path re-reads
+`__sx_default_context` mid-stack) does this become a compiler step.
 
 ## Known issues / capability gaps
 - **Orthogonal (not a B1 blocker):** default VALUES for comptime params don't bind on
@@ -117,4 +129,16 @@ x86_64 cross sibling `1802` (ir-only); add an `emit_llvm.test.zig` unit test ass
   corrupted the stack). Fixed generic.zig + pack.zig (set `is_pure` + asm-only `unreachable`
   cap); locked by `examples/1801-concurrency-pure-generic-bail.sx`. The review's `.pure`-
   lambda CRITICAL was a false positive (unparseable — `isLambda` breaks on `abi`). Suite
-  green (723/0). **Next: B1.0b (real `naked` emission).**
+  green (723/0).
+- **B1.0b** — real `naked` emission: emit_llvm declaration pass adds LLVM `naked`/`noinline`/
+  `nounwind` + skips `frame-pointer` for `func.is_pure`; Pass 2 emits the body verbatim (no
+  prologue). `1800` green aarch64-pinned (exit 42 + `.ir`); renamed `1801` → `-generic`
+  (generic `.pure` emits a naked body, exit 42); added x86_64 sibling `1802` (ir-only, `.ir`
+  locks `naked` + `movl $42, %eax`). Unit test asserts `naked` present + `frame-pointer`
+  absent. Suite green (724/0).
+- **B1.0c** — review-hardening: param-bearing `.pure` emitted invalid LLVM (loud verifier
+  error). Gated the param-alloca loop on `fd.abi != .pure` (decl.zig both paths + generic.zig)
+  — naked args stay in registers, read by the asm body (the B1.3 context-switch shape).
+  Locked by `examples/1803-concurrency-pure-asm-param.sx`. Pack `.pure` left unsupported
+  (loud, nonsensical). **B1.0 complete.** Suite green (725/0). **Next: B1.1 (per-fiber
+  context, probe-first).**

current/PLAN-FIBERS.md

@@ -1,8 +1,8 @@
 # PLAN-FIBERS — Stream B1 (fibers + Io + M:1 scheduler)
 
-> **STATUS: 🚧 in progress.** B1.0a (`abi(.pure)` lock commit) ✅ landed. Next step =
-> **B1.0b** (`abi(.pure)` real emission — LLVM `naked` attr). See the kickoff prompt at the
-> bottom.
+> **STATUS: 🚧 in progress.** B1.0 (`abi(.pure)` codegen) ✅ complete — emits a real LLVM
+> `naked` function end-to-end (decl / generic / pack paths; examples 1800/1801/1802 + unit
+> test). Next step = **B1.1** (per-fiber `context` root — probe-first, likely library-only).
 
 Carved from [PLAN-POST-METATYPE.md](PLAN-POST-METATYPE.md) Stream B (§B1) + the
 design-of-record [../design/execution-evolution-roadmap.md](../design/execution-evolution-roadmap.md)
@@ -152,23 +152,25 @@ B1.0 (`.pure`) forces these plumbing sites:
 
 ## Phases (xfail→green steps)
 
-### B1.0 — `abi(.pure)` codegen
-- **B1.0a (lock) — ✅ DONE** (commit pending). Carried `abi == .pure` into IR
-  `Function.is_pure`; threaded through `decl.zig` (`funcWantsImplicitCtx` skips `.pure` like
-  `.c`; both body-lowering paths bypass `lowerValueBody` for `.pure`, lowering the asm body +
-  capping with `unreachable`); `emit_llvm` Pass 2 **bails loudly** on `func.is_pure`
-  ("`abi(.pure)` function '…' LLVM emission not yet implemented", build-gating nonzero exit).
-  `examples/1800-concurrency-pure-asm.sx` (one host example, no `.build` pin — the bail is
-  host-independent) locked to the bail snapshot. Suite green (722/0).
-- **B1.0b (green) ← NEXT** — emit LLVM's `naked` attr
-  (`LLVMGetEnumAttributeKindForName("naked", 5)` + `LLVMCreateEnumAttribute` +
-  `LLVMAddAttributeAtIndex` at func index −1; shape per `nounwind` at emit_llvm.zig:1339);
-  emit the `.pure` body as the asm block only (no prologue/epilogue/ctx). Pin `1800`
-  aarch64 (`.build {"target":"aarch64-macos"}`) → runs end-to-end (exit 42) on this host,
-  ir-only on a mismatch; capture its `.ir` (asserts `naked` + the asm). Add an x86_64 cross
-  sibling `examples/1802-concurrency-pure-asm-x86.sx` (`.build {"target":"x86_64-linux"}`,
-  ir-only here). Add a unit test in `emit_llvm.test.zig` asserting the `naked` attribute is
-  present on a `.pure` function. Review the diff (no stray error text). Commit.
+### B1.0 — `abi(.pure)` codegen — ✅ COMPLETE
+- **B1.0a (lock) — ✅ DONE.** Carried `abi == .pure` into IR `Function.is_pure`; threaded
+  through `decl.zig` (`funcWantsImplicitCtx` skips `.pure` like `.c`; all body-lowering paths
+  bypass `lowerValueBody` for `.pure`, lowering the asm body + capping with `unreachable`) +
+  generic.zig + pack.zig; `emit_llvm` Pass 2 bailed loudly on `func.is_pure`. Locked by
+  `examples/1800-concurrency-pure-asm.sx` + the generic regression (review-found gap).
+- **B1.0b (green) — ✅ DONE.** `emit_llvm` declaration pass adds LLVM `naked` + `noinline` +
+  `nounwind` for `func.is_pure` and skips `frame-pointer=all` (incompatible with a frameless
+  function); Pass 2 emits the body normally (`naked` ⇒ verbatim asm + own `ret`, no
+  prologue). `1800` pinned aarch64 → exit 42 + `.ir`; `1801-concurrency-pure-generic.sx`
+  (renamed from `-bail`) proves the generic path emits a naked body (exit 42);
+  `1802-concurrency-pure-asm-x86.sx` x86_64 cross sibling (ir-only here, `.ir` locks `naked`
+  + `movl $42, %eax`). Unit test `emit: abi(.pure) function gets the naked attribute` asserts
+  `naked` present + `frame-pointer` absent. Suite green (724/0).
+- **B1.0c (review-hardening) — ✅ DONE.** A param-bearing `.pure` fn emitted invalid LLVM
+  (loud verifier error). Gated the param-alloca loop on `fd.abi != .pure` (decl.zig both
+  paths + generic.zig) so a naked fn's args stay in registers (read by the asm body) — this
+  *enables* B1.3's `swap_context(from, to)`. Locked by `1803-concurrency-pure-asm-param.sx`.
+  Pack `.pure` (variadic + naked, nonsensical) left unsupported → loud verifier error.
 
 ### B1.1 — per-fiber `context` root (probe-first; likely zero compiler change)
 - **B1.1a (probe + lock)** — write a probe (`.sx-tmp/`) + an `18xx` example that snapshots a

examples/1803-concurrency-pure-asm-param.sx

@@ -0,0 +1,25 @@
+// Stream B1 (fibers) — an `abi(.pure)` function with PARAMETERS reads its args
+// from ABI registers (the shape the fiber context-switch needs: `swap_context`
+// reads `from`/`to` from x0/x1).
+//
+// A naked function has no frame, so params are NOT spilled to stack slots — they
+// stay in their ABI registers and the asm body reads them directly. Here `a` is
+// in x0, `b` in x1 (aarch64 AAPCS), and the result returns in x0: `add x0, x0,
+// x1`. The lowering skips the param-alloca loop for `.pure` (decl.zig /
+// generic.zig); the LLVM args are declared-but-unused, which the verifier allows
+// (spilling them would emit `store i64 %0, …` → "cannot use argument of naked
+// function"). aarch64-pinned; runs end-to-end (exit 42), ir-only on a mismatch.
+//
+// Regression for an adversarial-review finding: before the param-alloca guard, a
+// param-bearing `.pure` fn emitted invalid LLVM (loud verifier error) instead of
+// a working naked function.
+add :: (a: i64, b: i64) -> i64 abi(.pure) {
+    asm volatile {
+        #string A
+        add x0, x0, x1
+        ret
+A
+    };
+}
+
+main :: () -> i64 { return add(40, 2); }

examples/expected/1803-concurrency-pure-asm-param.build

@@ -0,0 +1 @@
+{ "target": "macos" }

examples/expected/1803-concurrency-pure-asm-param.exit

@@ -0,0 +1 @@
+42

examples/expected/1803-concurrency-pure-asm-param.ir

@@ -0,0 +1,15 @@
+
+; Function Attrs: naked noinline nounwind
+define internal i64 @add(i64 %0, i64 %1) #0 {
+entry:
+  call void asm sideeffect "        add x0, x0, x1\0A        ret\0A", ""()
+  unreachable
+}
+
+; Function Attrs: nounwind
+define i32 @main() #1 {
+entry:
+  %call = call i64 @add(i64 40, i64 2)
+  %ca.tr = trunc i64 %call to i32
+  ret i32 %ca.tr
+}

examples/expected/1803-concurrency-pure-asm-param.stderr

@@ -0,0 +1 @@
+

examples/expected/1803-concurrency-pure-asm-param.stdout

@@ -0,0 +1 @@
+

src/ir/lower/decl.zig

@@ -2660,13 +2660,19 @@ pub fn lowerFunctionBodyInto(self: *Lowering, fd: *const ast.FnDecl, fid: FuncId
     const user_param_base: u32 = if (wants_ctx) 1 else 0;
     if (wants_ctx) self.current_ctx_ref = Ref.fromIndex(0);
 
-    for (fd.params, 0..) |p, i| {
+    // An `abi(.pure)` (naked) function has no frame: its params arrive in ABI
+    // registers and are read directly by the asm body (e.g. `swap_context`'s
+    // `from`/`to`). Spilling them to allocas would (a) need a frame and (b) emit
+    // `store i64 %0, …` — "cannot use argument of naked function" (LLVM verifier).
+    // Leave the LLVM args declared-but-unused (the verifier allows that); the asm
+    // references the registers.
+    if (fd.abi != .pure) for (fd.params, 0..) |p, i| {
         const pty = self.resolveParamType(&p);
         const slot = self.builder.alloca(pty);
         const param_ref = Ref.fromIndex(@intCast(i + user_param_base));
         self.builder.store(slot, param_ref);
         scope.put(p.name, .{ .ref = slot, .ty = pty, .is_alloca = true });
-    }
+    };
 
     // Inbound entry points + abi(.c) sx functions: bind current_ctx_ref
     // to the static default before any user code runs.
@@ -2812,7 +2818,10 @@ pub fn lowerFunction(self: *Lowering, fd: *const ast.FnDecl, name: []const u8, i
     const user_param_base_lf: u32 = if (wants_ctx_lf) 1 else 0;
     if (wants_ctx_lf) self.current_ctx_ref = Ref.fromIndex(0);
 
-    for (fd.params, 0..) |p, i| {
+    // `abi(.pure)` (naked): params arrive in registers, read directly by the asm
+    // body — no frame, no alloca/store (which the LLVM verifier rejects on a
+    // naked function). See the sibling guard in the other body-lowering path.
+    if (fd.abi != .pure) for (fd.params, 0..) |p, i| {
         const pty = self.resolveParamType(&p);
         // Allocate stack slot for param, store initial value.
         // Refs 0..N-1 are reserved for function parameters by beginFunction.
@@ -2820,7 +2829,7 @@ pub fn lowerFunction(self: *Lowering, fd: *const ast.FnDecl, name: []const u8, i
         const param_ref = Ref.fromIndex(@intCast(i + user_param_base_lf));
         self.builder.store(slot, param_ref);
         scope.put(p.name, .{ .ref = slot, .ty = pty, .is_alloca = true });
-    }
+    };
 
     // Inbound entry points + abi(.c) sx functions: bind
     // current_ctx_ref to &__sx_default_context. See companion comment

src/ir/lower/generic.zig

@@ -128,7 +128,10 @@ pub fn monomorphizeFunction(self: *Lowering, fd: *const ast.FnDecl, mangled_name
     defer scope.deinit();
     self.scope = &scope;
 
-    {
+    // `abi(.pure)` (naked): no frame — params arrive in registers, read by the
+    // asm body, never spilled to allocas (the LLVM verifier rejects a naked
+    // function that uses its arguments). Mirrors the decl-path guard.
+    if (fd.abi != .pure) {
         var param_idx: u32 = if (wants_ctx) 1 else 0;
         for (fd.params) |p| {
             if (isTypeParamDecl(&p, fd.type_params)) continue;