commit 0045e54d16cd90c3eae0448e3760c7aaeec8fc17
parent c0af21b7d2e5511016c8989f404945b0d697b8ba
Author: Ryan Sepassi <rsepassi@gmail.com>
Date: Sun, 26 Apr 2026 22:25:25 -0700
cc/cg: structured init for cg-emit-global; scalar global init (§E.1)
Reshape cg-emit-global's init argument from "bv or #f" to "list of
pieces or #f". Each piece is either a bytevector (raw bytes, emitted as
N×!(byte) entries in .data) or a (label-ref . <label-bv>) pair (emitted
as `&<label> %(0)` for an 8-byte ptr slot). #f remains the zero-init
sentinel.
This lets §E.2-§E.7 + §L.3 build initializer values compositionally
without each callsite reinventing the data-section grammar.
Existing call sites all pass #f today, so signature change is
backwards-compatible.
Diffstat:
4 files changed, 63 insertions(+), 6 deletions(-)
diff --git a/cc/cg.scm b/cc/cg.scm
@@ -793,16 +793,50 @@
;; --------------------------------------------------------------------
;; Globals and data
;; --------------------------------------------------------------------
-(define (cg-emit-global cg sym init-bv-or-false)
+;; cg-emit-global: emit a global symbol into either .data (initialized)
+;; or .bss (zero-init).
+;;
+;; init can be:
+;; #f — zero-init in .bss (size from sym's ctype).
+;; (piece ...) — initialized in .data; pieces concatenated.
+;;
+;; Each piece is either:
+;; <bytevector> — raw bytes; emitted as N×!(byte) entries.
+;; (label-ref . <label-bv>) — 8-byte pointer slot containing &label;
+;; emitted as `&<label> %(0)` (4B label ref +
+;; 4B zero pad).
+(define (%cg-init-piece->bv piece)
+ (cond
+ ((bytevector? piece)
+ (let ((n (bytevector-length piece)))
+ (let loop ((i 0) (acc '()))
+ (cond
+ ((= i n) (bv-cat (reverse acc)))
+ (else
+ (loop (+ i 1)
+ (cons (bv-cat (list "!("
+ (number->string
+ (bytevector-u8-ref piece i) 10)
+ ")\n"))
+ acc)))))))
+ ((and (pair? piece) (eq? (car piece) 'label-ref))
+ (bv-cat (list "&" (cdr piece) " %(0)\n")))
+ (else (die #f "cg-emit-global: bad init piece" piece))))
+
+(define (cg-emit-global cg sym init)
(let* ((nm (sym-name sym))
(lbl (%cg-mangle-global nm))
(sz (ctype-size (sym-type sym)))
(size (if (< sz 0) 8 sz)))
(cond
- (init-bv-or-false
- (buf-push! (cg-data cg)
- (bv-cat (list "\n:" lbl "\n"
- "\"" init-bv-or-false "\"\n"))))
+ (init
+ (buf-push! (cg-data cg) (bv-cat (list "\n:" lbl "\n")))
+ (let walk ((ps init))
+ (cond
+ ((null? ps) 0)
+ (else
+ (buf-push! (cg-data cg) (%cg-init-piece->bv (car ps)))
+ (walk (cdr ps))))))
(else
(buf-push! (cg-bss cg)
(bv-cat (list "\n:" lbl "\n"
diff --git a/docs/CC-INTERNALS.md b/docs/CC-INTERNALS.md
@@ -555,7 +555,10 @@ beat seven.)
### Globals and data
```scheme
-(cg-emit-global cg sym init-bv-or-#f) ; init-bv: bytes for .data, or #f for .bss
+(cg-emit-global cg sym init) ; init = #f (zero-init in .bss)
+ ; | (piece ...) in .data
+;; piece := <bytevector> — raw bytes
+;; | (label-ref . <label-bv>) — 8-byte slot holding &label
(cg-emit-extern cg sym) ; declare without defining
(cg-intern-string cg bv-content) -> bv-label ; idempotent; used internally by cg-push-string
```
diff --git a/tests/cc-cg/49-init-scalar-global.expected-exit b/tests/cc-cg/49-init-scalar-global.expected-exit
@@ -0,0 +1 @@
+42
diff --git a/tests/cc-cg/49-init-scalar-global.scm b/tests/cc-cg/49-init-scalar-global.scm
@@ -0,0 +1,19 @@
+;; tests/cc-cg/49-init-scalar-global.scm — emit a global int with a
+;; constant initializer; main returns its value. §E.1 of CC-PUNCHLIST.
+;;
+;; Models: int g = 42; int main(void) { return g; }
+;; Runtime: exits 42.
+
+(let* ((cg (cg-init))
+ (g (%sym "g" 'var 'static %t-i32 #f))
+ ;; structured init: a single 4-byte LE bytevector piece for 42.
+ (bv4 (make-bytevector 4 0))
+ (_ (bytevector-u8-set! bv4 0 42))
+ (init (list bv4)))
+ (cg-emit-global cg g init)
+ (cg-fn-begin cg "main" '() %t-i32)
+ (cg-push-sym cg g)
+ (cg-load cg)
+ (cg-return cg)
+ (cg-fn-end cg)
+ (write-bv-fd 1 (cg-finish cg)))
