cgen: store all eightbytes when an in-cap aggregate call returns into an array element (#31-G)

`arr[i] = mk()` where mk returns an in-cap (<=24B) struct/tuple/array
left the result in the #4 cgreturn registers (AX/DX/CX), but the
N_ASSIGN-into-N_INDEX path had no arm for an N_CALL rhs, so it fell to
the scalar store tail: only member 0 was written and the index scale
clobbered CX. Both stages emitted byte-identical wrong code (the
documented-but-silent #31-G gap), so the byte-id gate was blind to it.

Add an in-cap N_CALL-rhs arm: materialise the return into a frame scratch
first (keeping the CALL at the frame's natural 16B alignment), resolve
&arr[i], then word-copy the full eightbyte count + sub-8 tail -- mirroring
the #4 receive shape and the #270-1b copy. The eightbyte count derives
from the element size in the type table. Over-cap returns (#234),
non-call rhs (#270-1b) and tuple literals (#121) are unaffected; the
sibling field/deref shapes stay loud (#24).

Surfaced by the codegen miscompile hunt (finding C2c). Pinned by
test/lang/idx_aggret_recv_test.ww (10 value-asserting rows: 2/3-eightbyte
structs+tuples, array elem, sub-8 tail, const/runtime index, all four
base shapes; reddens on revert).
This commit is contained in:
2026-06-27 13:28:37 +09:00
parent 33295c41c6
commit c83a3403a4
3 changed files with 344 additions and 0 deletions

View File

@@ -6731,6 +6731,107 @@ cgexpr(Cg *c, Node *n, Local *locals)
cg_sret_dest_off = 0; cg_sret_dest_off = 0;
break; break;
} }
/* C2c / #31-G: an IN-CAP aggregate-returning CALL into an
* indexed element `a[i] = mk()`. The #234 arm above only
* fires for an OVER-cap (sret/memory) return; the #270-1b
* arm below excludes N_CALL (it copies from a source
* ADDRESS, which a call result has none). An in-cap (≤24B)
* struct/array/tuple return leaves AX/DX/CX per the #4
* cgreturn ABI but fell through to the 1-word scalar store
* (AX only) — dropping DX/CX (a silent field-drop, the
* documented-but-unfixed #31-G). Materialise the return into
* a frame scratch (the AX/DX/CX receive shape at the N_LET
* call-rhs site, cgen.c:3434), THEN compute &a[i] and word-
* copy scratch → dest. Scratch-first (not a dest spill across
* the call) keeps the call at the frame's natural alignment.
* In-cap only (cg_sret_retsize==0). */
if ((is_arr || is_sl || is_ptr) && n->op == TK_ASSIGN
&& esubu && (esubu->kind == TY_STRUCT
|| esubu->kind == TY_ARRAY
|| esubu->kind == TY_TUPLE)
&& esz > 8
&& n->rhs->kind == N_CALL
&& cg_sret_retsize(esub) == 0) {
int scr = cg_tagscr_slot(c, &locals, esz);
cgexpr(c, n->rhs, locals); /* call → AX/DX/CX */
/* AX/DX/CX → scratch (mirror cgen.c:3434 receive). */
int regs[3] = { D_AX, D_DX, D_CX };
int full = esz / 8;
int tail = esz % 8;
for (int i = 0; i < full; i++)
ins2(c, A_MOVQ, areg(regs[i]),
amem(D_BP, scr + i * 8));
if (tail > 0) {
int op = (tail == 4) ? A_MOVL
: (tail == 2) ? A_MOVW : A_MOVB;
ins2(c, op, areg(regs[full]),
amem(D_BP, scr + full * 8));
}
/* dest &a[i] → BX (mirror #121 / #270-1b resolve) */
cgexpr(c, n->lhs->rhs, locals); /* idx → AX */
if (esz > 1) {
ins2(c, A_MOVQ, aimm(esz), areg(D_CX));
ins2(c, A_IMULQ, areg(D_CX), areg(D_AX));
}
ins1(c, A_PUSHQ, areg(D_AX));
if (base->kind == N_IDENT) {
int off = localfind(locals, base->str);
int isglobal = (off == 0)
&& let_islet(base->str);
if (isglobal && is_arr)
ins2(c, A_LEAQ,
masym(c, base->str),
areg(D_BX));
else if (isglobal)
ins2(c, A_MOVQ,
masym(c, base->str),
areg(D_BX));
else if (is_arr)
ins2(c, A_LEAQ,
amem(D_BP, off), areg(D_BX));
else
ins2(c, A_MOVQ,
amem(D_BP, off), areg(D_BX));
} else if (cg_dotbase_addr(c, base, D_BX, locals)) {
/* N_DOT array-field base resolved inline. */
} else {
cgexpr(c, base, locals);
ins2(c, A_MOVQ, areg(D_AX), areg(D_BX));
}
ins1(c, A_POPQ, areg(D_AX));
ins2(c, A_ADDQ, areg(D_AX), areg(D_BX));
/* word-copy scratch → dest (tail-aware, mirror
* #270-1b copy). */
int k = 0;
for (; k + 8 <= esz; k += 8) {
ins2(c, A_MOVQ, amem(D_BP, scr + k),
areg(D_AX));
ins2(c, A_MOVQ, areg(D_AX),
amem(D_BX, k));
}
if (k + 4 <= esz) {
ins2(c, A_MOVL, amem(D_BP, scr + k),
areg(D_AX));
ins2(c, A_MOVL, areg(D_AX),
amem(D_BX, k));
k += 4;
}
if (k + 2 <= esz) {
ins2(c, A_MOVW, amem(D_BP, scr + k),
areg(D_AX));
ins2(c, A_MOVW, areg(D_AX),
amem(D_BX, k));
k += 2;
}
if (k + 1 <= esz) {
ins2(c, A_MOVB, amem(D_BP, scr + k),
areg(D_AX));
ins2(c, A_MOVB, areg(D_AX),
amem(D_BX, k));
k += 1;
}
break;
}
/* #121 (write-face of leg-b): a tuple-LITERAL rhs into an /* #121 (write-face of leg-b): a tuple-LITERAL rhs into an
* indexed element `a[i] = (3,4)`. A literal has no source * indexed element `a[i] = (3,4)`. A literal has no source
* ADDRESS, so the ident/dot/deref copy arm below can't reach * ADDRESS, so the ident/dot/deref copy arm below can't reach

View File

@@ -9123,6 +9123,137 @@ fn cgassign(c: *cgen, n: *syntax.node) void = {
c.sretdestoff = 0; c.sretdestoff = 0;
return; return;
}; };
// C2c / #31-G: an IN-CAP aggregate-returning CALL into
// an indexed element `a[i] = mk()`. The #234 arm above
// only fires for an OVER-cap (sret) return; the #270-1b
// arm below copies from a source ADDRESS (which a call
// result has none). An in-cap (<=24B) struct/array/tuple
// return leaves AX/DX/CX per the #4 cgreturn ABI but fell
// to the 1-word scalar store (AX only) — dropping DX/CX
// (the documented-but-unfixed #31-G). Materialise the
// return into a frame scratch (the AX/DX/CX receive shape,
// cstage cgen.c:3434), THEN compute &a[i] and word-copy
// scratch -> dest. Scratch-first (not a dest spill across
// the call) keeps the call at the frame's natural
// alignment. esz>8 non-str/non-slice IS struct/array/tuple
// here (tagged returned above). In-cap only
// (callsretsize==0). Mirror of cstage cgen.c C2c arm.
if (n.rhs.kind == syntax.nkind.N_CALL && esz > 8
&& !isstrtype(c, elemtn) && !isslicetype(c, elemtn)
&& callsretsize(c, n.rhs) == 0) {
let scrc2: i32 = tagscradd(c, esz);
cgexpr(c, n.rhs); // call -> AX/DX/CX
// AX/DX/CX -> scratch (mirror cstage cgen.c:3434).
let full: i32 = esz / 8;
let tail: i32 = esz % 8;
let wi: i32 = 0;
for (wi < full) {
let rn: str = "AX";
if (wi == 1) { rn = "DX"; }
else { if (wi == 2) { rn = "CX"; }; };
emitline("\tMOVQ\t");
emitline(rn);
emitline(", ");
emitoff((scrc2 + wi * 8): i64);
emitline("(BP)\n");
wi += 1;
};
if (tail > 0) {
let top: str = "MOVB";
if (tail == 4) { top = "MOVL"; }
else { if (tail == 2) { top = "MOVW"; }; };
let treg: str = "AX";
if (full == 1) { treg = "DX"; }
else { if (full == 2) { treg = "CX"; }; };
emitline("\t");
emitline(top);
emitline("\t");
emitline(treg);
emitline(", ");
emitoff((scrc2 + full * 8): i64);
emitline("(BP)\n");
};
// dest &a[i] -> BX (#121 / #270-1b base resolve)
cgexpr(c, idx);
if (esz > 1) {
emitline("\tMOVQ\t$");
emitint(esz: i64);
emitline(", CX\n");
emitline("\tIMULQ\tCX, AX\n");
};
emitline("\tPUSHQ\tAX\n");
if (isglobalarr) {
emitline("\tLEAQ\t");
emitsymname(c, globalname);
emitline("(SB), BX\n");
} else { if (isglobalptr) {
emitline("\tMOVQ\t");
emitsymname(c, globalname);
emitline("(SB), BX\n");
} else { if (baselocal != nil) {
let tn2: *syntax.node = baselocal.tnode;
let isarr2: bool = false;
if (tn2 != nil) { if (tn2.kind == syntax.nkind.N_TARRAY) { isarr2 = true; }; };
if (basealias) {
let bu60: *syntax.tinfo = tichase(base.type_: *syntax.tinfo);
if (bu60 != nil) { isarr2 = bu60.kind == syntax.tykind.TY_ARRAY; };
};
if (isarr2) {
emitline("\tLEAQ\t");
emitoff(baselocal.off: i64);
emitline("(BP), BX\n");
} else {
emitline("\tMOVQ\t");
emitoff(baselocal.off: i64);
emitline("(BP), BX\n");
};
} else { if (dotbaseaddr(c, base, "BX")) {
} else {
cgexpr(c, base);
emitline("\tMOVQ\tAX, BX\n");
};};};};
emitline("\tPOPQ\tAX\n");
emitline("\tADDQ\tAX, BX\n");
// word-copy scratch -> dest (tail-aware, #270-1b copy)
let kc: i32 = 0;
for (kc + 8 <= esz) {
emitline("\tMOVQ\t");
emitoff((scrc2 + kc): i64);
emitline("(BP), AX\n");
emitline("\tMOVQ\tAX, ");
emitoff(kc: i64);
emitline("(BX)\n");
kc += 8;
};
if (kc + 4 <= esz) {
emitline("\tMOVL\t");
emitoff((scrc2 + kc): i64);
emitline("(BP), AX\n");
emitline("\tMOVL\tAX, ");
emitoff(kc: i64);
emitline("(BX)\n");
kc += 4;
};
if (kc + 2 <= esz) {
emitline("\tMOVW\t");
emitoff((scrc2 + kc): i64);
emitline("(BP), AX\n");
emitline("\tMOVW\tAX, ");
emitoff(kc: i64);
emitline("(BX)\n");
kc += 2;
};
if (kc + 1 <= esz) {
emitline("\tMOVB\t");
emitoff((scrc2 + kc): i64);
emitline("(BP), AX\n");
emitline("\tMOVB\tAX, ");
emitoff(kc: i64);
emitline("(BX)\n");
kc += 1;
};
return;
};
// #121 (write-face of leg-b): a tuple-LITERAL rhs into // #121 (write-face of leg-b): a tuple-LITERAL rhs into
// an indexed element `a[i] = (3,4)`. A literal has no // an indexed element `a[i] = (3,4)`. A literal has no
// source ADDRESS, so the ident/dot/deref copy arm below // source ADDRESS, so the ident/dot/deref copy arm below

View File

@@ -0,0 +1,112 @@
// idx_aggret_recv_test — in-cap aggregate-returning CALL received into an
// INDEXED element `a[i] = mk()` (C2c / #31-G). Pre-fix the N_ASSIGN-into-
// N_INDEX path had no arm for an in-cap (<=24B, AX/DX/CX-return) struct /
// array / tuple call rhs: the #270-1b element-store arm gates its source to
// N_IDENT / N_DOT / STAR (a call result has no source address) and the #234
// arm only fires for an OVER-cap (sret) return, so the call fell to the
// 1-word scalar store — only AX (member 0) was written, DX/CX dropped, AND
// the index-scale clobbered CX. Both stages emitted byte-IDENTICAL wrong asm
// (gate-blind, the #263 both-wrong form). Each @test asserts EVERY member (a
// dropped word fails); covers 2- and 3-eightbyte structs/tuples, an array-of-
// array element, const + runtime index, and local + GLOBAL + N_DOT-field +
// slice base. T2 keeps the cs==ww net.
package idx_aggret_recv_test;
type t2 = struct { a: i64, b: i64 };
type t3 = struct { a: i64, b: i64, c: i64 };
type holder = struct { arr: [3]t3 };
let g3: [2]t3 = [t3{a=0i64,b=0i64,c=0i64}, t3{a=0i64,b=0i64,c=0i64}];
fn mk2(x: i64) t2 = { return t2{a=x, b=x+10i64}; };
fn mk3(x: i64) t3 = { return t3{a=x, b=x+1i64, c=x+2i64}; };
fn mktup2() (i64, i64) = { return (3i64, 4i64); };
fn mktup3() (i64, i64, i64) = { return (5i64, 6i64, 7i64); };
fn mkarr() [3]i64 = { return [8i64, 9i64, 10i64]; };
fn mkarr32() [3]i32 = { return [11i32, 22i32, 33i32]; };
fn idx1() i64 = { return 1i64; };
@test fn struct_2eb() void = {
let a: [2]t2;
a[1] = mk2(5i64);
assert(a[1].a == 5i64);
assert(a[1].b == 15i64);
};
@test fn struct_3eb() void = {
let a: [2]t3;
a[1] = mk3(5i64);
assert(a[1].a == 5i64);
assert(a[1].b == 6i64);
assert(a[1].c == 7i64);
};
@test fn runtime_index() void = {
let a: [2]t3;
a[idx1()] = mk3(7i64);
assert(a[1].a == 7i64);
assert(a[1].b == 8i64);
assert(a[1].c == 9i64);
};
@test fn global_dest() void = {
g3[1] = mk3(9i64);
assert(g3[1].a == 9i64);
assert(g3[1].b == 10i64);
assert(g3[1].c == 11i64);
};
@test fn dot_base() void = {
let h: holder;
h.arr[2] = mk3(3i64);
assert(h.arr[2].a == 3i64);
assert(h.arr[2].b == 4i64);
assert(h.arr[2].c == 5i64);
};
@test fn slice_base() void = {
let buf: [2]t3;
let sl: []t3;
sl.ptr = buf.ptr: *t3;
sl.len = 2;
sl.cap = 2;
sl[1] = mk3(4i64);
assert(buf[1].a == 4i64);
assert(buf[1].b == 5i64);
assert(buf[1].c == 6i64);
};
@test fn tuple_2eb() void = {
let a: [2](i64, i64);
a[1] = mktup2();
assert(a[1].0 == 3i64);
assert(a[1].1 == 4i64);
};
@test fn tuple_3eb() void = {
let a: [2](i64, i64, i64);
a[1] = mktup3();
assert(a[1].0 == 5i64);
assert(a[1].1 == 6i64);
assert(a[1].2 == 7i64);
};
@test fn array_elem() void = {
let a: [2][3]i64;
a[1] = mkarr();
assert(a[1][0] == 8i64);
assert(a[1][1] == 9i64);
assert(a[1][2] == 10i64);
};
// [3]i32 element = 12B: esz%8 != 0, so the receive-store and the word-copy
// both hit the sub-8 TAIL arm (MOVL of the DX low-half), which the
// 8-multiple rows above never exercise.
@test fn array_elem_tail() void = {
let a: [2][3]i32;
a[1] = mkarr32();
assert(a[1][0] == 11i32);
assert(a[1][1] == 22i32);
assert(a[1][2] == 33i32);
};