Hojun-Cho/ww
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

w6c+w6c_ww: global-base arm for indexed struct-element field read (fix #21)

Browse Source 
The `arr[i].field` N_DOT read branch in both stages was gated on a LOCAL
base lookup (cstage `localfind != 0`, wwstage `localfindnode != nil`). A
module-GLOBAL base (`let g: [2]pt = [...]`) missed it:

  - cstage fell to a generic index-load that drops f->offset — it read
    element[i] at offset 0, so `g[i].b` returned a's value (g[0].b -> 1,
    g[1].b -> 3 instead of 2, 4).
  - wwstage fell to the module-qualified SB fallback — garbage, no main.g
    load at all.

Silent, byte-id-divergent. This is the READ twin of #11 (the global
`g[i] = v` write fix) and the #15 sibling. Local `[N]struct` bases read
correctly (tests 680/681 cover only those), which is why it was never
caught.

Fix (both stages, converged byte-identical): resolve the global the same
way the N_INDEX arm does — cstage `let_islet || def_isarraydef`, wwstage
`letvartnode || defvartnode` — and dispatch the base load by shape: array
-> LEAQ name(SB) (the symbol IS the storage), slice/ptr -> MOVQ name(SB)
(the symbol's first word IS the .ptr). The field then loads at f->offset
exactly as the local arm does. esz (element stride) and f->offset both
come from the type table (rule 13). Mirrors #11's write-side global-base
resolution. combined.ww embeds (w6c + wwdump) regenerate.

688_global_arr_elem_field: global `[2]pt` reads of .a/.b on both elements
(the .b reads are the bug), a non-8-aligned `[2]rec {tag:u8,x:i32,y:i64}`
to stress f->offset + a u8 sub-word leaf, and a slice-base read
(`let g: []rec = arr;`) that exercises the MOVQ-deref .ptr arm. Runtime
(cstage build+run) + cstage==wwstage byte-id per row. The slice row is
byte-id ONLY: its read asm is correct and identical on both stages, but a
slice-of-struct module global does not data-emit a symbol yet (a separate,
pre-existing data-emission gap, sibling of #10/#20), so it cannot link/run.
This commit is contained in:
Hojun-Cho
2026-06-03 22:29:57 +09:00
parent d40224755a
commit 8dda8ea76c
6 changed files with 437 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
Makefile
View File

@@ -245,6 +245,7 @@ TESTS = $(BIN)/test_smoke $(BIN)/test_lex $(BIN)/test_parse $(BIN)/test_check \
$(BIN)/test_arr_infer_len \
$(BIN)/test_slice_str_global_zero \
$(BIN)/test_slice_literal_global \
$(BIN)/test_global_arr_elem_field \
$(BIN)/test_dot_str_chained_arg \
$(BIN)/test_dot_slice_arg \
$(BIN)/test_dot_tagged_source \
@@ -589,6 +590,12 @@ $(BIN)/test_slice_literal_global: test/wcc/687_slice_literal_global.c $(BIN)/ww
$(LIB)/libwwrt.a | $(BIN)
$(CC) $(CFLAGS) -o $@ $<
$(BIN)/test_global_arr_elem_field: test/wcc/688_global_arr_elem_field.c $(BIN)/ww \
$(BIN)/w6c $(BIN)/w6a $(BIN)/w6l \
$(BIN)/ww_ww $(BIN)/w6c_ww $(BIN)/w6a_ww $(BIN)/w6l_ww \
$(LIB)/libwwrt.a | $(BIN)
$(CC) $(CFLAGS) -o $@ $<
$(BIN)/test_dot_str_chained_arg: test/wcc/692_dot_str_chained_arg.c $(BIN)/ww \
$(BIN)/w6c $(BIN)/w6a $(BIN)/w6l \
$(BIN)/ww_ww $(BIN)/w6c_ww $(BIN)/w6a_ww $(BIN)/w6l_ww \

25
cmd/w6c/cgen.c
View File

@@ -8361,8 +8361,21 @@ cgexpr(Cg *c, Node *n, Local *locals)
int is_sl = bu && bu->kind == TY_SLICE;
int is_ptr = bu && bu->kind == TY_PTR;
int off = localfind(locals, idxbase->str);
/* #21 (READ twin of #11): a module-GLOBAL base
* makes localfind return 0, so the field-offset-
* aware branch was skipped and `g[i].field` fell to
* a generic index-load that drops f->offset (reads
* element[i] at offset 0). Resolve the global the
* same way the N_INDEX arm does (let_islet ||
* def_isarraydef) and dispatch the base load by
* shape: array -> LEAQ name(SB) (the symbol IS the
* storage), slice/ptr -> MOVQ name(SB) (the symbol's
* first word IS the .ptr). */
int isglobal = (off == 0)
&& (let_islet(idxbase->str)
|| def_isarraydef(idxbase->str));
if (f != NULL && (is_arr || is_sl || is_ptr)
&& off != 0) {
&& (off != 0 || isglobal)) {
int esz = (int)elemt->size;
cgexpr(c, n->lhs->rhs, locals);
if (esz > 1) {
@@ -8371,7 +8384,15 @@ cgexpr(Cg *c, Node *n, Local *locals)
ins2(c, A_IMULQ, areg(D_CX),
areg(D_AX));
}
if (is_arr)
if (isglobal && is_arr)
ins2(c, A_LEAQ,
masym(c, idxbase->str),
areg(D_BX));
else if (isglobal)
ins2(c, A_MOVQ,
masym(c, idxbase->str),
areg(D_BX));
else if (is_arr)
ins2(c, A_LEAQ,
amem(D_BP, off), areg(D_BX));
else

38
selfhost/cmd/w6c/main.combined.ww
View File

@@ -22306,8 +22306,28 @@ fn cgdot(c: *cgen, n: *node) void = {
let idxbase: *node = lhs.lhs;
if (idxbase != nil) { if (idxbase.kind == nkind.N_IDENT) {
let lc: *local = localfindnode(c, idxbase.str);
if (lc != nil) { if (lc.tnode != nil) {
let tn: *node = lc.tnode;
// #21 (READ twin of #11): a module-GLOBAL base makes
// localfindnode return nil, so this field-offset-aware
// branch was skipped and `g[i].field` fell through to
// the module-qualified fallback below (garbage — no
// main.g load at all). Resolve the global's tnode via
// letvartnode/defvartnode (the same source cgindex's
// global arm uses) and dispatch the base load by shape:
// array -> LEAQ name(SB) (the symbol IS the storage),
// slice/ptr -> MOVQ name(SB) (the symbol's first word
// IS the .ptr). Mirrors cstage cgen.c's #21 arm.
let tn: *node = nil;
let isglobal: bool = false;
let gname: str;
gname.ptr = nil; gname.len = 0;
if (lc != nil) {
tn = lc.tnode;
} else {
tn = letvartnode(c, idxbase.str);
if (tn == nil) { tn = defvartnode(c, idxbase.str); };
if (tn != nil) { isglobal = true; gname = idxbase.str; };
};
if (tn != nil) {
let elemt: *node = nil;
let baseisarray: bool = false;
let tk: nkind = tn.kind;
@@ -22342,6 +22362,17 @@ fn cgdot(c: *cgen, n: *node) void = {
emitline(", CX\n");
emitline("\tIMULQ\tCX, AX\n");
};
if (isglobal) {
if (baseisarray) {
emitline("\tLEAQ\t");
emitsymname(c, gname);
emitline("(SB), BX\n");
} else {
emitline("\tMOVQ\t");
emitsymname(c, gname);
emitline("(SB), BX\n");
};
} else {
if (baseisarray) {
emitline("\tLEAQ\t");
emitoff(lc.off: i64);
@@ -22351,6 +22382,7 @@ fn cgdot(c: *cgen, n: *node) void = {
emitoff(lc.off: i64);
emitline("(BP), BX\n");
};
};
emitline("\tADDQ\tAX, BX\n");
if (viaptr) {
emitline("\tMOVQ\t(BX), AX\n");
@@ -22415,7 +22447,7 @@ fn cgdot(c: *cgen, n: *node) void = {
};
};
};
};};
};
};};
};
};

38
selfhost/cmd/wcc/cgenexpr.ww
View File

@@ -2562,8 +2562,28 @@ fn cgdot(c: *cgen, n: *node) void = {
let idxbase: *node = lhs.lhs;
if (idxbase != nil) { if (idxbase.kind == nkind.N_IDENT) {
let lc: *local = localfindnode(c, idxbase.str);
if (lc != nil) { if (lc.tnode != nil) {
let tn: *node = lc.tnode;
// #21 (READ twin of #11): a module-GLOBAL base makes
// localfindnode return nil, so this field-offset-aware
// branch was skipped and `g[i].field` fell through to
// the module-qualified fallback below (garbage — no
// main.g load at all). Resolve the global's tnode via
// letvartnode/defvartnode (the same source cgindex's
// global arm uses) and dispatch the base load by shape:
// array -> LEAQ name(SB) (the symbol IS the storage),
// slice/ptr -> MOVQ name(SB) (the symbol's first word
// IS the .ptr). Mirrors cstage cgen.c's #21 arm.
let tn: *node = nil;
let isglobal: bool = false;
let gname: str;
gname.ptr = nil; gname.len = 0;
if (lc != nil) {
tn = lc.tnode;
} else {
tn = letvartnode(c, idxbase.str);
if (tn == nil) { tn = defvartnode(c, idxbase.str); };
if (tn != nil) { isglobal = true; gname = idxbase.str; };
};
if (tn != nil) {
let elemt: *node = nil;
let baseisarray: bool = false;
let tk: nkind = tn.kind;
@@ -2598,6 +2618,17 @@ fn cgdot(c: *cgen, n: *node) void = {
emitline(", CX\n");
emitline("\tIMULQ\tCX, AX\n");
};
if (isglobal) {
if (baseisarray) {
emitline("\tLEAQ\t");
emitsymname(c, gname);
emitline("(SB), BX\n");
} else {
emitline("\tMOVQ\t");
emitsymname(c, gname);
emitline("(SB), BX\n");
};
} else {
if (baseisarray) {
emitline("\tLEAQ\t");
emitoff(lc.off: i64);
@@ -2607,6 +2638,7 @@ fn cgdot(c: *cgen, n: *node) void = {
emitoff(lc.off: i64);
emitline("(BP), BX\n");
};
};
emitline("\tADDQ\tAX, BX\n");
if (viaptr) {
emitline("\tMOVQ\t(BX), AX\n");
@@ -2671,7 +2703,7 @@ fn cgdot(c: *cgen, n: *node) void = {
};
};
};
};};
};
};};
};
};

38
selfhost/cmd/wwdump/main.combined.ww
View File

@@ -22306,8 +22306,28 @@ fn cgdot(c: *cgen, n: *node) void = {
let idxbase: *node = lhs.lhs;
if (idxbase != nil) { if (idxbase.kind == nkind.N_IDENT) {
let lc: *local = localfindnode(c, idxbase.str);
if (lc != nil) { if (lc.tnode != nil) {
let tn: *node = lc.tnode;
// #21 (READ twin of #11): a module-GLOBAL base makes
// localfindnode return nil, so this field-offset-aware
// branch was skipped and `g[i].field` fell through to
// the module-qualified fallback below (garbage — no
// main.g load at all). Resolve the global's tnode via
// letvartnode/defvartnode (the same source cgindex's
// global arm uses) and dispatch the base load by shape:
// array -> LEAQ name(SB) (the symbol IS the storage),
// slice/ptr -> MOVQ name(SB) (the symbol's first word
// IS the .ptr). Mirrors cstage cgen.c's #21 arm.
let tn: *node = nil;
let isglobal: bool = false;
let gname: str;
gname.ptr = nil; gname.len = 0;
if (lc != nil) {
tn = lc.tnode;
} else {
tn = letvartnode(c, idxbase.str);
if (tn == nil) { tn = defvartnode(c, idxbase.str); };
if (tn != nil) { isglobal = true; gname = idxbase.str; };
};
if (tn != nil) {
let elemt: *node = nil;
let baseisarray: bool = false;
let tk: nkind = tn.kind;
@@ -22342,6 +22362,17 @@ fn cgdot(c: *cgen, n: *node) void = {
emitline(", CX\n");
emitline("\tIMULQ\tCX, AX\n");
};
if (isglobal) {
if (baseisarray) {
emitline("\tLEAQ\t");
emitsymname(c, gname);
emitline("(SB), BX\n");
} else {
emitline("\tMOVQ\t");
emitsymname(c, gname);
emitline("(SB), BX\n");
};
} else {
if (baseisarray) {
emitline("\tLEAQ\t");
emitoff(lc.off: i64);
@@ -22351,6 +22382,7 @@ fn cgdot(c: *cgen, n: *node) void = {
emitoff(lc.off: i64);
emitline("(BP), BX\n");
};
};
emitline("\tADDQ\tAX, BX\n");
if (viaptr) {
emitline("\tMOVQ\t(BX), AX\n");
@@ -22415,7 +22447,7 @@ fn cgdot(c: *cgen, n: *node) void = {
};
};
};
};};
};
};};
};
};

281
test/wcc/688_global_arr_elem_field.c Normal file
View File

@@ -0,0 +1,281 @@
/*
* 688_global_arr_elem_field — a module-GLOBAL indexed struct-element
* `.field` READ (`g[i].field`, g at module scope) lowers the field
* offset, not just the element stride. The READ twin of #11 (which
* fixed the global `g[i] = v` WRITE) and the #15 sibling.
*
* Pre-fix (#21): the `arr[i].field` N_DOT branch in both stages was
* gated on a LOCAL lookup (cstage localfind != 0, wwstage
* localfindnode != nil). A module-global base missed it and fell
* through — cstage to a generic index-load that drops f->offset
* (reads element[i] at offset 0, so g[0].b returned a's value), and
* wwstage to the module-qualified fallback (garbage — no g load).
* Byte-id-divergent silent miscompile.
*
* The fix (BOTH stages, converged byte-identical): resolve the global
* the same way the N_INDEX arm does (cstage let_islet || def_isarraydef;
* wwstage letvartnode || defvartnode) and dispatch the base load by
* shape — array -> LEAQ name(SB) (the symbol IS the storage),
* slice/ptr -> MOVQ name(SB) (the symbol's first word IS the .ptr) —
* then load the field at f->offset, exactly as the local arm does.
* esz = element stride and f->offset both come from the type table.
*
* Coverage:
* arr_a0/a1 | [2]pt struct array, g[0].a / g[1].a — control,
* | offset 0. RUNTIME + byte-id.
* arr_b0/b1 | [2]pt, g[0].b / g[1].b — THE bug (offset 8, the
* | reads that returned a's value pre-fix). RUNTIME.
* rec_* | [2]rec { tag:u8, x:i32, y:i64 } — non-8-aligned
* | fields (x@4, y@8) + a u8 sub-word leaf, to stress
* | f->offset and the load-width path. RUNTIME.
* slice_x1 | [2]rec backing + `let g: []rec = arr;`, g[1].x
* | read. Exercises the slice-base MOVQ-deref arm
* | (MOVQ g(SB),BX loads .ptr) — the asm is byte-id and
* | correct, but a slice-of-struct module global does
* | not data-emit a symbol yet (a separate, pre-existing
* | data-emission gap, sibling of #10/#20), so it cannot
* | LINK/run. Asserted byte-id ONLY; filed separately.
*
* Why this was never caught: 680/681 cover only LOCAL `[N]struct`
* bases (BP-relative). The global base routes through name(SB) and was
* a distinct, untested arm.
*
* Cstage and wwstage each on every runtime fixture; wwstage gated on
* access(X_OK). All fixtures additionally asm-byte-id checked.
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <sys/stat.h>
#include <sys/wait.h>
/* want == BYTEID_ONLY: the row cannot LINK/run (a slice-of-struct
* module global does not data-emit yet — separate gap), so only the
* cstage/wwstage asm-byte-id is asserted, never an exit value. */
#define BYTEID_ONLY (-2147483647 - 1)
static int
runwait(const char *cmd)
{
int rc = system(cmd);
if (rc == -1) return -1;
if (WIFEXITED(rc)) return WEXITSTATUS(rc);
return -1;
}
struct row { const char *label; const char *src; int want; };
static const struct row rows[] = {
/* [2]pt, .a field at offset 0 — control. */
{ "arr_a0",
"type pt = struct { a: i64, b: i64 };\n"
"let g: [2]pt = [pt { a = 1, b = 2 }, pt { a = 3, b = 4 }];\n"
"export fn main() i32 = { return g[0].a: i32; };\n",
1 },
{ "arr_a1",
"type pt = struct { a: i64, b: i64 };\n"
"let g: [2]pt = [pt { a = 1, b = 2 }, pt { a = 3, b = 4 }];\n"
"export fn main() i32 = { return g[1].a: i32; };\n",
3 },
/* [2]pt, .b at offset 8 — THE bug: pre-fix g[i].b returned a's
* value (cstage) or garbage (wwstage). */
{ "arr_b0",
"type pt = struct { a: i64, b: i64 };\n"
"let g: [2]pt = [pt { a = 1, b = 2 }, pt { a = 3, b = 4 }];\n"
"export fn main() i32 = { return g[0].b: i32; };\n",
2 },
{ "arr_b1",
"type pt = struct { a: i64, b: i64 };\n"
"let g: [2]pt = [pt { a = 1, b = 2 }, pt { a = 3, b = 4 }];\n"
"export fn main() i32 = { return g[1].b: i32; };\n",
4 },
/* [2]rec, non-8-aligned i32 field (x@4) on the second element —
* stresses f->offset combined with the element stride. */
{ "rec_x1",
"type rec = struct { tag: u8, x: i32, y: i64 };\n"
"let g: [2]rec = [rec { tag = 1u8, x = 100, y = 11 },"
" rec { tag = 2u8, x = 200, y = 22 }];\n"
"export fn main() i32 = { return g[1].x: i32; };\n",
200 },
/* [2]rec, i64 field (y@8) on the second element. */
{ "rec_y1",
"type rec = struct { tag: u8, x: i32, y: i64 };\n"
"let g: [2]rec = [rec { tag = 1u8, x = 100, y = 11 },"
" rec { tag = 2u8, x = 200, y = 22 }];\n"
"export fn main() i32 = { return g[1].y: i32; };\n",
22 },
/* [2]rec, u8 leaf (tag@0) — sub-word zero-extend load on a
* global element base. */
{ "rec_tag1",
"type rec = struct { tag: u8, x: i32, y: i64 };\n"
"let g: [2]rec = [rec { tag = 1u8, x = 100, y = 11 },"
" rec { tag = 2u8, x = 200, y = 22 }];\n"
"export fn main() i32 = { return g[1].tag: i32; };\n",
2 },
/* slice-base MOVQ-deref arm: `let g: []rec = arr;`, g[1].b read.
* Byte-id only — a slice-of-struct module global does not
* data-emit yet (separate gap), so it cannot link/run. The .s
* holds MOVQ g(SB),BX (the .ptr load) + the correct f->offset on
* both stages. */
{ "slice_x1",
"type rec = struct { tag: u8, x: i32, y: i64 };\n"
"let arr: [2]rec = [rec { tag = 1u8, x = 100, y = 11 },"
" rec { tag = 2u8, x = 200, y = 22 }];\n"
"let g: []rec = arr;\n"
"export fn main() i32 = { return g[1].x: i32; };\n",
BYTEID_ONLY },
};
static int
run_driver(const char *driver, const struct row *r, int i)
{
char src[64], tmpdir[64], cmd[1024];
snprintf(src, sizeof src, "/tmp/gae_%d_%d.ww", getpid(), i);
snprintf(tmpdir, sizeof tmpdir, "/tmp/gae_%d_d_%d", getpid(), i);
FILE *f = fopen(src, "wb");
if (!f) return -1;
fputs(r->src, f);
fclose(f);
mkdir(tmpdir, 0755);
snprintf(cmd, sizeof cmd, "cd %s && %s build %s 2>/dev/null",
tmpdir, driver, src);
if (runwait(cmd) != 0) {
fprintf(stderr, "row[%s]: build via %s failed\n",
r->label, driver);
unlink(src); rmdir(tmpdir);
return -1;
}
const char *base = strrchr(src, '/');
base = base ? base + 1 : src;
char outbin[128];
snprintf(outbin, sizeof outbin, "%s/%s", tmpdir, base);
char *dot = strrchr(outbin, '.');
if (dot && strcmp(dot, ".ww") == 0) *dot = '\0';
int got = runwait(outbin);
unlink(src); unlink(outbin); rmdir(tmpdir);
return got;
}
/* asm_byte_identical — generate .s via cstage's w6c and wwstage's
* w6c_ww and diff. Pins rule-10 convergence: pre-fix the global-base
* read diverged (cstage dropped f->offset, wwstage emitted no g load). */
static int
asm_byte_identical(const char *bin, const struct row *r, int i)
{
char src[64], cs[64], ws[64], cmd[1024];
snprintf(src, sizeof src, "/tmp/gae_asm_%d_%d.ww", getpid(), i);
snprintf(cs, sizeof cs, "/tmp/gae_asm_%d_%d_c.s", getpid(), i);
snprintf(ws, sizeof ws, "/tmp/gae_asm_%d_%d_w.s", getpid(), i);
FILE *f = fopen(src, "wb");
if (!f) return -1;
fputs(r->src, f);
fclose(f);
snprintf(cmd, sizeof cmd, "%s/w6c -o %s %s 2>/dev/null", bin, cs, src);
if (runwait(cmd) != 0) {
fprintf(stderr, "row[%s]: w6c errored\n", r->label);
unlink(src);
return -1;
}
snprintf(cmd, sizeof cmd, "%s/w6c_ww -o %s %s 2>/dev/null",
bin, ws, src);
if (runwait(cmd) != 0) {
fprintf(stderr, "row[%s]: w6c_ww errored\n", r->label);
unlink(src); unlink(cs);
return -1;
}
FILE *fc = fopen(cs, "rb");
FILE *fw = fopen(ws, "rb");
int rc = 0;
if (!fc || !fw) {
rc = -1;
} else {
for (;;) {
int a = fgetc(fc);
int b = fgetc(fw);
if (a != b) { rc = -1; break; }
if (a == EOF) break;
}
}
if (fc) fclose(fc);
if (fw) fclose(fw);
if (rc != 0)
fprintf(stderr, "row[%s]: cstage vs wwstage asm differs\n",
r->label);
unlink(src); unlink(cs); unlink(ws);
return rc;
}
int
main(void)
{
const char *bin = getenv("BIN");
if (!bin) bin = "out/bin";
char absbin[1024];
if (bin[0] != '/') {
char cwd[1024];
if (getcwd(cwd, sizeof cwd) == NULL) return 1;
snprintf(absbin, sizeof absbin, "%s/%s", cwd, bin);
bin = absbin;
}
char cdrv[1024];
snprintf(cdrv, sizeof cdrv, "%s/ww", bin);
char wdrv[1024];
snprintf(wdrv, sizeof wdrv, "%s/ww_ww", bin);
struct { const char *name; const char *path; int gated_on_existence; }
drivers[] = {
{ "cstage", cdrv, 0 },
{ "wwstage", wdrv, 1 },
{ NULL, NULL, 0 },
};
int n = (int)(sizeof rows / sizeof rows[0]);
int total = 0, fail = 0;
for (int d = 0; drivers[d].name; d++) {
if (drivers[d].gated_on_existence
&& access(drivers[d].path, X_OK) != 0) {
fprintf(stderr, "global_arr_elem_field: skip %s (no %s)\n",
drivers[d].name, drivers[d].path);
continue;
}
for (int i = 0; i < n; i++) {
if (rows[i].want == BYTEID_ONLY)
continue; /* byte-id only below */
int got = run_driver(drivers[d].path, &rows[i], i);
total++;
if (got != rows[i].want) {
fprintf(stderr,
"global_arr_elem_field[%s][%s]: exit=%d want=%d\n",
drivers[d].name, rows[i].label,
got, rows[i].want);
fail++;
}
}
}
if (access(wdrv, X_OK) == 0) {
for (int i = 0; i < n; i++) {
total++;
if (asm_byte_identical(bin, &rows[i], i) != 0)
fail++;
}
}
if (fail) {
fprintf(stderr,
"global_arr_elem_field: %d/%d fixtures failed\n", fail, total);
return 1;
}
printf("global_arr_elem_field: %d/%d ok\n", total, total);
return 0;
}