wcc/cgen: #66(b-i) wwstage inferred-literal scalar global — default annotation to int, emit DATA+load (align up)

An inferred-literal scalar module-global (`let s = 42;`) was wwstage
silent-wrong: the checker stamps the annotation N_TNAME("untyped_int"),
which letscalarprim does not recognise, so letemitsize returns 0 — the
global is dropped from collectlets (no DATA emitted) AND cgident falls to
the silent module-leaf (no load), running garbage. cstage defaults
untyped_int to an 8B int before emit (DATAW + MOVQ), which is correct.

Fix (wwstage-only, align up to cstage): defaultinferredlets in cgen.ww,
called from cgfile (cgendecl.ww) before collectlets, rewrites the
annotation "untyped_int" -> "int" (8B machine word, NOT i32 — the #108
truncation trap is the opposite polarity) for a module-level N_LET whose
rhs is N_INTLIT. All three consumers (letemitsize, emitletdataw, cgident
global-read) then resolve a concrete int. cstage is untouched.

Scope: N_INTLIT only. A const-expr inferred global (`let s = 7*6;`, N_BIN)
stays on its existing path — that is a separate live cs!=ww silent
miscompile tracked as #133, out of scope here.

Pin: 947_inferred_scalar_global_run — inferred `let s=42` (42, base
wwstage garbage) + typed control, cs==ww byte-id.

Makefile

@@ -460,6 +460,7 @@ TESTS	= $(BIN)/test_smoke $(BIN)/test_lex $(BIN)/test_parse $(BIN)/test_check \
 	  $(BIN)/test_checked_run \
 	  $(BIN)/test_floatarr_run \
 	  $(BIN)/test_deref_narrow_run \
+	  $(BIN)/test_inferred_scalar_global_run \
 	  $(BIN)/test_idx_compound_run \
 	  $(BIN)/test_ptrarr_index_run \
 	  $(BIN)/test_dotbase_arr_run \
@@ -2095,6 +2096,11 @@ $(BIN)/test_deref_narrow_run: test/wcc/947_deref_narrow_run.c $(BIN)/ww \
 		$(LIB)/libwwrt.a | $(BIN)
 	$(CC) $(CFLAGS) -o $@ $<
 
+$(BIN)/test_inferred_scalar_global_run: test/wcc/947_inferred_scalar_global_run.c $(BIN)/ww \
+		$(BIN)/w6c $(BIN)/w6c_ww $(BIN)/w6a $(BIN)/w6l \
+		$(LIB)/libwwrt.a | $(BIN)
+	$(CC) $(CFLAGS) -o $@ $<
+
 $(BIN)/test_idx_compound_run: test/wcc/948_idx_compound_run.c $(BIN)/ww \
 		$(BIN)/w6c $(BIN)/w6c_ww $(BIN)/w6a $(BIN)/w6l \
 		$(LIB)/libwwrt.a | $(BIN)

selfhost/cmd/w6c/main.combined.ww

@@ -38056,6 +38056,7 @@ export fn cgfile(c: *cgen, file: *node) void = {
 	collectfnrets(c, file);
 	fficollect(c, file);
 	collectmods(c, file);
+	defaultinferredlets(c, file);
 	collectlets(c, file);
 	let d: *node = file.list;
 	for (d != nil) {
@@ -39114,6 +39115,40 @@ fn letemitsize(c: *cgen, d: *node) i32 = {
 	return 0;
 };
 
+// defaultinferredlets — #66(b-i): an inferred module-global int-literal
+// `let s = 42;` is stamped by the checker with an N_TNAME("untyped_int")
+// annotation (d.lhs). letemitsize / emitletdataw / the cgident global-read arm
+// key on that annotation's name, which letscalarprim doesn't recognise → the
+// global is dropped from collectlets (no DATAW) and the read falls to the
+// silent module-leaf (no MOVQ, MOVSXD on stale AX → wrong). cstage instead
+// type_default's the untyped int to the 8B machine word BEFORE emit. Mirror
+// that here at the single global-decl pass: rewrite the annotation to the
+// concrete machine word `int` so all three consumers resolve it as an 8B int
+// global (DATAW + MOVQ, byte-identical to cstage). int (not i32) per
+// [[project_int_machine_word_derived_limits]] — i32 is the #108 truncation
+// trap, opposite polarity.
+// Guarded to N_INTLIT ONLY: a const-EXPR rhs (`let s = 7*6`, N_BIN) is also
+// stamped untyped_int but is #66(b-ii) — a SEPARATE loud both-stage gap (no
+// DATA → link-fail) — and must stay on its loud route, not be defaulted to a
+// silent value. Runs before collectlets in cgfile so the mutated d.lhs is
+// visible to letpreintern + emitletdataw too.
+fn defaultinferredlets(c: *cgen, file: *node) void = {
+	if (file == nil) { return; };
+	let d: *node = file.list;
+	for (d != nil) {
+		if (d.kind == nkind.N_LET) {
+			if (d.lhs != nil && d.rhs != nil) {
+				if (d.lhs.kind == nkind.N_TNAME
+				    && streq(d.lhs.str, "untyped_int")
+				    && d.rhs.kind == nkind.N_INTLIT) {
+					d.lhs.str = "int";
+				};
+			};
+		};
+		d = d.next;
+	};
+};
+
 fn collectlets(c: *cgen, file: *node) void = {
 	c.lets = nil;
 	if (file == nil) { return; };

selfhost/cmd/wcc/cgen.ww

@@ -1040,6 +1040,40 @@ fn letemitsize(c: *cgen, d: *node) i32 = {
 	return 0;
 };
 
+// defaultinferredlets — #66(b-i): an inferred module-global int-literal
+// `let s = 42;` is stamped by the checker with an N_TNAME("untyped_int")
+// annotation (d.lhs). letemitsize / emitletdataw / the cgident global-read arm
+// key on that annotation's name, which letscalarprim doesn't recognise → the
+// global is dropped from collectlets (no DATAW) and the read falls to the
+// silent module-leaf (no MOVQ, MOVSXD on stale AX → wrong). cstage instead
+// type_default's the untyped int to the 8B machine word BEFORE emit. Mirror
+// that here at the single global-decl pass: rewrite the annotation to the
+// concrete machine word `int` so all three consumers resolve it as an 8B int
+// global (DATAW + MOVQ, byte-identical to cstage). int (not i32) per
+// [[project_int_machine_word_derived_limits]] — i32 is the #108 truncation
+// trap, opposite polarity.
+// Guarded to N_INTLIT ONLY: a const-EXPR rhs (`let s = 7*6`, N_BIN) is also
+// stamped untyped_int but is #66(b-ii) — a SEPARATE loud both-stage gap (no
+// DATA → link-fail) — and must stay on its loud route, not be defaulted to a
+// silent value. Runs before collectlets in cgfile so the mutated d.lhs is
+// visible to letpreintern + emitletdataw too.
+fn defaultinferredlets(c: *cgen, file: *node) void = {
+	if (file == nil) { return; };
+	let d: *node = file.list;
+	for (d != nil) {
+		if (d.kind == nkind.N_LET) {
+			if (d.lhs != nil && d.rhs != nil) {
+				if (d.lhs.kind == nkind.N_TNAME
+				    && streq(d.lhs.str, "untyped_int")
+				    && d.rhs.kind == nkind.N_INTLIT) {
+					d.lhs.str = "int";
+				};
+			};
+		};
+		d = d.next;
+	};
+};
+
 fn collectlets(c: *cgen, file: *node) void = {
 	c.lets = nil;
 	if (file == nil) { return; };

selfhost/cmd/wcc/cgendecl.ww

@@ -619,6 +619,7 @@ export fn cgfile(c: *cgen, file: *node) void = {
 	collectfnrets(c, file);
 	fficollect(c, file);
 	collectmods(c, file);
+	defaultinferredlets(c, file);
 	collectlets(c, file);
 	let d: *node = file.list;
 	for (d != nil) {

selfhost/cmd/wwdump/main.combined.ww

@@ -38056,6 +38056,7 @@ export fn cgfile(c: *cgen, file: *node) void = {
 	collectfnrets(c, file);
 	fficollect(c, file);
 	collectmods(c, file);
+	defaultinferredlets(c, file);
 	collectlets(c, file);
 	let d: *node = file.list;
 	for (d != nil) {
@@ -39114,6 +39115,40 @@ fn letemitsize(c: *cgen, d: *node) i32 = {
 	return 0;
 };
 
+// defaultinferredlets — #66(b-i): an inferred module-global int-literal
+// `let s = 42;` is stamped by the checker with an N_TNAME("untyped_int")
+// annotation (d.lhs). letemitsize / emitletdataw / the cgident global-read arm
+// key on that annotation's name, which letscalarprim doesn't recognise → the
+// global is dropped from collectlets (no DATAW) and the read falls to the
+// silent module-leaf (no MOVQ, MOVSXD on stale AX → wrong). cstage instead
+// type_default's the untyped int to the 8B machine word BEFORE emit. Mirror
+// that here at the single global-decl pass: rewrite the annotation to the
+// concrete machine word `int` so all three consumers resolve it as an 8B int
+// global (DATAW + MOVQ, byte-identical to cstage). int (not i32) per
+// [[project_int_machine_word_derived_limits]] — i32 is the #108 truncation
+// trap, opposite polarity.
+// Guarded to N_INTLIT ONLY: a const-EXPR rhs (`let s = 7*6`, N_BIN) is also
+// stamped untyped_int but is #66(b-ii) — a SEPARATE loud both-stage gap (no
+// DATA → link-fail) — and must stay on its loud route, not be defaulted to a
+// silent value. Runs before collectlets in cgfile so the mutated d.lhs is
+// visible to letpreintern + emitletdataw too.
+fn defaultinferredlets(c: *cgen, file: *node) void = {
+	if (file == nil) { return; };
+	let d: *node = file.list;
+	for (d != nil) {
+		if (d.kind == nkind.N_LET) {
+			if (d.lhs != nil && d.rhs != nil) {
+				if (d.lhs.kind == nkind.N_TNAME
+				    && streq(d.lhs.str, "untyped_int")
+				    && d.rhs.kind == nkind.N_INTLIT) {
+					d.lhs.str = "int";
+				};
+			};
+		};
+		d = d.next;
+	};
+};
+
 fn collectlets(c: *cgen, file: *node) void = {
 	c.lets = nil;
 	if (file == nil) { return; };

test/wcc/947_inferred_scalar_global_run.c

@@ -0,0 +1,181 @@
+/*
+ * 947_inferred_scalar_global_run — runtime + byte-id net for #66(b-i): an
+ * INFERRED int-literal module-global `let s = 42;` (no type annotation) was
+ * wwstage SILENT-WRONG. The checker stamps such a global with an
+ * N_TNAME("untyped_int") annotation; letemitsize / emitletdataw / the cgident
+ * global-read arm key on that annotation's name, which letscalarprim doesn't
+ * recognise — so the global was dropped from collectlets (no DATAW emitted)
+ * AND the read fell to the silent module-leaf (no `MOVQ main.s(SB)`), leaving
+ * `MOVSXD` to sign-extend a STALE AX. cstage instead type_default's the untyped
+ * int to the 8B machine word before emit, so it produced the correct
+ * `DATAW main.s` + `MOVQ main.s(SB), AX` (exit 42). cs≠ww, ww silent-wrong.
+ *
+ * Fix (wwstage only, cgen.ww defaultinferredlets, one choke-point before
+ * collectlets): rewrite the untyped_int annotation to the concrete machine
+ * word `int` — NOT i32 (that's the #108 truncation trap, opposite polarity) —
+ * so all three consumers resolve it as an 8B int global. The emitted DATAW +
+ * MOVQ then match cstage byte-for-byte (align ww UP). Guarded to N_INTLIT rhs
+ * ONLY: a const-EXPR inferred global (`let s = 7*6`) is #66(b-ii), a SEPARATE
+ * loud both-stage no-DATA gap, and must stay on its loud route.
+ *
+ * Each row carries (a) a cstage `ww build` + run asserting the exit code (R1
+ * pre-fix: wwstage exit 136 / asm differs), and (b) a w6c vs w6c_ww `.s` cmp
+ * (rule-10 byte-id). R2 is the typed-annotation control (already cs==ww/42),
+ * locking no regression on the annotated path.
+ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/stat.h>
+#include <sys/wait.h>
+
+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_exit; };
+
+static const struct row rows[] = {
+	/* R1 — the #66(b-i) repro: inferred int-literal module-global, read
+	 * back from main. Pre-fix wwstage emitted neither the DATAW nor the
+	 * MOVQ → MOVSXD on stale AX → exit 136. Post-fix cs==ww, exit 42. */
+	{ "inferred",
+	  "package main;\n"
+	  "let s = 42;\n"
+	  "export fn main() i32 = {\n"
+	  "	return s: i32;\n"
+	  "};\n", 42 },
+	/* R2 — typed-annotation control: `let s: i64 = 42;` already worked
+	 * (cs==ww, DATA emitted). Regression guard for the annotated path. */
+	{ "typed_ctrl",
+	  "package main;\n"
+	  "let s: i64 = 42;\n"
+	  "export fn main() i32 = {\n"
+	  "	return s: i32;\n"
+	  "};\n", 42 },
+	{ NULL, NULL, 0 }
+};
+
+static int
+slurp_eq(const char *a, const char *b)
+{
+	FILE *fa = fopen(a, "rb");
+	FILE *fb = fopen(b, "rb");
+	if (!fa || !fb) { if (fa) fclose(fa); if (fb) fclose(fb); return -1; }
+	int rc = 0;
+	for (;;) {
+		int ca = fgetc(fa);
+		int cb = fgetc(fb);
+		if (ca != cb) { rc = -1; break; }
+		if (ca == EOF) break;
+	}
+	fclose(fa); fclose(fb);
+	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 w6c[1100], w6c_ww[1100];
+	snprintf(w6c, sizeof w6c, "%s/w6c", bin);
+	snprintf(w6c_ww, sizeof w6c_ww, "%s/w6c_ww", bin);
+	if (access(w6c_ww, X_OK) != 0) {
+		fprintf(stderr, "infglobal: w6c_ww missing — cannot run the "
+		    "cs==ww byte-id gate (the whole point of this test)\n");
+		return 1;
+	}
+
+	int n = 0, fail = 0;
+	for (int i = 0; rows[i].src; i++, n++) {
+		char src[64];
+		snprintf(src, sizeof src, "/tmp/wwisg_%d_%d.ww", getpid(), i);
+		FILE *f = fopen(src, "wb");
+		if (f == NULL) { fail++; continue; }
+		fputs(rows[i].src, f);
+		fclose(f);
+
+		/* (a) cstage build + run. */
+		char tmpdir[64];
+		snprintf(tmpdir, sizeof tmpdir, "/tmp/wwisg_%d_d_%d",
+		    getpid(), i);
+		mkdir(tmpdir, 0755);
+
+		char cmd[2048];
+		snprintf(cmd, sizeof cmd, "cd %s && %s/ww build %s",
+		    tmpdir, bin, src);
+		if (runwait(cmd) != 0) {
+			fprintf(stderr, "row[%s]: cstage build failed\n",
+			    rows[i].label);
+			fail++;
+			unlink(src); rmdir(tmpdir);
+			continue;
+		}
+
+		char outbin[128];
+		const char *base = strrchr(src, '/');
+		base = base ? base + 1 : src;
+		snprintf(outbin, sizeof outbin, "%s/%s", tmpdir, base);
+		char *dot = strrchr(outbin, '.');
+		if (dot && strcmp(dot, ".ww") == 0) *dot = '\0';
+
+		int got = runwait(outbin);
+		if (got != rows[i].want_exit) {
+			fprintf(stderr, "row[%s]: cstage exit %d, want %d\n",
+			    rows[i].label, got, rows[i].want_exit);
+			fail++;
+		}
+		unlink(outbin); rmdir(tmpdir);
+
+		/* (b) cs==ww byte-id gate. */
+		char cs_s[64], ws_s[64];
+		snprintf(cs_s, sizeof cs_s, "/tmp/wwisg_%d_%d_cs.s",
+		    getpid(), i);
+		snprintf(ws_s, sizeof ws_s, "/tmp/wwisg_%d_%d_ww.s",
+		    getpid(), i);
+
+		snprintf(cmd, sizeof cmd, "%s -o %s %s 2>/dev/null",
+		    w6c, cs_s, src);
+		if (runwait(cmd) != 0) {
+			fprintf(stderr, "row[%s]: w6c failed\n", rows[i].label);
+			fail++; unlink(src); continue;
+		}
+		snprintf(cmd, sizeof cmd, "%s -o %s %s 2>/dev/null",
+		    w6c_ww, ws_s, src);
+		if (runwait(cmd) != 0) {
+			fprintf(stderr, "row[%s]: w6c_ww failed\n",
+			    rows[i].label);
+			fail++; unlink(src); unlink(cs_s); continue;
+		}
+		if (slurp_eq(cs_s, ws_s) != 0) {
+			fprintf(stderr,
+			    "row[%s]: cstage/wwstage .s DIFFER (rule-10 "
+			    "byte-id violation)\n", rows[i].label);
+			fail++;
+		}
+		unlink(src); unlink(cs_s); unlink(ws_s);
+	}
+
+	if (fail) {
+		fprintf(stderr, "%d/%d inferred-scalar-global tests failed\n",
+		    fail, n);
+		return 1;
+	}
+	printf("infglobal: %d/%d ok (cstage run + cs==ww byte-id)\n", n, n);
+	return 0;
+}