/*
 * 798_tuple_sret_callee — project #10 Fold A (wide tuple-return / sret,
 * CALLEE side). An over-capacity tuple return (> 4 integer eightbytes or
 * > 2 SSE eightbytes) used to LOUD-STOP at the N_RETURN SEND site
 * ("tuple return exceeds ... register-return ABI capacity"). Fold A makes
 * the CALLEE emit such a return via the existing sret skeleton: the
 * prologue wires @sretarg (the caller-prealloc dest in RDI), and the
 * return stores each element through *(@sretarg) at its packed layout
 * offset, then returns @sretarg in RAX.
 *
 * THIS TEST IS COMPILE + BYTE-ID ONLY (no runtime row):
 *   (a) w6c (cstage) AND w6c_ww (wwstage) must now COMPILE the over-cap
 *       tuple-returning fn — no loud-stop. The pre-Fold-A behavior was a
 *       hard error; a green compile here proves the SEND emits sret.
 *   (b) the two .s outputs must be BYTE-IDENTICAL (rule-10). The classifier
 *       (cg_sret_retsize / sretretsize) and the SEND emitter share a single
 *       cap SSoT, so both stages classify + lay out the tuple the same way.
 *
 * WHY NO RUNTIME ROW: the CALL/receive side stays deliberately loud-stopped
 * — an N_MLET/N_MASSIGN destructure of an over-cap tuple still aborts
 * ("tuple destructure exceeds ... capacity"), so the over-cap callee is not
 * yet usefully callable. The end-to-end round-trip (allocate dest, call,
 * read the elements back) arrives with Fold B (#10-B), which wires the
 * receive. So this fn is compiled but never called/destructured here.
 *
 * ROWS exercise the layout arithmetic the SEND must get right:
 *   - ([]u8, []u8)        6 GP eightbytes, all-wide, offsets 0 / 24.
 *   - (str, str)          6 GP, str IS []u8 (24B header), offsets 0 / 24.
 *   - (str, str, i32)     7 GP; the trailing narrow i32 must store MOVL at
 *                         its NATURAL packed offset 48 (#169), not over-MOVQ.
 *   - (f64, f64, f64)     3 SSE eightbytes > the 2-wide SSE cap.
 *
 * GATE POLARITY: must stay GREEN. A non-zero compile means the SEND
 * loud-stop regressed (Fold A reverted); a byte-id FAIL means cstage and
 * wwstage diverged on the sret classification or the element layout
 * (rule-10 violation / cap-SSoT drift).
 */
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

static int
runwait(const char *cmd)
{
	int rc = system(cmd);
	if (rc == -1) return -1;
	return rc;
}

struct row { const char *label; const char *src; };

static const struct row rows[] = {
	{ "two_slices",
	  "package main;\n"
	  "export fn f(a: []u8, b: []u8) ([]u8, []u8) = { return (a, b); };\n" },
	{ "two_str",
	  "package main;\n"
	  "export fn f(a: str, b: str) (str, str) = { return (a, b); };\n" },
	{ "str_str_i32",
	  "package main;\n"
	  "export fn f(a: str, b: str, n: i32) (str, str, i32) = {\n"
	  "	return (a, b, n);\n"
	  "};\n" },
	{ "three_f64",
	  "package main;\n"
	  "export fn f(x: f64, y: f64, z: f64) (f64, f64, f64) = {\n"
	  "	return (x, y, z);\n"
	  "};\n" },
	{ NULL, NULL }
};

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, "tuple_sret_callee: w6c_ww missing — cannot run "
		    "the cs==ww byte-id gate\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/wwtsc_%d_%d.ww", getpid(), i);
		FILE *f = fopen(src, "wb");
		if (f == NULL) { fail++; continue; }
		fputs(rows[i].src, f);
		fclose(f);

		char cs_s[64], ws_s[64], cmd[2048];
		snprintf(cs_s, sizeof cs_s, "/tmp/wwtsc_%d_%d_cs.s", getpid(), i);
		snprintf(ws_s, sizeof ws_s, "/tmp/wwtsc_%d_%d_ww.s", getpid(), i);

		/* (a) cstage must COMPILE the over-cap tuple callee (no loud-stop). */
		snprintf(cmd, sizeof cmd, "%s -o %s %s 2>/dev/null", w6c, cs_s, src);
		if (runwait(cmd) != 0) {
			fprintf(stderr, "row[%s]: w6c loud-stopped (Fold A regressed)\n",
			    rows[i].label);
			fail++; unlink(src); continue;
		}
		/* (a') wwstage must compile it too. */
		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 loud-stopped (Fold A regressed)\n",
			    rows[i].label);
			fail++; unlink(src); unlink(cs_s); continue;
		}
		/* (b) cs==ww byte-id gate. */
		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 tuple-sret-callee tests failed\n", fail, n);
		return 1;
	}
	printf("tuple_sret_callee: %d/%d ok (compile + cs==ww byte-id)\n", n, n);
	return 0;
}