/*
 * 905_tupparam_run — runtime + byte-id net for #163, the tuple-PARAM ABI
 * (the param twin of #164's tuple RETURN).
 *
 * THE BUG (#163, LIVE drop on master): a tuple passed AS AN ARGUMENT was
 * unhandled in BOTH stages — no tuple arm in the cgcall arg push, the
 * cgcall arg pop, OR the callee cgfnparams receive. A tuple-typed call
 * result (`f(g())` where g returns a tuple) left its elements in the
 * return-ABI cursor (AX/DX/CX/R8 + X0/X1, per #164); the SEND fell to the
 * 1-GP-word `else` (PUSHQ AX / POPQ DI) so ALL BUT THE FIRST ELEMENT was
 * dropped, and the callee read its tuple param as a single GP word. This
 * broke INTEGER tuple params too; float elements were doubly lost (they
 * ride X0/X1, never AX).
 *
 * THE FIX: per-element SysV class placement reusing #164's helper. SEND —
 * cgexpr leaves the tuple in the return cursor; restage it into a frame
 * slot (@tupargscr) by class via tuple_store/tupstore, then push the slot
 * words high->low so the pop drains slot+0 first into the SysV ARG cursor
 * (DI/SI/.. + X0..X7). The frame slot decouples the return-class regs
 * (which OVERLAP the arg-class regs) from the arg placement. RECV — the
 * callee walks the tuple's elements over the arg cursor, storing each into
 * its frame slot positionally. Symmetric across cstage (cmd/w6c/cgen.c)
 * and wwstage (cgenutil.ww pushargsrev + cgenexpr.ww cgcall pop +
 * cgendecl.ww cgfnparams).
 *
 * SCOPE: register-class tuple ARGS produced by a CALL (the only form that
 * materialises a tuple value today — `let t = (1,2)` as a first-class
 * value is a separate unimplemented gap, so the SEND scopes to the N_CALL
 * producer and never pushes stale regs, rule 7). Arg-register overflow
 * loud-stops (the partial-spill stitch is out of scope, twin of #164's
 * cap); the loudstop row asserts BOTH stages ERROR.
 *
 * GATE-BLIND TO BYTE-ID ALONE: the bootstrap passes no tuple params, and
 * pre-fix both stages were symmetric-WRONG (both PUSHQ AX), so the cs==ww
 * .s gate HOLDS on master for the value rows — they diverge only at
 * RUNTIME. Each value row carries BOTH dimensions (modelled on 956):
 *   (a) cstage `ww build` + run, asserting the exit code (catches #163:
 *       master returns the wrong exit / dropped element).
 *   (b) w6c vs w6c_ww `.s` cmp (rule-10: both stages fixed identically).
 */
#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;
	int want_compile_fail;	/* loud-stop rows must NOT compile */
};

static const struct row rows[] = {
	/* HEADLINE — (f64, f64) arg. Pre-fix the SEND pushes only AX (the
	 * two floats stay stranded in X0/X1) and the callee reads one GP
	 * word; t.0+t.1 != 8.0 -> return 1. Post-fix each float rides the
	 * SSE arg cursor (X0,X1). */
	{ "f64f64_arg",
	  "package main;\n"
	  "fn pair(a: f64, b: f64) (f64, f64) = { return (a, b); };\n"
	  "fn add(t: (f64, f64)) f64 = { return t.0 + t.1; };\n"
	  "export fn main() i32 = {\n"
	  "\tif (add(pair(3.0, 5.0)) != 8.0) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* (i64, i64) arg — proves the broader INTEGER-tuple-param drop is
	 * fixed (master dropped the second i64 too). e0->DI, e1->SI. */
	{ "i64i64_arg",
	  "package main;\n"
	  "fn pair(a: i64, b: i64) (i64, i64) = { return (a, b); };\n"
	  "fn add(t: (i64, i64)) i64 = { return t.0 + t.1; };\n"
	  "export fn main() i32 = {\n"
	  "\tif (add(pair(3, 5)) != 8) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* (f64, i64) — class independent of position: f64@X0 (SSE cursor),
	 * i64@DI (INTEGER cursor), independent counters. */
	{ "f64i64_arg",
	  "package main;\n"
	  "fn mk(a: f64, b: i64) (f64, i64) = { return (a, b); };\n"
	  "fn add(t: (f64, i64)) i64 = { return (t.0: i64) + t.1; };\n"
	  "export fn main() i32 = {\n"
	  "\tif (add(mk(3.0, 5)) != 8) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* (i64, f64) — order-swap: i64@DI, f64@X0. Confirms the float lands
	 * in the next XMM regardless of its positional slot. */
	{ "i64f64_arg",
	  "package main;\n"
	  "fn mk(a: i64, b: f64) (i64, f64) = { return (a, b); };\n"
	  "fn add(t: (i64, f64)) i64 = { return t.0 + (t.1: i64); };\n"
	  "export fn main() i32 = {\n"
	  "\tif (add(mk(3, 5.0)) != 8) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* (f64, str) — SSE + wide (24B {ptr,len,cap}) coexist. The f64
	 * rides X0 (SSE, consuming no GP slot); the str rides DI/SI/DX
	 * (INTEGER cursor). f=4.0, s.len=5 -> 4+5 = 9. */
	{ "f64str_arg",
	  "package main;\n"
	  "fn mk(a: f64) (f64, str) = { return (a, \"hello\"); };\n"
	  "fn add(t: (f64, str)) i64 = {\n"
	  "\treturn (t.0: i64) + (t.1.len: i64);\n"
	  "};\n"
	  "export fn main() i32 = {\n"
	  "\tif (add(mk(4.0)) != 9) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* MULTI-TUPLE-ARG, ONE CALL — f(g(), h()) where BOTH args are
	 * tuple-producing calls. Proves @tupargscr (single-slot-per-fn) is
	 * REUSED per arg, not COLLIDED: pushargsrev evals right-to-left, so
	 * h() restages into the slot + drains it to the stack BEFORE g()
	 * restages into the SAME slot (h's words already pushed, safe to
	 * overwrite). Drain forward: s->DI,SI; t->DX,CX. 1+2+3+4 = 10. A
	 * collision (both restaged before either pushed) would corrupt the
	 * first-pushed tuple's words. */
	{ "two_tuple_args",
	  "package main;\n"
	  "fn pair(a: i64, b: i64) (i64, i64) = { return (a, b); };\n"
	  "fn add4(s: (i64, i64), t: (i64, i64)) i64 = {\n"
	  "\treturn s.0 + s.1 + t.0 + t.1;\n"
	  "};\n"
	  "export fn main() i32 = {\n"
	  "\tif (add4(pair(1, 2), pair(3, 4)) != 10) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* CONTROL — a tuple arg threaded through a chain of two calls,
	 * proving the SEND/RECV round-trips through the slot intact. */
	{ "f64f64_chain",
	  "package main;\n"
	  "fn pair(a: f64, b: f64) (f64, f64) = { return (a, b); };\n"
	  "fn id(t: (f64, f64)) f64 = { return t.0 * 10.0 + t.1; };\n"
	  "export fn main() i32 = {\n"
	  "\tif (id(pair(3.0, 5.0)) != 35.0) { return 1; };\n"
	  "\treturn 0;\n"
	  "};\n", 0 },
	/* LOUD-STOP — 5 i64 scalars + an (i64,i64) tuple = 7 INTEGER arg
	 * eightbytes, overflowing the 6 GP arg regs (DI/SI/DX/CX/R8/R9).
	 * The partial-spill stitch is out of scope (twin of #164's cap), so
	 * BOTH stages must FAIL TO COMPILE (rule 7: surface, never silently
	 * drop). Master has no tuple-arg arm (pushes the tuple as 1 word ->
	 * 6 GP, no overflow) and builds the miscompile, so want_compile_fail
	 * discriminates. */
	{ "gp_overflow_loudstop",
	  "package main;\n"
	  "fn pair(a: i64, b: i64) (i64, i64) = { return (a, b); };\n"
	  "fn f(a: i64, b: i64, c: i64, d: i64, e: i64, t: (i64, i64)) i64 = {\n"
	  "\treturn a + b + c + d + e + t.0 + t.1;\n"
	  "};\n"
	  "export fn main() i32 = {\n"
	  "\treturn (f(1, 2, 3, 4, 5, pair(6, 7)): i32);\n"
	  "};\n", 0, 1 },
	{ NULL, NULL, 0, 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, "tupparam: 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/wwtupp_%d_%d.ww", getpid(), i);
		FILE *f = fopen(src, "wb");
		if (f == NULL) { fail++; continue; }
		fputs(rows[i].src, f);
		fclose(f);

		char cmd[2048];

		/* LOUD-STOP rows: the arg-reg overflow must FAIL TO COMPILE in
		 * BOTH stages (rule 7). Assert (a) cstage `ww build` errors and
		 * (b) w6c AND w6c_ww each return non-zero. No .s is produced, so
		 * the byte-id cmp is skipped. */
		if (rows[i].want_compile_fail) {
			char ldir[64];
			snprintf(ldir, sizeof ldir, "/tmp/wwtupp_%d_l_%d",
			    getpid(), i);
			mkdir(ldir, 0755);
			snprintf(cmd, sizeof cmd,
			    "cd %s && %s/ww build %s >/dev/null 2>&1",
			    ldir, bin, src);
			if (runwait(cmd) == 0) {
				fprintf(stderr, "row[%s]: cstage build SUCCEEDED, "
				    "want loud-stop (arg-reg overflow)\n",
				    rows[i].label);
				fail++;
			}
			snprintf(cmd, sizeof cmd, "%s -o /dev/null %s 2>/dev/null",
			    w6c, src);
			if (runwait(cmd) == 0) {
				fprintf(stderr, "row[%s]: w6c emitted .s, want "
				    "loud-stop\n", rows[i].label);
				fail++;
			}
			snprintf(cmd, sizeof cmd, "%s -o /dev/null %s 2>/dev/null",
			    w6c_ww, src);
			if (runwait(cmd) == 0) {
				fprintf(stderr, "row[%s]: w6c_ww emitted .s, want "
				    "loud-stop\n", rows[i].label);
				fail++;
			}
			unlink(src); rmdir(ldir);
			continue;
		}

		/* (a) cstage build + run in a scratch dir. */
		char tmpdir[64];
		snprintf(tmpdir, sizeof tmpdir, "/tmp/wwtupp_%d_d_%d",
		    getpid(), i);
		mkdir(tmpdir, 0755);

		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: emit .s from both stages, cmp. */
		char cs_s[64], ws_s[64];
		snprintf(cs_s, sizeof cs_s, "/tmp/wwtupp_%d_%d_cs.s",
		    getpid(), i);
		snprintf(ws_s, sizeof ws_s, "/tmp/wwtupp_%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 tuple-param tests failed\n", fail, n);
		return 1;
	}
	printf("tupparam: %d/%d ok (cstage run + cs==ww byte-id)\n", n, n);
	return 0;
}