/*
 * 948_idx_compound_run — runtime + byte-id net for #133: a compound
 * assign on an indexed lvalue (`arr[i] OP= v`) must do a real
 * LOAD-OP-STORE, not silently demote `OP=` to `=` (cstage's prior
 * behaviour) and not silently emit NOTHING (wwstage's prior behaviour).
 *
 * Pre-fix:
 *   - cstage cgen.c N_ASSIGN N_INDEX-lhs branch (3699-3851) did NOT gate
 *     on `n->op == TK_ASSIGN`; every compound op fell through and
 *     emitted a plain store of the RHS — silent demotion of `arr[i] +=
 *     1` to `arr[i] = 1`.
 *   - wwstage cgenexpr.ww cgassign N_INDEX-lhs branch (4117-4344) had an
 *     explicit `if (n.op == tkind.TK_ASSIGN)` outer gate with no else —
 *     non-ASSIGN ops emitted nothing at all (silent no-op).
 *
 * Both stages were runtime-WRONG; the fix adds a real load-op-store arm
 * to both stages, mirroring the chained-pointer-field compound template
 * at cgen.c:3281-3317.
 *
 * Each row carries (a) a cstage `ww build` + run asserting the exit
 * code and (b) a w6c vs w6c_ww `.s` cmp (rule-10 byte-id). The wide
 * matrix exercises every compound op the template wires (+= -= *= &=
 * |= ^=) on the narrow/wide element widths the helpers must dispatch
 * (u8, i8, u32, i32, i64), with both bare-array and slice bases.
 * float/str/slice/tagged element compound stays explicitly unwired (the
 * cstage template never wired them either — separate follow-up). A
 * plain-assign control row asserts the byte-id surface for the ASSIGN
 * path is unchanged by the fix.
 */
#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;
}

/* builderr + experr: when builderr != 0 the cstage build MUST FAIL and
 * stderr MUST contain experr. wwstage builderr is verified via direct
 * w6c_ww invocation on the same source. Mirrors 945_tuple_nary's pattern
 * for loud-stop verification (rule 7 — never silent). */
struct row { const char *label; const char *src; int want_exit;
             int builderr; const char *experr; };

static const struct row rows[] = {
	/* u8 += : was silently demoted to `=` in cstage (exit 1, the
	 * post-init value), silently dropped in wwstage (exit 10, the
	 * init value untouched). Post-fix: 10 + 1 = 11. */
	{ "u8_pluseq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [10u8, 20u8, 30u8, 40u8];\n"
	  "	a[0] += 1u8;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 11, 0, NULL },
	/* u8 -= : 30 - 5 = 25. */
	{ "u8_minuseq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [10u8, 20u8, 30u8, 40u8];\n"
	  "	a[2] -= 5u8;\n"
	  "	return a[2]: i32;\n"
	  "};\n", 25, 0, NULL },
	/* u8 *= : 7 * 3 = 21. */
	{ "u8_stareq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [7u8, 0u8, 0u8, 0u8];\n"
	  "	a[0] *= 3u8;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 21, 0, NULL },
	/* u8 &= : 0xF3 & 0x0F = 0x03. Bitwise on the narrow width matters
	 * for the high-bit-set case the silent demote would clobber. */
	{ "u8_ampeq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [0xF3u8, 0u8, 0u8, 0u8];\n"
	  "	a[0] &= 0x0Fu8;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 3, 0, NULL },
	/* u8 |= : 0x10 | 0x07 = 0x17. */
	{ "u8_pipeeq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [0x10u8, 0u8, 0u8, 0u8];\n"
	  "	a[0] |= 0x07u8;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 23, 0, NULL },
	/* u8 ^= : 0xAA ^ 0xFF = 0x55. */
	{ "u8_careteq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [0xAAu8, 0u8, 0u8, 0u8];\n"
	  "	a[0] ^= 0xFFu8;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 85, 0, NULL },
	/* i32 += : exercises the MOVSXD load + MOVL store the localloadop
	 * dispatch picks for narrow signed elements (load extends, store
	 * truncates). 100 + 50 = 150. */
	{ "i32_pluseq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]i32 = [100, 200, 300, 400];\n"
	  "	a[1] += 50;\n"
	  "	return a[1];\n"
	  "};\n", 250, 0, NULL },
	/* i64 += : full-width control. 100 + 34 = 134 (kept <256 since
	 * Unix exit codes are 8-bit). */
	{ "i64_pluseq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]i64 = [100i64, 200i64, 300i64, 400i64];\n"
	  "	a[0] += 34i64;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 134, 0, NULL },
	/* Non-zero index: a[3] += 7 — exercises the scaled-index path
	 * (esz*idx) for u32 element. 100 + 7 = 107. */
	{ "u32_idx3_pluseq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u32 = [10u32, 20u32, 50u32, 100u32];\n"
	  "	a[3] += 7u32;\n"
	  "	return a[3]: i32;\n"
	  "};\n", 107, 0, NULL },
	/* Slice base: same shape through a []u8 view. 22 + 8 = 30. */
	{ "slice_pluseq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let buf: [4]u8 = [11u8, 22u8, 33u8, 44u8];\n"
	  "	let s: []u8 = buf[0:4];\n"
	  "	s[1] += 8u8;\n"
	  "	return s[1]: i32;\n"
	  "};\n", 30, 0, NULL },
	/* Plain `arr[i] = v` control: the ASSIGN path must be unchanged by
	 * the fix (byte-id preserved on the legacy surface). */
	{ "plain_assign_ctrl",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u8 = [10u8, 20u8, 30u8, 40u8];\n"
	  "	a[1] = 99u8;\n"
	  "	return a[1]: i32;\n"
	  "};\n", 99, 0, NULL },
	/* #133-expanded: SLASHEQ/PERCENTEQ/LSHIFTEQ/RSHIFTEQ on indexed
	 * scalar elements. Pre-expansion both stages silently identity-
	 * stored (load, pop RHS, store ORIGINAL — RHS discarded). Width-
	 * dispatch via fldloadop/fldstoreop; signedness via type_isunsigned
	 * picks IDIV/SAR (signed) vs DIV/SHR (unsigned). Note: signed
	 * RSHIFTEQ uses SHRQ in parity with cgen.c:4034 (A_SARQ not in w6a;
	 * pre-existing concern). */
	/* signed i32 /=: 100 / 4 = 25 (positive, no signed/unsigned diff). */
	{ "i32_slasheq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]i32 = [100, 200, 300, 400];\n"
	  "	a[0] /= 4;\n"
	  "	return a[0];\n"
	  "};\n", 25, 0, NULL },
	/* unsigned u32 /=: 100u32 / 4u32 = 25u32. Routes through DIVQ +
	 * zero-DX (NOT CQO+IDIVQ — signed-divide of a high-bit-set u32
	 * narrow-load would mis-treat the operand as negative). */
	{ "u32_slasheq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u32 = [100u32, 200u32, 300u32, 400u32];\n"
	  "	a[0] /= 4u32;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 25, 0, NULL },
	/* signed i32 %=: 17 % 5 = 2; result rides DX, MOVQ DX,AX to land
	 * in AX for the store. */
	{ "i32_percenteq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]i32 = [17, 0, 0, 0];\n"
	  "	a[0] %= 5;\n"
	  "	return a[0];\n"
	  "};\n", 2, 0, NULL },
	/* unsigned u32 %=: 100u32 % 7u32 = 2u32. */
	{ "u32_percenteq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u32 = [100u32, 0u32, 0u32, 0u32];\n"
	  "	a[0] %= 7u32;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 2, 0, NULL },
	/* i32 <<=: 3 << 4 = 48. SHLQ via CX. */
	{ "i32_lshifteq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]i32 = [3, 0, 0, 0];\n"
	  "	a[0] <<= 4;\n"
	  "	return a[0];\n"
	  "};\n", 48, 0, NULL },
	/* unsigned u32 >>=: 200u32 >> 2 = 50u32. SHRQ via CX. */
	{ "u32_rshifteq",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]u32 = [200u32, 0u32, 0u32, 0u32];\n"
	  "	a[0] >>= 2u32;\n"
	  "	return a[0]: i32;\n"
	  "};\n", 50, 0, NULL },
	/* signed i32 >>=: 200 >> 2 = 50 (positive, parity with SHRQ —
	 * A_SARQ unimplemented in w6a; pre-existing signed-RSHIFT on
	 * negatives subtle, out of scope for #133). */
	{ "i32_rshifteq_pos",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]i32 = [200, 0, 0, 0];\n"
	  "	a[0] >>= 2;\n"
	  "	return a[0];\n"
	  "};\n", 50, 0, NULL },
	/* HARD-ERROR rows (#133-expanded rule-7): float / str / slice /
	 * tagged element compound builds MUST FAIL LOUD on both stages
	 * with the cited diagnostic substring. Pre-expansion these were
	 * silent fall-through-to-default-break → silent miscompile.
	 * Verification: cstage build returns non-zero AND stderr carries
	 * the substring; wwstage same via w6c_ww direct invocation. */
	{ "he_float_indexed",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]f64 = [1.0, 2.0, 3.0, 4.0];\n"
	  "	a[0] += 1.5;\n"
	  "	return 0;\n"
	  "};\n", 0, 1, "indexed-lvalue compound on float element" },
	{ "he_str_indexed",
	  "package main;\n"
	  "export fn main() i32 = {\n"
	  "	let a: [4]str = [\"a\", \"b\", \"c\", \"d\"];\n"
	  "	a[0] += \"x\";\n"
	  "	return 0;\n"
	  "};\n", 0, 1, "indexed-lvalue compound on" },
	{ "he_float_chained_ptr",
	  "package main;\n"
	  "type t = struct { v: f64 };\n"
	  "type w = struct { i: *t };\n"
	  "fn bad(d: *w) void = { d.i.v += 1.5; };\n"
	  "export fn main() i32 = { return 0; };\n",
	  0, 1, "chained-ptr-field compound on float element" },
	{ NULL, NULL, 0, 0, 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, "idxcompound: 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/wwidx_%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, OR build-must-fail (builderr rows).
		 * For builderr: stderr captured to a temp file; pass iff
		 * exit-nonzero AND stderr contains experr substring. */
		char tmpdir[64];
		snprintf(tmpdir, sizeof tmpdir, "/tmp/wwidx_%d_d_%d",
		    getpid(), i);
		mkdir(tmpdir, 0755);

		char cmd[2048];
		if (rows[i].builderr) {
			char errf[96];
			snprintf(errf, sizeof errf, "/tmp/wwidx_%d_e_%d",
			    getpid(), i);
			snprintf(cmd, sizeof cmd,
			    "cd %s && %s/ww build %s >/dev/null 2>%s",
			    tmpdir, bin, src, errf);
			int brc = runwait(cmd);
			FILE *ef = fopen(errf, "rb");
			char ebuf[4096];
			size_t en = 0;
			if (ef) { en = fread(ebuf, 1, sizeof ebuf - 1, ef);
			    fclose(ef); }
			ebuf[en] = '\0';
			int ok = (brc != 0)
			    && (rows[i].experr == NULL
			        || strstr(ebuf, rows[i].experr) != NULL);
			if (!ok) {
				fprintf(stderr, "row[%s]: cstage expected "
				    "builderr+'%s' (brc=%d, stderr='%s')\n",
				    rows[i].label,
				    rows[i].experr ? rows[i].experr : "(any)",
				    brc, ebuf);
				fail++;
			}

			/* Also verify wwstage hard-errors with the SAME message
			 * (rule-10 — both stages identical diagnostic). Invoke
			 * w6c_ww directly on the .ww source. */
			snprintf(cmd, sizeof cmd,
			    "%s -o /dev/null %s >/dev/null 2>%s",
			    w6c_ww, src, errf);
			int wrc = runwait(cmd);
			ef = fopen(errf, "rb"); en = 0;
			if (ef) { en = fread(ebuf, 1, sizeof ebuf - 1, ef);
			    fclose(ef); }
			ebuf[en] = '\0';
			int wok = (wrc != 0)
			    && (rows[i].experr == NULL
			        || strstr(ebuf, rows[i].experr) != NULL);
			if (!wok) {
				fprintf(stderr, "row[%s]: wwstage expected "
				    "builderr+'%s' (wrc=%d, stderr='%s')\n",
				    rows[i].label,
				    rows[i].experr ? rows[i].experr : "(any)",
				    wrc, ebuf);
				fail++;
			}
			unlink(errf);
			unlink(src); rmdir(tmpdir);
			continue;
		}

		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/wwidx_%d_%d_cs.s",
		    getpid(), i);
		snprintf(ws_s, sizeof ws_s, "/tmp/wwidx_%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 idx-compound tests failed\n",
		    fail, n);
		return 1;
	}
	printf("idxcompound: %d/%d ok (cstage run + cs==ww byte-id)\n",
	    n, n);
	return 0;
}