/*
 * 767_match_variant_dispatch — project #179: wwstage cgmatch's
 * non-nullable arm gated its flat-tag dispatch on `pat.kind ==
 * N_TNAME` (with an `N_TSLICE` else-branch for task #19's untyped
 * elem fallback). N_TPTR / N_TFN / N_TPTR(N_TFN) case-patterns fell
 * through both arms, leaving the local `r` at -1 → `want = 0`, so
 * every variant past 0 silently collapsed to tag 0 — the kind of
 * miscompile that runtime-passes only when the value happens to be
 * variant 0 (zero-coincidence).
 *
 * Cstage cg_tag_for_variant (cmd/w6c/cgen.c:624) walks Type directly
 * and is kind-agnostic; it has never had this gap. Drew's settled
 * principle: match dispatch is resolved-type-only (harec
 * check.c:2527 stores `_case->type = ctype`). Project #66 Phase-N
 * already flipped the typeeq compare; #179 was the one site that
 * still keyed on AST kind. Memory: project_tinfo_lossy_nominal.
 *
 * Fix: selfhost/cmd/wcc/cgenexpr.ww cgmatch non-nullable arm —
 * route through `flatvariantidxt(scrutt.type_, pat.type_)` directly,
 * guarded only by `istaggedtype(scrutt)` + `typeisslice(pattype)`
 * for the slice axis. No new helper; existing flatvariantidxt /
 * flatslicevariantidx wire up unchanged.
 *
 * Coverage (5 rows):
 *   1. nullable_ptr_fn        — `(*fn(i32) i32 | void)`, branched
 *                              store (#105 lesson: two distinct
 *                              fns picked at runtime), match via
 *                              indirect call. KEN'S VERIFY GATE:
 *                              row 1 byte-id proves wwstage's
 *                              nullable-fold ordering matches
 *                              cstage cg_tag_for_variant — fail
 *                              here flags a separate-signoff fold.
 *                              Goes through the nullable disc path
 *                              (line 1484-1502, untouched by #179).
 *   2. ptr_variants_past_zero — `(*i32 | *i64)` storing &i64.
 *                              Pre-fix wwstage emits CMPQ $0 for
 *                              the *i64 arm; post-fix CMPQ $1.
 *   3. fn_ptr_variants_via_local — `(*fn(i32) i32 | *fn(i64) i64)`
 *                              built from a local. Direct N_TPTR
 *                              over N_TFN case-pattern is the new-
 *                              shape exercise; store path goes via
 *                              ident (not `&fn` inline) to dodge
 *                              the fn-variant store-side sibling
 *                              bug (see below).
 *   4. stored_variant1_roundtrip — `(*i32 | *i64)` with branched
 *                              runtime variant choice (pick = 1
 *                              forces variant 1; defeats const-
 *                              fold to variant 0). Mirrors #105.
 *   5. aliased_ptr_variants   — `type pa = *i32; type pb = *i64;
 *                              (pa | pb)`. typeeq through TY_NAMED
 *                              wrappers — the alias-aware path that
 *                              the surface-name compare couldn't
 *                              key on. OR-pattern was the original
 *                              row 5, dropped because the wwstage
 *                              parser doesn't accept `case T1 | T2`.
 *
 * Per-row gates: cstage runtime exit, wwstage runtime exit,
 * cs.s == ww.s byte-identical (rule-10 stage symmetry).
 *
 * Sibling bug surfaced (not fixed here, filed inline): widening
 * `&fn` INLINE into a `(*fn(...) | *fn(...))` slot computes the
 * wrong tag in wwstage — taggedvariantindext / typeeq on the
 * synthetic fn-ptr tinfo built from N_UN TK_AMP IDENT-of-FN doesn't
 * match the variant tinfo, so tag collapses to 0. Storing via an
 * intermediate ident-typed local (row 3 shape) routes through a
 * different store path and tags correctly. cstage cg_widen_tagged_
 * store hits the same node with the same Type and tags correctly,
 * so the divergence is wwstage-only on the inline-amp shape. Out
 * of scope for #179 (which is the match-DISPATCH side, not the
 * widen-STORE side).
 *
 * GATE POLARITY: must stay GREEN. A red here means the cgmatch
 * non-nullable dispatch regressed, the nullable arm got perturbed,
 * or stage symmetry drifted on the new typeeq route.
 */
#include <stdio.h>
#include <stdlib.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;
}

#define STAGE_CS	1
#define STAGE_WW	2

struct row {
	const char *label;
	const char *src;
	int expected_exit;
	int stage_mask;
};

static const struct row rows[] = {
	/* KEN'S VERIFY GATE: nullable disc path must byte-id even
	 * though #179 didn't touch it. A divergence here flags a
	 * separate flatvariantidxt nullable-fold ordering issue. */
	{ "nullable_ptr_fn",
	  "fn ai(x: i32) i32 = { return x + 1; };\n"
	  "fn bi(x: i32) i32 = { return x + 100; };\n"
	  "fn indi(f: *fn(x: i32) i32, x: i32) i32 = { return (*f)(x); };\n"
	  "export fn main() i32 = {\n"
	  "    let pick: i32 = 1;\n"
	  "    let v: (*fn(x: i32) i32 | void) = &ai;\n"
	  "    if (pick != 0) { v = &bi; };\n"
	  "    let r: i32 = match (v) {\n"
	  "    case void => yield 99: i32;\n"
	  "    case let f: *fn(x: i32) i32 => yield indi(f, 7);\n"
	  "    };\n"
	  "    return r;\n"
	  "};\n",
	  107,
	  STAGE_CS | STAGE_WW },
	{ "ptr_variants_past_zero",
	  "export fn main() i32 = {\n"
	  "    let a: i64 = 42;\n"
	  "    let v: (*i32 | *i64) = &a;\n"
	  "    let r: i32 = match (v) {\n"
	  "    case *i32 => yield 11: i32;\n"
	  "    case *i64 => yield 22: i32;\n"
	  "    };\n"
	  "    return r;\n"
	  "};\n",
	  22,
	  STAGE_CS | STAGE_WW },
	{ "fn_ptr_variants_via_local",
	  "fn ai(x: i32) i32 = { return x + 1; };\n"
	  "fn bi(x: i64) i64 = { return x + 1; };\n"
	  "fn indl(f: *fn(x: i64) i64, x: i64) i64 = { return (*f)(x); };\n"
	  "export fn main() i32 = {\n"
	  "    let g: *fn(x: i64) i64 = &bi;\n"
	  "    let v: (*fn(x: i32) i32 | *fn(x: i64) i64) = g;\n"
	  "    let r: i32 = match (v) {\n"
	  "    case let p: *fn(x: i32) i32 => yield 1: i32;\n"
	  "    case let q: *fn(x: i64) i64 => yield indl(q, 9): i32;\n"
	  "    };\n"
	  "    return r;\n"
	  "};\n",
	  10,
	  STAGE_CS | STAGE_WW },
	{ "stored_variant1_roundtrip",
	  "export fn main() i32 = {\n"
	  "    let a: i32 = 5;\n"
	  "    let b: i64 = 9;\n"
	  "    let pick: i32 = 1;\n"
	  "    let v: (*i32 | *i64) = &a;\n"
	  "    if (pick != 0) { v = &b; };\n"
	  "    let r: i32 = match (v) {\n"
	  "    case *i32 => yield 11: i32;\n"
	  "    case *i64 => yield 22: i32;\n"
	  "    };\n"
	  "    return r;\n"
	  "};\n",
	  22,
	  STAGE_CS | STAGE_WW },
	{ "aliased_ptr_variants",
	  "type pa = *i32;\n"
	  "type pb = *i64;\n"
	  "export fn main() i32 = {\n"
	  "    let a: i64 = 100;\n"
	  "    let g: pb = &a;\n"
	  "    let v: (pa | pb) = g;\n"
	  "    let r: i32 = match (v) {\n"
	  "    case pa => yield 11: i32;\n"
	  "    case pb => yield 22: i32;\n"
	  "    };\n"
	  "    return r;\n"
	  "};\n",
	  22,
	  STAGE_CS | STAGE_WW },
};

static int
write_source(const char *src_path, const char *src)
{
	FILE *f = fopen(src_path, "wb");
	if (!f) return -1;
	fputs(src, f);
	fclose(f);
	return 0;
}

static void
cleanup_tmp(const char *tmpdir, const char *base)
{
	char p[512];
	snprintf(p, sizeof p, "%s/%s.ww", tmpdir, base); unlink(p);
	snprintf(p, sizeof p, "%s/%s.o",  tmpdir, base); unlink(p);
	snprintf(p, sizeof p, "%s/%s.combined.ww", tmpdir, base); unlink(p);
	snprintf(p, sizeof p, "%s/%s",    tmpdir, base); unlink(p);
	/* #93: the sep scratch dir + the tmpdir-pinned pkgcache. */
	snprintf(p, sizeof p, "rm -rf %s/%s.sepwork %s/pkgc", tmpdir, base, tmpdir);
	if (system(p)) {} /* best-effort */
	rmdir(tmpdir);
}

static int
build_via_driver(const char *driver, const char *tmpdir, const char *src)
{
	char cmd[1024];
	/* #93 sep layout: emit asm to <src-stem>.sepwork/__root.s; pin
	 * WW_PKGCACHE under tmpdir so out/.pkgcache is untouched. */
	snprintf(cmd, sizeof cmd,
	    "cd %s && b='%s'; WW_PKGCACHE=%s/pkgc timeout 180 %s build --sep "
	    "-o \"${b%%.ww}\" \"$b\" 2>/dev/null",
	    tmpdir, src, tmpdir, driver);
	return runwait(cmd);
}

static int
run_row(const char *driver, const struct row *r, int seq)
{
	char tmpdir[256], src[256], base[64], outbin[512];
	snprintf(tmpdir, sizeof tmpdir, "/tmp/mvd_%d_d_%d", getpid(), seq);
	snprintf(src,    sizeof src,    "%s/main767.ww", tmpdir);
	snprintf(base,   sizeof base,   "main767");
	mkdir(tmpdir, 0755);
	if (write_source(src, r->src) != 0) {
		cleanup_tmp(tmpdir, base);
		return -1;
	}
	int rc = -1;
	if (build_via_driver(driver, tmpdir, src) == 0) {
		snprintf(outbin, sizeof outbin, "%s/%s", tmpdir, base);
		rc = runwait(outbin);
	}
	cleanup_tmp(tmpdir, base);
	return rc;
}

/* asm_byte_identical — diff cstage vs wwstage .s. Parallel trees so
 * ww_ww writing intermediates next to the source doesn't clobber the
 * cstage .s (filed bug per CLAUDE.md rule 14 phase split). */
static int
asm_byte_identical(const char *cdrv, const char *wdrv,
    const struct row *r, int seq)
{
	char src[256], tdc[256], tdw[256], base[64], cs[512], ws[512];
	snprintf(tdc, sizeof tdc, "/tmp/mvd_%d_c_%d", getpid(), seq);
	snprintf(tdw, sizeof tdw, "/tmp/mvd_%d_w_%d", getpid(), seq);
	snprintf(base, sizeof base, "main767");
	mkdir(tdc, 0755);
	mkdir(tdw, 0755);
	snprintf(src, sizeof src, "%s/main767.ww", tdc);
	if (write_source(src, r->src) != 0) { cleanup_tmp(tdc, base); cleanup_tmp(tdw, base); return -1; }
	int rc = -1;
	if (build_via_driver(cdrv, tdc, src) != 0) goto out;
	snprintf(cs, sizeof cs, "%s/%s.sepwork/__root.s", tdc, base);

	snprintf(src, sizeof src, "%s/main767.ww", tdw);
	if (write_source(src, r->src) != 0) goto out;
	if (build_via_driver(wdrv, tdw, src) != 0) goto out;
	snprintf(ws, sizeof ws, "%s/%s.sepwork/__root.s", tdw, base);

	FILE *fc = fopen(cs, "rb");
	FILE *fw = fopen(ws, "rb");
	if (fc && fw) {
		rc = 0;
		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);
out:
	cleanup_tmp(tdc, base);
	cleanup_tmp(tdw, base);
	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], wdrv[1024];
	snprintf(cdrv, sizeof cdrv, "%s/ww",    bin);
	snprintf(wdrv, sizeof wdrv, "%s/ww_ww", bin);

	int n = (int)(sizeof rows / sizeof rows[0]);
	int total = 0, fail = 0;
	int wwpresent = (access(wdrv, X_OK) == 0);
	int seq = 0;

	for (int i = 0; i < n; i++) {
		if (rows[i].stage_mask & STAGE_CS) {
			total++;
			int got = run_row(cdrv, &rows[i], seq++);
			if (got != rows[i].expected_exit) {
				fprintf(stderr,
				    "match_variant_dispatch[cstage run][%s]: exit=%d want=%d\n",
				    rows[i].label, got, rows[i].expected_exit);
				fail++;
			}
		}
		if (wwpresent && (rows[i].stage_mask & STAGE_WW)) {
			total++;
			int got = run_row(wdrv, &rows[i], seq++);
			if (got != rows[i].expected_exit) {
				fprintf(stderr,
				    "match_variant_dispatch[wwstage run][%s]: exit=%d want=%d\n",
				    rows[i].label, got, rows[i].expected_exit);
				fail++;
			}
			total++;
			if (asm_byte_identical(cdrv, wdrv, &rows[i], seq++) != 0) {
				fprintf(stderr,
				    "match_variant_dispatch[byte-id][%s]: cstage vs wwstage asm differs\n",
				    rows[i].label);
				fail++;
			}
		}
	}

	if (!wwpresent)
		fprintf(stderr, "match_variant_dispatch: skip wwstage (no %s)\n", wdrv);

	if (fail) {
		fprintf(stderr, "match_variant_dispatch: %d/%d fixtures failed\n",
		    fail, total);
		return 1;
	}
	printf("match_variant_dispatch: %d/%d ok\n", total, total);
	return 0;
}