selfhost/cmd/w6a+w6l: rt.malloc/amalloc β grow → alloc([], n)! (β-2)

Phase 0 second β-batch. 6 alloc sites across 3 β grow loops:
 - selfhost/cmd/w6a/obj.ww:100,108  buf bufinit/bufgrow (amalloc)
 - selfhost/cmd/w6a/asm.ww:26,63    asm_ emitbyte/emitdatabyte text+data (rt.malloc)
 - selfhost/cmd/w6l/obj.ww:113,135  lnk emittext/emitdata text+data (rt.malloc)

Same β shape as d9f0972: `*u8 = (rt.malloc|amalloc)(_, n): *u8`
→ `[]u8 = alloc([], n)!`, inner copy bounded by the manual length
counter (b.n / textlen / datalen) unchanged, struct field stays
*u8, writeback via .ptr. Bear-trap N/A (slice .len = 0 never read).

w6a/asm.ww `import rt;` (line 15) is now dead — deferred to a
post-Phase-0 import-hygiene sweep, mirrors 368b85e.

Verified 132/132 + 995_self_rebuild byte-identity.

selfhost/cmd/w6a/asm.ww

@@ -23,10 +23,10 @@ export fn emitbyte(a: *asm_, b: u8) void = {
 		let nc: u64 = a.textcap;
 		if (nc == 0u64) { nc = 4096u64; };
 		nc = nc * 2u64;
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < a.textlen) { nb[i] = a.text[i]; i += 1u64; };
-		a.text = nb;
+		a.text = nb.ptr;
 		a.textcap = nc;
 	};
 	a.text[a.textlen] = b;
@@ -60,10 +60,10 @@ export fn emitdatabyte(a: *asm_, b: u8) void = {
 		let nc: u64 = a.datacap;
 		if (nc == 0u64) { nc = 256u64; };
 		nc = nc * 2u64;
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < a.datalen) { nb[i] = a.data[i]; i += 1u64; };
-		a.data = nb;
+		a.data = nb.ptr;
 		a.datacap = nc;
 	};
 	a.data[a.datalen] = b;

selfhost/cmd/w6a/main.combined.ww

@@ -1760,10 +1760,10 @@ export fn emitbyte(a: *asm_, b: u8) void = {
 		let nc: u64 = a.textcap;
 		if (nc == 0u64) { nc = 4096u64; };
 		nc = nc * 2u64;
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < a.textlen) { nb[i] = a.text[i]; i += 1u64; };
-		a.text = nb;
+		a.text = nb.ptr;
 		a.textcap = nc;
 	};
 	a.text[a.textlen] = b;
@@ -1797,10 +1797,10 @@ export fn emitdatabyte(a: *asm_, b: u8) void = {
 		let nc: u64 = a.datacap;
 		if (nc == 0u64) { nc = 256u64; };
 		nc = nc * 2u64;
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < a.datalen) { nb[i] = a.data[i]; i += 1u64; };
-		a.data = nb;
+		a.data = nb.ptr;
 		a.datacap = nc;
 	};
 	a.data[a.datalen] = b;
@@ -2687,17 +2687,18 @@ fn bufinit(b: *buf, a: *arena) void = {
 	b.a = a;
 	b.cap = 256u64;
 	b.n = 0u64;
-	b.p = amalloc(a, b.cap): *u8;
+	let np: []u8 = alloc([], b.cap)!;
+	b.p = np.ptr;
 };
 
 fn bufgrow(b: *buf, need: u64) void = {
 	if (b.n + need <= b.cap) { return; };
 	let nc: u64 = b.cap;
 	for (nc < b.n + need) { nc = nc * 2u64; };
-	let np: *u8 = amalloc(b.a, nc): *u8;
+	let np: []u8 = alloc([], nc)!;
 	let i: u64 = 0u64;
 	for (i < b.n) { np[i] = b.p[i]; i += 1u64; };
-	b.p = np;
+	b.p = np.ptr;
 	b.cap = nc;
 };
 

selfhost/cmd/w6a/obj.ww

@@ -97,17 +97,18 @@ fn bufinit(b: *buf, a: *arena) void = {
 	b.a = a;
 	b.cap = 256u64;
 	b.n = 0u64;
-	b.p = amalloc(a, b.cap): *u8;
+	let np: []u8 = alloc([], b.cap)!;
+	b.p = np.ptr;
 };
 
 fn bufgrow(b: *buf, need: u64) void = {
 	if (b.n + need <= b.cap) { return; };
 	let nc: u64 = b.cap;
 	for (nc < b.n + need) { nc = nc * 2u64; };
-	let np: *u8 = amalloc(b.a, nc): *u8;
+	let np: []u8 = alloc([], nc)!;
 	let i: u64 = 0u64;
 	for (i < b.n) { np[i] = b.p[i]; i += 1u64; };
-	b.p = np;
+	b.p = np.ptr;
 	b.cap = nc;
 };
 

selfhost/cmd/w6l/main.combined.ww

@@ -1084,13 +1084,13 @@ fn emittext(l: *lnk, src: *u8, n: u64) void = {
 		// Grow by mmap'ing a fresh region and copying. The old buffer
 		// is leaked into the page allocator; for a linker run this is
 		// trivial waste.
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < l.textlen) {
 			nb[i] = l.text[i];
 			i += 1u64;
 		};
-		l.text = nb;
+		l.text = nb.ptr;
 		l.textcap = nc;
 	};
 	let i: u64 = 0u64;
@@ -1106,13 +1106,13 @@ fn emitdata(l: *lnk, src: *u8, n: u64) void = {
 		let nc: u64 = l.datacap;
 		if (nc == 0u64) { nc = 256u64; };
 		for (nc < l.datalen + n) { nc = nc * 2u64; };
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < l.datalen) {
 			nb[i] = l.data[i];
 			i += 1u64;
 		};
-		l.data = nb;
+		l.data = nb.ptr;
 		l.datacap = nc;
 	};
 	let i: u64 = 0u64;

selfhost/cmd/w6l/obj.ww

@@ -110,13 +110,13 @@ fn emittext(l: *lnk, src: *u8, n: u64) void = {
 		// Grow by mmap'ing a fresh region and copying. The old buffer
 		// is leaked into the page allocator; for a linker run this is
 		// trivial waste.
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < l.textlen) {
 			nb[i] = l.text[i];
 			i += 1u64;
 		};
-		l.text = nb;
+		l.text = nb.ptr;
 		l.textcap = nc;
 	};
 	let i: u64 = 0u64;
@@ -132,13 +132,13 @@ fn emitdata(l: *lnk, src: *u8, n: u64) void = {
 		let nc: u64 = l.datacap;
 		if (nc == 0u64) { nc = 256u64; };
 		for (nc < l.datalen + n) { nc = nc * 2u64; };
-		let nb: *u8 = rt.malloc(nc): *u8;
+		let nb: []u8 = alloc([], nc)!;
 		let i: u64 = 0u64;
 		for (i < l.datalen) {
 			nb[i] = l.data[i];
 			i += 1u64;
 		};
-		l.data = nb;
+		l.data = nb.ptr;
 		l.datacap = nc;
 	};
 	let i: u64 = 0u64;