w6l: purge dead elf branches; pin rx extent

selfhost/cmd/w6l/obj.ww

@@ -14,8 +14,8 @@ def SHT_SYMTAB:    i32 = 2;
 def SHT_STRTAB:    i32 = 3;
 def SHT_RELA:      i32 = 4;
 
-// w6a/w6l use straight LE on amd64. Reading via byte offsets keeps us off
-// the cgen's u16 field-load story for now (MOVZBQ exists; MOVZWQ doesn't).
+// ELF64/x86-64 is little-endian; bytewise decoding also avoids alignment
+// and host-layout assumptions about input buffers.
 
 fn rdu16(p: *u8, off: u64) u16 = {
 	let b0: u16 = p[off]: u16;
@@ -507,9 +507,8 @@ fn loadimage(l: *lnk, path: *u8, buf: *u8, len: u64) i32 = {
 		};
 	};
 
-	// We don't keep a per-object map[] of *lsym. Instead the reloc
-	// loop re-walks symtab and re-interns by name. Simpler than
-	// dancing around the cgen's u64-shift gaps.
+	// Rewalking avoids retaining an nsyms-sized per-object map solely
+	// for relocation lookup.
 	let si: u64 = 1u64;	// skip index 0 (always undef sentinel)
 	for (si < nsyms) {
 		let symp: u64 = symoff + si * SYM_SIZE;
@@ -526,7 +525,6 @@ fn loadimage(l: *lnk, path: *u8, buf: *u8, len: u64) i32 = {
 				if (idxdata >= 0) {
 					indt = (stshndx: i32) == idxdata;
 				};
-				if (intext && indt) { indt = false; };
 				if (intext) {
 					if (gs.defined != 0) {
 						let m: str = "w6l: duplicate symbol\n";