Memory64Lowering: table.get/table.set keep i64 index on lowered table64, output fails validation #9245

--memory64-lowering turns an i64 table into an i32 table but leaves the i64 index of table.get and table.set unwrapped, so the output fails validation.

(module
 (table $t i64 1 funcref)
 (func $f (param $r funcref)
  (table.set $t (i64.const 0) (local.get $r))
  (drop (table.get $t (i64.const 0)))))

wasm-opt input.wat -all --memory64-lowering -o /dev/null

[wasm-validator error in function f] i64 != i32: table.set index must match the table index type., on 
(table.set $t
 (i64.const 0)
 (local.get $0)
)
[wasm-validator error in function f] i64 != i32: table.get index must match the table index type., on 
(table.get $t
 (i64.const 0)
)
Fatal: error after opts

I expected the pass to wrap these indices to i32 the way it does for the other table instructions, giving a valid module. The input itself validates (wasm-opt -all input.wat -o /dev/null exits 0).

Found at commit 93d6e9d and still reproduces at 611832f. A normal release build shows it as above (exit 1). With BINARYEN_PASS_DEBUG=1 the same errors are followed by Fatal: Last pass (memory64-lowering) broke validation.

Looks like src/passes/Memory64Lowering.cpp wraps the i64 operands of table.size/grow/fill/copy/init through wrapTableAddress64 but has no visitTableGet/visitTableSet, so those indices stay i64.

Found with an LLM-based testing tool; I used Claude to reduce it and look for the cause.