Incorrect handling of int128 values about arocc

[ehaas]

Possibly relevant:

void foo(void) {
    unsigned __int128 x = ((unsigned __int128)18446744073709551615UL)<< 64;
}

test.c:2:70: warning: overflow in expression; result is '-18446744073709551616' [-Winteger-overflow]
    unsigned __int128 x = ((unsigned __int128)18446744073709551615UL)<< 64;
                                                                     ^
1 warning generated.

(no diagnostics on clang)

[Vexu]

Bug is in Type.signedness:

diff --git a/src/aro/Type.zig b/src/aro/Type.zig
index 7e691bc..1369d7b 100644
--- a/src/aro/Type.zig
+++ b/src/aro/Type.zig
@@ -609,7 +609,7 @@ pub fn signedness(ty: Type, comp: *const Compilation) std.builtin.Signedness {
     return switch (ty.specifier) {
         // zig fmt: off
         .char, .complex_char => return comp.getCharSignedness(),
-        .uchar, .ushort, .uint, .ulong, .ulong_long, .bool, .complex_uchar, .complex_ushort,
+        .uchar, .ushort, .uint, .ulong, .ulong_long, .uint128, .bool, .complex_uchar, .complex_ushort,
         .complex_uint, .complex_ulong, .complex_ulong_long, .complex_uint128 => .unsigned,
         // zig fmt: on
         .bit_int, .complex_bit_int => ty.data.int.signedness,