#define BIT(n) Shift Operand (<<) that's translated from C is not working as expected

[ShoshinX]

Hi,

I was working on importing a set of C libraries into zig to be used in zig. However, I found a bug either in the translate-c code or in the zig compiler. The following is the smallest example I found as I was exploring the compilation errors.

This is using Zig 0.8.0 release.

I was exporting a C preprocessor BIT(n) defined as follows:

#define BIT(n) (1ul << n)

into zig. Zig translated it into

pub inline fn BIT(n:anytype) @TypeOf(@as(c_ulong,1) << n) {
  return 1 << n;
}

When I tried using it through const c = @cInclude(...) or the translated cimport.zig
I get an error as the following

error: expected type 'u6', found 'c_ulong'
fn bit(n: anytype) @TypeOf(@as(c_ulong, 1) << n) {
                                              ^
...
note: unsigned 6-bit int cannot represent all possible unsigned 64-bit values
fn bit(n: anytype) @TypeOf(@as(c_ulong, 1) << n) {
                                              ^

I found out from Spex_guy#0444 that this could be a bug in the zig translate-c.
Here's the following additional info Spex_guy provided me

"Zig has extra rules about shift operands, which we are planning to change to be less restrictive. The current rule is that the rhs of a shift must be an unsigned int type with ceil(log2(number of bits in lhs type)) bits. If you don't know the lhs type statically, you can use %%std.math.Log2Int to calculate the rhs type at comptime.

translate-c should probably translate shifts to have the same semantics as C, and add an implicit cast of the rhs, I think this does count as a C translation bug."

[ShoshinX]

Further investigations, I found out that adding comptime before the parameter n removes the error, so pub inline fn BIT(comptime n:anytype) @TypeOf(@as(c_ulong,1) << n) now works. However, I think this is just a bandaid fix for now i.e I will have to rewrite the #define BIT(n)in zig and somehow create a function that handles it during runtime instead of the current comptime bandaid fix

[AssortedFantasy]

This bug is because zig requires an awkward, and in my opinion unessasary cast because it doesn't implicitly @intCast to the correct type.

There are proposals related to this. See my comment in #7605 for this.

The TL;DR is that a << b should have semantics of a << @intCast(@import("std").math.Log2Int(a), b).

Anyway doing this also fixes the macro.

Edit: lmao didn't read, Already mentioned in issue by Spex.