feat: bitfield struct translate

[GreyElaina]

Basically copied from https://github.com/mahkoh/repr-c/tree/master/repc/impl/src/builder.

Required to build a cross-verifier to fit existing behavior.

[Vexu]

C structs cannot be translated as packed structs, instead bitfields that are part of the same host byte should be translated as a packed struct that is part of the larger struct:

struct S {
    int arr[4];
    int a: 12;
    int b: 12;
    int :0;
    int c: 12;
};

pub const struct_S = extern struct {
    arr: [4]c_int = @import("std").mem.zeroes([4]c_int),
    bitfield: packed struct(c_int) {
        a: i12 = 0,
        b: i12 = 0,
        _pad: u8 = 0,
    } = .{},
    bitfield1: packed struct(c_int) {
        c: i12 = 0,
        _pad: u20 = 0,
    } = .{},
};

[GreyElaina]

Even just the work of creating documentation for what the C ABI for bit fields is for the various compilers would be 60%+ of the work of solving this issue.

So is it. Any helpful reference could we have? or more testcases.

[GreyElaina]

ok i'm stupid, arocc already ship it.

[steeve]

@GreyElaina incredible work, what is the status of that draft PR ?
For context, we use DPDK heavily and it relies a lot on bitfields. Even ignoring them instead of turning the whole struct opaque would go a long way (at least for us)

[GreyElaina]

@GreyElaina incredible work, what is the status of that draft PR ? For context, we use DPDK heavily and it relies a lot on bitfields. Even ignoring them instead of turning the whole struct opaque would go a long way (at least for us)

eh, i just borrowed these values from arocc itself. it's hard (and too busy for me) to write a comprehensive verifier to, at least, fuzz test it.