Looks like the CI errors for miri occur on master also

Thanks for fixing the case for an offset of 4, but it seems like using an offset of 2 still causes it to fail:

let test_data: &[u8] = [0b11000000, 0b00111111].as_ref();

#[derive(Debug, PartialEq, DekuRead)]
#[deku(endian = "little")]
struct Container {
    #[deku(bits = 2)]
    field1: u8,
    field2: u8,
    #[deku(bits = 6)]
    field3: u8,
}

assert_eq!(
    Container::from_bytes((test_data, 0)).unwrap().1,
    Container {
        field1: 0b11,
        field2: 0,
        field3: 0b111111,
    }
);

Adding in an explicit #[deku(bits = 8)] to field2 causes this case to work as well, so I believe it is likely related.

Thanks for fixing the case for an offset of 4, but it seems like using an offset of 2 still causes it to fail:

let test_data: &[u8] = [0b11000000, 0b00111111].as_ref();

#[derive(Debug, PartialEq, DekuRead)]
#[deku(endian = "little")]
struct Container {
    #[deku(bits = 2)]
    field1: u8,
    field2: u8,
    #[deku(bits = 6)]
    field3: u8,
}

assert_eq!(
    Container::from_bytes((test_data, 0)).unwrap().1,
    Container {
        field1: 0b11,
        field2: 0,
        field3: 0b111111,
    }
);

Adding in an explicit #[deku(bits = 8)] to field2 causes this case to work as well, so I believe it is likely related.

I'll fix this tonight, turns out I was over eager with removing some code and need to align the bits properly. Thanks for catching this.

I'd like to add an aligned attribute or optimization in the future..