Re: [PATCH v2 16/16] rust: io: register: unify handling of register with/without bitfields

On Fri Aug 14, 2026 at 2:56 PM BST, Alexandre Courbot wrote:
> On Thu Aug 6, 2026 at 1:35 AM JST, Gary Guo wrote:
> <...>
>> @@ -577,6 +547,9 @@ macro_rules! register {
>>      (base: $reg_base:ty;) => {};
>>  
>>      // Creates a register at a fixed offset of the MMIO space with provided type.
>> +    //
>> +    // This handles all of the fixed offset `@ offset`, alias of register `=> alias` and alias of
>> +    // register array element `=> alias[idx]` cases.
>>      (
>>          base: $reg_base:ty;
>>          // `$ty` cannot be `:ty` due to follow-set restrictions.
>> @@ -593,10 +566,29 @@ macro_rules! register {
>>          $crate::register!(base: $reg_base; $($rest)*);
>>      };
>>  
>> +    // `#[unique]` indicates that this is the only register of this type in this given register.
>
> Sentence doesn't parse (in this given register?).

Oops, I meant "in the given region" (i.e. base type).

>
>> +    // Thus generate a `FixedIoLoc` impl for it as well.
>> +    (
>> +        base: $reg_base:ty;
>> +        $(#[$attr:meta])* $vis:vis $name:ident: #[unique] $ty: ident $(:: $path_frag:ident)*
>> +            $(@ $offset:literal)?
>> +            $(=> $alias:path $([$alias_idx:expr])? )?;
>> +        $($rest:tt)*
>
> This rule looks like it should be private, or am I mistaken?

The idea is that if you create bitfield yourself you can still use `#[unique]`
to bring back the `FixedIoLoc` impl so you're not losing any expressiveness
compared to let `register!()` macro generating it.

Best,
Gary