Re: [PATCH v3 08/13] gpu: nova-core: Add bindings and accessors for GspSystemInfo

["Alexandre Courbot" <acourbot@nvidia.com>]

On Fri Oct 3, 2025 at 8:38 AM JST, Alistair Popple wrote:
> On 2025-10-02 at 23:49 +1000, Alexandre Courbot <acourbot@nvidia.com> wrote...
>> Hi Alistair, (+Benno as this concerns the `init!` macros)
>> 
>> On Tue Sep 30, 2025 at 10:16 PM JST, Alistair Popple wrote:
>> > Adds bindings and an in-place initialiser for the GspSystemInfo struct.
>> >
>> > Signed-off-by: Alistair Popple <apopple@nvidia.com>
>> >
>> > ---
>> >
>> > It would be good to move to using the `init!` macros at some point, but
>> > I couldn't figure out how to make that work to initialise an enum rather
>> > than a struct as is required for the transparent representation.
>> 
>> Indeed we have to jump through a few (minor) hoops.
>> 
>> First the `init!` macros do not seem to support tuple structs. They
>> match a `{` after the type name, which is not present in
>> `GspSystemInfo`. By turning it into a regular struct with a single
>> field, we can overcome this, and it doesn't affect the layout the
>> `#[repr(transparent)]` can still be used.
>
> I was thinking we should fix the `init!` macro to support tuple structs. Is
> there some fundamental reason `init!` couldn't be modified to support tuple
> structs? It seems like it would be nicer to fix that limitation rather than work
> around it here.

I took a look at it and quickly got lost in the macros's internals. :)
Let's see what Benno has to say about this. In the meantime, using a
single member is just as valid a constructs as a tuple struct for us.