Re: [PATCH v2 09/13] rust: clean up define_property macro

[Kevin Wolf <kwolf@redhat.com>]

Am 21.10.2024 um 18:35 hat Paolo Bonzini geschrieben:
> Use the "struct update" syntax to initialize most of the fields to zero,
> and simplify the handmade type-checking of $name.
> 
> Reviewed-by: Junjie Mao <junjie.mao@hotmail.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  rust/qemu-api/src/device_class.rs | 29 ++++++-----------------------
>  1 file changed, 6 insertions(+), 23 deletions(-)
> 
> diff --git a/rust/qemu-api/src/device_class.rs b/rust/qemu-api/src/device_class.rs
> index 2219b9f73d0..5aba426d243 100644
> --- a/rust/qemu-api/src/device_class.rs
> +++ b/rust/qemu-api/src/device_class.rs
> @@ -29,44 +29,27 @@ macro_rules! device_class_init {
>  macro_rules! define_property {
>      ($name:expr, $state:ty, $field:expr, $prop:expr, $type:expr, default = $defval:expr$(,)*) => {
>          $crate::bindings::Property {
> -            name: {
> -                #[used]
> -                static _TEMP: &::core::ffi::CStr = $name;
> -                _TEMP.as_ptr()
> -            },
> +            // use associated function syntax for type checking
> +            name: ::core::ffi::CStr::as_ptr($name),

I like this part.

>              info: $prop,
>              offset: ::core::mem::offset_of!($state, $field)
>                  .try_into()
>                  .expect("Could not fit offset value to type"),
> -            bitnr: 0,
> -            bitmask: 0,
>              set_default: true,
>              defval: $crate::bindings::Property__bindgen_ty_1 { u: $defval.into() },
> -            arrayoffset: 0,
> -            arrayinfo: ::core::ptr::null(),
> -            arrayfieldsize: 0,
> -            link_type: ::core::ptr::null(),
> +            ..unsafe { ::core::mem::MaybeUninit::<$crate::bindings::Property>::zeroed().assume_init() }

But is it really worth introducing unsafe code just for a more compact
notation? If the compiler doesn't actually understand the pattern, it
might even be less efficient than what we had (i.e. if it really creates
the zeroed object and copies stuff over).

Kevin