On 8/25/25 5:17 AM, Alexandre Courbot wrote:
>> + /// Convert a raw 24-bit class code value to a `Class`.
>> + impl From<u32> for Class {
>> + fn from(value: u32) -> Self {
>> + match value {
>> + $(x if x == Self::$variant.0 => Self::$variant,)+
>> + _ => Self::UNKNOWN,
>> + }
>> + }
>
> Should we normalize `value` to 24 bits (i.e. call `to_24bit_class`)
> before doing the match? The constants we compare against are all
> normalized, but if we pass a 16-bit class to this method the result will
> be `UNKNOWN`, unless I missed something.
>
> Being able to store a class as either a 16-bit or 24-bit representation
> in the same type also opens the door to bugs, which we can avoid if we
> always normalize to 24-bit and make the class/subclass representation
> accessible through a convenience method only.
It's constrained by the PCI ID patterns, and should only be used in
order to get values from the C bindings, actually.
>
>> + }
>> + };
>> +}
>> +
>> +/// Once constructed, a `Class` contains a valid PCI Class code.
>> +impl Class {
>> + /// Create a new Class from a raw 24-bit class code.
>> + pub fn new(class_code: u32) -> Self {
>> + Self::from(class_code)
>> + }
>
> Do we need a `new` method when the `From` implementation does exactly
> the same thing and has the same signature?
>
Looks like I'll remove the From, and only provide new, based on the other
thread.
thanks,
--
John Hubbard
