Re: [PATCH] dt-bindings: clock: si5351: Make compatible string required property

[Krzysztof Kozlowski <krzk@kernel.org>]

On 03/10/2024 07:57, Michal Simek wrote:
> 
> 
> On 10/2/24 23:41, Rob Herring wrote:
>> On Wed, Oct 02, 2024 at 02:17:22PM +0200, Krzysztof Kozlowski wrote:
>>> On 02/10/2024 12:31, Michal Simek wrote:
>>>>
>>>>
>>>> On 10/2/24 10:24, Michal Simek wrote:
>>>>>
>>>>>
>>>>> On 10/2/24 10:19, Krzysztof Kozlowski wrote:
>>>>>> On 02/10/2024 09:51, Michal Simek wrote:
>>>>>>> Compatible property is likely also required property.
>>>>>>>
>>>>>>> Signed-off-by: Michal Simek <michal.simek@amd.com>
>>>>>>> ---
>>>>>>
>>>>>> That's a convention but not necessary, a no-op.
>>>>>
>>>>> But how do you identify device then?
>>>>> Or are you saying that device description is valid even if there is no
>>>>> compatible string?
>>>>
>>>> One more thing
>>>> commit 524dfbc4e9fc ("dt-bindings: clock: si5351: convert to yaml") is showing
>>>> that compatible property was required in txt file.
>>>>
>>>> -Required properties:
>>>> -- compatible: shall be one of the following:
>>>> -       "silabs,si5351a" - Si5351a, QFN20 package
>>>> -       "silabs,si5351a-msop" - Si5351a, MSOP10 package
>>>>
>>>> I can update commit message to describe it too.
>>>
>>> Devices do not work without compatible, so this is obvious... and like
>>> said - it is already required, so the change is redundant. Does not
>>> harm, though.
>>
>> To put it another way, by the time the schema is applied, we already
>> know that compatible is present because that is *how* the schema gets
>> applied in the first place.
> 
> I get that argument but then based on this we should remove all records about 
> compatible string as required property.

We could... but we have a style of keeping it. What is the harm in
having it in 99% of bindings and missing in a few?

Best regards,
Krzysztof