Re: [PATCH v7 02/15] dt-bindings: media: qcom,x1e80100-camss: Convert from inline PHY definitions to PHY handles

[Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>]

On 13/07/2025 11:14, Bryan O'Donoghue wrote:
> On 13/07/2025 09:20, Krzysztof Kozlowski wrote:
>> On 13/07/2025 10:18, Krzysztof Kozlowski wrote:
>>> On 11/07/2025 14:57, Bryan O'Donoghue wrote:
>>>> We currently do not have an upstream user of the x1e CAMSS schema which
>>>
>>> On first glance there is, in Linus tree:
>>>
>>> git grep qcom,x1e80100-camss
>>> drivers/media/platform/qcom/camss/camss.c
>>>
>>> If this wasn't released mention it.
>> ... and then this should be marked as fixes and picked up fast, because
>> you have only like 2 weeks to fix it.
>>
>> Best regards,
>> Krzysztof
> 
> I thought schema changes were acceptable so long as we haven't applied 
> dts, which we haven't done yet.


Accepted DTS is just one story, but following your argumentation that
docs do not define ABI break, then accepted DTS also does not matter,
because it is always in the kernel sources synced with the ABI.
Following your argument about "accepted DTS", what is different between:
1. accepted DTS, then changed DT binding and changed DTS,
2. not accepted DTS and changed DT binding?

Why can't you accept DTS and then change it?

Lack of in-kernel DTS is a good argument in your case, but you must
mention ALL OTHER USERS:
1. All drivers in Linux
2. All other upstream projects, BSD, U-boot, everywhere upstream
3. ... all possible other users of the ABI, so out of tree DTS and out
of tree kernel folks. This one is close to impossible to prove...
Luckily we assume this point does not apply here at all. No one out of
upstream trees uses these new bindings.

Best regards,
Krzysztof