Re: [PATCH v2 24/28] dt-bindings: pinctrl: samsung: convert to dtschema

[Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxxxxx>]

On 17/01/2022 21:26, Olof Johansson wrote:
> On Sun, Jan 16, 2022 at 11:45 PM Krzysztof Kozlowski
> <krzysztof.kozlowski@xxxxxxxxxxxxx> wrote:
>>
>> On 16/01/2022 22:38, Linus Walleij wrote:
>>> On Sun, Jan 16, 2022 at 6:10 PM Krzysztof Kozlowski
>>> <krzysztof.kozlowski@xxxxxxxxxxxxx> wrote:
>>>
>>>> Anyway DTS and dtschema will have to wait for one release, because they
>>>> depend on samsung pinctrl driver change (patch #2).
>>>
>>> What about I put that (and maybe this schema) on an immutable
>>> branch so you can pull the commit into your for-arm-soc branch and
>>> put the DTS changes on top?
>>
>> That would be a solution if not a policy for arm-soc of keeping DTS
>> separate. Arnd and Olof since some time are not happy when DTS branch
>> receives any driver updates.
>>
>> Arnd, Olof,
>> This is a set of dtschema conversion + DTS alignment with new schema:
>> 1. Driver change necessary to accept new DTS (driver depends on node
>> names and this has to change because of dtschema),
>> 2. DTS commits depending on above, which convert node name to new format,
>> 3. Finally dtschema requiring new naming of the GPIO nodes.
>>
>> If I got correctly, the policy of not mixing drivers and DTS requires
>> that #2 above (DTS changes) will wait for one more release. During the
>> time, if dtschema (#3 above) is applied, there will be new warnings
>> about non-compliant DTS.
>>
>> Do you see any chance of merging driver + DTS + dtschema via same tree
>> in same release?
> 
> Our general guidance to separate DTS and driver changes is to avoid
> large entangled changes between the two, and to discourage a developer
> mentality of "the implementation is the binding".
> 
> I think this is a good example of when it makes sense to bring in what
> is a fairly small and clean driver change to deal with this. So the
> right answer here is to stage such a stable branch and merge into both
> arm-soc and the pinctrl subsystem trees as proposed.

Thanks for clarification, I'll go with this approach.

Best regards,
Krzysztof