Re: [PATCH] dt-bindings: can: convert tcan4x5x.txt to DT schema

[Krzysztof Kozlowski <krzk@kernel.org>]

On 05/11/2024 13:59, Sean Nyekjaer wrote:
> On Tue, Nov 05, 2024 at 01:41:26PM +0100, Krzysztof Kozlowski wrote:
> 
> NOW I get it :)
> 
> diff --git a/Documentation/devicetree/bindings/net/can/ti,tcan4x5x.yaml b/Documentation/devicetree/bindings/net/can/ti,tcan4x5x.yaml
> index f1d18a5461e0..4fb5e5e80a03 100644
> --- a/Documentation/devicetree/bindings/net/can/ti,tcan4x5x.yaml
> +++ b/Documentation/devicetree/bindings/net/can/ti,tcan4x5x.yaml
> @@ -169,7 +169,7 @@ examples:
>          #size-cells = <0>;
> 
>          can@0 {
> -            compatible = "ti,tcan4552", "ti,tcan4x5x";
> +            compatible = "ti,tcan4552";
>              reg = <0>;
>              clocks = <&can0_osc>;
>              pinctrl-names = "default";
> 
> Would result in a schema check fail, but the driver will never be probed.
> 
>>

Yeah, but we dont' talk about this case.


>>> Agree that is kinda broken.
>>> If I have time I can try to fix that later.
>>
>> No, the fix is to drop the wildcard alone, as I said in your RFC.
> 
> @Mark, would you be okay with fixing the wildcard in this series?
> We have some out-of-tree dtb's that will need fixing, but I get it would be
> prefered to get this fixed.

Out of tree DTB will not need any fixes per-se. They will work 100%
fine. They will however report dtbs_check warnings, but before there was
no DT schema validation for them, so you replace one warning into
another warning.

You can of course improve out of tree users by dropping all warnings,
but that's kind of optional. The point is that nothing gets worse.

> 
>>
>>>
>>> Please explain one more time for me. Is this a comment on the if
>>> sentence or the broken behavior of the driver?
>>
>> This is just generic comment, nothing to change here because you decided
>> not to fix that wildcard from old binding.
> 
> Thanks for the clarification!
> 
> @Mark, @Krzysztof: What to do from here?

I will give Rb tag for next version fixing commented issues regardless
whether you drop usage of the wildcard-like 4x5x compatible alone. IOW,
I will be fine with pure conversion and keeping 4x5x, even though I
would prefer to improve the binding based on arguments before: there
will be no changes needed for in-kernel users, no out-of-tree users will
be affected, no breakage and using wildcard compatible alone is
discouraged. Sometimes strongly discouraged, depending on the case.

Best regards,
Krzysztof