On 22.05.2025 15:48, Krzysztof Kozlowski wrote: > On 22/05/2025 14:46, Krzysztof Kozlowski wrote: >>>>> >>>>> + - if: >>>>> + properties: >>>>> + compatible: >>>>> + contains: >>>>> + const: renesas,usb2-phy-r9a08g045 >>>>> + then: >>>>> + required: >>>>> + - renesas,sysc-signals >>>> >>>> That's ABI break. >>> >>> There is no in kernel device tree users of "renesas,usb2-phy-r9a08g045" >>> compatible. It is introduced in patch 11/12 from this series. With this do >>> you still consider it ABI break? >> >> Then this patch cannot be split from binding introducing the user. Don't >> add unused/undocumented compatibles. >> > Or you meant DTS? Yes, I meant in tree DTS. > I asked about ABI which is not about in-kernel users. > You can always change in-kernel users, so what would be any point of a > binding and its ABI? > > Best regards, > Krzysztof
