On Wed, Aug 28, 2024 at 02:52:25PM +0200, Krzysztof Kozlowski wrote: > On 28/08/2024 14:45, Laurent Pinchart wrote: > > On Sun, Aug 18, 2024 at 08:48:54PM +0200, Krzysztof Kozlowski wrote: > >> On 18/08/2024 19:51, Laurent Pinchart wrote: > >>> On Sun, Aug 18, 2024 at 07:44:22PM +0200, Krzysztof Kozlowski wrote: > >>>> On 18/08/2024 19:41, Laurent Pinchart wrote: > >>>>> On Sun, Aug 18, 2024 at 07:30:02PM +0200, Krzysztof Kozlowski wrote: > >>>>>> Each variable-length property like interrupts or resets must have fixed > >>>>>> constraints on number of items for given variant in binding. The > >>>>>> clauses in "if:then:" block should define both limits: upper and lower. > >>>>> > >>>>> I thought that, when only one of minItems or maxItems was specified, the > >>>>> other automatically defaulted to the same value. I'm pretty sure I > >>>>> recall Rob asking me to drop one of the two in some bindings. Has the > >>>>> rule changes ? Is it documented somewhere ? > >>>> > >>>> New dtschema changed it and, even if previous behavior is restored, the > >>>> size in if:then: always had to be constrained. You could have skipped > >>>> one side of limit if it was equal to outer/top-level limit, e.g: > >>>> > >>>> properties: > >>>> clocks: > >>>> minItems: 1 > >>>> maxItems: 2 > >>>> > >>>> > >>>> if:then:properties: > >>>> clocks: > >>>> minItems: 2 > >>> > >>> Where can I find a description of the behaviour of the new dtschema > >>> (hopefully with some documentation) ? > >> > >> No clue, but I feel there is some core concept missing. Your earlier > >> statement: > >> "I thought that, when only one of minItems or maxItems was specified, the" > >> > >> was never logically correct for the "if:then", except for the case I > >> mentioned above. That's why all schema used as examples had it explicit: > >> > >> My talk from 2022, page 30: > >> https://static.sched.com/hosted_files/osseu2022/bd/How%20to%20Get%20Your%20DT%20Schema%20Bindings%20Accepted%20in%20Less%20than%2010%20Iterations%20-%20Krzysztof%20Kozlowski%2C%20Linaro.pdf?_gl=1*kmzqmt*_gcl_au*MTU2MzQ1MjY0Mi4xNzIxNzE0NDc1 > >> all constraints defined,. > >> > >> My talk from 2023, page 34: > >> https://static.sched.com/hosted_files/eoss2023/a8/How%20to%20Get%20Your%20DT%20Schema%20Bindings%20Accepted%20in%20Less%20than%2010%20Iterations%20-%20Krzysztof%20Kozlowski%2C%20Linaro%20-%20ELCE%202023.pdf?_gl=1*1jgx6d3*_gcl_au*MTU2MzQ1MjY0Mi4xNzIxNzE0NDc1 > >> > >> Recently, I started using other example as "useful reference": > >> https://elixir.bootlin.com/linux/v6.8/source/Documentation/devicetree/bindings/ufs/qcom,ufs.yaml#L132 > >> > >> That's nothing. All three above reference examples I keep giving are > >> already there and repeated in emails all the time. > >> > >> So aren't you confusing the entire "skip one limit" for top-level > >> properties? This patch is not about it all and dtschema did not change. > > > > There must have been a misunderstanding indeed, I interpreted "New > > dtschema changed it" as meaning there were now new rules. Is that > > incorrect ? > > For the binding with a property defined only in top-level properties: no > changes, no new rules. > > For the binding with top-level and if:then:else: dtschema since few > months changed interpretation. OK, that's what I didn't understand correctly. > > If you don't mind clarifying, what is the current recommendation to > > indicate that a property has a fixed number of items ? Which of the > > following three options is preferred ? > > Answer below assumes we have clocks defined in top-level properties and > there is no if:then:else customizing it. > > > properties: > > clocks: > > minItems: 2 > > That's wrong, because items are unconstrained. > > > properties: > > clocks: > > maxItems: 2 > > This one is preferred. > > > properties: > > clocks: > > minItems: 2 > > maxItems: 2 > > This one is correct, but less preferred. Thank you, that is clear now. -- Regards, Laurent Pinchart
