On 30/09/2022 23:14, Rob Herring wrote:
>>>>> + dc@54200000 {
>>>>> + status = "okay";
>>>>
>>>> You should override by labels, not by full path.
>>>
>>> Why exactly is that? I've always stayed away from that (and asked others
>>> not to do so, at least on Tegra) because I find it impossible to parse
>>> for my human brain. Replicating the original full hierarchy makes it
>>> much more obvious to me where the changes are happening than the
>>> spaghetti-like mess that you get from overriding by label reference.
>>
>> Sure, it's entirely up to you. I forgot your preference.
>>
>> But it is a really nice way to have duplicated nodes and mistakes (which
>> happen from time to time).
>
> We could have a schema or dtc check for that. We already warn for
> duplicate unit-addresses which would catch some typos. Checking for a
> node with only 'status' would probably work when that's the only
> addition. Maybe status without a compatible would be better? We also
> check for nodes without a specific schema, but child nodes in schemas
> aren't handled.
Usually these are overrides of few properties and status=okay, so
looking for nodes without a compatible would work. Except for all the
cases where we do not have schema yet...
Best regards,
Krzysztof
