Re: [PATCH 1/2] ARM: dts: qcom: ipq8064: reorganize node order and sort them

From: Konrad Dybcio
Date: Tue Jul 19 2022 - 06:27:38 EST




On 19.07.2022 10:19, Christian Marangi wrote:
> On Tue, Jul 19, 2022 at 12:22:24PM +0200, Konrad Dybcio wrote:
>>
>>
>> On 19.07.2022 12:16, Krzysztof Kozlowski wrote:
>>> On 19/07/2022 11:59, Dmitry Baryshkov wrote:
>>>> On Tue, 19 Jul 2022 at 12:56, Krzysztof Kozlowski
>>>> <krzysztof.kozlowski@xxxxxxxxxx> wrote:
>>>>>
>>>>> On 18/07/2022 17:38, Christian Marangi wrote:
>>>>>> Reorganize node order and sort them by address.
>>>>>>
>>>>>> Signed-off-by: Christian Marangi <ansuelsmth@xxxxxxxxx>
>>>>>> ---
>>>>>>
>>>>>> This was picked from for-next qcom branch [1]. Reorganize dtsi as requested.
>>>>>>
>>>>>> [1] https://git.kernel.org/pub/scm/linux/kernel/git/qcom/linux.git/?h=for-next
>>>>>
>>>>> If this is picked by qcom branch, no need to resend it.
>>>>>
>>>>> I don't see value in such reshuffle. Reviewing is not possible and you
>>>>> did not mention tests (results should be equal).
>>>>
>>>> The value is usual for all the cleanups: make it follow the
>>>> established practice.
>>>
>>> Are you sure this is established practice?
>> Yes.
>>
>> New DTSI files (see SC8280XP,
>>> sm8450 although sc7280 looked ordered) do not always follow it, so why
>>> imposing it for existing code?
>> Perhaps it slipped through review.. Partially my bad.
>>
>>
>> Such reshuffle can cause conflicts thus
>>> stops parallel development. Review is close to impossible...
>> Almost any addition or removal also causes conflicts, because git is
>> not as smart as we would like it to be. If the commit is structured
>> properly (i.e. it *only* changes the order and nothing else),
>> decompiling the dtbs before and after applying it and using a tool
>> like meld that can find similar chunks of text at different offsets
>> review is definitely possible, though not very pleasant (you can't
>> just diff them, as order is preserved & phandles change due to that)
>> as you have to look at it manually and can't tell much by just taking
>> a look at the email.
>>
>
> Can you give me an example of such tool? So I can put these data in the
> commit description. I have to rebase this anyway as more changes got
> merged so it might be a good idea to add more info about how this won't
> make actualy changes.
dtc -I dtb filename.dtb -O dts > outfile.dts

meld old.dts new.dts

Then you have to look at them side by side.

Alternatively, python has libfdt and one could make a tool for that
particular usecase..

Konrad
>
>> Konrad
>>>
>>> Best regards,
>>> Krzysztof
>