Re: [PATCH v3 1/4] ARM: dts: qcom: msm8960: expressatt: Sort node references and includes

[Krzysztof Kozlowski]

On 11/04/2026 06:01, Rudraksha Gupta wrote:
> On 4/7/26 14:46, David Heidelberg wrote:
>> On 07/04/2026 23:04, Krzysztof Kozlowski wrote:
>>> On 07/04/2026 22:39, Rudraksha Gupta wrote:
>>>> On 4/7/26 12:59, Krzysztof Kozlowski wrote:
>>>>> On 01/04/2026 22:32, Rudraksha Gupta via B4 Relay wrote:
>>>>>> From: Rudraksha Gupta <guptarud@xxxxxxxxx>
>>>>>>
>>>>>> Reorganize the DTS file for consistency with other msm8960 board 
>>>>>> files.
>>>>>>
>>>>>> Assisted-by: Claude:claude-opus-4.6
>>>>>> Signed-off-by: Rudraksha Gupta <guptarud@xxxxxxxxx>
>>>>>> ---
>>>>>>    .../dts/qcom/qcom-msm8960-samsung-expressatt.dts   | 408 
>>>>>> +++++++++++----------
>>>>>>    1 file changed, 207 insertions(+), 201 deletions(-)
>>>>>>
>>>>> Sorry, but no. We are not taking Claude as one determining coding 
>>>>> style.
>>>>> Are we going to do the work again the moment we come with proper tool?
>>>>
>>>> There is no tool currently to auto format DTS, and doesn't seem to be
>>>> coming for a while:
>>>>
>>>> https://www.youtube.com/watch?v=cvoIbTL_ZQA
>>>>
>>>>
>>>> Claude didn't determine the coding style. I did based on sony-huashan,
>>>> which is already upstream:
>>>>
>>>> https://github.com/torvalds/linux/blob/master/arch/arm/boot/dts/qcom/qcom-msm8960-sony-huashan.dts 
>>>>
>>>>
>>>>
>>>> I just used Claude to do the manual work for me. In v2, I made sure the
>>>> diff before and after the change was nill. v3 included additional
>>>> changes requested by Konrad and some comments that I remembered during
>>>> prior attempts mainlining patch series for this device.
>>>
>>> IMO, it is just too risky to let Claude reorganize the nodes, but I
>>> assume reviewers of your code did run dtx_diff.
>>
>> I think it depends on the prompt. Since I’m performing many of the 
>> same tasks repeatedly across multiple sdm845 devices, asking an LLM to 
>> do node-by-node reorganization can be reasonably reviewable (at least 
>> when reviewing incremental progress, not just the final diff).
>>
>> I would prefer to do more of the sorting myself, but I find it quite 
>> tedious. The diff tool struggles when similar or identical lines 
>> appear in different nodes, which often results in a messy final diff 
>> (I noticed this in Sajattack’s sdm845 LG patchset).
>>
>> This leads me to an idea:
>>
>> For these sorting cleanups, perhaps we could introduce a “squash mode”?
>>
>> Contributors could submit commits per node, making the reorganization 
>> clearly visible (and ensuring nothing is accidentally lost), and then 
>> the maintainer could squash them into a single commit to avoid 
>> cluttering the git log.
>>
>> What do you think?
> 
> Easiest solution would be to get Claude to make a DTS auto formatter. I 
> estimate it would likely take a couple iterations to get a functional 
> prototype and max a week to get it into a mergable state, if the style 
> is agreed beforehand. Simply provide DTS'es that follow the pattern you 
> like to Claude, then tell Claude that you want to make a Python script 

Yeah, and who wants to review the Claude code?

I already have a formatter to review in the pipeline...

> to auto format DTS files and make functions for each different common 
> style pattern identified in the DTS'es. I assume it would give a good 
> enough base to work off of. The most painful part will be determining 
> what the correct style for all DTS'es as I'm sure others will have 
> opinions on that.
> 

The biggest problem is that no way maintainers will run untrusted 3rd
party code, co-contributed by random people with Claude.


Best regards,
Krzysztof