Hi Bjorn,
On 11/11/2023 5:17 AM, Bjorn Andersson wrote:
On Mon, Oct 23, 2023 at 11:48:13AM +0530, Jishnu Prakash wrote:
Hi Krzysztof,
On 7/9/2023 10:56 PM, Krzysztof Kozlowski wrote:
On 08/07/2023 09:28, Jishnu Prakash wrote:
Update ADC dt-bindings file paths in QCOM devicetree files to
match the dt-bindings change moving the files from 'iio' to
'iio/adc' folder.
Signed-off-by: Jishnu Prakash<quic_jprakash@xxxxxxxxxxx>
---
Your order of patches is incorrect. DTS cannot be before bindings,
because this indicates you broke ABI. Please keep entire patchset
bisectable and split DTS to separate patchset (linking the driver and
bindings).
Best regards,
Krzysztof
OK....so you mean I can move the dtbinding files and update documentation
and driver files for this, all in one patch and then make a second patch for
updating devicetree files for the file paths?
I'll make this change in the next patchset if it looks fine, although the
two patches would need to be picked together to avoid kernel breakage. I
have asked Jonathan about this too in my reply for the previous patchset 9.
1) We need to be able to use git bisect to identify regressions, and as
such the kernel need to be buildable and functional after each applied
patch.
2) DeviceTree source (dts) changes enters the kernel through different
maintainer paths than the DeviceTree binding and driver changes, and
therefore at different times. In some cases the .dtb file is also
delivered separately from the kernel, which means people might be
running an old .dtb for a considerable amount of time after your change.
As such your driver changes needs to be compatible with both the new and
the old DeviceTree binding.
Thanks for your comments. For now, Jonathan and Dmitry have asked me to
move the binding files and fix all errors in one commit for their
review, so I'll be doing as they requested. Based on their review, I'm
assuming they will also suggest what other changes are needed for
compatibility with older devicetrees as you mentioned.
Regards,
Bjorn
