Re: [PATCH v5 4/4] iio: ABI: document accel and roc event attributes

[Jonathan Cameron <jic23@kernel.org>]

On Sat, 30 Aug 2025 22:58:17 +0300
Andy Shevchenko <andy.shevchenko@gmail.com> wrote:

> On Sat, Aug 30, 2025 at 8:09 PM Jonathan Cameron <jic23@kernel.org> wrote:
> > On Sat, 30 Aug 2025 18:05:34 +0100
> > Jonathan Cameron <jic23@kernel.org> wrote:  
> > > On Sat, 30 Aug 2025 15:49:50 +0300
> > > Andy Shevchenko <andy.shevchenko@gmail.com> wrote:  
> > > > On Sat, Aug 30, 2025 at 2:58 PM Gustavo Silva <gustavograzs@gmail.com> wrote:  
> > > > >
> > > > > Add accelerometer and rate of change event-related sysfs attributes
> > > > > exposed by the bmi270 driver.  
> > > >
> > > > Seems to me like the absent attributes that are already in the kernel,
> > > > should be added in the separate patch.  
> > > Agreed that would be ideal.  
> >
> > Actually what did you mean by absent attributes?  
> 
> Absent in the documentation, but present in the code. That's what this
> patch adds mainly, no?
> 
> > This is documenting ABI that is part of the general 'scope' of the full
> > IIO ABI but which hasn't turned up before in this particular combination
> > (or possibly we missed updating docs when it did!)
> >
> > Whether it is worth separating out any we know are in another driver is
> > an open question, but Gustavo hasn't called out any as being like that.
> > It's possible that these are all surfacing for the first time in this driver.  
> 
> Hmm... But if the code handles the attribute which is not documented,
> that needs to be fixed.
> 
The core code handles a massive number of combinations that make no sense
because the attributes are constructed from 5+ parameters, each with many
values. So what matters here is whether a combination is in use, not whether it
could be instantiate by a driver using standard IIO structures.

Sometimes we miss things in reviews so gaps occur. Those absolutely should
be fixed, but we shouldn't documents stuff on basis it 'might' be ABI in
future.

Jonathan