Re: [PATCH v2 2/2] iio: adc: ti-ads112c14: add gpio support

[Andy Shevchenko <andriy.shevchenko@intel.com>]

+Cc: Rob. Here is a discussion how properly to check if the specific node has
gpio-controller flag set. I tend to say that in the case of a simple branch
(in opposite to if-else) we should use _property_present() and not
_property_read_bool(). Can you be an arbiter here and tell the correct way
of doing this?

On Thu, Oct 15, 2026 at 09:04:58AM -0500, David Lechner wrote:
> On 10/15/26 8:37 AM, Andy Shevchenko wrote:
> > On Thu, Oct 15, 2026 at 08:01:14AM -0500, David Lechner wrote:
> >> On 10/13/26 9:05 AM, Andy Shevchenko wrote:
> >>> On Sat, Oct 10, 2026 at 08:46:30AM -0500, David Lechner wrote:
> >>>> On 10/10/26 4:00 AM, Bartosz Golaszewski wrote:
> >>>>> On Sat, 10 Oct 2026 01:25:52 +0200, "David Lechner (TI)"
> >>>>> <dlechner@baylibre.com> said:

...

> >>>>>> +	if (device_property_read_bool(dev, "gpio-controller")) {
> >>>>>
> >>>>> I think the right one to use is device_property_present().
> >>>>
> >>>> I thought this was the one case where device_property_read_bool() is
> >>>> actually correct because gpio-controller property is a
> >>>> /schemas/types.yaml#/definitions/flag.
> >>>>
> >>>> Otherwise, we should just remove device_property_read_bool() from the tree
> >>>> because there would never be a reason to use it.
> >>>>
> >>>> My understanding is we should use device_property_present() for non-boolean
> >>>> properties when we want to know if it is present or not before reading the
> >>>> value of it.
> >>>
> >>> That's not true. The _property_present() checks any type of property, while
> >>> _property_read_bool() returns an actual value of the flag, while they might
> >>> overlap in the behaviour, they are semantically completely different.
> >>
> >> That is exactly what I said.
> > 
> > But you made a code different. As I read it it checks for "is the
> > gpio-controller property present?" For the question that you may
> > think off "is the gpio-controller flag is true?" I consider rather
> > if-else branch as we *either* option.
> > 
> > So, when we do
> > 
> > 	if (...check_flag...)
> > 		do_smth();
> > 
> > I assume we check the presence of the property (flag). If we do
> > 
> > 	x = ...read_flag...;
> > 
> > OR
> > 
> > 	if (...read_flag... == true)
> > 		do_smth();
> > 	else
> > 		do_smth_else();
> > 
> > I consider the appropriate is reading the property.
> > 
> 
> Are you saying flag != boolean DT property type?

No, I'm saying that semantics is defined also in the context of use. In your
case you check for the property presence and do not use it as a value.

> type: boolean
> 
> vs.
> 
> $ref: /schemas/types.yaml#/definitions/flag
> 
> I know we have both names in dt-bindings but I don't know the history on it
> and if we consider them semantically different.

-- 
With Best Regards,
Andy Shevchenko