Re: [PATCH v6 2/2] iio: amplifiers: adl8113: add driver support

[Jonathan Cameron <jic23@kernel.org>]

On Sat, 6 Dec 2025 22:48:19 +0200
Andy Shevchenko <andriy.shevchenko@intel.com> wrote:

> On Sat, Dec 06, 2025 at 07:03:32PM +0000, Jonathan Cameron wrote:
> > On Fri, 5 Dec 2025 16:40:41 +0200
> > Antoniu Miclaus <antoniu.miclaus@analog.com> wrote:  
> 
> ...
> 
> > > +	/* Always include internal amplifier (14dB) */
> > > +	st->gain_configs[i].path = ADL8113_INTERNAL_AMP;
> > > +	st->gain_configs[i].gain_db = 14;  
> > 
> > Could do this as something like:  
> 
> I remember a discussion where it was against this approach due to compiler
> warnings or so. However, IIRC, there was slightly different pattern, i.e.
> 
> 	foo[i].bar = ...
> 	foo[i++].baz = ...
> 
> That said, I have no objection to your proposal, but we need to use it with
> a good compile test coverage (clang with `make W=1` for a starter.

That's my standard 1st build choice now because it seems to be fussiest ;)
I haven't tried this one however.

Googling came up with a link saying GCC gives warnings on cases where
not all structure elements are set.  I think they always should be here
though which may make this fine (and act as a correct warning if they aren't).
That only works if the two uninitialized structure elements are dropped
obviously!

> 
> OTOH, the original code is robust enough...
> 
> > 	st->gain_configs[i++] = (struct adl8113_gain_config) {
> > 		.path = ADL8113_INTERNAL_AMP,
> > 		.gain_db = 14,
> > 	};  
> 
> ...and doesn't require a good understanding of differences between designated
> initialisers and compound literals.

Fair enough.  Original is better than the i++ in only the second in my view.

Jonathan

> 
> > 	st->gain_configs[i++] = (struct adl8113_gain_config) {
> > 		.path = ADL8113_INTERNAL_BYPASS,
> > 		.gain_db = -2,
> > 	};  
> 
>