Re: [PATCH v3 1/3] iio: add struct declarations for iio types

From: Jonathan Cameron <jic23@kernel.org>
To: Michael Riesch <michael.riesch@wolfvision.net>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
	Gerald Loacker <gerald.loacker@wolfvision.net>,
	linux-iio@vger.kernel.org, devicetree@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	Lars-Peter Clausen <lars@metafoo.de>,
	Rob Herring <robh+dt@kernel.org>,
	Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
	Jakob Hauser <jahau@rocketmail.com>,
	Linus Walleij <linus.walleij@linaro.org>,
	Nikita Yushchenko <nikita.yoush@cogentembedded.com>
Subject: Re: [PATCH v3 1/3] iio: add struct declarations for iio types
Date: Sat, 3 Dec 2022 17:11:31 +0000	[thread overview]
Message-ID: <20221203171131.6d096078@jic23-huawei> (raw)
In-Reply-To: <b56cdcf1-459b-63a2-a060-49564e11ce9f@wolfvision.net>

On Mon, 28 Nov 2022 15:26:51 +0100
Michael Riesch <michael.riesch@wolfvision.net> wrote:

> Hi Andy,
> 
> On 11/28/22 15:05, Andy Shevchenko wrote:
> > On Mon, Nov 28, 2022 at 02:48:48PM +0100, Michael Riesch wrote:  
> >> On 11/28/22 14:27, Andy Shevchenko wrote:  
> >>> On Mon, Nov 28, 2022 at 01:18:04PM +0100, Gerald Loacker wrote:  
> >>>> Am 25.11.2022 um 12:01 schrieb Andy Shevchenko:  
> > 
> > ...
> >   
> >>> It's a rule to use _t for typedef:s in the kernel. That's why
> >>> I suggested to leave struct definition and only typedef the same structures
> >>> (existing) to new names (if needed).  
> >>
> >> Andy, excuse our ignorance but we are not sure how this typedef approach
> >> is supposed to look like...
> >>  
> >>>> or  
> >>>  
> >>>> 	typedef iio_val_int_plus_micro iio_val_int_plus_micro_db;  
> >>
> >> ... because
> >>
> >> #include <stdio.h>
> >>
> >> struct iio_val_int_plus_micro {
> >> 	int integer;
> >> 	int micro;
> >> };
> >>
> >> typedef iio_val_int_plus_micro iio_val_int_plus_micro_db;
> >>
> >> int main()
> >> {
> >>   struct iio_val_int_plus_micro a = { .integer = 100, .micro = 10, };
> >>   struct iio_val_int_plus_micro_db b = { .integer = 20, .micro = 10, };
> >>   return 0;
> >> }
> >>
> >> won't compile.  
> > 
> > I see. Thanks for pointing this out.
> > 
> > Then the question is why do we need the two same structures with different
> > names?  
> 
> Most probably we don't need "struct iio_val_int_plus_micro_db" at all
> since IIO_VAL_INT_PLUS_MICRO_DB and IIO_VAL_INT_PLUS_MICRO get the same
> treatment in industrialio-core.c. At least it should not be introduced
> in the scope of this series. In the end this is up to whoever writes the
> first driver using the common data structures and IIO_VAL_INT_PLUS_MICRO_DB.

They get same treatment today because we don't attempt to deal with
IIO_VAL_INT_PLUS_MICRO_DB in conjunction with any of the analog circuit type
front ends yet. Mind you, even though the handling in iio-rescale.c will be
different if anyone ever adds support for the DB form (I shudder at the maths
of combining this with other scale factors), I don't see the possibility meaning
we need a different structure.  

Jonathan

> 
> Best regards,
> Michael
> 