Re: [PATCH v2 2/2] iio: pressure: ms5637: Add variant specific temperature compensation

[Andy Shevchenko]

On Wed, Aug 05, 2026 at 01:06:51PM -0400, Louis Adamian wrote:
> On Thu, 11 Jun 2026 22:32:02 +0300
> Andy Shevchenko <andriy.shevchenko@xxxxxxxxx> wrote:
> > On Thu, Jun 11, 2026 at 03:10:26PM -0400, Louis Adamian wrote:
> > > On Wed, 10 Jun 2026 21:37:58 +0300
> > > Andy Shevchenko <andriy.shevchenko@xxxxxxxxx> wrote:  
> > > > On Tue, Jun 09, 2026 at 10:04:58PM -0400, Louis Adamian wrote:  

...

> > > > >  struct ms_tp_dev {
> > > > >  	struct i2c_client *client;    
> > > >   
> > > > >  	const struct ms_tp_hw_data *hw;
> > > > >  	u16 prom[MS_SENSORS_TP_PROM_WORDS_NB];
> > > > >  	u8 res_index;
> > > > > +	const struct ms_tp_comp_consts *comp_consts;    
> > > > 
> > > > Please, check with `pahole` if this is the best layout.  
> > > 
> > > I will correct the packing on this per pahole's finding.  
> > > > >  struct ms_tp_data {
> > > > >  	const char *name;
> > > > >  	const struct ms_tp_hw_data *hw;
> > > > > +	const struct ms_tp_comp_consts *comp_consts;
> > > > >  };    
> > > > 
> > > > Can this be simply embedded into ms_tp_dev (and copied there if
> > > > required)?  
> > > 
> > > Copying in ms_tp_comp_consts provides no benefit over
> > > referencing the pointer; the values are per-descriptor, not
> > > per-instance and never modified. I'll keep the pointer (same as
> > > ms_tp_hw_data).  
> > 
> > Okay, let me ask differently then: Can we use the same data type in
> > ms_tp_dev instead of adding basically the same fields as in
> > ms_tp_data?
> > 
> Sorry for the very slow reply. Yes, for v3 ms_tp_dev will hold a single
> const ms_tp_data *data instead of its own hw and comp_consts. This
> drops the duplicated hw that predates this series. 
> 
> One thing worth noting before I do that: ms_tp_data is currently
> private to ms5637.c and will need to be moved to ms_sensors_i2c.h for
> ms_tp_dev to point to it. I don't think this is a problem; it matches
> how ms_tp_hw_data is defined now, but let me know if you would like to
> avoid it.

The pointer may be opaque, in that case it's not needed to be exposed.

-- 
With Best Regards,
Andy Shevchenko