Re: [PATCH V3 3/5] iio: accel: sca3300: modified to support multi chips

[Jonathan Cameron <jic23@kernel.org>]

On Thu, 5 May 2022 20:01:22 +0200
Andy Shevchenko <andy.shevchenko@gmail.com> wrote:

> On Thu, May 5, 2022 at 4:12 PM LI Qingwu
> <qing-wu.li@leica-geosystems.com.cn> wrote:
> > > -----Original Message-----
> > > From: Andy Shevchenko <andy.shevchenko@gmail.com>
> > > Sent: Wednesday, May 4, 2022 10:39 PM
> > > On Wed, May 4, 2022 at 4:35 PM LI Qingwu
> > > <qing-wu.li@leica-geosystems.com.cn> wrote:  
> > > > > From: Andy Shevchenko <andy.shevchenko@gmail.com>
> > > > > Sent: Wednesday, May 4, 2022 10:20 PM On Wed, May 4, 2022 at 3:36 PM
> > > > > LI Qingwu <Qing-wu.Li@leica-geosystems.com.cn> wrote:  
> 
> ...
> 
> > > > > > +struct sca3300_chip_info {
> > > > > > +       const struct iio_chan_spec *channels;
> > > > > > +       const int (*accel_scale_table)[2];
> > > > > > +       const int *accel_scale_modes_map;
> > > > > > +       const unsigned long *scan_masks;
> > > > > > +       const int *avail_modes_table;
> > > > > > +       const int *freq_modes_map;
> > > > > > +       const int *freq_table;
> > > > > > +       const u8 num_accel_scales;
> > > > > > +       const u8 num_avail_modes;
> > > > > > +       const u8 num_channels;
> > > > > > +       const u8 num_freqs;
> > > > > > +       const u8 chip_id;  
> > > > >
> > > > > Why do you have const qualifier on all members?  The last one is
> > > > > understandable, but the rest, esp. pointers should be justified.  
> > > > Because I thought it was static and has fix value for each chip, unacceptable  
> > > for you?
> > >
> > > But why const qualifier? What is the point of it for example for u8 members if
> > > the entire object is qualified as const below in the same patch?
> > >
> > > On top of that, please explain what in your opinion the "const ...
> > > *foo" gives us, and what we will lose if we remove the "const" part out of them.  
> >
> > Ah, you are right, those const are unnecessary for nonpointer members.
> > for the pointers, the contexts that the pointer points to are still writable.
> > what about if I remove all the const from nonpointer and keep it for the pointers?
> > Like：
> > const struct iio_chan_spec *channels;
> > const int (*accel_scale_table)[2];
> > const int (*incli_scale_table)[2];
> > const int *accel_scale_modes_map;
> > const int *incli_scale_modes_map;
> > const unsigned long *scan_masks;
> > const int *avail_modes_table;
> > const int *freq_modes_map;
> > const int *freq_table;
> > const char *name;
> > u8 num_accel_scales;
> > u8 num_incli_scales;
> > u8 num_avail_modes;
> > u8 num_channels;
> > u8 num_freqs;
> > u8 chip_id;
> > bool angle;  
> 
> It's better, but you still need to justify the rest with explanation
> in the commit message.
> And I leave this to maintainers to say if the const:s are needed or not.
Where they are being set to point to
static const int array[] 
then to my mind it makes sense to have them as const as makes it nice and obvious
what is going on.

Jonathan

> 
> > > > > > +       const char *name;
> > > > > > +};  
>