Re: [PATCH v4 2/3] pinctrl: tps6594: Add driver for TPS6594 pinctrl and GPIOs

["Esteban Blanc" <eblanc@xxxxxxxxxxxx>]

On Wed May 17, 2023 at 5:04 PM CEST, Andy Shevchenko wrote:
> On Wed, May 17, 2023 at 5:43 PM Esteban Blanc <eblanc@xxxxxxxxxxxx> wrote:
> > On Wed May 17, 2023 at 3:51 PM CEST, Andy Shevchenko wrote:
> > > On Wed, May 17, 2023 at 12:58 PM Esteban Blanc <eblanc@xxxxxxxxxxxx> wrote:
> > > > On Tue May 16, 2023 at 6:48 PM CEST, Andy Shevchenko wrote:
> > > > > On Tue, May 16, 2023 at 4:05 PM Esteban Blanc <eblanc@xxxxxxxxxxxx> wrote:
> > > > > > On Fri May 12, 2023 at 7:07 PM CEST,  wrote:
> > > > > > > Fri, May 12, 2023 at 04:17:54PM +0200, Esteban Blanc kirjoitti:
>
> ...
>
> > > > > > > > -#define TPS6594_REG_GPIOX_CONF(gpio_inst)          (0x31 + (gpio_inst))
> > > > > > > > +#define TPS6594_REG_GPIO1_CONF                             0x31
> > > > > > > > +#define TPS6594_REG_GPIOX_CONF(gpio_inst)  (TPS6594_REG_GPIO1_CONF + (gpio_inst))
> > > > > > >
> > > > > > > Why? The original code with parameter 0 will issue the same.
> > > > > >
> > > > > > I felt that replacing 0x31 with a constant would make the computation
> > > > > > in TPS6594_REG_GPIOX_CONFIG more understandable. What do you think?
> > > > >
> > > > > The question is why that register is so special that you need to have
> > > > > it as a constant explicitly?
> > > >
> > > > It is not special, it's just the first one of the serie of config
> > > > registers. I felt like just having 0x31 without context was a bit weird
> > >
> > > I'm not sure I understand what 'context' you are talking about.
> > I was trying to convey the fact that 0x31 was representing
> > TPS6594_REG_GPIO1_CONF address. This way when looking at
> > TPS6594_REG_GPIOX_CONF(...), one will better understand that this macro
> > is just about offsetting from the first GPIO_CONF register.
>
> You can add a comment on top of the macro, so anybody can read and see
> what this macro is doing.

> > > This is pretty normal to have two kind of definitions (depending on the case):
> > > 1/
> > >
> > >   #define FOO_1 ...
> > >   #define FOO_2 ...
> > >
> > > and so on
> > >
> > > 2/
> > >
> > >   #define FOO(x)  (... (x) ...)
> > >
> > > Having a mix of them seems quite unusual.
> > I did not know that. I will revert this change for next version then.
>
> Don't get me wrong, it's possible to have, but since it's unusual it
> needs to be well justified. In the change you proposed you have
> changed that, but I haven't seen where the new definition is used  (in
> *.c files).

Actualy it used in 2 places:
- In the switch case of `tps6594_gpio_regmap_xlate`
- In `tps6594_pinctrl_probe` when setting `reg_dir_out_base`

I already sent a v5 with this change but I managed to fail my .config
and this driver was not compiled... and it is not compiling... I feel so
stupid.

I need to send a v6 now anyway. Should I convert all
TPS6594_REG_GPIO1_CONF to TPS6594_REG_GPIOX_CONF(0)?