On Mon, Mar 30, 2020 at 11:12 PM Michael Auchter <michael.auchter@xxxxxx> wrote:
> On Sun, Mar 22, 2020 at 01:46:21AM +0200, Andy Shevchenko wrote:
> > On Tue, Mar 17, 2020 at 4:53 PM Michael Auchter <michael.auchter@xxxxxx> wrote:
...
> > > + chip->reg = devm_regulator_get_optional(&client->dev, "vref");
> > > + if (!IS_ERR(chip->reg)) {
> >
> > Why not to go with usual positive conditional?
>
> I took this pattern from ad7266.c which Lars pointed me to. I agree that
> a positive conditional here would probably be more natural. I'll change
> that if you'd prefer.
Yes, please do.
...
> > > + .of_match_table = of_match_ptr(ad7291_of_match),
> >
> > No need to use of_match_ptr(). Haven't you got a compiler warning in !OF case?
>
> Hm, no warning as far as I can see with !OF...
Have you used `make W=1 ...`? With it you should get a warning that
table defined but not used.
> but agreed that this
> doesn't make much sense as-is.
>
> Is dropping of_match_ptr() the preferred route here? The driver doesn't
> depend on OF, so it seems like keeping of_match_ptr and instead guarding
> the ad7291_of_match table with #ifdef CONFIG_OF would be preferred. Of
> course, maybe that's not worth it for saving some bytes from the final
> image.
You need either both (of_match_ptr() _and_ ugly ifdeffery, and note
you will need of.h for that) or none (mod_devicetable.h maybe needed,
though).
> Let me know which route would be preferred.
If we would like to use this in non-DT environment, then drop all that
OF-specific stuff.
--
With Best Regards,
Andy Shevchenko
