Re: [RFC PATCH] hwmon: remove tautological dependency expressions from Kconfig

[Guenter Roeck <linux@roeck-us.net>]

On 7/29/26 14:54, Laszlo Ersek wrote:
> Commit 554b3529fe01 ("thermal/drivers/core: Remove the module Kconfig's
> option") [1] obviated some of the "Optional dependencies" (as described in
> "Documentation/kbuild/kconfig-language.rst") in "drivers/hwmon/Kconfig".
> 
> Namely, said commit restricted the valid/possible subset of the full
> Cartesian product "CONFIG_THERMAL x CONFIG_THERMAL_OF" to
> 
>    { (n, n), (y, n), (y, y) }
> 
> through "drivers/thermal/Kconfig". Numerically (using tristate values),
> this set can be expressed as:
> 
>    { (0, 0), (2, 0), (2, 2) }
> 
> Over this set, the "drivers/hwmon/Kconfig" tristate expression
> 
>    THERMAL || !THERMAL_OF
> 
> or equivalently,
> 
>    max(THERMAL, 2-THERMAL_OF)
> 
> invariably evaluates to "y" (2).
> 
> Similarly, with THERMAL having been restricted to
> 
>    { n, y }
> 
> or numerically,
> 
>    { 0, 2 }
> 
> the "drivers/hwmon/Kconfig" tristate expression
> 
>    THERMAL || THERMAL=n
> 
> or equivalently,
> 
>    max(THERMAL, THERMAL=0)
> 
> invariably evaluates to "y" (2). (Note that "THERMAL=0" returns either "y"
> (2) or "n" (0).)
> 
> Commit f37353320ee9 ("hwmon/drivers/core: Simplify complex dependency")
> [2], which is the final commit of the series that [1] is the beginning of,
> cleaned up the hwmon C source code in fact; however, it missed the
> now-collapsed dependencies in Kconfig. At the time of [2], six of today's
> such dependencies were present in Kconfig. Since then, two more have been
> added (such that they persist, anyway): one by commit 4f8d374b770a
> ("hwmon: (max6650) add thermal cooling device capability"), which was
> independent of the above-noted series, and another by commit 9855caf5d4eb
> ("hwmon: add driver for the hwmon parts of qnap-mcu devices"), which was a
> descendant of the series.
> 
> Remove these always-"y" dependencies.
> 

Dependencies such as "THERMAL || THERMAL=n" were supposed to enable the
driver to be built as module or into the kernel if THERMAL=n, and to
force it to be built as module if THERMAL=m. However, both THERMAL
and THERMAL_OF are boolean, so this really does not make sense.

Applied.

Thanks,
Guenter