Re: awful kconfig help texts.

[Lars-Peter Clausen <lars@metafoo.de>]

On 08/01/2012 02:46 PM, Steven Rostedt wrote:
> On Wed, 2012-08-01 at 12:56 +0200, Lars-Peter Clausen wrote:
> 
>> You could do that by using a construct like below, but well...
>>
>> config HAS_PWM
>>     bool
>>
>> config PWM
>>     bool "PWM
>>     depends on HAS_PWM
>>
>> config PWM_CAN_BUILD_DRIVER_X
>>     defbool I2C
>>     select HAS_PWM
>>
>> config PWM_DRIVER_X
>>     tristate "PWM chip X support"
>>     depends on PWM && PWM_CAN_BUILD_DRIVER_X
>>
>> config PWM_CAN_BUILD_DRIVER_Y
>>     defbool ARCH_Y
>>     select HAS_PWM
>>
>> config PWM_DRIVER_Y
>>     tristate "PWM chip Y support"
>>     depends on PWM && PWM_CAN_BUILD_DRIVER_Y
> 
> What selects the 'PWM_CAN_BUILD_DRIVER_FOO'?
> 

Its def_bool statement, which lists the prerequisites to build the driver.
E.g. for a I2C PWM expander I2S support, for a on-SoC PWM core support for
the SoC family, etc.. So it will be true if the driver can actually be built
and false otherwise. If one of the PWM_CAN_BUILD_DRIVER_FOO symbols is true
also HAS_PWM will be true and PWM becomes selectable.

But it seems to be a bid tedious to have these extra lines for each driver
and I guess it is not a PWM subsystem specific issue. There are other
subsystems where this probably applies as well, e.g. the MFD subsystem.
Also such a solution would rule out out-of-tree PWM driver modules, since it
is not possible to get CONFIG_PWM selected.

- Lars