Re: [PATCH 1/2] hwmon: gpio-fan: mark OF related data as maybe unused

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

On 3/12/23 12:32, Krzysztof Kozlowski wrote:
> On 12/03/2023 19:48, Guenter Roeck wrote:
>> On 3/12/23 11:08, Krzysztof Kozlowski wrote:
>>> On 12/03/2023 19:06, Guenter Roeck wrote:
>>>> On Sat, Mar 11, 2023 at 12:16:09PM +0100, Krzysztof Kozlowski wrote:
>>>>> The driver can be compile tested with !CONFIG_OF making certain data
>>>>> unused:
>>>>>
>>>>>     drivers/hwmon/gpio-fan.c:484:34: error: ‘of_gpio_fan_match’ defined but not used [-Werror=unused-const-variable=]
>>>>>
>>>>
>>>> How do you trigger that ? The driver depends on OF_GPIO which
>>>> in turn depends on OF. Arguably that means that of_match_ptr()
>>>> doesnot really make sense, but still I don't see how you can
>>>> trigger the above error message.
>>>
>>> I can drop of_match_ptr, it's indeed in many cases not needed.
>>>
>>> I just build x86_64 allyesconfig minus OF, with W=1.
>>>
>>
>> x86_64 allyesconfig minus OF minus results in SENSORS_GPIO_FAN being
>> deselected (it doesn't show up in the configuration anymore at all
>> after "make olddefconfig").
>>
>> $ make allyesconfig
>> $ grep SENSORS_GPIO_FAN .config
>> CONFIG_SENSORS_GPIO_FAN=y
>> $ sed -i -e 's/CONFIG_OF=y/# CONFIG_OF is not set/' .config
>> $ make olddefconfig
>> #
>> # configuration written to .config
>> #
>> $ grep SENSORS_GPIO_FAN .config
>> $
>>
>> I suspect what you did is to disable CONFIG_OF, and then you built
>> the driver without realizing that is was deselected.
> 
> The driver would not built. I mean, I did not cheat here to built it
> with incorrect config.
> 
> I rather suspect that config is broken due to:
> WARNING: unmet direct dependencies detected for OF_GPIO
> WARNING: unmet direct dependencies detected for GPIO_SYSCON
> WARNING: unmet direct dependencies detected for MFD_STMFX
> 
> This was next-20230308
> 
Interesting. That has been fixed in next-20230310, where
SENSORS_GPIO_FAN is again deselected if CONFIG_OF is disabled.

>>
>>> Do you want to me to drop of_match_ptr?
>>>
>>
>> Yes, sure, that makes more sense. The reason though is that the
>> driver depends on CONFIG_OF=y, not because of the build failure.
> 
> I'll send v2 of both patches because anyway dropping of_match_ptr is
> preferred.
> 
Ok.

Thanks,
Guenter