Re: [PATCH] watchdog: Do not enable by default during compile testing

[Arnd Bergmann]

On Fri, Apr 4, 2025, at 14:59, Krzysztof Kozlowski wrote:
> On 04/04/2025 14:57, Krzysztof Kozlowski wrote:
>> On 04/04/2025 14:49, Guenter Roeck wrote:
>>> On 4/4/25 05:39, Krzysztof Kozlowski wrote:
>>>> Enabling the compile test should not cause automatic enabling of all
>>>> drivers.
>>>>
>>>
>>> Sorry, I seem to be missing something.
>>>
>>> Isn't that what COMPILE_TEST is all about, that it enables everything ?
>> 
>> No. Compile test *allows* to compile test, but it does not mean you want
>> immediately compile everything. allyesconfig is for everything. Maybe
>> you want to compile some subset of drivers.
>> 
>> BTW, I am aligning with the most frequent pattern (quickly judging), so
>> of course I also accept argument that we should revert that other
>> pattern and use "default y" everywhere.
>
> I also dug out old recommendation from Arnd (so I am not making this
> stuff up):

Right, I agree with your patch here: COMPILE_TEST=y should make a
lot more options visible but not generally enable those.

Note that COMPILE_TEST=y also turns *off* some other options in
order to speed up the (allmodconfig) build process and avoid
some options that are unhelp for finding build issues:

lib/Kconfig.kasan-config KASAN_STACK
lib/Kconfig.kasan:      bool "Stack instrumentation (unsafe)" if CC_IS_CLANG && !COMPILE_TEST
security/Kconfig.hardening-config GCC_PLUGIN_STACKLEAK_VERBOSE
security/Kconfig.hardening-     bool "Report stack depth analysis instrumentation" if EXPERT
security/Kconfig.hardening-     depends on GCC_PLUGIN_STACKLEAK
security/Kconfig.hardening:     depends on !COMPILE_TEST        # too noisy
kernel/gcov/Kconfig-config GCOV_PROFILE_ALL
kernel/gcov/Kconfig-    bool "Profile entire Kernel"
kernel/gcov/Kconfig:    depends on !COMPILE_TEST

     Arnd