Re: [GIT] kbuild and kconfig fixes for 2.6.37-rc5

[Randy Dunlap <randy.dunlap@oracle.com>]

On 12/06/10 10:01, Catalin Marinas wrote:
> On 5 December 2010 16:51, trapDoor <trapdoor6@gmail.com> wrote:
>> On Sun, Dec 5, 2010 at 4:11 PM, Randy Dunlap <randy.dunlap@oracle.com> wrote:
>>> On Sun, 5 Dec 2010 15:42:46 +0000 trapDoor wrote:
>>>
>>>> Hello,
>>>> I couldn't wait and I merged the changes into my copy of mainline
>>>> tree. The merge was successful, no conflicts and the merge-diffstat
>>>> was exactly the same as above. However, when  running oldconfig on the
>>>> patched kernel I still get lots of warnings:
>>>>
>>>> warning: (GPIO_RDC321X && PCI && GPIOLIB) selects MFD_RDC321X which
>>>> has unmet direct dependencies (MFD_SUPPORT && PCI)
>>>> warning: (I2C_ISCH && I2C && PCI || GPIO_SCH && GPIOLIB && PCI ||
>>>> GPIO_VX855 && GPIOLIB || GPIO_RDC321X && PCI && GPIOLIB ||
> [...]
>>>> MFD_JZ4740_ADC && MFD_SUPPORT && MACH_JZ4740 || MFD_TPS6586X &&
>>>> MFD_SUPPORT && I2C=y && GPIOLIB && GENERIC_HARDIRQS || MFD_VX855 &&
>>>> MFD_SUPPORT && PCI) selects MFD_CORE which has unmet direct
>>>> dependencies (MFD_SUPPORT)
>>>
>>> That's 2 warnings.  I have sent a patch for the first one.
>>> Looks like I need to resend it.
>>
>> I was actually impressed by the amount of lines produced by those two
>> warnings :) I'd call it 'a lot'.
> 
> This could be simplified to only show the first option of every '&&'
> expression, something like below:
> 
> warning: I2C_ISCH selects MFD_CORE...
> warning: GPIO_SCH selects MFD_CORE...
> 
> But I couldn't figure out whether the expression reduction code in
> kbuild reorders the options or not.

I'm pretty sure that I have seen cases where the faulty kconfig is not the
first one listed, so all of them need to be available.

-- 
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***