Re: reasonable for a Kconfig variable to be tested only for "_MODULE"?

[Randy Dunlap <rdunlap@xenotime.net>]

On 04/12/2012 04:30 AM, Robert P. J. Day wrote:

> 
>   playing with my cleanup scripts this morning and ran across this
> output looking for unused CONFIG_ variables under drivers/ata:
> 
> $ ../s/find_unused_configs.sh drivers/ata
> ===== PATA_QDI
> drivers/ata/Kconfig:831:config PATA_QDI
> ===== PATA_WINBOND_VLB
> drivers/ata/Kconfig:865:config PATA_WINBOND_VLB
> $
> 
>   normally, i would flag that as "variables that are defined in a
> Kconfig file but are never tested anywhere", but upon further
> examination:
> 
> $ $ grep -r PATA_QDI *
> drivers/ata/Kconfig:config PATA_QDI
> drivers/ata/pata_legacy.c:#ifdef CONFIG_PATA_QDI_MODULE
> $
> 
> so "CONFIG_PATA_QDI" isn't tested, but "CONFIG_PATA_QDI_MODULE" *is*,
> which is not something i'd considered.  the Kconfig entry:
> 
> config PATA_QDI
>         tristate "QDI VLB PATA support"
>         depends on ISA
>         select PATA_LEGACY
>         help
>           Support for QDI 6500 and 6580 PATA controllers on VESA local bus.
> 
> and the relevant source:
> 
> /* Set to probe QDI controllers */
> #ifdef CONFIG_PATA_QDI_MODULE
> static int qdi = 1;
> #else
> static int qdi;
> #endif
> 
>   is that combination reasonable?  to test *only* for
> "CONFIG_whatever_MODULE"?  if it is, i'll add that check to avoid
> reporting false positives.

Yes, sometimes modules want to behave differently (e.g., initialize
differently) if they are built as a loadable module.
I've seen this other places.


>   all of the above can be said for PATA_WINBOND_VLB as well.
> 
> rday
> 



-- 
~Randy