On Tue, Feb 23, 2016 at 8:29 AM, Jeremiah Mahler <[email protected]> wrote: > Hi Andy, > > Running the latest linux-next I am getting a Kconfig recursive > dependency detected message that was not present before.
Here's a heavily edited version, because the actual output is unreadable. > > jeri@hudson:~/linux-next$ make oldconfig > scripts/kconfig/conf --oldconfig Kconfig > drivers/gpio/Kconfig:34:error: recursive dependency detected! > drivers/gpio/Kconfig:34: symbol GPIOLIB is selected by GEOS > arch/x86/Kconfig:2572: symbol GEOS depends on DMI > arch/x86/Kconfig:810: symbol DMI is selected by DELL_LAPTOP > drivers/platform/x86/Kconfig:104: symbol DELL_LAPTOP depends on > BACKLIGHT_CLASS_DEVICE > drivers/video/backlight/Kconfig:158: symbol BACKLIGHT_CLASS_DEVICE is > selected by FB_BACKLIGHT > drivers/video/fbdev/Kconfig:192: symbol FB_BACKLIGHT is selected by > FB_SSD1307 > drivers/video/fbdev/Kconfig:2463: symbol FB_SSD1307 depends on GPIOLIB This is still unreadable because "selected by" is confusingly backwards. Let's rewrite where --> means "depends" and ==> means "selects". GEOS ==> GPIOLIB GEOS --> DMI DELL_LAPTOP ==> DMI DELL_LAPTOP --> BACKLIGHT_CLASS_DEVICE FB_BACKLIGHT ==> BACKLIGHT_CLASS_DEVICE FB_SSD1307 ==> FB_BACKLIGHT FB_SSD1307 --> GPIOLIB Which has no recursive dependencies. Unfortunately, it seems that actual limitation isn't that kconfig can't handle recursive dependencies -- it's that it can't handle cycles in the graph in which all the "select" edges are *backwards*. Presumably we should either make DELL_LAPTOP depend on DMI instead of selecting it or we should complain to the kconfig maintainers. Kconfig people: what's the actual problem here? --Andy
