On Fri, 29 Mar 2019, Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote: > On 3/28/19 8:18 PM, Stephen Rothwell wrote: >> Hi all, >> >> Changes since 20190328: >> >> The pidfd tree lost its build failures. >> > > on x86_64, when # CONFIG_ACPI is not set/enabled: > > ld: drivers/gpu/drm/i915/intel_panel.o: in function `intel_backlight_device_register': > intel_panel.c:(.text+0x2c49): undefined reference to `backlight_device_register' > ld: drivers/gpu/drm/i915/intel_panel.o: in function `intel_backlight_device_unregister': > intel_panel.c:(.text+0x2d01): undefined reference to `backlight_device_unregister' > > > > Full randconfig file is attached. # CONFIG_ACPI is not set CONFIG_DRM_I915=y CONFIG_BACKLIGHT_CLASS_DEVICE=m drivers/gpu/drm/i915/Kconfig: config DRM_I915 # ... select BACKLIGHT_CLASS_DEVICE if ACPI The above pretty much sums it up. It's been reported before, the earliest I could find was [1]. I've previously suggested patches that would make backlight a dependency instead of selecting it, in line with [2]. However we can't do that in i915 alone without recursive dependencies, and changing all backlight users was met with opposition because "select is convenient". Maybe I'll look at this one of these days, but don't hold your breath. BR, Jani. [1] http://mid.mail-archive.com/CA+r1ZhhmT4DrWtf6MbRQo5EqXwx+LxCqh15Vsu_d9WpftLhnxw@xxxxxxxxxxxxxx [2] Documentation/kbuild/kconfig-language.txt "In general use select only for non-visible symbols (no prompts anywhere) and for symbols with no dependencies." -- Jani Nikula, Intel Open Source Graphics Center
