Hi. On 25/05/10 08:47, Dave Airlie wrote: > On Mon, 2010-05-24 at 07:17 -0700, Randy Dunlap wrote: >> On 05/24/10 06:59, Matthew Garrett wrote: >>> On Mon, May 24, 2010 at 06:53:51AM -0700, Randy Dunlap wrote: >>>> On 05/24/10 05:56, Matthew Garrett wrote: >>>>> Won't this result in a behavioural difference? The desirable outcome is >>>> >>>> It could, yes. >>>> >>>>> that that configuration be impossible, not for that configuration to >>>>> build but be buggy. >>>> >>>> so nouveau should depend on (or select, if ACPI is enabled) ACPI_BUTTON? >>> >>> There's an argument that it doesn't need to depend on it, but if button >>> is a module then nouveau has to be. Except the inverse isn't true. >>> Kconfig is hard, let's weep gently. >> >> Maybe Dave can weep with us when he is back at work... > > Yeah I've had problems like this a few times lately with the drm, I'm > torn between just adding select all over the place, or assuming someone > sane is configuring the kernel. > > I'm sort of erring on the someone sane is configuring the kernel just > because Linus's objects to "default y" things seems to point at that we > can't really give pointers to the people who haven't done it before. So > I'm quite happy to leave it having different behaviour depending on the > configuration and simply ignoring bug reports from incompetents.
'scuse me for butting in, but would you at least consider adding something to the help for the Nouveau and/or ACPI button options? You might have a sane person configuring the kernel, but that doesn't mean a dependency like this would necessarily be obvious to them. Personally, I'd argue for the select. Regards, Nigel
