On 11.02.2008 03:53, Peter Stuge wrote: > On Mon, Feb 11, 2008 at 02:39:15AM +0100, Carl-Daniel Hailfinger wrote: > >>> The config system is not there to set code internal only >>> variables. >>> > > Hmm. > >>> That's what the arch/x86 cpu dependent include files are there >>> for. >>> > > Which files? >
Stefan? >> How do you define "code internal only variables"? I think variables >> with arch-dependent value (as opposed to arch-dependent existence) >> should be set by Kconfig. >> > > ..also valid point. This is why I liked it - the values always need > to be set. But Stefan is right in that we are overloading a > configuration mechanism. > > Linux solves this by using Kconfig to pick what objects to build, and > two objects can provide the same functions but with different > implementation. Do we just want better code abstraction? Could a > #define be used to #include one .S from another .S ? > CARBASE is like SMP (yes, it's in Kconfig) and DEFAULT_LOGLEVEL (in Kconfig as well) in that all of these variables do not select which objects to build. Should we remove all these variables from Kconfig? >> There is no reason to have stuff like LX_NUM_CACHELINES in Kconfig >> because it doesn't even exist on other subarches. >> > > Yes, agreed. > Of course. LX_NUM_CACHELINES is just a symbolic constant with no dependency on any configuration variable. CARBASE is a config variable depending on another configuration variable. It's that simple. Regards, Carl-Daniel -- http://www.hailfinger.org/ -- coreboot mailing list [email protected] http://www.coreboot.org/mailman/listinfo/coreboot
