Re: [PATCH] bring sanity to div64.h and do_div usage

[Jeff Garzik <jgarzik@mandrakesoft.com>]

Jes Sorensen wrote:
> 
> >>>>> "Jeff" == Jeff Garzik <jgarzik@mandrakesoft.com> writes:
> 
> Jeff> Jes Sorensen wrote:
> >>  Why? CONFIG_$ARCH only makes sense if you can enable two
> >> architectures in the same build. What does CONFIG_M68K give you that
> >> __mc68000__ doesn't provide?
> 
> Jeff> 1) it is a Linux kernel standard.  all arches save two define
> Jeff> CONFIG_$arch.
> 
> Ehm, what standard? the standard way has always been ARCH==, CONFIG_PPC
> used to be the only place using this and all it did was to make things
> uglier and inconsistent.

*shrug*  I've known about this for years.  And every arch but m68k and
cris follows the standard.


> Jeff> 2) you have two tests, "ARCH==m68k" in config.in and "__mc68000__"
> Jeff> in C code.  CONFIG_M68K means you only test one symbol, the same
> Jeff> symbol, in all code.
> 
> If you want to do that, then one should use CONFIG_<ARCH> in the
> Makefiles as well.

I'm betting a few special makefiles require $ARCH information when
.config is info is not available.  Otherwise, agreed.


> Jeff> 3) as this thread shows, due to #1, users -expect- that
> Jeff> CONFIG_M68K will exist
> 
> Ehm, most kernel developers will expect ARCH== in Config.in as thats
> how it's always been.

and I forgot about this flat-out bug:

4) you cannot test ARCH==m68k as a component of dep_xxx declarations. 
And yes, this is a requirement, this is used currently in */config.in
all over the kernel tree.

	Jeff


-- 
Jeff Garzik      | "UNIX enhancements aren't."
Building 1024    |           -- says /usr/games/fortune
MandrakeSoft     |