On Mon, 3 Jul 2006 10:13:12 -0700 (PDT) Linus Torvalds wrote:
[snip]
> The whole "ARCH_HAS_XYZZY" is nothing but crap. It's totally unreadable,
...
> WE SHOULD GET RID OF ARCH_HAS_XYZZY. It's a disease.
Using Kconfig symbols for some of them seems more appropriate to me,
i.e., moving the symbol definitions from "random" header files to
Kconfig files. After all, these are just hidden config settings,
so use the Kconfig system and namespace for them.
This series of 9 patches converts:
__ARCH_HAS_DO_SOFTIRQ --> CONFIG_ARCH_DO_SOFTIRQ
ARCH_HAS_NMI_WATCHDOG --> CONFIG_ARCH_NMI_WATCHDOG.
ARCH_HAS_READ_CURRENT_TIMER --> CONFIG_ARCH_READ_CURRENT_TIMER.
ARCH_HAS_SCHED_WAKE_IDLE --> CONFIG_SCHED_SMT.
ARCH_HAS_SOCKET_TYPES --> CONFIG_ARCH_SOCKET_TYPES.
ARCH_HAS_VALID_PHYS_ADDR_RANGE --> CONFIG_ARCH_VALID_PHYS_ADDR_RANGE.
__ARCH_HAS_NO_PAGE_ZERO_MAPPED --> CONFIG_NO_PAGE_ZERO_MAPPED.
ARCH_HAS_POWER_INIT --> CONFIG_ACPI_POWER_INIT.
---
~Randy
On Monday 07 August 2006 23:13, Randy.Dunlap wrote:
> On Mon, 3 Jul 2006 10:13:12 -0700 (PDT) Linus Torvalds wrote:
>
> [snip]
>
> > The whole "ARCH_HAS_XYZZY" is nothing but crap. It's totally unreadable,
> ...
> > WE SHOULD GET RID OF ARCH_HAS_XYZZY. It's a disease.
>
> Using Kconfig symbols for some of them seems more appropriate to me,
IMHO that's no better.
Instead one should add dummy inlines/macros or asm-generic dummy inlines/macros
to all the architectures.
-Andi
On Tue, 8 Aug 2006 04:05:11 +0200 Andi Kleen wrote:
> On Monday 07 August 2006 23:13, Randy.Dunlap wrote:
> > On Mon, 3 Jul 2006 10:13:12 -0700 (PDT) Linus Torvalds wrote:
> >
> > [snip]
> >
> > > The whole "ARCH_HAS_XYZZY" is nothing but crap. It's totally unreadable,
> > ...
> > > WE SHOULD GET RID OF ARCH_HAS_XYZZY. It's a disease.
> >
> > Using Kconfig symbols for some of them seems more appropriate to me,
>
> IMHO that's no better.
>
> Instead one should add dummy inlines/macros or asm-generic dummy inlines/macros
> to all the architectures.
They are config characteristics (in many cases). Hiding them in
header files with dummy inlines or with ARCH_HAS_XYZZY is still
hiding them IMO, although the inlines are better than the
ARCH_HAS_XYZZY method. I prefer to see them in the config space
since I think that's where they belong.
---
~Randy
> They are config characteristics (in many cases). Hiding them in
> header files with dummy inlines or with ARCH_HAS_XYZZY is still
> hiding them IMO, although the inlines are better than the
> ARCH_HAS_XYZZY method. I prefer to see them in the config space
> since I think that's where they belong.
That's just a different way to write a #define. You could as
well keep the originals then.
-Andi
On Tue, 8 Aug 2006 05:15:10 +0200 Andi Kleen wrote:
>
> > They are config characteristics (in many cases). Hiding them in
> > header files with dummy inlines or with ARCH_HAS_XYZZY is still
> > hiding them IMO, although the inlines are better than the
> > ARCH_HAS_XYZZY method. I prefer to see them in the config space
> > since I think that's where they belong.
>
> That's just a different way to write a #define. You could as
> well keep the originals then.
yes, except that
1. these aren't buried in random header files
2. these are in the kconfig system & namespace where they belong
---
~Randy