Re: CVS commit: src/sys/arch/x86/x86

On Wed, Dec 12, 2007 at 06:30:27PM -0500, Perry E. Metzger
wrote:
> 
> Pavel Cahyna <pavelnetbsd.org> writes:
> >> I've noticed that Luke is churning a bunch of
the kernel with
> >> __KERNEL_RCSID changes, which are apparently
not causing enough
> >> trouble to make any of the branch owners
complain.
> >
> > The difference, AIUI, is that the rcsids are
actually useful for
> > something.
> 
> Being able to other compilers on the tree is also
useful for
> something. We've got lots of gccisms around that aren't
serving any
> purpose -- the idea is to get rid of them so that any
C99 compiler
> will work. __FUNCTION__ is one of several things I was
going to get
> rid of. The next step (as requested by Christos) was to
get rid of the
> unwrapped use of __attribute__ etc. -- you cut that out
of my message,
> I'll note, as well as my mention of *why* this is
wanted.

in the section 5.40 of the gcc manual you'll find an example
code that
you can put e.g. in cdefs.h for compilers that don't
support
__FUNCTION__.

Pavel

Pavel Cahyna <pavelNetBSD.org> writes:
> in the section 5.40 of the gcc manual you'll find an
example code that
> you can put e.g. in cdefs.h for compilers that don't
support
> __FUNCTION__.

All C99 compilers including gcc support __func__, which is
C99 standard.

Other compilers do not support __attribute__ and that is not
quite so
easy to macro away. There are macros in header files that
are supposed
to be used instead, but they aren't used everywhere.

In any case, the churn from this is pretty minimal compared
to
everything else going on, and I see no reason to believe
that this
will cause serious pain to anyone.

-- 
Perry E. Metzger		perrypiermont.com