On Wed, May 21, 2008 at 10:00:34AM +0100, David Woodhouse wrote:
> On Wed, 2008-05-21 at 01:43 -0700, Andrew Morton wrote:
> > On Wed, 21 May 2008 04:36:30 -0400 (EDT) "Robert P. J. Day" <rpjday@xxxxxxxxxxxxxx> wrote:
> >
> > > On Tue, 20 May 2008, Andrew Morton wrote:
> > >
> > > > On Sat, 17 May 2008 20:18:07 -0400 (EDT) "Robert P. J. Day" <rpjday@xxxxxxxxxxxxxx> wrote:
> > > >
> > > > >
> > > > > Move exported header files under include/linux that don't check the
> > > > > __KERNEL__ preprocessor from unifdef-y to header-y.
> > > > >
> > > >
> > > > Changelog fails to tell us why this change is being made.
> > > >
> > > > Perhaps it's because these headers just don't need unifdef processing?
> > > >
> > > > If so, that seems fragile. If we later add a __KERNEL__ section to
> > > > a header we need to remember to move the file to unifdef-y, and
> > > > we'll forget. It'd be better to process all files with unifdef.
> > > >
> > > > Or something. Or not.
> > >
> > > i don't see a problem with simply running all exported files through
> > > unifdef -- i've never understood the two categories since the unifdef
> > > process is not exactly CPU-intensive and it can't possibly hurt for
> > > some of those operations to be redundant.
> > >
> > > but as long as the two categories exist, might as well keep them
> > > clean.
> > >
> >
> > Well... not to let this go for now - nuking the header-y stuff is an
> > attractive cleanup. David, are we missing something here?
>
> I don't recall unifdef being anywhere near as fast as simply copying the
> headers, but a quick test shows that it is now -- it makes virtually no
> difference to the time if I unifdef them all. Maybe I was just making it
> up, or maybe the unifdef implementation we were using before was a lot
> less efficient.
>
> The other reason for having header-y separate from unifdef-y was that it
> highlights the difference between headers which are clean and ifdef-free
> (which we like), and headers with ifdefs (which, in general, we don't).
>
> Since we've removed the gratuitous __KERNEL__ from unexported headers,
> though, I don't think that matters much either -- we can see the ifdefs
> quite happily.
>
> So I see no particular reason why we shouldn't drop header-y completely
> and just unifdef the lot.
Sam already ACK'ed this idea when I suggested it a few weeks ago and
I'll get my ass up and cook a patch today or tomorrow.
> dwmw2
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
--
To unsubscribe from this list: send the line "unsubscribe linux-kbuild" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
