Re: anyone using sanitized headers generated from the kernel source?

["Robert P. J. Day" <rpjday at mindspring dot com>]

On Fri, 1 Dec 2006, Kumar Gala wrote:

>
> On Dec 1, 2006, at 2:22 AM, Robert P. J. Day wrote:
>
> >
> >   been away from crosstool for a bit, is anyone using the sanitized
> > headers you get from the kernel from running "make headers_install"
> > rather than the linux-libc-headers package?  thanks.
>
> I just posted a patch for this.
>
> http://sourceware.org/ml/crossgcc/2006-11/msg00055.html

well, that's good timing.  :-)  and on a related note, what is the
precise *technical* definition of the sanitized headers?  i realize
that the primary property is that all "__KERNEL__" preprocessor
content should be removed because the headers are meant for user
space.  but if you had to explain what these headers represent in one
sentence or less, what sentence would that be?

rday

p.s.  i also recently noticed that, when you create the sanitized
headers using "make headers_install", not *all* of the "__KERNEL__"
protected content is removed.  there are a small number of header
files in include/linux that still have some content simply because
they're not listed in the appropriate Kbuild file.  i've already
submitted a kernel patch to take care of that.

on a related note, though, i did notice and report this:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=217840

because of what seems to be a fairly obvious shortcoming in the
"unifdef" utility, some "__KERNEL__" content remains in the sanitized
headers just because unifdef seems incapable of handling fairly simple
compound expressions in preprocessor directives.

to see what i'm talking about, just grep for "__KERNEL__" in the
generated sanitized headers to see what remains.

rday

--
For unsubscribe information see http://sourceware.org/lists.html#faq