kernel-headers or glibc-kernheaders

[David Woodhouse]

On Fri, 2006-08-25 at 22:02 -0500, Chris Adams wrote:
> I thought the idea of having a relatively static set of kernel headers
> used was because you were supposed to use the kernel headers that your
> glibc was built against, and that you were not supposed to just update
> the kernel headers along with the kernel.  Or has the kernel headers
> cleanup/export made that a thing of the past?

You don't _have_ to update the kernel headers along with the kernel,
although it should be perfectly safe to do so -- the ABI represented
therein should not be changing in incompatible ways.

It's perfectly feasible that we'll release kernel errata for FC6
_without_ releasing the kernel-headers packages to match.

Having kernel headers come directly from the kernel in a usable form,
instead of being maintained separately, was the whole _point_ of the
cleanup/export in the kernel tree. It means we get new syscalls, new
ioctls, etc. immediately rather than only when someone notices they're
missing and files a bug for them.

-- 
dwmw2