Improve consistency of --enable-kernel and default builds

[Carlos O'Donell]

On Mon, May 14, 2012 at 11:40 AM, Joseph S. Myers
<joseph@codesourcery.com> wrote:
> If the user passes --enable-kernel, that version (or the
> architecture-specific minimum, if the --enable-kernel version was too
> old) is always used to set __LINUX_KERNEL_VERSION and
> __ABI_TAG_VERSION.  If the user does not, those are set only if the
> architecture's minimum is greater than the global minimum (currently
> 2.2.0).  This works as long as patches removing old kernel support do
> what my pre-2.2-removal patch did and remove every conditional on the
> old kernel version; it doesn't allow multi-stage removal of support
> for old kernel versions (first increase the default, then gradually
> remove the conditionals).
>
> I propose this patch to make things simpler and more consistent by
> always setting minimum_kernel (and so those two macros, and so causing
> the ABI tags always to give the configured minimum kernel whether or
> not it's the global minimum).  Tested x86.
>
> 2012-05-14  Joseph Myers  <joseph@codesourcery.com>
>
>        * sysdeps/unix/sysv/linux/configure.in (minimum_kernel): Always
>        set if not set by the user.  Do not allow for being unset.
>        * sysdeps/unix/sysv/linux/configure: Regenerated.

I'm OK with this.

This looks like a step in the right direction.

The less variance there is in the build process the better.

Cheers,
Carlos.