svn commit: r299086 - head

[Ngie Cooper (yaneurabeya)]

> On May 4, 2016, at 17:31, Gleb Smirnoff <glebius at FreeBSD.org> wrote:
> 
> On Wed, May 04, 2016 at 02:06:33PM -0700, Ngie Cooper (yaneurabeya) wrote:
> N> 
> N> > On May 4, 2016, at 14:00, Garrett Cooper <ngie at FreeBSD.org> wrote:
> N> > 
> N> > Author: ngie
> N> > Date: Wed May  4 21:00:41 2016
> N> > New Revision: 299086
> N> > URL: https://svnweb.freebsd.org/changeset/base/299086
> N> > 
> N> > Log:
> N> >  Default NO_INSTALLEXTRAKERNELS to "no" to unbreak the build
> N> > 
> N> >  MFC after: soon (was insta-MFCed -_-..)
> N> >  Pointyhat to: glebius
> N> >  Sponsored by: EMC / Isilon Storage Division
> N> > 
> N> > Modified:
> N> >  head/Makefile.inc1
> 
> Defaulting it to "no" you changed the behaviour backwards, which is
> wrong.

Agreed. That’s why I inverted it back to “yes” in r299088.

> N> This broke for me on 11.0-CURRENT because I use:
> N> 
> N> KERNCONFS=	GENERIC GENERIC-NODEBUG
> N> 
> N> and use installkernel with INSTKERNNAME.
> 
> That's quite specific setup. Probably NO_INSTALLEXTRAKERNELS should
> be played with conditionally.

It’s not an uncommon setup though. I have used it on all my CURRENT machines for some time because I might want to be able to use INVARIANTS kernels sometimes when doing kernel changes, and boot !INVARIANTS kernels all of the time.

I chose "?= yes” because it’s better syntactic sugar than `defined(NO_INSTALLEXTRAKERNELS) && ${NO_INSTALLKERNELS} != “yes”` and it works with pre-bmake (the other idiom that bmake allows that would work here… although with more duplicity is `:Uyes`, which would break installkernel on FreeBSD 9 machines or with fmake as the system make).

Thanks!
-Ngie