[OE-core] Is a dependency on hostperl-runtime-native actually useful?

[richard.purdie at linuxfoundation.org]

On Mon, 2018-07-30 at 18:06 -0700, Andre McCurdy wrote:
> On Mon, Jul 30, 2018 at 3:23 PM, Richard Purdie
> <richard.purdie at linuxfoundation.org> wrote:
> > On Mon, 2018-07-30 at 14:44 -0700, Andre McCurdy wrote:
> > > Currently the openssl 1.0 recipe defines a dependency on
> > > hostperl-runtime-native and the openssl 1.1 recipe does not. Both
> > > run
> > > "perl ./Configure ..." as part of do_configure().
> > > 
> > > Since hostperl-runtime-native is included in ASSUME_PROVIDED, is
> > > it
> > > really useful for the openssl 1.0 recipe to list it in DEPENDS?
> > > 
> > > ie is the openssl 1.0 recipe being unnecessarily complex or is
> > > the
> > > openssl 1.1 recipe being too simplistic?
> > 
> > It is useful for things to list their dependencies and we did have
> > an
> > effort to actually list things out so we know ASSUME_PROVIDED is
> > correct. This means we can spot areas we might be able to trim back
> > dependencies (amongst other reasons).
> > 
> > With the introduction of HOSTTOOLS, its perhaps less needed than it
> > was
> > but in principle it is still useful to know which things need a
> > given
> > item, particularly where its more unusual. I'd still be interested
> > in
> > trying to cut back HOSTTOOLS a bit more.
> > 
> > The 1.1 recipe was pretty heavily cut back, probably too much so
> > based
> > on some of the patches we've been getting...
> 
> Thanks. Is there an obvious reason why both recipes shouldn't be
> using perl-native rather than hostperl-runtime-native? Building with
> perl-native seems to work fine.

Build time overhead. openssl-native is one of the early dependencies
(or it certainly used to be due to git-native before we ASSUME_PROVIDED
it). Allowing openssl-native to use host perl cut down build time
significantly.

Cheers,

Richard