On Wed, 2020-12-16 at 11:09 -0500, Sinan Kaya wrote:
> On 12/16/2020 6:50 AM, Richard Purdie wrote:
> > > +++ b/meta/recipes-connectivity/iproute2/iproute2.inc
> > > @@ -46,10 +46,11 @@ do_install () {
> > > # The .so files in iproute2-tc are modules, not traditional
> > > libraries
> > > INSANE_SKIP_${PN}-tc = "dev-so"
> > >
> > > -PACKAGES =+ "\
> > > +IPROUTE2_PACKAGES =+ "\
> > > ${PN}-devlink \
> > > ${PN}-genl \
> > > ${PN}-ifstat \
> > > + ${PN}-ip \
> > > ${PN}-lnstat \
> > > ${PN}-nstat \
> > > ${PN}-rtacct \
> > > @@ -58,12 +59,17 @@ PACKAGES =+ "\
> > > ${PN}-tipc \
> > > "
> > >
> > > +PACKAGES += "${IPROUTE2_PACKAGES}"
> > > +PROVIDES += "${IPROUTE2_PACKAGES}"
> > This doesn't look correct. Why do we need to add all the individual
> > package splits to PROVIDES?
> >
>
> I want to add iproute2-ip into my IMAGE_INSTALL and not take the rest
> of the iproute2-$foo packages. If I don't add PROVIDES, build stage
> fails saying nobody provides iproute2-ip.
IMAGE_INSTALL is a runtime namespace, the same as RPROVIDES/RDEPENDS,
not PROVIDES/DEPENDS so using iproute2-ip there should be fine.
> > or is there some other reason for this? This is basically the issue
> > with some of the other package splits too, its not using the build
> > time verses runtime namespaces correctly.
>
> If I misunderstood something or am not doing it right, let me know.
I think IMAGE_INSTALL should work (behind the scenes it adds to
RDEPENDS for the image).
Cheers,
Richard
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#145768):
https://lists.openembedded.org/g/openembedded-core/message/145768
Mute This Topic: https://lists.openembedded.org/mt/78992706/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
