[OE-core] [PATCH] linux-dummy: Add package kernel

Bruce Ashfield bruce.ashfield at gmail.com
Fri Nov 1 10:10:16 UTC 2019


On Thu, Oct 31, 2019 at 10:01 PM He Zhe <zhe.he at windriver.com> wrote:
>
>
>
> On 10/31/19 6:57 PM, Bruce Ashfield wrote:
> > On Thu, Oct 31, 2019 at 6:20 AM <zhe.he at windriver.com> wrote:
> >> From: He Zhe <zhe.he at windriver.com>
> >>
> >> Some package like packagegroup-core-boot may ask for package kernel. Let
> >> linux-dummy rprovide package kernel to fix the following build failure.
> >>
> >> ERROR: Nothing RPROVIDES 'kernel' (but
> >> .../meta/recipes-core/packagegroups/packagegroup-core-boot.bb RDEPENDS on or
> >> otherwise requires it)
> > Can you expand on what the higher level use case is that something is
> > using packagegroup-core-boot (or whatever), but also linux-dummy ?
> >
> > It seems like this is going to hide misconfigurations .. since you may
> > want something bootable, but yet are not properly configured.
> >
> > If there's part of that packagroup you need, why not split it up,
> > rather than masking the symptom seen here ?
>
> It's the "efi" in MACHINE_FEATURES who asks for "kernel".
> https://git.openembedded.org/openembedded-core/tree/meta/recipes-core/packagegroups/packagegroup-core-boot.bb#n31
>

Right. But the question is, why is the image that is being built
trying to install packagegroup-core boot when it has linux-dummy as
the kernel ? It's basically a misconfiguration, and I'm trying to
understand the details.

If this was provided by linux-dummy, I'd think that we'd at least want
a bbwarning to let the user know that they've asked for something
bootable, but will not end up with something that can boot.

> I thought the user would take care of the kernel himself, so simply added it to
> linux-dummy. I'm not quite sure what you mean by "split it up".

Meaning, make that part of the packagegroup into a new/separate
packagegroup that could be conditionally included by a variable (and
that way the configuration you are talking about could still depend on
packagegroup-core-boot, and not get the kernel dependency ... hence no
need for linux-dummy to provide it at all). Or alternately, just have
that entry in the package group be a variable, that another layer can
bbappend and clear (and again, remove the dependency).

i.e. if this (packagroup-core-boot) is being pulled in by something
like a container image build, there are already variables that control
the kernel dependency and other ways to avoid it.

Bruce

>
> Regards,
> Zhe
>
> >
> > Bruce
> >
> >> Signed-off-by: He Zhe <zhe.he at windriver.com>
> >> ---
> >>  meta/recipes-kernel/linux/linux-dummy.bb | 7 +++++--
> >>  1 file changed, 5 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/meta/recipes-kernel/linux/linux-dummy.bb b/meta/recipes-kernel/linux/linux-dummy.bb
> >> index 62cf6f5ea6..20d7ed815d 100644
> >> --- a/meta/recipes-kernel/linux/linux-dummy.bb
> >> +++ b/meta/recipes-kernel/linux/linux-dummy.bb
> >> @@ -8,19 +8,22 @@ LICENSE = "GPLv2"
> >>  LIC_FILES_CHKSUM = "file://${WORKDIR}/COPYING.GPL;md5=751419260aa954499f7abaabaa882bbe"
> >>
> >>  PROVIDES += "virtual/kernel"
> >> +RPROVIDES_${PN} += "kernel lib32-kernel"
> >>
> >>  PACKAGES_DYNAMIC += "^kernel-module-.*"
> >>  PACKAGES_DYNAMIC += "^kernel-image-.*"
> >>  PACKAGES_DYNAMIC += "^kernel-firmware-.*"
> >>
> >> -PACKAGES += "kernel-modules kernel-vmlinux"
> >> +PACKAGES += "kernel-modules kernel-vmlinux kernel"
> >>  FILES_kernel-modules = ""
> >>  ALLOW_EMPTY_kernel-modules = "1"
> >>  DESCRIPTION_kernel-modules = "Kernel modules meta package"
> >>  FILES_kernel-vmlinux = ""
> >>  ALLOW_EMPTY_kernel-vmlinux = "1"
> >>  DESCRIPTION_kernel-vmlinux = "Kernel vmlinux meta package"
> >> -
> >> +FILES_kernel = ""
> >> +ALLOW_EMPTY_kernel = "1"
> >> +DESCRIPTION_kernel = "Kernel meta package"
> >>
> >>  INHIBIT_DEFAULT_DEPS = "1"
> >>
> >> --
> >> 2.17.1
> >>
> >> --
> >> _______________________________________________
> >> Openembedded-core mailing list
> >> Openembedded-core at lists.openembedded.org
> >> http://lists.openembedded.org/mailman/listinfo/openembedded-core
> >
> >
>


-- 
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II


More information about the Openembedded-core mailing list