On Fri, May 20, 2011 at 11:29 AM, Richard Purdie
<richard.pur...@linuxfoundation.org> wrote:
> Yes please :)
>
>> >> --- a/meta/classes/image.bbclass
>> >> +++ b/meta/classes/image.bbclass
>> >> @@ -11,8 +11,45 @@ INHIBIT_DEFAULT_DEPS = "1"
>> >>
>> >> # "export IMAGE_BASENAME" not supported at this time
>> >> IMAGE_BASENAME[export] = "1"
>> >> -export PACKAGE_INSTALL ?= "${IMAGE_INSTALL}"
>> >> -PACKAGE_INSTALL_ATTEMPTONLY ?= ""
>> >> +
>> >> +PACKAGE_INSTALL = "${@'
>> >> '.join(oe.packagegroup.required_packages('${IMAGE_FEATURES}'.split(),
>> >> d))}"
>> >> +PACKAGE_INSTALL_ATTEMPTONLY = "${@'
>> >> '.join(oe.packagegroup.optional_packages('${IMAGE_FEATURES}'.split(),
>> >> d))}"
>> >> +RDEPENDS += "${@'
>> >> '.join(oe.packagegroup.active_packages('${IMAGE_FEATURES}'.split(), d))}"
>> >
>> > I also noticed this patch changes things so PACKAGE_INSTALL_ATTEMPTONLY
>> > is used for the dev/doc/dbg packages. I'm not sure its a major issue but
>> > it is a change in behaviour and I'd have expected it in the commit
>> > message.
>>
>> Ah, that's my mistake then, I didn't realize the behavior was
>> different in the current implementation. I just figured some packages
>> might not have dev/doc/dbg, so it should be nonfatal to miss them. We
>> may also need to make sure rpm/deb both handle attemptonly properly,
>> as upstream's did not. I'll add this to the commit message, unless you
>> think they shouldn't be optional? I'm inclined to prefer it this way,
>> as, iirc, the depchain stuff uses recommends rather than depends.
>> (though i may be remembering wrong?)
>
> I just checked and all the rootfs package backends have code which looks
> like it makes that work. I'm fine with the behaviour change as long as
> we document it.
>
> FWIW, I merged the other patches in the series since they were not
> directly related.
Thanks, I'll resend this commit shortly. Aside: can anyone think of a
better name than PACKAGE_GROUP for defining named groups of packages?
I rather dislike it, but can't think of anything better. PACKAGES is
clearly already used, and isn't sufficiently explicit. PACKAGESET or
something could work, but I don't know that it's an improvement..
maybe this is best, but I thought I'd ask. I'm terrible at naming just
about everything, projects included (*cough* OE *cough*) :)
--
Christopher Larson
clarson at kergoth dot com
Founder - BitBake, OpenEmbedded, OpenZaurus
Maintainer - Tslib
Senior Software Engineer, Mentor Graphics
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core
