[meta-intel] [PATCHv2 4/5] meta-valleyisland: add kernel recipe for valleyisland

[Paul Eggleton]

On Monday 24 March 2014 15:16:56 Darren Hart wrote:
> On 3/24/14, 15:11, "Paul Eggleton" <paul.eggleton at linux.intel.com> wrote:
> >On Monday 24 March 2014 15:07:10 Darren Hart wrote:
> >> On 3/24/14, 10:57, "rebecca.swee.fun.chang at intel.com"
> >> 
> >> <rebecca.swee.fun.chang at intel.com> wrote:
> >> >+#############################
> >> >+# MACHINE = valleyisland-32 #
> >> >+#############################
> >> >+COMPATIBLE_MACHINE_valleyisland-32 = "valleyisland-32"
> >> >+KMACHINE_valleyisland-32 = "valleyisland-32"
> >> >+KBRANCH_valleyisland-32 = "standard/base"
> >> >+KERNEL_FEATURES_valleyisland-32 =
> >> >"features/valleyisland-io/valleyisland-io"
> >> >+
> >> >+LINUX_VERSION_valleyisland-32 = "3.10.32"
> >> >+SRCREV_machine_pn-linux-yocto_valleyisland-32 ?=
> >> >"61dde96f97bb5b1ed4c11caf9a857d55ad8f6e17"
> >> >+SRCREV_meta_pn-linux-yocto_valleyisland-32 ?=
> >> >"99c503a92885060bebf2bba6747735e8e9346a40"
> >> 
> >> With respect the need for pn-linux-yocto here, this is because all the
> >> other SRCREV_machine* in meta-intel are using = instead of ?=. This
> >>
> >>won't
> >>
> >> get set because the base linux-yocto recipe specifies a SRCREV, so the
> >>
> >>?=
> >>
> >> will no override it. The pn-linux-yocto provides a more specific value
> >> which has not yet been set, so it sticks and gets applied.
> >> 
> >> Just for consistency, the right answer is probably to use = instead of
> >>
> >>?=
> >>
> >> and drop the pn-linux-yocto.
> >> 
> >> I do not really care for have "=" in the bbappend honestly. I wonder if
> >> Paul would have another recommendation here...
> >
> >What concerns you about using = here? I would have thought it was OK in
> >this
> >situation; ?= is only really helpful where you want to apply a default
> >where a
> >value that should override the default may already have been set (e.g. in
> >local.conf where a desired value might have been set from the command
> >line).
> 
> That is basically the concern, do we want people to be able to override it
> easily? Perhaps the answer is no. I suspect SRCREV's are really not
> something we want people messing with. We can still override in local.conf
> using = right?

We can if an additional override that is not present in the recipe/append is 
used (e.g. pn-linux-yocto, assuming that is removed from the append as you 
suggest). 

If the recipe is setting it with = then that does make it a little harder to 
override, but I think with something like SRCREV that is a good thing as you 
say.

Cheers,
Paul

-- 

Paul Eggleton
Intel Open Source Technology Centre