Re: [PATCH V1] kernel-devsrc: add 'make scripts' postinstall

[Bruce Ashfield <bruce.ashfield@windriver.com>]

On 15-11-24 04:32 PM, Ming Liu wrote:
>
>
> On 11/24/2015 05:29 PM, Bruce Ashfield wrote:
>> On 15-11-23 04:58 PM, Ming Liu wrote:
>>> From: Ming Liu <peter.x.liu@external.atlascopco.com>
>>>
>>> Run 'make scripts' inside qemu in order to create script binaries that
>>> needed by external modules build on the target.
>>>
>>> Add RDEPENDS on gcc, make, which are needed by the compilation.
>>
>> This version looks fine to me .. and by "fine", I mean that it probably
>> is still less than ideal, but with our current packaging and separation
>> it makes sense to make it work, with all dependencies being explicit.
>>
>> As we (re)slim down the devsrc/headers split, we can revisit this as
>> necessary.
>>
>> One question below though ..
>>
>>>
>>> [YOCTO #6630]
>>>
>>> Signed-off-by: Ming Liu <peter.x.liu@external.atlascopco.com>
>>> ---
>>>   meta/recipes-kernel/linux/kernel-devsrc.bb | 14 ++++++++++++--
>>>   1 file changed, 12 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/meta/recipes-kernel/linux/kernel-devsrc.bb
>>> b/meta/recipes-kernel/linux/kernel-devsrc.bb
>>> index 196c8c7..87826ad 100644
>>> --- a/meta/recipes-kernel/linux/kernel-devsrc.bb
>>> +++ b/meta/recipes-kernel/linux/kernel-devsrc.bb
>>> @@ -11,7 +11,7 @@ LIC_FILES_CHKSUM =
>>> "file://${COMMON_LICENSE_DIR}/GPL-2.0;md5=801f80980d171dd6425
>>>   inherit linux-kernel-base
>>>
>>>   # Whilst not a module, this ensures we don't get multilib extended
>>> (which would make no sense)
>>> -inherit module-base
>>> +inherit module-base qemu
>>
>> Shouldn't this be qemu-native ? Since it is running on the host ? Or
>> maybe I'm completely misunderstanding something :)
> Indeed it is calling qemu-native when inheriting this class, to make a

That is probably my misunderstanding .. I didn't go look at the inherit,
and I read this more as a RDEPENDS, which would be qemu for the target.

> certain postinstall scriptlet run at do_rootfs time through a
> architecture dependent qemu.
> I am not sure if I had answered your question, maybe you mean the name
> is not proper? qemu-native.bbclass would make more sense?

I think you have. If the class is called qemu .. then that's what
you should use.

Bruce

>
> //Ming Liu
>>
>> Bruce
>>
>>>
>>>   # We need the kernel to be staged (unpacked, patched and
>>> configured) before
>>>   # we can grab the source and make the source package. We also need
>>> the bits from
>>> @@ -71,6 +71,16 @@ do_install() {
>>>   # Ensure we don't race against "make scripts" during cpio
>>>   do_install[lockfiles] = "${TMPDIR}/kernel-scripts.lock"
>>>
>>> +pkg_postinst_${PN} () {
>>> +    if [ -n "$D" ]; then
>>> +        ${@qemu_run_binary(d, '$D', '${bindir}/make')} \
>>> +            -C $D${KERNEL_SRC_PATH} O=$D${KERNEL_SRC_PATH} scripts
>>> +    else
>>> +        make -C ${KERNEL_SRC_PATH} O=${KERNEL_SRC_PATH} scripts
>>> +    fi
>>> +}
>>> +
>>>   PACKAGES = "kernel-devsrc"
>>>   FILES_${PN} = "${KERNEL_SRC_PATH}"
>>> -RDEPENDS_${PN} = "bc"
>>> +DEPENDS = "qemu-native"
>>> +RDEPENDS_${PN} = "bc gcc make"
>>>
>>
>