Re: [poky] One recipe for fetch a bigger r epo, multiple recipes to build parts of it

Well, the idea with work-shared sounds good. But I didn't figure out what param makes fetcher to checkout into work-shared?
Just modify $S doesn't help, it is checked out still to tmp/work/<my-arch>/my-fw/${TAG}+git${SRCPV}/git
To workaround that for the moment I create a link, but that doesn't work reliable.

Recipe my-fw.bb looks like:

SRC_URI = "gitsm://[email protected];protocol=ssh;branch=${BRANCH_FW};tag=${TAG};lfs=0"
PV = "${TAG}+git${SRCPV}"

S = "${TMPDIR}/work-shared/${PN}-${TAG}"

do_configure () {
}

do_compile () {
}

do_install () {
rm -rf ${S}
ln -sf ${WORKDIR}/git ${S}

}

FILES:${PN}:append = " \
"


A recipe my-test using that and really creates a package that would be installed looks like that:


S = "${TMPDIR}/work-shared/my-fw-${TAG}/ "

RDEPENDS:${PN} = "net-snmp-lib-netsnmp net-snmp-lib-agent net-snmp-lib-mibs"
RDEPENDS:${PN} += "my-fw"

do_configure () {
}

do_compile () {
}

do_install:microzed-zynq7 () {
install -d ${D}/opt/lib
install -m 666 ${S}/lib/arm/release/yyyy.so ${D}/opt/lib/
${TARGET_PREFIX}strip ${D}/opt/lib/yyyy.so
}

But what not happened is, that calling bitbake my-test is causing to run bitbake my-fw before....
You see, there are some open points.


Am 5. November 2024 14:17:36 MEZ schrieb "Bruce Ashfield via lists.yoctoproject.org" <[email protected]>:
>On Tue, Nov 5, 2024 at 7:48 AM Steve via lists.yoctoproject.org <star=
>[email protected]> wrote:
>
>> On 11/5/24 13:06, Alexander Kanavin via lists.yoctoproject.org wrote:
>> > On Tue, 5 Nov 2024 at 13:01, Steve via lists.yoctoproject.org
>> > <[email protected]> wrote:
>> >> well download may indeed happened just once, but isn't in each recipe
>> >> the git folder with it's complete content?
>> >> Maybe I described it misleading:
>> >> So it is not about download, but about the amount of storage needed for
>> >> each recipe. Assuming 10 recipes using same main repo, in each of the
>> >> ../build/tmp/<recipe-name>/<tag>+gitAUTOINC+<git-hash>/git
>> >> I see the full checked out repo.
>> >
>> > If you want to use individual submodules, then the best would be
>> > probably to have an external script that determines SRCREV for each of
>> > them from what is recorded in the parent repo, and writes that into
>> > recipes.
>> >
>> > Alex
>>
>> True - that is one way.
>> But that doesn't sound easy with external scripts.
>> And the approach to use one recipe for checkout and multiple recipes to
>> access this cloned git is more effective in terms of storage usage,
>> isn't it?
>> How a recipeB, C, ... can know the path below - where the git-has is the
>> unknown component in this case.
>> To workaround the path
>> ../build/tmp/>arch>/<recipe-nameA>/<tag>+gitAUTOINC+<git-hash>/git
>>
>
>Have you looked at the work-shared technique we use for the kernel
>and gcc ?
>
>Bruce
>
>
>
>>
>> I guess an DEPENDS=recipeA needs also be part of all recipes using it.
>> Well, the recipes B,C,D might need to evaluate that git-hash, but I
>> don't know how in a recipe.
>>
>> Thanks
>>
>>
>> 
>>
>>
>
>-- 
>- Thou shalt not follow the NULL pointer, for chaos and madness await thee
>at its end
>- "Use the force Harry" - Gandalf, Star Trek II