Re: [PATCH v2 2/6] dpkg-base: introduce an "apt-get source" fetch/unpack step

[Henning Schild <henning.schild@siemens.com>]

On Wed, 30 Jan 2019 08:38:34 +0100
"[ext] Jan Kiszka" <jan.kiszka@siemens.com> wrote:

> On 30.01.19 07:54, Maxim Yu. Osipov wrote:
> > From: Henning Schild <henning.schild@siemens.com>
> > 
> > This creates a new task to fetch/unpack debian sources from debian
> > source mirrors. This is done using "apt-get source" and the new
> > variable APT_SRC to control the arguments.
> > 
> > An example where a original debian package gets rebuild will follow
> > in a later commit.
> > 
> > Signed-off-by: Cedric Hombourger <Cedric_Hombourger@mentor.com>  
> 
> This SOB is still wrong.
> 
> > Signed-off-by: Henning Schild <henning.schild@siemens.com>
> > ---
> >   meta/classes/dpkg-base.bbclass | 19 +++++++++++++++++++
> >   1 file changed, 19 insertions(+)
> > 
> > diff --git a/meta/classes/dpkg-base.bbclass
> > b/meta/classes/dpkg-base.bbclass index f1b127c..2f324f9 100644
> > --- a/meta/classes/dpkg-base.bbclass
> > +++ b/meta/classes/dpkg-base.bbclass
> > @@ -20,6 +20,25 @@ do_adjust_git[stamp-extra-info] =
> > "${DISTRO}-${DISTRO_ARCH}" inherit patch
> >   addtask patch after do_adjust_git before do_build
> >   
> > +SRC_APT ?= ""
> > +
> > +do_apt_fetch[depends] = "buildchroot-target:do_build"
> > +
> > +do_apt_fetch() {
> > +	if [ -z "${@d.getVar("SRC_APT", True).strip()}" ]; then
> > +		exit
> > +	fi
> > +	dpkg_do_mounts
> > +	E="${@ bb.utils.export_proxies(d)}"
> > +	sudo -E chroot --userspec=$( id -u ):$( id -g )
> > ${BUILDCHROOT_DIR} \
> > +		sh -c 'cd ${PP} && apt-get -y source ${SRC_APT}'
> > +	dpkg_undo_mounts
> > +}
> > +
> > +addtask apt_fetch after do_unpack before do_patch
> > +do_apt_fetch[lockfiles] += "${REPO_ISAR_DIR}/isar.lock"
> > +do_apt_fetch[stamp-extra-info] = "${DISTRO}-${DISTRO_ARCH}"
> > +
> >   def get_package_srcdir(d):
> >       s = d.getVar("S", True)
> >       workdir = d.getVar("WORKDIR", True)
> >   
> 
> ...and I'm still wondering if we cannot extend bitbake (without
> patching it) instead.

I have one more trick in mind to at least get the interface right and
keep the implementation that relies on another task instead of do_fetch.

In fact the debian fetcher is an unpacker as well and has funny
buildchroot dependencies ... not a "simple" bb fetcher.

Henning

> Jan
>