Re: [PATCH] fix(sstate): don't fail if there are no packages

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

Am Tue, 11 Jan 2022 11:23:41 +0100
schrieb "Schmidt, Adriaan (T CED SES-DE)" <adriaan.schmidt@siemens.com>:

> Schild, Henning, 11. Januar 2022 10:14:
> > Am Tue, 11 Jan 2022 09:02:00 +0100
> > schrieb Adriaan Schmidt <adriaan.schmidt@siemens.com>:
> >  
> > > The code to put generated deb packages into the sstate cache
> > > currently uses the pattern
> > >   test <condition> && do_stuff
> > > which, if the condition is not met, not only skips do_stuff,
> > > but also returns failure.
> > >
> > > The consequence is that in cases where there are no packages
> > > found in ${S}/../*.deb, the sstate caching fails completely.
> > >
> > > This changes that pattern to use an explicit "if" instead.
> > >
> > > Signed-off-by: Adriaan Schmidt <adriaan.schmidt@siemens.com>
> > > ---
> > >  meta/classes/dpkg-base.bbclass | 6 ++++--
> > >  1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/meta/classes/dpkg-base.bbclass
> > > b/meta/classes/dpkg-base.bbclass index cb5ce4a..2add0b2 100644
> > > --- a/meta/classes/dpkg-base.bbclass
> > > +++ b/meta/classes/dpkg-base.bbclass
> > > @@ -222,13 +222,15 @@ do_dpkg_build[sstate-plaindirs] =
> > > "${DPKG_SSTATE}" do_dpkg_build[sstate-interceptfuncs] =
> > > "dpkg_build_sstate_prepare"
> > >  dpkg_build_sstate_prepare() {
> > > -    test -n "$(find ${S}/.. -maxdepth 1 -name '*.deb' -print
> > > -quit)" &&
> > > +    if [ -n "$(find ${S}/.. -maxdepth 1 -name '*.deb' -print
> > > -quit)" ]; then ln -f ${S}/../*.deb -t ${DPKG_SSTATE}
> > > +    fi  
> >
> > I think we can skip the "-print" while at it. And is that hard
> > linking not problematic when DPKG_SSTATE and S are in different
> > filesystems?  
> 
> The "-print" is needed. Otherwise "-quit" will terminate the find
> before anything is printed.

Sure ... need to skip that "-n" as well and just use the return value
of find.

> Both DPKG_SSTATE and S are (in all recipes in the Isar repo) always
> below WORKDIR, which needs to be on one filesystem so it can be
> bind-mounted into the buildchroot.
> 
> (From looking at other code in Isar, my impression is that many things
> would fail if the complete Isar build tree is spread across multiple
> filesystems, e.g. the buildchroot, sdkchroot, and bootstrap recipes
> create hard links from deploy dir to work dir.)

I see, was assuming DPKG_STATE to be under SSTATE_DIR where different
fs could be likely i guess.

Henning

> Adriaan
> 
> >  
> > >  }
> > >
> > >  dpkg_build_sstate_finalize() {
> > > -    test -n "$(find ${DPKG_SSTATE} -maxdepth 1 -name '*.deb'
> > > -print -quit)" &&
> > > +    if [ -n "$(find ${DPKG_SSTATE} -maxdepth 1 -name '*.deb'
> > > -print -quit)" ]; then ln -f ${DPKG_SSTATE}/*.deb -t ${S}/..
> > > +    fi
> > >  }
> > >
> > >  python do_dpkg_build_setscene() {  
>