Re: [PATCH v2 3/6] meta/dpkg: pull out actual build command from 'do_build'

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

Am Wed, 13 Sep 2017 16:40:50 +0200
schrieb Claudius Heine <ch@denx.de>:

> Hi,
> 
> On 13.09.2017 15:41, Henning Schild wrote:
> > Am Wed, 13 Sep 2017 15:44:00 +0300
> > schrieb Alexander Smirnov <asmirnov@ilbers.de>:
> >   
> >> Hi,
> >>
> >> On 09/12/2017 06:58 PM, Henning Schild wrote:  
> >>> Issue:
> >>> The 'do_build' task in dpkg does more than just building, it also
> >>> prepares the build environment and cleans it up after building.
> >>> Other classes building debian packages might need the same steps
> >>> while implementing the actual build step in a different way.
> >>>
> >>> Change:
> >>> Pull the actual building out into a function that gets called from
> >>> 'do_build' between the pre and post stuff.
> >>>
> >>> Impact:
> >>> This patch does not change the behaviour of Isar. It addresses the
> >>> issue and prepares for another patch.
> >>>
> >>> Signed-off-by: Henning Schild <henning.schild@siemens.com>
> >>> ---
> >>>   meta/classes/dpkg.bbclass | 7 ++++++-
> >>>   1 file changed, 6 insertions(+), 1 deletion(-)
> >>>
> >>> diff --git a/meta/classes/dpkg.bbclass b/meta/classes/dpkg.bbclass
> >>> index abd4cef..5c3c1f8 100644
> >>> --- a/meta/classes/dpkg.bbclass
> >>> +++ b/meta/classes/dpkg.bbclass
> >>> @@ -15,10 +15,15 @@ BUILDROOT = "${BUILDCHROOT_DIR}/${PP}"
> >>>   do_build[stamp-extra-info] = "${DISTRO}-${DISTRO_ARCH}"
> >>>   
> >>>   # Build package from sources using build script
> >>> +dpkg_runbuild() {
> >>> +    sudo chroot ${BUILDCHROOT_DIR} /build.sh ${PP}/${SRC_DIR}
> >>> +}    
> >>
> >> That's one of the weak places in Isar (also there are several
> >> others with the same issue). If chroot failed - we will have some
> >> folders mounted. Can we somehow track return value and perform
> >> clean-up if build failed?  
> > 
> > I found that issue as well. But i decided to ignore it for now,
> > because i did not introduce it.
> > 
> > It would be easy to deal with it. i.e.
> > 
> > mkdir
> > touch
> > mount
> > do
> > umount
> > rm
> > rmdir
> > 
> > would become
> > 
> > ret=0
> > mkdir
> > touch
> > mount
> > do || ret=$?
> > umount
> > rm
> > rmdir
> > return ret  
> 
> Another solution might be to look into events and write some cleanup
> code there [1].

Sounds like a good idea. It is basic error/exception handling. I am
sure bitbake has some mechanism for that, my example above was just
made up without any research in existing mechanism.

Henning

> Cheers,
> Claudius
> 
> [1]
> https://www.yoctoproject.org/docs/1.6/bitbake-user-manual/bitbake-user-manual.html#events
>