Re: [PATCH v3 4/4] dpkg-base: Update git alternates

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

Am Mon, 11 Dec 2017 22:45:15 +0300
schrieb Alexander Smirnov <asmirnov@ilbers.de>:

> On 12/11/2017 08:56 PM, Henning Schild wrote:
> > Well i thought about this again. This is a very nasty workaround
> > for a problem this actually is in bitbake.  
> 
> That's an attempt to solve here and now git-3.0 issue reported by 
> Christian. So if in general it's not critical problem for now, I
> could drop this series.
> 
> > 
> > https://github.com/openembedded/bitbake/blob/ebce92bf8d71f8a6e8af1c6cf6ba335faf9d67c8/lib/bb/fetch2/git.py#L179
> > 
> > That problem should be fixed in bitbake, if they accept patches. The
> > "-s" needs to depend on some condition.
> >   
> 
> I'm not sure that this is the best solution. The main idea of using
> '-s' was to avoid unnecessary copying of git repos within one build,
> it costs disk space and extra time for copying (for example linux.git
> is about 1GB). So usage of '-s' is quite reasonable here.

I know why it is there, all i am saying it should maybe not be there
unconditionally. We could probably try and fix that upstrem and
introduce BB_GIT_SHARED or a new param "shared".

And now you could say
BB_GIT_SHARED=0
SRC_URI=git:// ....git
or
SRC_URI=git:// ....git;shared=0

Let me know what you think about that idea and i will prepare patches.
If they get merged we could cherry-pick them into Isar and forget about
the alternate-patching.

We could also agree that "shared" is an important feature we want in
Isar, in that case the alternates should get patched. And we need an
updated version of the proposed patch.
Thinking about that ... again ;) This is would i would now vote for.

> Another solution could be deprecate 'downloads' directory and fetch
> gits directly to package WORKDIR.

I think having the downloads dir might eventually help with the
reproducible builds.

> 
> > For Isar i would tell people to fetch tarballs from git servers and
> > avoid git://.  
> 
> This should not be the requirement, because it limits Isar
> application area. Not all git repo could be fetched as tarball.

Agreed, i am just talking about temporary workarounds until we have a
solution. Fetching with git is something that, to me, does not sound
like a high prio feature.

Henning

> Alex
> 
> > 
> > Henning
> > 
> > Am Mon, 27 Nov 2017 01:00:25 +0300
> > schrieb Alexander Smirnov <asmirnov@ilbers.de>:
> >   
> >> Update git alternates to work in buildchroot.
> >>
> >> Signed-off-by: Alexander Smirnov <asmirnov@ilbers.de>
> >> ---
> >>   meta/classes/dpkg-base.bbclass | 12 ++++++++++++
> >>   1 file changed, 12 insertions(+)
> >>
> >> diff --git a/meta/classes/dpkg-base.bbclass
> >> b/meta/classes/dpkg-base.bbclass index 5d7f8b6..6a78a26 100644
> >> --- a/meta/classes/dpkg-base.bbclass
> >> +++ b/meta/classes/dpkg-base.bbclass
> >> @@ -18,12 +18,24 @@ dpkg_runbuild() {
> >>   
> >>   # Wrap the function dpkg_runbuild with the bind mount for
> >> buildroot do_build() {
> >> +    if [ -d ${WORKDIR}/git/.git ]; then
> >> +        OBJ_PATH=$(cat
> >> ${WORKDIR}/git/.git/objects/info/alternates)
> >> +        REPO_PATH=$(dirname $OBJ_PATH)
> >> +        REPO_NAME=$(basename $REPO_PATH)
> >> +        echo "/git/$REPO_NAME/objects" >
> >> ${WORKDIR}/git/.git/objects/info/alternates
> >> +    fi
> >> +
> >>       mkdir -p ${BUILDROOT}
> >>       sudo mount --bind ${WORKDIR} ${BUILDROOT}
> >>       _do_build_cleanup() {
> >>           ret=$?
> >>           sudo umount ${BUILDROOT} 2>/dev/null || true
> >>           sudo rmdir ${BUILDROOT} 2>/dev/null || true
> >> +
> >> +        if [ -d ${WORKDIR}/git/.git ]; then
> >> +            echo $OBJ_PATH >
> >> ${WORKDIR}/git/.git/objects/info/alternates
> >> +        fi
> >> +
> >>           (exit $ret) || sudo umount ${BUILDCHROOT_DIR}/git
> >>           (exit $ret) || bb_exit_handler
> >>       }  
> >   
>