The osc copypac (I assume that you have installed the osc package on your workstation) has an option -t which enables copying towards a remote target OBS instance. osc help and osc help command will advise on how to use these. First you need to import the project configuration.

$ export PROJECT=MeeGo-test:0.1
$ osc -A http://api-url-source-obs meta prjconf $PROJECT > my_project.conf
$ osc -A http://api-url-target-obs meta prjconf -F my_project.conf $PROJECT

Then import the project. As you might have some Links in the project that you import, it is a good idea to keep the source and target project names identical.

$ PRJ=ProjectToCopy; for i in `osc -A http://api.source.obs.domain ls $PRJ`; do \\
osc -A http://api.source.obs.domain \\
copypac -t http://api.target.obs.domain $PRJ $i $PRJ ;done

If you have access only to the repositories of your source reference target, then your life will be a bit more difficult. My advice would be to recheck if you find you cannot get a login on a public OBS service - such as provided by openSUSE or MeeGo - before proceeding this way. You will not be able to import the project config and you will have to create it by hand. This is too long to be covered in this HowTo. For more explanation about Build Service project config, see http://en.opensuse.org/openSUSE:Build_Service_prjconf.

Then you need to download all your rpm source on to a local machine and import it into your project with the command.

$ osc importsrcpkg

To initiate the build process, we will copy the rpm binary from the source OBS of the source project. These binary RPMs, from which we will remove any reference to release and version, will be used to trigger the first build. The OBS appliance will recompile all the RPMs until all RPMs in the project have been compiled only with packages compiled from their source code. Some base packages (e.g. tool chains) will be compiled several times during that process. Alternatively, you can at first build against a target which is similar to the base that you need to bootstrap in lieu of building against your own base and change the build reference to your bootstrap base once that the first build has been successful. Remember that you can also build against remote baseline. Double check that the preliminary step have been executed correctly. You must have already: copied a Linux base distribution in an OBS project defined a build target for that base project.

If you have not defined a build target, the necessary directory structure will not exist. This is a mandatory step of preparation. Stop the scheduler as it will create a mess if the system is not stable:

# rcobsscheduler stop

* Add binaries to the :full directory of the Project ssh onto the OBS server. Now go to the project’s build directory, and create a directory called “:full”. Note : standard is the default name of your Build repository as defined in your project. It might change depending on who created the initial build repo.

# cd /obs/build/$PROJECT/standard/i586

This directory structure should already exist. If not, there is a problem (note that /obs is link and the target may vary with your implementation). Now create the “:full” directory. $ mkdir :full Copy over all the binary RPMs of the project you are trying to build from scratch. These RPMs should have the release and version numbers stripped from them. e.g. alsa-utils-1.0.22-2.7.i586.rpm -- should be – alsa-utils.rpm Note : If the original project has a :full directory you can copy from there to avoid the issue of stripping version and release numbers. * Add binaries to the :full directory of the Project. Change all user/group privileges under /srv/obs/build/ to “obsrun”

# chown -R obsrun:obsrun /srv/obs/build

If you leave root as owner of :full, it will still build but the scheduler will fail (almost silently) to upgrade :full with the latest built packages. Except in very special cases, it is very unlikely that you want to do so. * Start the OBS scheduler

# rcobsscheduler start

* Force the obs to reindex your new :full directory. It will send an event to the scheduler which will create a file named :full.solv

# obs_admin --rescan-repository $OBS-PROJECT $REPO $ARCH

#find /obs/build ! -user obsrun
#chown -R obsrun:obsrun  /obs/build   (will correct ownership issue)

$obs_admin --rescan-repository $OBS-PROJECT $REPO $ARCH

#cd /obs/build/$PROJECT/standard/i586:full
#for I in `ls *.rpm` ; do rpm -qlp $I >/dev/null; if [ $? -ne 0 ] ;then echo $I >>../error.lst ; fi ; done
#cat ../error.lst (must be empty, all rpm in error needs re-installation)

#tail -f /obs/log/scheduler_i586.log