docker-rpmbuild

A minimal docker rpmbuilder image.

Based on centos, includes only rpmdevtools and yum-utils and a couple of scripts that automate building RPM packages.

The scripts take care of installing build dependencies (using yum-builddep), building the package (using rpmbuild) and placing the resulting RPMs in output directory.

The setup is based on Fedora packaging how-to: http://fedoraproject.org/wiki/How_to_create_an_RPM_package

Supported tags and respective Dockerfile links

• 6 (centos:6 Dockerfile) - based on centos:6
• 7, latest (centos:7 Dockerfile) - based on centos:7

Usage

The image expects that work directory will be set to directory containing the sources (mounted from the host).

Typical usage:

docker run --rm --volume=$PWD:/src --workdir=/src \
  jitakirin/rpmbuild MYPROJ.spec

This will build the project MYPROJ in current directory, placing results in RPMS/${ARCH}/ and SRPMS/ subdirectories under current directory.

You can also specify to place the results in a subdirectory:

docker run --rm --volume=$PWD:/src --workdir=/src \
  jitakirin/rpmbuild MYPROJ.spec OUTDIR

This will create OUTDIR if necessary and place the results in OUTDIR/RPMS/${ARCH}/ and OUTDIR/SRPMS/.

If your package requires something from a non-core repo to build, you can add that repo using a PRE_BUILDDEP hook. It is an env variable that should contain an inline script or command to add the repo you need. E.g. for EPEL do:

docker run --rm --volume=$PWD:/src --workdir=/src \
  --env=PRE_BUILDDEP="yum install -y epel-release" \
  jitakirin/rpmbuild MYPROJ.spec

Debugging

There are two options to aid with debugging the build. One is to set VERBOSE option in the environment (with -e VERBOSE=1 option to docker run) which will enable verbose output from the scripts and rpmbuild. The other is to pass an --sh option to the image, which will drop to the shell instead of running rpmbuild, e.g.:

docker run -it -e VERBOSE=1 --rm --volume=$PWD:/src --workdir=/src \
  jitakirin/rpmbuild --sh MYPROJ.spec

From there you can inspect the environment and you can run the build manually either by switching to rpmbuild user:

su - rpmbuild
rpmbuild -ba rpmbuild/SPECS/MYPROJ.spec

or by running the same script the image uses:

runuser -u rpmbuild /usr/local/bin/docker-rpm-build.sh \
  ~rpmbuild/SPECS/MYPROJ.spec

Jenkins

To use this from a Jenkins builder which itself is running under docker (assuming it has access to host's docker socket), use something like:

docker run --rm \
  --volumes-from=JENKINS-VOLUME-CONTAINER --workdir="${WORKSPACE}" \
  jitakirin/rpmbuild MYPROJ.spec

This will build RPMs and place the results back in Jenkins' workspace directory.