Skip to content


Repository files navigation

Spark Overlay

Scan ebuilds with pkgcheck Test ebuilds in Docker container

The Spark overlay is a Gentoo ebuild repository that offers Gentoo packages supporting big data infrastructures based on the Java platform. It was originally created for distributing Apache Spark to Gentoo users (hence the name "Spark overlay"), but it has been expanded later to include packages for H2O and Kotlin core libraries that can be built from source.

Provided Packages

Apache Spark

  • dev-java/spark-core: The core package for Spark
  • dev-java/spark-demo: A demo program for the Spark packages in this overlay, which can be run with command spark-demo-2.12

Note: Some dependencies of Spark require Kotlin compiler 1.4. When being pulled as a dependency, Kotlin compiler 1.4 might not be fully-automatically installable by emerge. If emerge comes up with any error pertaining to Kotlin packages, please install Kotlin compiler 1.4 separately before retrying:

# emerge --ask --oneshot dev-lang/kotlin-bin:1.4


  • dev-java/h2o: The meta package for H2O
    • Most H2O sub-packages are slotted based on the major version of H2O (e.g. 3.32, 3.34), so it is possible to install many major versions of H2O in parallel
    • The name of H2O's executable for a slot is h2o-${SLOT}; for example, the executable for H2O 3.32.x.y is h2o-3.32
  • dev-python/h2o-py: The Python module for H2O
  • dev-java/h2o-flow: The web-based interactive computational environment for H2O
    • To use H2O Flow, please enable the flow USE flag of dev-java/h2o


For information regarding the Kotlin packages, please consult the relevant page on Gentoo Wiki.

Initial Configuration

The Spark overlay needs to be added to the system before the packages in it can be installed. This can be done with the following steps:

  1. Because the Spark overlay is offered as a Git repository, Git must be installed before the repository's contents can be downloaded to the system. The following command ensures that Git is installed on the system:

    # emerge --ask --noreplace dev-vcs/git
  2. Add the Spark overlay to the system with any of the following method:

    • Enable the Spark overlay from eselect-repository:

      # emerge --ask --noreplace app-eselect/eselect-repository
      # eselect repository enable spark-overlay
    • Manually add the repository definition to /etc/portage/repos.conf:

      location = /var/db/repos/spark-overlay
      sync-type = git
      sync-uri =
  3. Download the contents of the Spark overlay to the system:

    # emerge --sync spark-overlay


The tests directory under the Spark overlay's Git tree contains scripts and test cases for installation tests of the ebuilds in the Spark overlay. These tests are automatically run by a GitHub Actions workflow once a day to capture issues preventing ebuilds in the Spark overlay from being installed caused by both problems in those ebuilds themselves and changes in the official Gentoo ebuild repository (like removal of packages needed by ebuilds in the Spark overlay).

Running the Tests Locally

The tests can be run in a local environment to replicate the jobs executed by GitHub Actions:

  1. Install ebuild-commander -- the tool used to run the installation test cases -- in the local environment. Please make sure every dependency required by ebuild-commander is installed too, because some of the dependencies are required during the runtime only, and missing runtime dependencies will not cause any errors in the installation process.

  2. Change working directory to the top directory of the Spark overlay, then:

    • Run tests/ to run all test cases stored under tests/test-cases.
    • Run tests/ TESTCASE... to run one or more specific TESTCASEs. For example:
      • tests/ tests/test-cases/ runs the test case for H2O packages.
      • tests/ tests/test-cases/ tests/test-cases/ runs the test case for Kotlin packages for the latest feature release and then the test case for H2O packages.


  • Some test cases can take hours to run.

  • The tests/ script may indirectly invoke Docker via ebuild-commander, which might require tests/ to be run as root and ebuild-commander to be installed globally for all users.

Working with the Test Cases

Test cases are stored in the tests/test-cases directory under the Spark overlay's Git tree. A test case's format is similar to an ebuild:

  • A test case is written in Bash syntax, which will allow tests/ to source it.

  • A test case needs to have a run_test Bash function whose body contains the list of commands to run in the Docker container created by ebuild-commander for the test case.

    • Please note that ebuild-commander executes each command in the run_test function in a separate shell, even if the commands are separated by a semicolon instead of a newline. This means that in run_test, variable declarations, if statements, for loops and so on do not work as expected. There exist some workarounds:

      • Put any parts of the test case involving variable declarations and control flows into standalone scripts, and call the scripts in run_test. The Spark overlay's Git tree will be available at /var/db/repos/spark-overlay in the Docker container created for the test.

      • Write variable values to files, and read those files to retrieve them later.

  • A test case may define additional Bash variables recognized by tests/ to control command-line options of ebuild-commander.

    • To find out which variables are supported, please refer to the content of tests/

Besides and test-cases, the tests directory contains some other sub-directories for files supporting test cases:

  • portage-configs: Portage configurations for test cases
    • default: The default configuration, used by all test cases
    • unstable: A configuration that accepts the ~arch keyword for all ebuilds, including those in ::gentoo
      • This configuration is enabled automatically by tests/ when the UNSTABLE environment variable's value is not empty.
    • features-test: A configuration that sets FEATURES="test" for ebuilds in the Spark overlay
    • binary: A configuration that sets USE="binary" for ebuilds in the Spark overlay
  • resources: Scripts, programs and other miscellaneous files used by test cases
    • This is the best location to place any scripts that need to be called in the run_test function for a test case. The path to this directory in the Docker container is /var/db/repos/spark-overlay/tests/resources.

See Also


The Spark overlay could still use some improvements. Any contributions that help resolve the following issues are welcome!

  • Update Apache Spark packages to the latest upstream releases
  • Update dev-java/hadoop-* and dev-java/netty-* packages to the latest upstream releases that do not have known security vulnerabilities
  • Expand the family of H2O packages by adding packages for the H2O XGBoost and AutoML extension
  • Improve Kotlin packages and eclasses
  • Fix other documented issues in the issue tracker


No description, website, or topics provided.






No releases published