Skip to content
Branch: master
Commits on Apr 16, 2019
  1. Add ability to configure separate host/port for FDW connection strings,

    milesrichardson committed Apr 16, 2019
    which might be different from one engine to another, vs. from sgr client
    to a foreign engine
Commits on Apr 2, 2019
  1. Add branch coverage to tests, fix (#52)

    milesrichardson committed Apr 2, 2019
    * Add branch coverage to tests, fix, "fix" flaky test
    Add branch coverage to tests:
    - Collect and report branch coverage when running tests, whether from
      within docker container (where tests are configured with,
      or on the host (where tests are configured with setup.cfg)
    - Fix bug in entrypoint of sgr dev container where pycache files were cleaned
      only after installing sgr, when they should be cleaned before (and might as
      well be cleaned after, too, so as not to cause the same problem when
      importing later from the host). Point is, with bind mounting, we don't
      know at any given time where the pycache (+ .pyc) files came from -- the
      host or the container -- so they could be wrong at any given time, and
      thus we err on the side of caution by clearing them aggressively.
    - Add down -v directive to to make sure tests all start from
      fresh volumes (e.g. fix problem where existing volume was created with pg10
      and new engine is pg11 complaining about incompatible format)
    "Fix" Flaky test:
    - Increase timeout by 25 seconds in from 50 to 75
Commits on Apr 1, 2019
  1. Merge pull request #47 from splitgraph/dev/ci-cleanup

    milesrichardson committed Apr 1, 2019
    CI Cleanup: Refactor multiline travis blocks into scripts,
  2. Add ./ file for running tests from fresh state on dev ma…

    milesrichardson committed Mar 28, 2019
    Run a "clean" test. Intended for running in local development to give a close
    simulation of the test suite running in travis. Also accounts for the
    possibility that some e.g. splitgraphci images are outdated.
      - Destroy and recreate docker environment (using down -v)
      - Run tests from within container
      - Exit 0 on all tests pass
      - Exit 1 otherwise
    Give up on caching for now. Well, we can keep the pip cache, it
    saves a couple seconds.
  3. Cleanup travis.yml, add .ci/*.sh, refactor CI image docker-compose

    milesrichardson committed Mar 28, 2019
    - Replace multi-line directives in travis with .sh scripts that exit 0
      or 1 located in .ci. If one step fails, the whole script fails.
    - Directives turned into scripts:
      - (create single binary, configure pypi)
      - (install docker-compose and poetry)
      - (clear env, start test architecture, wait for it)
    - Add entries to /etc/hosts manually in rather than via travis
    Separate the CI images into their own file
    - Ideally, travis should be able to run docker-compose build, and not
      need to specify every individual service that is *not* a CI image
      (because they take so long to build)
    - So, take the CI images out of the main docker-compose.yml file, so that
      when travis (or the user) runs docker-compose build, it does not build
      the CI containers
    - Also change and to keep up with
      the change
  4. Refactor

    milesrichardson committed Mar 28, 2019
    - Move logic to file test/architecture/, and call
      test/architecture/ from $ROOT/
    - Suppress error output while waiting for architecture (we know there will be failures,
      no need to spam the logs)
    - Be more explicit about the target of sgr init for local_engine
    - Implement various bashisms and best practices
You can’t perform that action at this time.