The manage_externals package has an automated test suite. All pull requests are expected to pass 100% of the automated tests, as well as be pep8 and lint ‘clean’ and maintain approximately constant (at a minimum) level of code coverage.

Quick Start

Do nothing approach

When you create a pull request on GitHub, Travis-CI continuous integration testing will run the test suite in both python2 and python3. Test results, lint results, and code coverage results are available online.

Do something approach

In the test directory, run:

make env
make lint
make test
make coverage

Automated Testing

The manage_externals manic library and executables are developed to be python2 and python3 compatible using only the standard library. The test suites meet the same requirements. But additional tools are required to provide lint and code coverage metrics and generate documentation. The requirements are maintained in the requirements.txt file, and can be automatically installed into an isolated environment via Makefile.

Bootstrap requirements:

  • python2 - version 2.7.x or later

  • python3 - version 3.6 tested other versions may work

  • pip and virtualenv for python2 and python3

Note: all make rules can be of the form make python=pythonX rule or make rule depending if you want to use the default system python or specify a specific version.

The Makefile in the test directory has the following rules:

  • make python=pythonX env - create a python virtual environment for python2 or python3 and install all required packages. These packages are required to run lint or coverage.

  • make style - runs autopep8

  • make lint - runs autopep8 and pylint

  • make test - run the full test suite

  • make utest - run jus the unit tests

  • make stest - run jus the system integration tests

  • make coverage - run the full test suite through the code coverage tool and generate an html report.

  • make readme - automatically generate the README files.

  • make clean - remove editor and pyc files

  • make clobber - remove all generated test files, including virtual environments, coverage reports, and temporary test repository directories.

Unit Tests

Unit tests are probably not ‘true unit tests’ for the pedantic, but are pragmatic unit tests. They cover small practicle code blocks: functions, class methods, and groups of functions and class methods.

System Integration Tests

NOTE(bja, 2017-11) The systems integration tests currently do not include svn repositories.

The manage_externals package is extremely tedious and error prone to test manually.

Combinations that must be tested to ensure basic functionality are:

  • container repository pulling in simple externals

  • container repository pulling in mixed externals with sub-externals.

  • mixed repository acting as a container, pulling in simple externals and sub-externals

Automatic system tests are handled the same way manual testing is done:

  • clone a test repository

  • create an externals description file for the test

  • run the executable with the desired args

  • check the results

  • potentially modify the repo (checkout a different branch)

  • rerun and test

  • etc

The automated system stores small test repositories in the main repo by adding them as bare repositories. These repos are cloned via a subprocess call to git and manipulated during the tests.