As BME grows more complex, the need for unit tests arises. While there are some frameworks for testing bash/shell scripts, I preferred building one customized for this project's needs (this may change in the future).
- Tests should be files matching the 'test_*.sh' pattern and should have the execution bit set (i.e.:
chmod ug+x). - Tests are run by the maketests.sh script.
- maketests.sh script without parameters will run all tests it finds under the 'tests/' directory and subdirectories within.
- maketests.sh also accepts paths (either absolute or relative to current directory) and in that case it will run tests found within those.
i.e.:./maketests.sh coreor./maketests.sh ../../alternate_dir/modules/something/test_sometest.sh core. - maketests.sh runs each script on a clean subshell with helper functions and some environment variables:
- helper functions: the helper functions file is sourced into the tests' environment so its included functions can be used by scripts. Review its contents.
- environment variables: only those needed to launch the test scripts.
- PATH: PATH is mangled so it includes the directory where BME code can be found. That means that in order to load BME's entrypoint you just need to
source bash-magic-enviro(just like it's expected for end-users to do on their Bash environments). - HOME: an empty directory that the tests may use (i.e.: it can mimic a user's home directory under which you can set a test project, etc.). This scratch dir is deleted after each test unless it fails; in that case it is preserved for diagnosis.
- CURRENT_TESTFILE_NUMBER: "internal" counter that helps tests' presentation output from the test_name() helper function. Do not modify.
- PATH: PATH is mangled so it includes the directory where BME code can be found. That means that in order to load BME's entrypoint you just need to
Each test script runs on a clean subshell (except for the environment variables described above) and its return code is checked: 0 for success, any other code for failure.
NOTES:
- version tests require internet access since it tries to compare local versions with highest published at GitHub.
- modules' tests: review each module requirements or else their related tests may fail.