A script for regression testing of command line executables that output text (E.g. compilers).

All it does is run small command line files that end in .case.sh, manage their outputs and detect changes. It's a very simple Python program but has proven to be useful. It was first written to test a compiler built for a university course.

1. Copy, submodule or otherwise get crystalmountain.py to be callable from a directory in your project
2. In such directory, create a cases subdirectory
3. Populate cases with *.case.sh files that run the program you want to test

And you are al set. The suite is operational. You can run crystalmountain.py -g to generate initial outputs. Which will be written to corresponding *.spec files.

Note that cases can also be in nested subdirectories

Assuming the above setup:

• crystalmountain.py: run the test suite. This will record if there are any changes in the outupts
• crystalmountain.py cases/my/test.case1.sh cases/my/test.case2.sh: run those two particular test cases
• crystalmountain.py -g: generate new outputs, will not override existing ones
• crystalmountain.py -s: show the diffs from the last run
• crystalmountain.py -l -f: list all the test cases that are failing

usage: crystalmountain.py [-h] [-p | -c | -s | -r | -g | -l] [-f | -n | -a] [CASE ...]

Run a regression test suite for liblsqecc

positional arguments:
  CASE                Specify one or more particular test cases

optional arguments:
  -h, --help          show this help message and exit
  -p, --displayonly   display the results of the execution without writing anything
  -c, --copyover      Instead of running, copy over the results from the previous run
  -s, --showdiffs     Show diffs after failed tests
  -r, --removeoutput  Removes outputs for this test
  -g, --generate      Generates tests if they don't exist yet, otherwise do nothing
  -l, --list          List available tests
  -f, --failing       Only consider cases that failed in the latest run
  -n, --nospec        Only consider cases that don't have a spec defined
  -a, --passing       Only consider cases that passed in the last run

The requirements are pretty minimal:

• A shell to run commands (E.g. bash or zsh)
• A Python >= 3.7 installation
• A diff command for rendering changes (Optional)

Only tested on Linux systems, but probably works on MAC WSL/GitBash.

Details of the license in the top of crystalmountain.py