shunit requires cmdarg (and the install script requires versioners). Install them first:
shunit is a bash script for running tests scripts that are written with the shunit library. To use it, first install it:
make install
... If you want to install it somewhere other than /usr (not recommended), you can use:
PREFIX=/some/other/path make install
... and it will install to /some/other/path/usr/...
... then run it with the name of a single bash test script or a whole directory of scripts:
shunit -t test_script.sh
shunit -t test_directory/
... The default output format is junit. You can select tunit with the '-f' flag:
shunit -t test_script.sh -f tunit
- In your project, create a tests/ directory. This directory will contain a number of .sh files. You can have any number of test functions in them, named "shunittest_**".
- Only functions in the *sh files named "shunittest_" will be executed.
- Doing anything other than "source" at the top level of your *sh test file is considered Very Bad Form (tm)
- We don't care what your shunittest_ functions do, except that they MUST:
- Exit 0 to indicate that the test has succeeded, - OR -
- Print something informative to stderr, and Exit 1 to indicate that the test failed
shunit is just a set of bash functions for producing unit test output from bash scripts. I wrote this library because I often found myself writing things in bash whose output I wanted to consume as discrete pass/fail tasks into Bamboo, so I wrote the junit functions. Then later on, I got sick and tired of (as a human) reading junit output, so I wrote the tunit functions, so my scripts could output tunit or junit depending on which flags I passed.
Complete documentation is inside the functions (each function has --help) in ./lib/*unit.sh. A pair of complete examples are in tests/*unit.sh.
akesterson@akesterson-pc ~/source/upstream/git/shunit
$ bash tests/tunit.sh
[super::class] someTest .... [OK]
[super::class] someOtherTest .... [FAILED]
generic failure : Yo dawg, I heard you like failures
This is some raw data, please read it
==== 2 TESTS in 0 SECONDS : 1 ERRORS, 1 FAILURES ====
akesterson@akesterson-pc ~/source/upstream/git/shunit
$ bash tests/junit.sh
<?xml version="1.0" encoding="UTF-8"?>
<testsuite failures="1" time="0" timestamp="Wed Dec 19 22:51:02 EST 2012" errors="1" tests="2">
<properties/>
<testcase classname="super::class" time="31337" name="someTest">
</testcase>
<testcase classname="super::class" time="31337" name="someOtherTest">
<failure type="generic failure" message="Yo dawg, I heard you like failures">
<![CDATA[
This is some raw data, please read it
]]>
</failure>
</testcase>
</testsuite>
This library should not be confused with shunit2 (http://shunit2.googlecode.com/svn/trunk/source/2.1/doc/shunit2.html). They are completely unrelated! I didn't even know shunit2 existed until long after I made this.
shunit2 seems neat, but I prefer shunit's style. shunit2 looks like it's trying to be a java-ish unit testing library, and doesn't operate very much like bash. shunit is much simpler; you don't need to know any assert commands, special constants, or how to report failure. You do your tests however you want, report error on stderr, return non-zero to indicate failure. It doesn't get much simpler than that!
The only really useful thing I can see in shunit2 vs shunit, I will admit, is the ability to skip tests. shunit does not currently know how to skip tests, and that is because I didn't fully understand junit when I wrote it, so it does not report on skipped tests even if it did skip them! I will add some mechanism for this at some point.