Buildpack Testrunner

A simple unit testing framework for testing buildpacks based on shUnit2. It provides utilities for loading buildpacks and capturing and asserting their behavior. It can be run locally, part of a continuous integration system, or even directly on Heroku as a buildpack itself. To run the testrunner locally, see the Setup and Usage sections. To run on Heroku, see the Running Buildpack Tests on Heroku section.

Setup

To use the testrunner locally, first clone this repository:

git clone git://github.com/ryanbrainard/heroku-buildpack-testrunner.git

If you do not already have shUnit2 installed, either download it or checkout it out from SVN:

svn checkout http://shunit2.googlecode.com/svn/trunk/ shunit2

Do not use apt-get for obtaining shUnit2 because it is the wrong version.

Once you have shUnit2, set an SHUNIT_HOME environment variable to the root of the version you wish to use. For example:

export SHUNIT_HOME=/usr/local/bin/shunit/source/2.1

Usage

To run the tests for one or more buildpacks, execute:

bin/run [-c] [-s single_suite_test.sh] buildpack_1 [buildpack_2 [...]]

where buildpack_n can either be a local directory or a remote Git repository ending in .git. Each buildpack must have a test directory and files matching the *_test.sh pattern to be run. The -s flag sets a single test suite to run in the test directories of the buildpacks. The -c flag enables persistent caching of files downloaded with cUrl. See lib/magic_curl/README.md for more info.

For example, the following command:

bin/run ~/a_local_buildpack [email protected]:ryanbrainard/heroku-buildpack-gradle.git

Would first run the tests in the buildpack at ~/a_local_buildpack and then clone the Git repository at [email protected]:rbrainard/heroku-buildpack-gradle.git into a temp directory and run the tests there too.

Running Buildpack Tests on Heroku

The testrunner is itself a buildpack and can be used to run tests for your buildpack on Heroku. This can be very helpful for testing your buildpack on a real Heroku dyno before pushing it to a public repo. To do this, create a Cedar app out of your buildpack and set the testrunner as its buildpack:

$ cd your_buildpack_dir
$ heroku create --stack cedar --buildpack git://github.com/ryanbrainard/heroku-buildpack-testrunner.git

Creating deep-thought-1234... done, stack is cedar
http://deep-thought-1234.herokuapp.com/ | [email protected]:deep-thought-1234.git
Git remote heroku added

Once the testrunner is set as your buildpack's buildpack, push it to Heroku. This will automatically download and install shUnit2 and create a tests process for you:

$ git push heroku master

Counting objects: 425, done.
Delta compression using up to 8 threads.
Compressing objects: 100% (271/271), done.
Writing objects: 100% (425/425), 48.08 KiB, done.
Total 425 (delta 126), reused 396 (delta 113)

-----> Heroku receiving push
-----> Fetching custom buildpack... done
-----> Buildpack Test app detected
-----> Downloading shunit2-2.1.6..... done
-----> Installing shunit2-2.1.6.... done
-----> Installing Buildpack Testrunner.... done
-----> Discovering process types
   Procfile declares types          -> (none)
   Default types for Buildpack Test -> tests
-----> Compiled slug size is 108K
-----> Launching... done, v5
   http://deep-thought-1234.herokuapp.com deployed to Heroku

Now, you can run your tests on Heroku in their own dyno:

heroku run tests

If you would like caching to be enabled run tests-with-caching instead. Note, the cache will only live for the life of the dyno (i.e. one test run).

herouk run tests-with-caching

When running tests on a dyno, the exit code is not returned correctly to the local shell. To workaround this limitation, pipe the output to the report script, which will parse the output and return the correct exit code. For example:

heroku run tests | bin/report

Writing Unit Tests for a Buildpack

Writing tests for a buildpack is similar to any other xUnit framework, but the steps below summarize what you need to get started testing a buildpack. In addition, to the steps below, its advised to familarize yourself with the shUnit2 documentation before starting.

1. Create a test directory in the root of the buildpack.

2. Create test scripts in the test directory ending in _test.sh. They can be grouped any way you like, but creating a test script for each buildpack script is recommended. For example the detect script should have a corresponding detect_test.sh test script.

3. It is recommended (but not required) to source in the test_utils.sh script at the beginning of your test script. This contains common functions for setup, teardown, and asserting buildpack behavior.

   . ${BUILDPACK_TEST_RUNNER_HOME}/lib/test_utils.sh

4. Each test case in the script should be contained a function starting with test. Like testing with other xUnit frameworks, the test cases should be fairly granular and try not to depend on outside factors or upon each other.

If you are using test_util.sh, at the beginning of each test case, you will be provided empty ${BUILD_DIR} and ${CACHE_DIR} directories for use with buildpack scripts. These directories are deleted after each test case completes. You will also be provided a ${BUILDPACK_HOME} value to deterministically find the root of your buildpack.

When running buildpack scripts, it is recommended to use the detect, compile, and release functions from test_utils.sh, which will provide the correct parameters and capture the stdout, stderr, and return values of the scripts. If you need to manually capture a command, the capture function is also available to you by just calling capture before your command, but use the pre-defined functions whenever possible. Either way you capture, you will then have access to the ${STD_OUT} file, ${STD_ERR} file, and ${RETURN} value after the capture completes. To inspect these files and values, there are a few helpful assertions:

• assertCapturedSuccess: captured command exited with 0 and stderr is empty
• assertCapturedError [[expectedErrorCode] expectedValue]: captured command exited with non-0 value (or optional specified error code), stderr is empty, and stdout contains expected value
• assertCaptured [[assertionMessage] expectedValue]: captured stdout contains an expected value
• assertNotCaptured [[assertionMessage] expectedValue]: captured stdout does not contain an expected value
• assertCapturedEquals [[assertionMessage] expectedValue]: captured stdout exactly equals the expected value
• assertCapturedNotEquals [[assertionMessage] expectedValue]: captured stdout does not exactly equals the expected value
• assertAppDetected appName: stdout only contains app name
• assertNoAppDetected: stdout only contains "no"

For example, to test that compile completes successfully and contains something in the logs:

compile
assertCapturedSuccess
assertCaptured "A string that should be in the output"

An example of asserting an error:

compile
assertCapturedError "An error message we're expecting"

Manually capturing is also available, which can be helpful when debugging tests, but is generally not needed. Use the assertions above whenever possible:

capture ${BUILDPACK_HOME}/bin/compile ${BUILD_DIR} ${CACHE_DIR} 

Manually asserting on the raw captured values is also available, but is generally not needed. Use the assertions above whenever possible:

assertEquals 0 "${RETURN}"
assertContains "expected output" "$(cat ${STD_OUT})"
assertEquals "" "$(cat ${STD_ERR})"

All captured data is cleared betweeen test cases and before every capture.

If you are downloading files in tests, it is highly recommended to use

assertFileMD5 expectedHash filename 

to make sure you actually downloaded the correct file. This assertion is more portable between platforms rather than computing the MD5 yourself.

In addition, please see the shUnit2 documentation for information on additional asssertions available.

Metatesting

The tests for the testrunner itself work just like any other buildpack. To test the testrunner itself, just run:

bin/run .

This can be helpful to make sure all the testrunner libraries work on your platform before testing any real buildpacks.

One caveat about negative tests for assertions is that they need to be captured and wrapped in paraenthesis to supress the assertion failure from causing the metatest to fail. For example, if you want to test that assertContains prints out the proper failure message, capture, wrap, and then assert on the captured output.

( capture assertContains "xxx" "zookeeper" )
assertEquals "ASSERT:Expected <zookeeper> to contain <xxx>" "`cat ${STD_OUT}`"