Jenkins Pipeline Unit is a testing framework for unit testing Jenkins pipelines, written in Groovy Pipeline DSL.
If you use Jenkins as your CI workhorse (like us @ lesfurets.com) and you enjoy writing pipeline-as-code, you already know that pipeline code is very powerful but can get pretty complex.
This testing framework lets you write unit tests on the configuration and conditional logic of the pipeline code, by providing a mock execution of the pipeline. You can mock built-in Jenkins commands, job configurations, see the stacktrace of the whole execution and even track regressions.
Maven:
<dependency>
<groupId>com.lesfurets</groupId>
<artifactId>jenkins-pipeline-unit</artifactId>
<version>0.12</version>
<scope>test</scope>
</dependency>Gradle:
testCompile group:'com.lesfurets', name:'jenkins-pipeline-unit', version:'0.12'You can write your tests in Groovy or Java 8, using the test framework you prefer.
The easiest entry point is extending the abstract class BasePipelineTest, which initializes the framework with JUnit.
Let's say you wrote this awesome pipeline script, which builds and tests your project :
def execute() {
node() {
def utils = load "src/main/jenkins/lib/utils.jenkins"
stage('Checkout') {
checkout scm
String revision = utils.currentRevision()
gitlabBuilds(builds: ["build", "test"]) {
stage("build") {
gitlabCommitStatus("build") {
sh "mvn clean package -DskipTests -DgitRevision=$revision"
}
}
stage("test") {
gitlabCommitStatus("test") {
sh "mvn verify -DgitRevision=$revision"
}
}
}
}
}
}
return thisNow using the Jenkins Pipeline Unit you can unit test if it does the job :
import com.lesfurets.jenkins.helpers.BasePipelineTest
class TestExampleJob extends BasePipelineTest {
//...
@Test
void should_execute_without_errors() throws Exception {
def script = loadScript("job/exampleJob.jenkins")
script.execute()
printCallStack()
}
}
This test will print the call stack of the execution :
exampleJob.run()
exampleJob.execute()
exampleJob.node(groovy.lang.Closure)
exampleJob.load(src/main/jenkins/lib/utils.jenkins)
utils.run()
exampleJob.stage(Checkout, groovy.lang.Closure)
exampleJob.checkout({$class=GitSCM, branches=[{name=feature_test}]})
utils.currentRevision()
utils.sh({returnStdout=true, script=git rev-parse HEAD})
exampleJob.gitlabBuilds({builds=[build, test]}, groovy.lang.Closure)
exampleJob.stage(build, groovy.lang.Closure)
exampleJob.gitlabCommitStatus(build, groovy.lang.Closure)
exampleJob.sh(mvn clean package -DskipTests -DgitRevision=bcc19744fc4876848f3a21aefc92960ea4c716cf)
exampleJob.stage(test, groovy.lang.Closure)
exampleJob.gitlabCommitStatus(test, groovy.lang.Closure)
exampleJob.sh(mvn verify -DgitRevision=bcc19744fc4876848f3a21aefc92960ea4c716cf)
You can register interceptors to mock Jenkins commands, which may or may not return a result.
@Override
@Before
void setUp() throws Exception {
super.setUp()
helper.registerAllowedMethod("sh", [Map.class], {c -> "bcc19744fc4876848f3a21aefc92960ea4c716cf"})
helper.registerAllowedMethod("timeout", [Map.class, Closure.class], null)
helper.registerAllowedMethod(method("readFile", String.class), { file ->
return Files.contentOf(new File(file), Charset.forName("UTF-8"))
})
}The test helper already includes some mocks, but the list is far from complete. You need to register allowed methods if you want to override these mocks and add others. Note that you need to provide a method signature and a callback (closure or lambda) in order to allow a method. Any method call which is not recognized will throw an exception.
You can take a look at the BasePipelineTest class to have the short list of allowed methods.
Some tricky methods such as load and parallel are implemented directly in the helper.
If you want to override those, make sure that you extend the PipelineTestHelper class.
The helper registers every method call to provide a stacktrace of the mock execution.
@Test
void should_execute_without_errors() throws Exception {
loadScript("Jenkinsfile")
assertThat(helper.callStack.findAll { call ->
call.methodName == "sh"
}.any { call ->
callArgsToString(call).contains("mvn verify")
}).isTrue()
assertJobStatusSuccess()
}
This will check as well mvn verify has been called during the job execution.
One other use of the callstacks is to check your pipeline executions for possible regressions.
You have a dedicated method you can call if you extend BaseRegressionTest:
@Test
void testNonReg() throws Exception {
def script = loadScript("job/exampleJob.jenkins")
script.execute()
super.testNonRegression("example", false)
}This will compare the current callstack of the job to the one you have in a text callstack reference file.
To overwrite this file with new callstack, just set the updateReference to true when calling testNonRegression:
super.testNonRegression("example", true)You then can go ahead and commit this change in your SCM to check in the change.
The abstract class BasePipelineTest configures the helper with useful conventions:
- It looks for pipeline scripts in your project in root (
./.) andsrc/main/jenkinspaths. - Jenkins pipelines let you load other scripts from a parent script with
loadcommand. Howeverloadtakes the full path relative to the project root. The test helper mock successfully theloadcommand to load the scripts. To make relative paths work, you need to configure the path of the project where your pipeline scripts are, which defaults to.. - Pipeline script extension, which defaults to jenkins (matches any
*.jenkinsfile)
Overriding these default values is easy:
class TestExampleJob extends BasePipelineTest {
@Override
@Before
void setUp() throws Exception {
helper.baseScriptRoot = 'jenkinsJobs'
helper.roots += 'src/main/groovy'
helper.extension = 'pipeline'
super.setUp()
}
}
This will work fine for such a project structure:
jenkinsJobs
└── src
├── main
│ └── groovy
│ └── ExampleJob.pipeline
└── test
└── groovy
└── TestExampleJob.groovy
With Shared Libraries Jenkins lets you share common code
on pipelines across different repositories of your organization.
Shared libraries are configured via a settings interface in Jenkins and imported
with @Library annotation in your scripts.
Testing pipeline scripts using external libraries is not trivial because the shared library code is checked in another repository. JenkinsPipelineUnit lets you test shared libraries and pipelines depending on these libraries.
Here is an example pipeline using a shared library:
@Library('commons')
import net.courtanet.jenkins.Utils
sayHello 'World'
node() {
stage ('Checkout') {
def utils = new Utils()
checkout "${utils.gitTools()}"
}
stage ('Build') {
sh './gradlew build'
}
stage ('Post Build') {
String json = libraryResource 'net/courtanet/jenkins/request.json'
sh "curl -H 'Content-Type: application/json' -X POST -d '$json' ${acme.url}"
}
}This pipeline is using a shared library called commons.
Now lets test it:
String clonePath = 'path/to/clone'
def library = library()
.name('commons')
.retriever(gitSource('git@gitlab.admin.courtanet.net:devteam/lesfurets-jenkins-shared.git'))
.targetPath(clonePath)
.defaultVersion("master")
.allowOverride(true)
.implicit(false)
.build()
helper.registerSharedLibrary(library)
loadScript("job/library/exampleJob.jenkins")
printCallStack()Notice how we defined the shared library and registered it to the helper. Library definition is done via a fluent API which lets you set the same configurations as in Jenkins Global Pipeline Libraries.
The retriever and targetPath fields tell the framework how to fetch the sources of the library, in which local path.
The framework comes with two naive but useful retrievers, gitSource and localSource.
You can write your own retriever by implementing the SourceRetriever interface.
Note that properties defaultVersion, allowOverride and implicit are optional with
default values master, true and false.
Now if we execute this test, the framework will fetch the sources from the Git repository and load classes, scripts, global variables and resources found in the library. The callstack of this execution will look like the following:
Loading shared library commons with version master
libraryJob.run()
libraryJob.sayHello(World)
sayHello.echo(Hello, World.)
libraryJob.node(groovy.lang.Closure)
libraryJob.stage(Checkout, groovy.lang.Closure)
Utils.gitTools()
libraryJob.checkout({branch=master})
libraryJob.stage(Build, groovy.lang.Closure)
libraryJob.sh(./gradlew build)
libraryJob.stage(Post Build, groovy.lang.Closure)
libraryJob.libraryResource(net/courtanet/jenkins/request.json)
libraryJob.sh(curl -H 'Content-Type: application/json' -X POST -d '{"name" : "Ben"}' http://acme.com)
If you already fiddled with Jenkins pipeline DSL, you experienced strange errors during execution on Jenkins. This is because Jenkins does not directly execute your pipeline in Groovy, but transforms the pipeline code into an intermediate format to in order to run Groovy code in Continuation Passing Style (CPS).
The usual errors are partly due to the the sandboxing Jenkins applies for security reasons, and partly due to the serializability Jenkins imposes.
Jenkins requires that at each execution step, the whole script context is serializable, in order to stop and resume the job execution. To simulate this aspect, CPS versions of the helpers transform your scripts into the CPS format and check if at each step your script context is serializable.
To use this experimental feature, you can use the abstract class BasePipelineTestCPS instead of BasePipelineTest.
You may see some changes in the call stacks that the helper registers.
Note also that the serialization used to test is not the same as what Jenkins uses.
You may find some incoherence in that level.
JenkinsPipelineUnit aims to help devops code and test Jenkins pipelines with a shorter development cycle. It addresses some of the requirements traced in JENKINS-33925. If you are willing to contribute please don't hesitate to discuss in issues and open a pull-request.