SeamFramework.orgCommunity Documentation

Technology Compatibility Kit Reference Guide for JSR-299: Context and Dependency Injection for Java EE

Specification Lead: Red Hat Inc.

1. Who Should Use This Book
2. Before You Read This Book
3. How This Book Is Organized
I. Getting Acquainted with the TCK
1. Introduction (CDI TCK)
1.1. TCK Primer
1.2. Compatibility Testing
1.2.1. Why Compatibility Is Important
1.3. About the CDI TCK
1.3.1. CDI TCK Specifications and Requirements
1.3.2. CDI TCK Components
2. Appeals Process
2.1. Who can make challenges to the TCK?
2.2. What challenges to the TCK may be submitted?
2.3. How these challenges are submitted?
2.4. How and by whom challenges are addressed?
2.5. How accepted challenges to the TCK are managed?
3. Installation
3.1. Obtaining the Software
3.2. The TCK Environment
3.3. Eclipse Plugins
3.3.1. TestNG Plugin
3.3.2. Maven 2 Plugin (m2eclipse)
4. Configuration
4.1. TCK Harness Properties
4.2. The Porting Package
4.3. Using the CDI TCK with the Java EE Web Profile
4.4. Using custom deployment descriptors
4.5. Configuring TestNG to execute the TCK
4.6. Configuring your build environment to execute the TCK
4.7. Configuring your application server to execute the TCK
5. Reporting
5.1. CDI TCK Coverage Metrics
5.2. CDI TCK Coverage Report
5.2.1. CDI TCK Assertions
5.2.2. Producing the Coverage Report
5.2.3. TestNG Reports
II. Executing and Debugging Tests
6. Running the Signature Test
6.1. Obtaining the sigtest tool
6.2. Running the signature test
6.3. Forcing a signature test failure
7. Executing the Test Suite
7.1. The Test Suite Runner
7.2. Running the Tests In Standalone Mode
7.3. Running the Tests In the Container
7.4. Dumping the Test Artifacts
8. Running Tests in Eclipse
8.1. Leveraging Eclipse's plugin ecosystem
8.2. Readying the Eclipse workspace
8.3. Running a test in standalone mode
8.4. Running integration tests
9. Debugging Tests in Eclipse
9.1. Debugging a standalone test
9.2. Debugging an integration test
9.2.1. Attaching the IDE debugger to the container
9.2.2. Launching the test in the debugger
III. JBoss Test Harness
10. Introduction (JBoss Test Harness)
10.1. Negotiating the execution of an in-container test
11. Configuration
11.1. JBoss Test Harness Properties
12. Executing a Test Suite
12.1. Building a test suite runner using Maven 2
12.2. Dumping the Test Artifacts to Disk

This guide describes how to download, install, configure, and run the Technology Compatibility Kit (TCK) used to verify the compatibility of an implementation of the JSR-299: Context and Dependency Injection for Java EE (CDI) specification.

The CDI TCK is built atop the JBoss Test Harness, a portable and configurable automated test suite for authoring unit and integration tests in a Java EE environment. The CDI TCK 1.0.0 uses the JBoss Test Harness version 1.0.0 to execute the test suite.

The CDI TCK is provide under the Apache Public License 2.0.

Before reading this guide, you should familiarize yourself with the Java EE programming model, specifically the Enterprise JavaBeans (EJB) 3.0 and the Contexts and Dependency Injection for Java EE 1.0 specifications. A good resource for the Java EE programming model is the JCP web site.

The CDI TCK is based on the Context and Dependency Injection for Java EE technology specification 1.0 (JSR-299). Information about the specification, including links to the specification documents, can be found on the JSR-299 JCP page.

Before running the tests in the CDI TCK, read and become familiar with the JBoss Test Harness Reference Guide (pending), which describes how the test harness functions.

If you are running the CDI TCK for the first time, read Chapter 1, Introduction (CDI TCK) and Chapter 10, Introduction (JBoss Test Harness) completely for the necessary background information about the TCK and the JBoss Test Harness, respectively. Once you have reviewed that material, perform the steps outlined in the remaining chapters.

  • Chapter 1, Introduction (CDI TCK) gives an overview of the principles that apply generally to all Technology Compatibility Kits (TCKs), outlines the appeals process and describes the CDI TCK architecture and components. It also includes a broad overview of how the TCK is executed and lists the platforms on which the TCK has been tested and verified.

  • Chapter 2, Appeals Process explains the process to be followed by an implementor should they wish to challenge any test in the TCK.

  • Chapter 3, Installation explains where to obtain the required software for the CDI TCK and how to install it. It covers both the primary TCK components as well as tools useful for troubleshooting tests.

  • Chapter 4, Configuration details the configuration of the JBoss Test Harness, how to create a TCK runner for the TCK test suite and the mechanics of how an in-container test is conducted.

  • Chapter 5, Reporting explains the test reports that are generated by the TCK test suite and introduces the TCK audit report as a tool for measuring the completeness of the TCK in testing the JSR-299 specification and in understanding how testcases relate to the specification.

  • Chapter 7, Executing the Test Suite documents how the TCK test suite is executed. It covers both modes supported by the TCK, standalone and in-container, and shows how to dump the generated test artifacts to disk.

  • Chapter 8, Running Tests in Eclipse shows how to run individual tests in Eclipse and advises the best way to setup your Eclipse workspace for running the tests.

  • Chapter 9, Debugging Tests in Eclipse builds on Chapter 8, Running Tests in Eclipse by detailing how to debug individual tests in Eclipse.

  • Part III, “JBoss Test Harness” includes excerpts from the JBoss Test Harness Reference Guide. How to configure the JBoss Test Harness as it relates to the CDI TCK is presented in Chapter 4, Configuration. However, to aid in debugging or configuring the TCK in your environment, you may want to read in more detail how to use the JBoss Test Harness.

The CDI TCK must be used to ensure that your implementation conforms to the CDI specification. This part introduces the TCK, gives some background about its purpose, states the requirements for passing the TCK and outlines the appeals process.

In this part you will learn where to obtain the CDI TCK and supporting software. You are then presented with recommendations of how to organize and configure the software so that you are ready to execute the TCK.

Finally, it discusses the reporting provided by the TCK.

This chapter explains the purpose of a TCK and identifies the foundation elements of the CDI TCK.

A TCK, or Technology Compatibility Kit, is one of the three required pieces for any JSR (the other two being the specification document and the reference implementation). The TCK is a set of tools and tests to verify that an implementation of the technology conforms to the specification. The tests are the primary component, but the tools serve an equally critical role of providing a framework and/or set of SPIs for executing the tests.

The tests in the TCK are derived from assertions in the written specification document. The assertions are itemized in an XML document, where they each get assigned a unique identifier, and materialize as a suite of automated tests that collectively validate whether an implementation complies with the aforementioned assertions, and in turn the specification. For a particular implementation to be certified, all of the required tests must pass (i.e., the provided test suite must be run unmodified).

A TCK is entirely implementation agnostic. Ideally, it should validate assertions by consulting the specficiation's public API. However, when the information returned by the public API is not low-level enough to validate the assertion, the implementation must be consulted directly. In this case, the TCK provides an independent API as part of a porting package that enables this transparency. The porting package must be implemented for each CDI implementation. Section 1.3.4 introduces the porting package and section 4.1 covers the requirements for implementing it.

The goal of any specification is to eliminate portability problems so long as the program which uses the implementation also conforms to the rules laid out in the specification.

Executing the TCK is a form of compatibility testing. It's important to understand that compatibility testing is distinctly different from product testing. The TCK is not concerned with robustness, performance or ease of use, and therefore cannot vouch for how well an implementation meets these criteria. What a TCK can do is to ensure the exactness of an implementation as it relates to the specification.

Compatibility testing of any feature relies on both a complete specification and a complete reference implementation. The reference implementation demonstrates how each test can be passed and provides additional context to the implementor during development for the corresponding assertion.

The CDI TCK is designed as a portable, configurable and automated test suite for verifying the compatibility of an implementation of the JSR-299: Contexts and Dependency Injection for Java EE specification. The test suite is built atop TestNG and provides a series of extensions that allow runtime packaging and deployment of JEE artifacts for in-container testing (JBoss Test Harness).

Each test class in the suite acts as a deployable unit. The deployable units, or artifacts, are defined declaratively using annotations. The artifact produced can be either a WAR or an EAR.

The declarative approach allows many of the tests to be executed in a standalone implementation of CDI, accounting for a boast in developer productivity. However, an implementation is only valid if all tests pass using the in-container execution mode. The standalone mode is merely a developer convenience.

This section lists the applicable requirements and specifications for the CDI TCK.

The CDI TCK includes the following components:

The CDI TCK has been tested run on following platforms:

CDI supports Java EE 5, Java EE 6, Java EE 6 Web Profile and the Embeddable EJB 3.1. The TCK will execute on any of these runtimes, but is only part of the CTS for Java EE 6 and Java EE 6 Web Profile.

While the CDI TCK is rigourous about enforcing an implementation's conformance to the JSR-299 specification, it's reasonable to assume that an implementor may discover new and/or better ways to validate the assertions. This chapter covers the appeals process, defined by the Specification Lead, Red Hat Middleware LLC., which allows implementors of the JSR-299 specification to challenge one or more tests defined by the CDI TCK.

The appeals process identifies who can make challenges to the TCK, what challenges to the TCK may be submitted, how these challenges are submitted, how and by whom challenges are addressed and how accepted challenges to the TCK are managed.

Following the recent adoption of transparency in the JCP, implementors are encouraged to make their appeals public, which this process facilitates. The JCP community should recognize that issue reports are a central aspect of any good software and it's only natural to point out shortcomings and strive to make improvements. Despite this good faith, not all implementors will be comfortable with a public appeals process. Instructions about how to make a private appeal are therefore provided.

To submit a challenge, a new issue should be created in the CDITCK project of the JBoss JIRA using the Issue Type: Bug. The appellant should complete the Summary, Component (TCK Appeal), Environment and Description Field only. Any communication regarding the issue should be pursed in the comments of the filed issue for accurate record.

To submit an issue in the JBoss JIRA, you must have a (free) member account. You can create a member account using the on-line registration.

If you wish to make a private challenge, you should follow the above procedure, setting the Security Level to Private. Only the issue reporter, TCK Project Lead and designates will be able to view the issue.

This chapter explains how to obtain the TCK and supporting software and provides recommendations for how to install/extract it on your system.

You can obtain a release of the CDI TCK project from the from the download page on Seam Framework website. The CDI TCK is distributed as a ZIP file, which contains the TCK artifacts (the test suite binary and source, porting package API binary and source, the test suite descriptor, the audit source and report) in /artifacts, the TCK library dependencies in /lib and documentation in /lib.

You can also download the current source code from JBoss SVN repository.

The TCK project is available in the JBoss Maven 2 repository as org.jboss.jsr299.tck:jsr299-tck-impl; the POM defines all dependencies required to run the TCK.

Executing the TCK requires a Java EE 5 or better runtime environment (i.e., application server), to which the test artifacts are deployed and the individual tests are invoked. The TCK does not depend on any particular Java EE implementation.

The JSR-299: Contexts and Dependency Injection for Java EE 1.0 reference implementation (RI) project is named Weld. You can obtain the latest release from the download page on Seam Framework website.


Weld is not required for running the CDI TCK, but it can be used as a reference for familiarizing yourself with the TCK before testing your own CDI implementation.

Naturally, to execute Java programs, you must have a Java SE runtime environment. The TCK requires Java 5 or better, which you can obtain from the Java Software website.

The TCK requires the following two Java runtime environments:

You should refer to vendor instructions for how to install the runtime.

The rest of the TCK software can simply be extracted. It's recommended that you create a folder named jsr299 to hold all of the jsr299-related projects. Then, extract the TCK distribution into a subfolder named tck. If you have downloaded the Weld distribution, extract it into a sibling folder named weld. The resulting folder structure is shown here:


Each test class is treated as an individual artifact (hence the @Artifact annotation on the class). All test methods (i.e., methods annotated with @Test) in the test class are run in the application, meaning bean discovery occurs exactly once per artifact and the same BeanManager is used by each test method in the class.

Running the TCK against the CDI RI (Weld) and JBoss AS

JBoss AS 6.0.0 M1 and above releases bundle Weld. To run the TCK with JBoss AS, you must use JBoss AS 6.0.0.M1 or above.

To install JBoss AS 6:

  • First, you should download JBoss AS 6 from the JBoss AS project page.

  • Set the JBOSS_HOME environment variable to the location of the JBoss AS software.

  • Change to the weld directory in the TCK.

  • Create a file called in weld/jboss-as and make sure the jboss.home property references a JBoss AS 5.2 installation:


The CDI TCK includes a TCK runner that executes the TCK using Weld as the CDI implementation and JBoss AS as the Java EE runtime. To run the tck:

  • You need to install Maven 2.0.10. You can find documention on how to install Maven 2 in the Maven: The Definitive Guide book published by Sonatype.

  • Next, instruct Maven to run the TCK:

    cd weld/jboss-tck-runner
    mvn test -Dincontainer
  • TestNG will report, via Maven, the outcome of the run, and report any failures on the console. Details can be found in target/surefire-reports/TestSuite.txt.

Eclipse, or any other IDE, is not required to execute or pass the TCK. However an implementor may wish to execute tests in an IDE to aid debugging the tests. This section introduces two essential Eclipse plugins, TestNG and Maven 2, and points you to resources explaining how to install them.

Another useful plugin is m2eclipse. Both the TCK project and are use Maven 2. Therefore, to work with these projects in Eclipse, you may wish to have native support for Maven 2 projects, which the m2eclipse plugin provides. Instructions for using the m2eclipse update site to add the m2eclipse plugin to Eclipse are provided on the m2eclipse home page. For more, read the m2eclipse reference guide.

m2eclipse is still a rather young project dealing with a complex domain and you may run into problems using it. If that is the case, you can alternatively use the Eclipse plugin for Maven 2 to generate native Eclipse projects from Maven 2 projects.

If you have Maven 2 installed, you have everything you need. Just execute the following command from any Maven 2 project to produce the Eclipse project files.

mvn eclipse:eclipse

Again, the Eclipse plugins are not required to execute the TCK, but can be very helpful when validating an implementation against the TCK test suite and especially when using the modules from the project.

This chapter lays out how to configure the TCK Harness by specifying the SPI implementation classes, defining the target container connection information, and various other switches. You then learn how to setup a TCK runner project that executes the the TCK test suite, putting these settings into practice. Finally, a detailed account of how the JBoss Test Harness negotiates the execution of the tests in the container is given.

This chapter does not discuss in detail how to use the TCK in standalone mode. The JBoss Test Harness guide provides more on running in standalone mode.

The JBoss Test Harness allows the test suite to be launched in a pluggable fashion. In order to execute the TCK, the JBoss Test Harness must be configured by specifying implementations of the test launcher and container APIs.

System properties and/or the resource META-INF/, a Java properties file, are used to configure the JBoss Test Harness. You can read more about configuring the JBoss Test Harness in Section 11.1, “JBoss Test Harness Properties”.

You should set the following properties:

To run the full TCK you must additionally implement org.jboss.testharness.spi.Containers, which handles deploying the test artifact to the container. An implementations of this API is already available for JBoss AS 5.2. Therefore, you only need to implement this part of the porting package if you wish to use another container.

The JBoss Test Harness is built atop TestNG, and it's TestNG that is responsible for selecting the tests to execute, the order of execution, and reporting the results. Detailed TestNG documentation can be found at

The tck-tests.xml artifact provided in the TCK distribution must be run by TestNG 5.9 (described by the TestNG documentation as "with a testng.xml file") unmodified for an implementation to pass the TCK. This file also allows tests to be excluded from a run:

<suite name="JSR-299 TCK" verbose="2">
   <test name="JSR-299 TCK">
         <class name="org.jboss.jsr299.tck.tests.context.application.ApplicationContextTest">
               <exclude name="testApplicationScopeActiveDuringServiceMethod"/>

TestNG provides extensive reporting information. Depending on the build tool or IDE you use, the reporting will take a different format. Please consult the TestNG documentation and the tool documentation for more information.

This chapter covers the two types of reports that can be generated from the TCK, an assertion coverage report and the test execution results. The chapter also justifies why the TCK is good indicator of how accurately an implementation conforms to the JSR-299 specification.

A specification can be distilled into a collection of assertions that define the behavior of the software. This section introduces the CDI TCK coverage report, which documents the relationship between the assertions that have been identified in the JSR-299 specification document and the tests in the TCK test suite.

The structure of this report is controlled by the assertion document, so we'll start there.

The CDI TCK developers have analyzed the JSR-299 specification document and identified the assertions that are present in each chapter. Here's an example of one such assertion found in section 2.3.3:

The assertions are listed in the XML file impl/src/main/resources/tck-audit.xml in the CDI TCK distribution. Each assertion is identified by the section of the specification document in which it resides and assigned a unique paragraph identifier to narrow down the location of the assertion further. To continue with the example, the assertion shown above is listed in the tck-audit.xml file using this XML fragment:

<section id="2.3.3" title="Declare the bindings of a bean">
  <assertion id="d">
     <text>A bean may declare multiple binding types.</type>

The strategy of the CDI TCK is to write a test which validates this assertion when run against an implementation. A test case (a method annotated with @Test in an @Artifact class) is correlated with an assertion using the @org.jboss.test.audit.annotations.SpecAssertion annotation as follows:


@SpecAssertion(section = "2.3.3", id = "d")
public void testMultipleBindings()
  Bean<?> model = getBeans(Cod.class, new ChunkyBinding(true), new WhitefishBinding()).iterator().next();
  assert model.getBindings().size() == 3;

To help evaluate the distribution of coverage for these assertions, the TCK provides a detailed coverage report. This report is also useful to help implementors match tests with the language in the specification that supports the behavior being tested.

The coverage report is an HTML report generated as part of the TCK project build. Specifically, it is generated by an annotation processor that attaches to the compilation of the classes in the TCK test suite, another tool from the JBoss Test Utils project. The report is only generated when using Java 6 or above, as it requires the annotation processor.

mvn clean install

The report is written to the file target/coverage.html in the same project. The report has five sections:

The coverage report is color coded to indicate the status of an assertion, or group of assertions. The status codes are as follows:

For reasons provided in the tck-audit.xml document and presented in the coverage report, some assertions are not testable.

The coverage report does not give any indication as to whether the tests are passing. That's where the TestNG reports come in.

As you by now, the CDI TCK test suite is really just a TestNG test suite. That means an execution of the CDI TCK test suite produces all the same reports that TestNG produces. This section will go over those reports and show you were to go to find each of them.

When the CDI TCK test suite is executed during the Maven 2 test phase of the TCK runner project, TestNG is invoked indirectly through the Maven Surefire plugin. Surefire is a test execution abstraction layer capable of executing a mix of tests written for JUnit, TestNG, and other supported test frameworks.

Why is this relevant? It means two things. First, it means that you are going to get a summary of the test run on the commandline. Here's the output generated when the tests are run using standalone mode.

 T E S T S
Running TestSuite
[XmlMethodSelector] CLASSNAME:org.jboss.testharness.impl.testng.DisableIntegrationTestsMethodSelector
[XmlMethodSelector] SETTING PRIORITY:0
[XmlMethodSelector] CLASSNAME:org.jboss.testharness.impl.testng.ExcludeIncontainerUnderInvestigationMethodSelector
[XmlMethodSelector] SETTING PRIORITY:0
Tests run: 441, Failures: 0, Errors: 0, Skipped: 0, Time elapsed: 22.816 sec

Results :

Tests run: 441, Failures: 0, Errors: 0, Skipped: 0

If the Maven reporting plugin that compliments Surefire is configured properly, Maven will also generate a generic HTML test result report. That report is written to the file test-report.html in the target/surefire-reports directory of the TCK runner project. It shows how many tests were run, how many failed and the success rate of the test run.

The one drawback of the Maven Surefire report plugin is that it buffers the test failures and puts them in the HTML report rather than outputting them to the commandline. If you are running the test suite to determine if there are any failures, it may be more useful to get this information in the foreground. You can prevent the failures from being redirected to the report using the following commandline switch:

mvn test -Dsurefire.useFile=false

The information that the Surefire provides is fairly basic and the detail pales in comparison to what the native TestNG reports provide.

After running a test in Eclipse, as described in Section 8.3, “Running a test in standalone mode” and Section 8.4, “Running integration tests”, the test results are displayed in the TestNG plugin view, as shown below.

The view offers two lists. The first is a list of all methods (tests) in the class flagged as either passed or failed. The second is a list of methods (tests) in the class that failed. If there is a test failure, you can click on the method name to get the stacktrace leading up to the failure to display in the lower frame.

You can also find the raw output of the TestNG execution in the IDE console view. In that view, you can click on a test in the stacktrace to open it in the editor pane.

One of the nice features of TestNG is that it can keep track of which tests failed and offer to run only those tests again. You can also rerun the entire class. Buttons are available for both functions at the top of the view.

In this part you learn how to execute the CDI TCK on the CDI reference implementation (Weld). First, you are walked through the steps necessary to execute the test suite on Weld. Then you discover how to modify the TCK runner to execute the test suite on your own implementation. Finally, you learn how to debug tests from the test suite in Eclipse.

One of the requirements of an implementation passing the TCK is for it to pass the CDI signature test. This section describes how the signature file is generated and how to run it against your implementation.

Just for fun (and to confirm that the signature test is working correctly), you can try the following:

1) Edit cdi-api.sig

2) Modify one of the class signatures - in the following example we change one of the constructors for BusyConversationException - here's the original:

CLSS public javax.enterprise.context.BusyConversationException
cons public BusyConversationException()
cons public BusyConversationException(java.lang.String)
cons public BusyConversationException(java.lang.String,java.lang.Throwable)
cons public BusyConversationException(java.lang.Throwable)
supr javax.enterprise.context.ContextException
hfds serialVersionUID

Let's change the default (empty) constructor parameter to one with a java.lang.Integerparameter instead:

CLSS public javax.enterprise.context.BusyConversationException
cons public BusyConversationException(java.lang.Integer)
cons public BusyConversationException(java.lang.String)
cons public BusyConversationException(java.lang.String,java.lang.Throwable)
cons public BusyConversationException(java.lang.Throwable)
supr javax.enterprise.context.ContextException
hfds serialVersionUID

3) Now when we run the signature test using the above command, we should get the following errors:

Missing Constructors

javax.enterprise.context.BusyConversationException:         constructor public javax.enterprise.context.BusyConversationException.BusyConversationException(java.lang.Integer)

Added Constructors

javax.enterprise.context.BusyConversationException:         constructor public javax.enterprise.context.BusyConversationException.BusyConversationException()

STATUS:Failed.2 errors

This chapter explains how to run the TCK on Weld as well as your own implementation. The CDI TCK uses the Maven 2 TestNG plugin and the JBoss Test Harness to execute the test suite. Learning to execute the test suite from Maven 2 is prerequisite knowlege for running the tests in an IDE, such as Eclipse.

To execute the TCK test suite against Weld, first switch to the jboss-tck-runner directory in the extracted Weld distribution:

cd weld/jboss-tck-runner

Then execute the Maven 2 life cycle through the test phase:

mvn test

Without any command-line flags, the test suite is run in standalone mode, which means that any test class with the @org.jboss.testharness.impl.packaging.IntegrationTest annotation is skipped. This mode uses the StandaloneContainers SPI to invoke the test artifact within a mock Java EE life cycle and capture the results of the test. However, passing the suite in this mode is not sufficient to pass the TCK as a whole. The suite must be passed while executing using the in-container mode.

To execute the test suite using in-container mode with the JBoss TCK runner, you first have to setup JBoss AS as described in the Running the TCK against the CDI RI (Weld) and JBoss AS callout.

Then, execute the TCK runner with Maven 2 as follows:

mvn test -Dincontainer

The presence of the incontainer property activates a Maven 2 profile that assigns the org.jboss.testharness.standalone system property to false and the org.jboss.testharness.runIntegrationTests system property to true, hence activating the in-container test mode. This time, all the test artifacts in the test suite are executed.

The in-container profile will also start and stop the application server automatically, a feature which the profile activates by setting the org.jboss.testharness.container.forceRestart to true.

The in-container mode uses the Containers SPI to deploy the test artifact to the container and execute the test in a true Java EE life cycle. The JBoss TCK runner has a dependency on the library that provides an implementation of this interface for JBoss AS 5.2.

Since in-container tests are executed in a remote JVM, the results of the test must be communicated back to the runner over a container-supported protocol. The JBoss Test Harness provides servlet-based communication over HTTP as described in ???.

This chapter explains how to run individual tests using the Eclipse TestNG plugin. It covers running non-integration tests in standalone mode and integration tests (as well as non-integration tests) in in-container mode. You should be able to use the lessons learned here to debug tests in an alternate IDE as well.

Using an existing test harness (TestNG) allows the tests to be executed and debugged in an Integrated Development Environment (IDE) using available plugins. Using an IDE is also the easiest way to execute a test class in isolation.

The TCK can be executed in any IDE for which there is a TestNG plugin available. Running a test from the CDI TCK test suite using the Eclipse TestNG plugin is almost as simple as running any other TestNG test. You can also use the plugin to debug a test, which is described in the next chapter.

Before running a test from the TCK test suite in Eclipse, you must have the Eclipse TestNG plugin and either the m2eclipse plugin or an Eclipse project generated use the Maven 2 Eclipse plugin (maven-eclipse-plugin). Refer to Section 3.3, “Eclipse Plugins” for more information on these plugins.

With the m2eclipse plugin installed, Eclipse should recognize the CDI TCK projects as valid Eclipse projects (or any Weld project for that matter). Import them into the Eclipse workspace at this time. You should also import the Weld projects if you want to debug into that code, which is covered later.


If you choose to use the Maven 2 Eclipse plugin (maven-eclipse-plugin), you should execute the plugin in both the tck and weld projects:

cd tck
mvn clean eclipse:clean eclipse:eclipse -DdownloadSources -DdownloadJavadocs
cd ../weld
mvn clean eclipse:clean eclipse:eclipse -DdownloadSources -DdownloadJavadocs

When setting up your Ecilpse workspace, we recommended creating three workings sets:

The dependencies between the projects will either be established automatically by the m2eclipse plugin, based on the dependency information in the pom.xml files, or as generated by the mvn eclipse:eclipse command.

Your workspace should appear as follows:

Weld JBoss TCK Runner

The tests in the TCK test suite are located in the jsr299-tck-impl project. You'll be working within this project in Eclipse when you are developing tests. However, as you learned earlier, there are no references to a CDI implementation in the TCK. So how can you execute an individual test in Eclipse? The secret is that you need to establish a link in Eclipse (not in Maven) between the jsr299-tck-impl project and your TCK runner project, which in this case is weld-jboss-tck-runner (the project in the jboss-tck-runner directory).

Here are the steps to establish the link:

Of course, the weld-jboss-tck-runner also depends on the jsr299-tck-impl at runtime (so it can actually find the tests to execute). But Eclipse doesn't distinguish between build-time and runtime dependencies. As a result, we've created a circular dependency between the projects. In all likelihood, Eclipse will struggle (if not fail) to compile one or more projects. How can we break this cycle?

As it turns out, the TCK runner doesn't need to access the tests to build. It only needs its classes, configurations and other dependenices at runtime (when the TestNG plugin executes). Therefore, we can remove the TCK runner's dependence on the jsr299-tck-impl project by following these steps:

You are now ready to execute an individual test class (or artifact). Let's start with a test artifact capable of running in standalone mode.

As you have learned, the JBoss test harness determines how to behave based on the values of numerous system properties or properties defined in META-INF/ classpath resources. If the property named org.jboss.testharness.standalone is not defined, the harness assumes that the test is to be run in standalone mode. In order to run the tests in the container, you need to add a properties file to the classpath that sets the standalone property to false and provides values for other properties required to run an in-container test.

The JBoss TCK runner project conveniently provides the properties file src/test/debug-resources/META-INF/ that contains all of the necessary properties for in-container testing in Eclipse. Assuming you followed the directory structure recommended in Section 3.2, “The TCK Environment”, you are good to go. Otherwise, you may have to tune the org.jboss.testharness.container.extraConfigurationDir and org.jboss.testharness.libraryDirectory properties to point to the relative location of the related projects. The properties should be defined as follows:

  • org.jboss.testharness.container.extraConfigurationDir - the relative path from the jboss-tck-impl project to a directory that contains a or file defining the location of a JBoss AS 5.2 installation in the jboss.home property

  • org.jboss.testharness.libraryDirectory - the relative path from the jboss-tck-impl project to the target/dependency/lib directory in the TCK runner project.

The other properties in that file are defined as follows:


You're now ready to execute an integration test. Select an integration test (a class that extends AbstractJSR299Test and is annotated with both @Artifact and @IntegrationTest) and open it in your Eclipse editor. Follow these steps to execute the class with the TestNG plugin:

  1. Right click in the editor view and select Run As > TestNG Test

  2. Observe the test fail because of missing dependencies

  3. Select the Run > Run Configurations... menu from the main menubar

  4. Select the name of the test class under the TestNG category

  5. Select the Classpath tab

  6. Select User Entries in the tree

  7. Click the Advanced... button on the right

  8. Select Add Folders and click the OK button

  9. Select the weld-jboss-tck-runner/src/test/debug-resources folder

  10. Click the OK button on the Folder Selection dialog window

  11. Click on the weld-jboss-tck-runner entry

  12. Move the weld-jboss-tck-runner to the first entry using the Up button

  13. Click the Run button on the Run Configurations dialog window

When you run the test this time, it should pass. If you get a failure that the container (e.g., JBoss AS 5.2) must be run with assertions enabled, you need to stop the container and start it with the -ea flag (or just leave it stopped and the test will start it appropriately).

You can simply right click and select Run As > TestNG Test for all subsequent runs for the reason cited earlier, the run configuration for a class is retained indefinitely.

Alternatively, you can configure TestNG to execute all tests in-container by default by adding the properties file in the debug-resources folder to the project's classpath as follows:

  1. Right click on the jsr299-tck-impl project

  2. Select Build Path > Configure Build Path...

  3. Click on the Libraries tab

  4. Click the Add Class Folder... button on the right

  5. Check the weld-jboss-tck-runner/src/test/debug-resources folder

  6. Click the OK button on the Class Folder Selection dialog window

  7. Click the OK button on the Java Build Path window

Now you don't have to do any special configuration per test class.

You can stop the individual tests from running in-container by reversing the steps above to remove the debug-resources folder from the Eclipse classpath.

You have now mastered running the CDI TCK against Weld using both Maven 2 and within Eclipse. Now you're likely interested in how to debug a test so that you can efficiently investigate test failures.

This chapter explains how to debug standalone and integration tests from the TCK test suite in Eclipse. You should be able to use the lessons learned here to debug tests in an alternate IDE as well.

In order to debug an integration test, or any test run using in-container mode, the test must be configured to run in-container, as described in Section 8.4, “Running integration tests”, and you must attach the IDE debugger to the container. That puts the debugger on both sides of the fence, so to speak.

Since setting up a test to run in-container has already been covered, we'll look at how to attach the IDE debugger to the container, and then move on launching the test in debug mode.

This chapter explains the purpose of the test harness and describes its key features.

The JBoss Test Harness is a testing framework based on TestNG that provides a series of extensions that allow runtime packaging and deployment of Java EE artifacts (EAR or WAR) for in-container testing. It's important to note that the JBoss Test Harness has no relation with, or dependency on, the JBoss Application Server (JBoss AS).


You'll often see the term in-container used in this reference guide. This term refers to running the test suite in any of the aforementioned environments, whilst standalone refers to running the tests outside the container via an implementation-specific standalone bootstrap. The standalone mode only runs those tests which the CDI RI can run without deployment in a Java EE container.

The last thing Java developers want is yet another testing framework to make their life more complicated. That's why the JBoss Test Harness is built entirely upon TestNG. TestNG is one of two prominent test frameworks for Java (the other being JUnit). Furthermore, what developers want is a good integration with their Integrated Development Environment (IDE). These days, if a tool doesn't have an IDE plugin, then it won't get the attention it deserves. TestNG plugins are available for all major IDEs and build tools (Ant and Maven 2). Again, a motivating factor for extending TestNG.

Because it leverages the existing TestNG ecosystem, there is no need for a special test launcher for the JBoss Test Harness. You simply use the IDE or build tool of your choice (so long as it has TestNG support). You also get reporting and debugging for free (various reporting plugins are provided for TestNG).

You can read more about TestNG at

The JBoss Test Harness supports the following features:

  • Test activation via any method supported by the TestNG configuration descriptor (package, group, class)

  • Exclusion of in-container tests in standalone mode

  • Exclusion of individual tests labeled as under investigation

  • Integration with any TestNG plugin (Eclipse, IntelliJ, Ant, Maven)

  • Automated reporting capability as provided by TestNG

  • Standalone and in-container test mode

  • Container pluggability

  • Declarative packaging of additional resources and classes in artifact

  • Declarative deployment exception trapping

  • Artifact dumping for failure and packaging analysis

A test is designated by a method annotated with @org.testng.annotations.Test in a class which extends org.jboss.testharness.AbstractTest and is annotated with @org.jboss.testharness.impl.packaging.Artifact.


Test suites may often choose to extend AbstractTest and require tests to extend that base class. In fact, both the CDI TCK and the Bean Validation TCK provide base classes that extend AbstractTest to provide functionality specific to the needs of the TCK.

The @Test annotation is provided by TestNG, the @Artifact annotation is provided by the JBoss Test Harness and the AbstractTest is part of the JBoss Test Harness. There is a one-to-one mapping between a TestNG test class and an artifact. The packaging type is defined by the @org.jboss.testharness.impl.packaging.Packaging annotation on the test class, defaulting to a WAR if not specified.

Prior to executing the tests for a given class, the JBoss Test Harness packages the class as a deployable artifact (EAR or WAR), along with any extra resources specified, and deploys the artifact to the container. The harness provides test execution and result reporting via HTTP communication to a simple Servlet using a thin layer over the TestNG test launcher. The test harness can also catch and enforce expected deployment exceptions. This setup and tear down activity is provided by the super class org.jboss.testharness.AbstractTest, which all test classes must extend (directly or indirectly).

If the annotation @org.jboss.testharness.impl.packaging.IntegrationTest is not present on the test class, then it means the test class can be executed in standalone mode. In standalone mode, the deployable artifact is assembled on the local classpath and the tests execute in the same JVM as the launcher, just as though it were a regular TestNG test case. The standalone mode is provided for convenience and efficiency, allowing you the speed of mock-based testing and the confidence of an in-container test, using the same test objects and tests.

The basic procedure of an in-container test is as follows. The JBoss Test Harness produces a deployable artifact from an @Artifact test class and any declared dependent classes, descriptors or other resources. Then it deploys the artifact to the container using the Containers SPI, negotiates with the container to execute the test and return the result and, finally, undeploys the artifact. TestNG collects the results of all the tests run in the typical way and produces a report.

The question is, how does the JBoss Test Harness negotiate with the container to execute the test when TestNG is being invoked locally? Technially the mechanism is pluggable, but JBoss Test Harness provides a default implementation that uses HTTP communication that you will likely use. Here's how the default implementation works.

The artifact generator bundles and registers (in the web.xml descriptor) an HttpServlet, org.jboss.testharness.impl.runner.servlet.ServletTestRunner, that responds to test execution GET requests. TestNG running on the client side delegates to a test launcher (more on that in a moment) which originates these text execution requests to transfer control to the container JVM. The name of the test method to be executed is specified in a request query parameter named methodName.

When the test execution request is received, the servlet delegates to an instance of org.jboss.testharness.impl.runner.TestRunner, passing it the name of the test method. TestRunner reads the name of the test class from the resource META-INF/, which is bundled in the artifact by the artifact generator. It then combines the class name and the method name to produce a TestNG test suite and runs the suite (within the context of the container).

TestNG returns the results of the run as an ITestResult object. ServletTestRunner translates this object into a org.jboss.testharness.api.TestResult and passes it back to the test launcher on the client side by encoding the translated object into the response. The object gets encoded as either html or a serialized object, depending on the value of the outputMode request parameter that was passed to the servlet. Once the result has been transfered to the client-side TestNG, TestNG wraps up the run of the test as though it had been executed in the same JVM.

There's one piece missing. How does TestNG on the client side know to submit a request to the ServletTestRunner servlet to get TestNG to execute the test in the container JVM? That's the role of the test launcher.

The test launcher is the API that allows test suite to launch the test in a pluggable fashion. AbstractTest, the super class of AbtractJSR299Test, implements IHookable, a TestNG interface which allows the execution of the test method to be intercepted. Using that mechanism, AbstractTest delegates execution of the test method (a method annotated with @Test in an @Artifact class) to an implementation of org.jboss.testharness.api.TestLauncher if the tests are being executed in-container. As you might anticipate, the implementation is specified using a property with the same name as the interface in a META-INF/ resource. The JBoss Test Harness provides a default implementation, org.jboss.testharness.impl.runner.servlet.ServletTestLauncher, that hooks into the HTTP communication infrastructure described above. It invokes the ServletTestRunner servlet for each method annotated with @Test in the @Artifact that is not otherwise disabled.

If you wish to implement the runner yourself, you must return a TestResult as a result of executing the method in the container. You must also ensure that any exception which occurs during deployment is wrapped as a org.jboss.testharness.api.DeploymentException, and that any communication problem is rethrown as an IOException. The deployment exception may be transformed by an implementation of the org.jboss.testharness.api.DeploymentExceptionTransformer interface, which is specified using the org.jboss.testharness.container.deploymentExceptionTransformer property. The default implementation passes on the original exception unchanged. The implementation for JBoss AS used with the CDI TCK, on the other hand, deciphers the exception thrown by the JBoss deployer and converts it to one of the catagory exceptions defined in the CDI TCK API.

So in short, JBoss Test Harness takes care of all the interfaces you need to execute tests in-container except for the implementation of the Containers SPI. That is, unless you are deploying to one of the containers supported by the JBoss Test Harness (TODO we need a table showing supported containers).

This chapter lays out how to configure the JBoss Test Harness by specifying the API implementation classes, defining the target container connection information, and various other switches. Finally, a detailed account of how the JBoss Test Harness negotiates the execution of the tests in the container is given.

The JBoss Test Harness allows the test suite to be launched in a pluggable fashion. In order to execute a test suite, the JBoss Test Harness must be configured by specifying implementations of the test launcher and container APIs.

System properties and/or the resource META-INF/, a Java properties file, are used to configure the JBoss Test Harness. The bootstrap configuration builder looks to the property org.jboss.testharness.api.ConfigurationBuilder, the first property listed in table 3.1, for the fully qualified class name (FQCN) of a concrete configuration builder implementation to get started. This implementation loads the remaining configuration settings and produces a JBoss Test Harness configuration.

For you convenience, the default configuration builder implementation org.jboss.testharness.impl.PropertiesBasedConfigurationBuilder is provided, which collates all the JBoss Test Harness configuration settings from system and Java properties. It does so by aggregating the system properties with the properties defined in the META-INF/ resource in any classpath entry under a single properties map, allowing you to partition the configuration settings as needed.

A complete list of configuration properties for the JBoss Test Harness has been itemized in Table 11.1, “JBoss Test Harness Configuration Properties”, accompanied by the default value (if any) and a description for each property.

Table 11.1. JBoss Test Harness Configuration Properties

Property = Default ValueDescription
org.jboss.testharness.api.ConfigurationBuilder= org.jboss.testharness.impl.PropertiesBasedConfigurationBuilder The configuration bootstrap class for the JBoss Test Harness.
org.jboss.testharness.testPackage= The top-level Java package containing classes to be tested. Used to determine which artifacts to dump to disk only; not used during running of a suite.
org.jboss.testharness.libraryDirectory= Directory containing extra JARs which should be deployed in the artifact (for example in WEB-INF/lib.
org.jboss.testharness.standalone=true Tests are run using standalone mode if true or using in-container mode if false.
org.jboss.testharness.runIntegrationTests=false If true, integration tests are run. In-container mode must be activated.
org.jboss.testharness.spi.Containers= The deployment implementation for setting up and tearing down the container and deploying and undeploying in-container tests. The host and port on which the container is running.
org.jboss.testharness.connectDelay=5000 The timeout (ms) when attempting to connect to the container (e.g. via http).
org.jboss.testharness.api.TestLauncher= The in-container test launcher, the built in org.jboss.testharness.impl.runner.servlet.ServletTestLauncher is provided and suitable for any Servlet environment.
org.jboss.testharness.container.\ deploymentExceptionTransformer= Provides an interception feature for deployment exceptions, allowing them to be inspected and altered before reporting to the test harness for validation by the test case.
org.jboss.testharness.container.forceRestart=false Whether the container should be restarted before the tests are executed.
org.jboss.testharness.container.extraConfigurationDir= A directory containing a or files that define additional properties. Can be used to provide runtime specific properties.
org.jboss.testharness.spi.StandaloneContainers= The container implementation for executing standalone tests.
dumpArtifacts=false Whether the test artifacts should be written to disk for inspection.
org.jboss.testharness.outputDirectory= Directory where test artifacts will be written to disk, if dumpArtifacts is true.

This chapter explains how to execute and debug a test suite built using the JBoss Test Harness.

The test suite runner project is the magic that makes everything come together and allows you to execute the test suite. If you fully understand how the JBoss Test Harness functions, and have a good grasp on Maven 2, then it's not to difficult to understand how the test suite runner project works. Regardless of your background, this guide covers what you need to know to get up and running by studying the test suite runner used to run the CDI TCK against the CDI RI, Weld.

The TCK runner for the Weld can be found in the jboss-tck-runner directory in the Weld distribution. The dependencies of the TCK runner project for Weld are listed in Table 12.1, “Weld JBoss TCK Runner Dependencies”. You must also pull in the transitve dependencies of these projects.

You can find all of these artifacts in the JBoss Maven repository.

You should substituate the weld-core and weld-porting-package artifacts from table 2.2.3 with your own artifacts. You'll also need to replace the jboss-test-harness-jboss-as-52 artifact if you are not testing your implementation on JBoss AS 5.2. The jboss-test-harness-jboss-as-52 artifact contains implementations of the Containers SPI for the JBoss Test Harness for JBoss AS 5.2.


When running the test suite in the in-container mode, the tests will run against libraries installed into the container. In this project, Weld is only declared as a Maven dependency for when the TCK test suite is being executed in standalone mode.


The target folder for the copies of the dependencies (i.e., the JAR files) is declared as the JBoss Test Harness library directory; this results in these libraries being added to the test artifact using the following property assignment:


We also copy the test suite configuration from the local Maven repository (groupId=org.jboss.jsr299.tck, artifactId=jsr299-tck-impl, classifier=suite, type=xml, version=1.0.0-SNAPSHOT) to a local repository as the TestNG Maven plugin expects a local file.

The TCK is executed using the Maven TestNG plugin. Maven 2 profiles are used to control the properties that are set at the time of the execution. For instance, the incontainer profile enables integration tests and disables standalone mode, changing the default settings.

The jboss-tck-runner project also defines the JBoss Test Harness extra configuration directory using the following property:


The JBoss Test Harness looks in this directory for either a or file that declares additional configuration properties. In particular, the JBoss AS Containers implementation looks here to find the jboss.home property so it knows where the scripts are located to start and stop JBoss AS.