2.0-M5 (2021-03-23)

Breaking Changes

  • The ReportLogExtension vestiges were removed. As this extension was mostly used for an unreleased Spock module, this won’t affect many users. If you are using a Spock Configuration File with a report section, then you must delete everything from this section except for issueNamePrefix and issueUrlPrefix. These two properties are still supported and used by the @Issue extension.

Misc

  • Add support for injection into @Shared fields in spock-spring module which users can opt-in for by adding @EnableSharedInjection to the specification. #76

  • Add new displayName via INameable for SpecInfo, FeatureInfo, and IterationInfo. This field can be set via extensions to change the reported name. The existing iteration NameProvider now also sets the displayName instead of the name. Modifying the name instead of displayName is now considered deprecated for extensions. #1236

  • Add support for constructor injection for extensions

  • Improve final field handling. Final fields are now transformed similar to shared fields, so that we can still delay the initialization but keep them unmodifiable to user code. #1011

  • Improve parallel extensions to support inheritance #1245

  • Improve PollingConditions

  • Improve some AST transformation code regarding error handling

  • Deprecate AbstractGlobalExtension and replace with IGlobalExtension

  • Remove unnecessary try-finally construct for assertions

  • Fix ErrorSpecNode throw Exception in prepare instead of execute this fixes an issue that interceptors for prepare, before, and around were still executed and any Exception they throw would hide the actual cause.

  • Fix #1294 swallowing of unrecoverable Errors

  • Fix #1260 Change InteractionRewriter to keep casting of ListExpressions intact

  • Fix #1282 Make TempDirInterceptor safe for parallel invocation of iterations

  • Fix #1267 Retry extension behavior for unrolled tests

  • Fix #1267 Retry extension behavior for unrolled tests

  • Fix #1229 @TempDir not working for inherited fields

  • Fix #1232 compile error for nested conditions without top-level condition

  • Fix #1256 handling of is as getter for boolean properties on Mocks

  • Fix #1270 handling of is as getter for boolean properties on GroovyMocks

  • Fix ErrorSpecNode to re-throw Exception in prepare instead of execute, this fixes an issue that interceptors for prepare, before, and around were still executed and any Exception they throw would hide the actual cause.

  • Fix #1263 ExceptionAdapterExtension to also handle inherited fixture methods

  • Fix #1279 Cast data variables with type coercion to the declared parameter type

Thanks to all the contributors to this release: Marcin Erdmann, Björn Kautler, Alexander Kriegisch, Josh Soref, Vaidotas Valuckas

2.0-M4 (2020-11-01)

  • Added Parallel Execution support

  • Add a way to register ConfigurationObject globally without the need for a global extension.

  • Annotations for local extensions can now be defined as @Repeatable and applied multiple times to the same target. Spock will handle this appropriately and call the extensions visitSpecAnnotations method with all annotations. If these methods are not overwritten, they forward to the usual visitSpecAnnotation methods once for each annotation.

  • @ConfineMetaClassChanges, @Issue, @IgnoreIf, @PendingFeatureIf, @Requires, @See, @Subject, @Use, and @UseModules are now repeatable annotations

  • @Requires, @IgnoreIf and @PendingFeatureIf can now access instance fields, shared fields and instance methods by using the instance. qualifier inside the condition closure.

  • AbstractAnnotationDrivenExtension is now deprecated and its logic was moved to default methods of IAnnotationDrivenExtension which should be implemented directly now instead of extending the abstract class.

  • Add @TempDir built-in extension

  • @PendingFeature and @PendingFeatureIf can now be used together

  • Fix #1158 Fix strange bug with setter/getter handling of mocks in groovy

  • Fix #1216 perform argument coercion for GroovyMock method arguments

  • Fix #1169 check skipped state in Node.prepare and do nothing if already skipped

  • Fix #1200 name clashes where variables that are named like method calls destroy the method call

  • Fix #1202 NullPointerException with array initializers

  • Fix #994 don’t treat nested closures in argument constraints as implicit assertions anymore

  • Replace hamcrest-core dependency by hamcrest

Thanks to all the contributors to this release: Björn Kautler, Marcin Zajączkowski, DQYuan, Tom Wieczorek, Alexander Kriegisch, Jasper Vandemalle

2.0-M3 (2020-06-11)

Breaking Changes

Sputnik Runner removed (an alternative)

In 2.0-M1 the Sputnik runner was removed and the 2.0-M2 release notes explicitly mentioned that enhancements that extend Sputnik or use it as a delegate like for example the PowerMockRunnerDelegate will not work anymore.

This is not the full truth though, but be aware that the information in this section is just for informational purpose. This is not a solution the Spock maintainers explicitly support or maintain. It is just mentioned as hint for those that have no other choice right now. It is strongly recommended to instead use native solutions or integrations. If those are not available, you might consider asking about proper integration with Spock for a proper long-term solution. As long as this is not available, the work-around described here is at least usable as fallback and mid-term solution as long as the JUnit team decides to maintain this legacy support module.

There is a JUnit 4 runner provided by JUnit 5 called JUnitPlatform that can run any JUnit platform based tests - like Spock 2+ based tests - in a JUnit 4 environment. This is intended for situations where an IDE, build tool, CI server, or similar does not yet natively support JUnit platform. It is provided in the artifact org.junit.platform:junit-platform-runner.

This means, if you make sure your tests are launched with JUnit 4 and you use JUnitPlatform where you before used Sputnik, all should work out properly. In case of PowerMock this means you annotate your specification with @RunWith(PowerMockRunner) and @PowerMockRunnerDelegate(JUnitPlatform) and make sure your tests are launched using JUnit 4.

Removal of JUnit 4 dependency from spock-core

Spock 2.0 from the beginning has leveraged JUnit Platform to execute tests. However, starting with 2.0-M3 junit4.jar is no longer a transitive dependency of spock-core. This might affect people using @Before/@After/…​ instead of Spock-native setup/cleanup/…​ fixture methods. To keep it work the spock-junit4 dependency has to be added.

As a side effect of the removal, the order of the fixture methods execution has changed from:

beforeClass, setupSpec, before, setup, cleanup, after, setup, before, cleanup, after, cleanupSpec, afterClass

to:

setupSpec, beforeClass, setup, before, cleanup, after, setup, before, cleanup, after, cleanupSpec, afterClass

which should not be a problem in the majority of cases.

At the same time, using JUnit 4’s annotations is discouraged (and considered deprecated), the same using the other elements of JUnit 4 (as `(Class)Rule`s).

Reduce spock-core direct groovy dependencies

spock-core now only depends on groovy.jar. All other Groovy dependencies have been removed, this should make dependency management a bit easier. If you relied on other groovy dependencies transitively, you will need to add them directly.

Upgrade to JUnit 5.6 (and JUnit Platform 1.6)

JUnit Platform 1.6 deprecated methods (from experimental API) in EngineExecutionResults that Spock was using. To keep runtime compatibility with JUnit 5.6 and incoming 5.7 the implementation has been switched to the new methods. As a result, Spock 2.0-M3 cannot work with JUnit 5.5 and lower. The problem might only occur if a project overrides default JUnit Platform version provided by Spock.

New meaning of >> _

The meaning of >> _ has changed from "use the default response" to "return a stubbed value" (Docs). The original behavior was only ever documented in the Javadocs and was basically the same to just omitting it. The only use-case was chained responses >> "a" >> _ >> "b", but even here it is clearer to just use null or { callRealMethod() } explicitly. With the new behavior, you can have a Mock or Spy return the same value as a Stub would.

subscriber.receive(_) >> _

Renamed iterationCount token

The token #iterationCount in unroll patterns was renamed to #iterationIndex. If you use it somewhere, you have to manually change it to the new name or the test will fail unless you disabled expression asserting, then you will get an #Error:iterationCount rendering instead.

No access to data variables in data pipes anymore

It is not possible anymore to access any data variable from a data pipe or anything else but a previous data table column in a data table cell. This access was partly possible, but could easily prematurely drain iterators, access data providers sooner as expected, behaved differently depending on the concrete code construct used. All these points are more confusing than necessary. If you want to calculate a data variable from others, you can always use a derived data variable that has full access to all previous data variables and can also call helper methods for more complex logic.

If you switch your tests that are fully green to use Spock 2.0 and get any MissingPropertyExceptions, you are probably hitting this change, you should then change to a derived data variable there instead of a data pipe.

If you for example had:

where:
a << [1, 2]
b << a

what you want instead is:

where:
a << [1, 2]
b = a

Assert unroll expressions by default

The system property spock.assertUnrollExpressions is not supported anymore. Instead the new default behavior is equal to having this property set to true. This means tests that were successful but had an #Error: name rendering will now fail. It can be set back to the old pre Spock 2.0 behaviour by setting unroll { validateExpressions false } in the Spock configuration file.

For extension developers

  • FeatureInfo#getDataVariables() and FeatureInfo#getParameterNames() used to return the same value, the parameter names, in the order of the method parameters. This can disturb some calculations like method argument determination and so on and is plainly wrong, as some parameters could be injected by extensions like injecting mock objects or test proxies or similar. FeatureInfo#getDataVariables() now only returns the actual data variables and in the order how they are defined in the where block.

  • Method arguments in MethodInfo now have a value of MethodInfo.MISSING_ARGUMENT if no value was set so far, for example by some extensions or from data variables. If any of these is not replaced by some value, an exception will be thrown at runtime.

Ant support removed

SpecClassFileSelector was removed, which was the only class that required ant. If you are still using spock with ant, then you can just copy the class from the spock source code into your build.

Misc

  • A new MutableClock utility class to support time related testing, see docs.

  • Defining interactions on property getters within Mock instantiation closures or with closures works now without the need to explicitly qualify the property access with it.:

    Foo foo = Stub {
        bar >> 'my stubbed property'
    }
  • A sequence of two or more underscores can be used to separate multiple data tables in one where block.

  • Type casts in conditions are now properly carried over to properly disambiguate method calls. For more information see issue #1022.

  • Data tables now support any amount of semicolons to separate data table columns as alternative to pipes and double pipes, but cannot be mixed with them in one table:

    where:
    a ; b ;; c
    1 ; 2 ;; 3
  • The default unroll pattern changed from the rather generic #featureName[#iterationIndex] to a more fancy version that lists all data variables and their values additionally to the feature name and iteration index. If you prefer to retain the old behaviour, you can set the setting unroll { defaultPattern '#featureName[#iterationIndex]' } in the Spock configuration file and you will get the same result as previously.

  • If neither a parameter to the @Unroll annotation is given, nor the method name contains a #, now the configuration file setting unroll { defaultPattern } is inspected. If it is set to a non-null string, this value is used as unroll pattern.

  • IterationInfo now has a property for the iterationIndex and one for a map of data variable names to values. This is typically not interesting for the average user, but might be helpful for authors of Spock extensions.

  • constructorArgs now properly transport type cast information to the constructor selection, for example to disambiguate multiple candidate constructors:

    Spy(constructorArgs: [null as String, (Pattern) null])
  • Accessing previous data table columns was broken in some cases, now it should work properly, even cross-table and without being disturbed by previous derived data variables.

  • The order of parameters in a data-driven feature does no longer have to be identical to the declaration order in the where block. Data variables are now injected by name.

  • Data driven features do no longer have the requirement that either none or all data variables are declared as parameters and that all parameters are also data variables. Now you can either declare none, some, or all data variables as parameters and also have additional method parameters that are no data variables. Those additional parameters must be provided by some Spock extension though. If they are not set at execution time, an exception will be thrown.

  • Condition closures used in @Requires, @IgnoreIf and @PendingFeatureIf now also get the context passed as argument, so you can use this, typed as org.spockframework.runtime.extension.builtin.PreconditionContext to enable IDE support like code completion:

    @IgnoreIf({ PreconditionContext it -> it.os.windows })
    def "I'll run everywhere but on Windows"() { ... }
  • Execution with an unsupported Groovy version can be allowed with -Dspock.iKnowWhatImDoing.disableGroovyVersionCheck=true (#1164)

  • Relax maximum allowed Groovy version in snapshot builds (#1108)

  • Upgrade Groovy to 2.5.12 (improved Java 14+ support) and 3.0.4 (fixes #1127)

  • Upgrade JUnit 4 to 4.13 (in spock-junit4)

  • Upgrade Hamcrest to 2.2 (from 1.3 provided previously by junit4.jar)

  • Derived data variables (assignments in where blocks) now also support multi-assignment syntax, including ignoring some values with the wildcard expression.

    (a, b, _, c) = row
  • Multi-variable data pipes now also support nesting.

    [a, [_, b]] << [
      [1, [5, 1]],
      [2, [5, 2]]
    ]
  • The condition closures for @IgnoreIf, @Requires and @PendingFeatureIf can now access data variables if applied to a data driven feature. This has further implications that are documented in the respective documentation parts and JavaDocs.

  • Data driven features are now unrolled by default. @Unroll can still be used to specify a custom naming pattern. A simple @Unroll without argument is not needed anymore except when undoing a spec-level @Rollup annotation or if unrolling by default is disabled, so any simple @Unroll annotations can be removed from existing code. You can verify this by looking at the test count which should not have been changed after you removed the simple @Unroll annotations.

  • @Rollup can now be used on feature and spec level to explicitly roll up any feature where the reporting of single iterations is not wanted.

  • The setting unroll { unrollByDefault false } in the Spock configuration file can be set to roll up all features by default if not overwritten by explicit @Unroll annotations and thus reinstate the pre Spock 2.0 behaviour.

  • Updated OSGI support with using bnd (#1154, #1175)

  • Fix @PendingFeature logic (#1103)

  • Do not strip type information from arguments (#1134)

  • Simplify work-around to get the reference to the current closure (#1131)

  • Set source position for return statement in data provider method (#1116)

  • Add Spy(T obj, Closure interactions) (#1115)

  • Reduce number of Groovy dependencies to just groovy.jar (#1109)

  • Improve ExceptionUtil.sneakyThrow declaration (#1077)

  • Print up to 5 last mock invocations for a wrong order error (#1093)

  • Fix EmptyOrDummyResponse returning mock instance for Object (#1092)

  • Remove unnecessary reflection for Java 8 types in EmptyOrDummyResponse (#1091)

  • Update old Issue annotations to point to migrated github issues (#1003)

  • Test Spock with Java 14 (#1155)

Thanks to all the contributors to this release: Björn Kautler, Marcin Zajączkowski, Raymond Augé, Roman Tretiak, Camilo Jorquera, Shil Sinha

2.0-M2 (2020-02-10)

Groovy-3.0 Support

The main feature of this milestone is support for Groovy 3. To use Spock in your Groovy 3 project just select the spock-\*-2.0-M2 artifact(s) ending with -groovy-3.0.

Note
As Groovy 3 is not backward compatible with Groovy 2, there is a layer of abstraction in Spock to allow to build (and use) the project with both Groovy 2 and 3. As a result, an extra artifact spock-groovy2-compat is (automatically) used in projects with Groovy 2. It is very important to do not mix the spock-*-2.x-groovy-2.5 artifacts with the groovy-*-3.x artifacts on a classpath. This may result in weird runtime errors.

Breaking Changes

Sputnik Runner removed

Although already in 2.0-M1 it wasn’t explicitly mentioned: All enhancements that either extended Sputnik or used it as a delegate runner will not work anymore, e.g, PowerMockRunnerDelegate.

Misc

  • Add @PendingFeatureIf annotation (Docs)

  • Fail-fast for invalid Stub interactions, added new validation that catches more invalid usage of Stub

  • Forbid spying on Spy instances, as this doesn’t work anyway and leads to wrong expectations (#1029)

  • Update Jvm helper utility to include new Java versions, removed pre 8 versions

  • Update docs regarding ByteBuddy, cglib, objenesis

  • Provide compatibility to both JUnit 5.5.2 and 5.6.0

Thanks to all the contributors to this release: Marcin Zajączkowski (Groovy 3 support), Björn Kautler, Ryan Gardner

2.0-M1 (2019-12-31)

This is the first milestone release to version 2.0. This means that we have migrated to the new JUnit Platform and all internal tests pass. We have tried to keep the API as compatible as possible and if you’ve only used the public spock API then there is a high possibility that all you have to do is to update the spock version and configure gradle/maven to use the JUnit Platform.

However, this doesn’t mean that the API is finalized yet, the goal for the first milestone was just to get it running on the new platform. The next milestones will focus on improvements like the much requested parallel execution support.

Please try it out and report any new bugs so that we can fix them for the final 2.0 release.

Breaking Changes

New JUnit Platform

Switch from JUnit 4 to the JUnit Platform. See https://junit.org/junit5/docs/current/user-guide/#running-tests-build on how to configure maven and gradle to use the JUnit Platform.

JUnit 4 Rules are not supported by spock-core anymore, however, there is a new spock-junit4 module that provides best effort support to ease migration.

Misc

  • Spock now requires at least Java 8

  • All data-driven test iterations are always reported (unrolled), while @Unroll is not necessary anymore it can still be used to define the name template

  • @Retry.Mode.FEATURE didn’t work anymore and has been removed

  • spock-report module has been removed, it was never officially released

  • The SpockReportingExtension has been disabled until we can integrate it with JUnitPlatform

  • Removed testing for spring-2.x as it is incompatible

  • Fix some Javadocs

Special thanks goes to Marc Philipp who helped a lot with the integration of the JUnit Platform.

Thanks to all the contributors to this release: Marcin Zajączkowski, k3v1n0x90

1.3 (2019-03-05)

No functional changes

1.3-RC1 (2019-01-22)

The theme for this release is to increase the information that is provided when an assertion failed.

Potential breaking changes

code argument constraints are treated as implicit assertions

Before this release the code argument constrains worked by returning a boolean result. This was fine if you just wanted to do a simple comparison, but it breaks down if you need to do 5 comparisons. Users also often assumed that it worked like the assertions in then blocks and didn’t add && to chain multiple assertions together, so their constraint ignored all before the next line.

1 * mock.foo( { it.size() > 1
                it[0].length == 2 })

This would only use the length comparison, to make it work you had to add &&. Another problem arises by having more than one comparison inside the constraints, you don’t know which of the 5 comparisons failed. If you just expected one method call you could use an explicit assert as a workaround, but since it immediately breaks, you can’t use it if you want to have multiple different calls to the same mock.

With 1.3 the above code will actually work as intended, and even more important it will give actual feedback what didn’t match.

So what can break?

If you used the code argument constraint as a way of capturing the argument value, then this will most likely not work anymore, since assignments to already declared variables are forbidden in implicit assertion block. If you still need access to the argument, you can use the response generator closure instead.

def extern = null

1 * mock.foo( { extern = it; it.size() > 0 })  // old
1 * mock.foo( { it.size() > 0 }) >> { extern = it[0] } // new

The added benefit of this changes is, that it clearly differentiates the condition from the capture.

Another consequence of the change is, that the empty {} assertion block will now pass instead of fail, since no assertion error is being treated as passing, while it required a true result beforehand.

It is advised, that if you have multiple conditions joined by &&, that you remove it to get individual assertions reports instead of a large joined block.

assertions with explicit messages now include power assertions output.

Given
def a = 1
def b = 2
assert a == b : "Additional message"
Before
a == b

Additional message
Now
a == b
| |  |
1 |  2
  false

Additional message

If you relied on this behavior to hide some output, or to prevent a stack overflow due to a self referencing data structure, then you need to move the condition into a separate method that just returns the boolean result.

What’s New In This release

  • Add implicit assertions for CodeArgument constraints (#956)

  • Add power assertion output to asserts with explicit message (#928)

  • Add support for mixed named and positional arguments in mocks (#919)

  • Add NamedParam support for groovy-2.5 with backport to 2.4 (#921)

  • Add special rendering for Set comparisons (#925)

  • Add identity hash code to type hints in comparison failures if they are identical

  • Fix erroneous regex where an optional colon was defined instead of a non-capturing group (#931)

  • Improve CodeArgumentConstraint by supporting assertions (#918)

  • Improve IDE type inference in MockingApi (#920)

  • Improve reporting of TooFewInvocationsError (#912)

  • Improve render class loader for classes in comparison failures (#932)

  • Improve record class literal values to display FQCN in comparison failures (#935)

  • Improve filter Java 9+ reflection stack frames

  • Improve show stacktrace of throwables in comparison failure result

  • Improve use canonical class name in comparison failure results if present

  • Improve render otherwise irrelevant expressions if they get a type hint in comparison failure (#936)

  • Fix do not convert implicit "this" expression like when calling the constructor of a non-static inner class (#930)

  • Fix class expression recording when there are comments with dots in the same line (#937)

Thanks to all the contributors to this release: Björn Kautler, Marc Philipp, Marcin Zajączkowski, Martin Vseticka, Michael Kutz, Kacper Bublik

1.2 (2018-09-23)

Breaking Changes: Spock 1.2 drops support for Java 6, Groovy 2.0 and Groovy 2.3

What’s New In This release

  • Add support for Java 11+ (#895, #902, #903)

  • Add Groovy 2.5.0 Variant for better Java 10+ Support

  • Add @SpringBean and @SpringSpy inspired by @MockBean, Also add @StubBeans (Docs)

  • Add @UnwrapAopProxy to make automatically unwrap SpringAopProxies

  • Add @AutoAttach extension (Docs)

  • Add @Retry extension (Docs)

  • Add flag to UnrollNameProvider to assert unroll expressions (set the system property spock.assertUnrollExpressions to true) (#767)

  • Add automatic module name descriptors for Java 9

  • Add configurable condition to @Retry extension to allow for customizing when retries should be attempted (Docs)

  • Improve @PendingFeature to now have an optional reason attribute (#907)

  • Improve @Retry to be declarable on a spec class which will apply it to all feature methods in that class and subclasses (Docs)

  • Improve StepwiseExtension mark only subsequent features as skipped in case of failure (#893)

  • Improve in assertions Spock now uses DefaultGroovyMethods.dump instead of toString if a class doesn’t override the default Object.toString.

  • Improve verifyAll can now also have a target same as with

  • Improve static type hints for verifyAll and with

  • Improve reporting of exceptions during cleanup, they are now properly reported as suppressed exceptions instead of hiding the real exception

  • Improve default responses for stubs, Java 8 types like Optional and Streams now return empty, CompletableFuture completes with null result

  • Improve support for builder pattern, stubs now return themselves if the return type matches the type of the stub

  • Improve tapestry support with by supporting @ImportModule

  • Improve constructorArgs for spies can now accept a map directly without the need to wrap it in a list

  • Improve Guice Module now automatically attaches detached mocks

  • Improve unmatched mock messages by using dump instead of inspect for classes which don’t provide a custom toString

  • Improve spying on concrete instances to enable partial mocking

  • Fix use String renderer for Class instances (#909)

  • Fix mark new Spring extensions as @Beta (#890)

  • Fix exclude groovy-groovysh from compile dependencies (#882)

  • Fix Retry.Mode.FEATURE and Retry.Mode.SETUP_FEATURE_CLEANUP to make a test pass if a retry was successful.

  • Fix issue with @SpringBean mocks throwing InvocationTargetException instead of actual declared exceptions (#878, #887)

  • Fix void methods with implicit targets failing in with and verifyAll (#886)

  • Fix SpockAssertionErrors and its subclasses now are properly Serializable

  • Fix Spring injection of JUnit Rules, due to the changes in 1.1 the rules where initialized before Spring could inject them, this has been fixed by performing the injection earlier in the process

  • Fix SpringMockTestExecutionListener initializes lazy beans

  • Fix OSGi Import-Package header

  • Fix re-declare recorder variables (#783), this caused annotations such as @Slf4j to break Specifications

  • Fix MissingFieldException in DiffedObjectAsBeanRenderer

  • Fix problems with nested with and verifyAll method calls

  • Fix assertion of mock invocation order with nested invocations (#475)

  • Fix ignore inferred type for Spies on existing instance

  • General dependency update

Thanks to all the contributors to this release: Marc Philipp, Rob Elliot, jochenberger, Jan Papenbrock, Paul King, Marcin Zajączkowski, mrb-twx, Alexander Kazakov, Serban Iordache, Xavier Fournet, timothy-long, John Osberg, AlexElin, Benjamin Muschko, Andreas Neumann, geoand, Burk Hufnagel, signalw, Martin Vseticka, Tilman Ginzel

1.2-RC3 (2018-09-16)

What’s New In This release

  • Add support for Java 11+ (#895, #902, #903)

  • Improve @PendingFeature to now have an optional reason attribute (#907)

  • Fix use String renderer for Class instances (#909)

  • Fix mark new Spring extensions as @Beta (#890)

  • Fix exclude groovy-groovysh from compile dependencies (#882)

Thanks to all the contributors to this release: Marc Philipp, Marcin Zajączkowski, signalw

1.2-RC2 (2018-09-04)

What’s New In This release

  • Add configurable condition to @Retry extension to allow for customizing when retries should be attempted (Docs)

  • Fix Retry.Mode.FEATURE and Retry.Mode.SETUP_FEATURE_CLEANUP to make a test pass if a retry was successful.

  • Improve @Retry to be declarable on a spec class which will apply it to all feature methods in that class and subclasses (Docs)

  • Improve StepwiseExtension mark only subsequent features as skipped in case of failure (#893)

  • Fix issue with @SpringBean mocks throwing InvocationTargetException instead of actual declared exceptions (#878, #887)

  • Fix void methods with implicit targets failing in with and verifyAll (#886)

Thanks to all the contributors to this release: Marc Philipp, Tilman Ginzel, Marcin Zajączkowski, Martin Vseticka

1.2-RC1 (2018-08-14)

Breaking Changes: Spock 1.2 drops support for Java 6, Groovy 2.0 and Groovy 2.3

What’s New In This release

  • Add Groovy 2.5.0 Variant for better Java 10 Support

  • Add @SpringBean and @SpringSpy inspired by @MockBean, Also add @StubBeans

  • Add @UnwrapAopProxy to make automatically unwrap SpringAopProxies

  • Add flag to UnrollNameProvider to assert unroll expressions (set the system property spock.assertUnrollExpressions to true) #767

  • Add automatic module name descriptors for Java 9

  • Add @AutoAttach extension (Docs)

  • Add @Retry extension (Docs)

  • Fix SpockAssertionErrors and its subclasses now are properly Serializable

  • Fix Spring injection of JUnit Rules, due to the changes in 1.1 the rules where initialized before Spring could inject them, this has been fixed by performing the injection earlier in the process

  • Fix SpringMockTestExecutionListener initializes lazy beans

  • Fix OSGi Import-Package header

  • Fix re-declare recorder variables (#783), this caused annotations such as @Slf4j to break Specifications

  • Fix MissingFieldException in DiffedObjectAsBeanRenderer

  • Fix problems with nested with and verifyAll method calls

  • Fix assertion of mock invocation order with nested invocations (#475)

  • Fix ignore inferred type for Spies on existing instance

  • Improve in assertions Spock now uses DefaultGroovyMethods.dump instead of toString if a class doesn’t override the default Object.toString.

  • Improve verifyAll can now also have a target same as with

  • Improve static type hints for verifyAll and with

  • Improve reporting of exceptions during cleanup, they are now properly reported as suppressed exceptions instead of hiding the real exception

  • Improve default responses for stubs, Java 8 types like Optional and Streams now return empty, CompletableFuture completes with null result

  • Improve support for builder pattern, stubs now return themselves if the return type matches the type of the stub

  • Improve tapestry support with by supporting @ImportModule

  • Improve constructorArgs for spies can now accept a map directly without the need to wrap it in a list

  • Improve Guice Module now automatically attaches detached mocks

  • Improve unmatched mock messages by using dump instead of inspect for classes which don’t provide a custom toString

  • Improve spying on concrete instances to enable partial mocking

  • General dependency update

Thanks to all the contributors to this release: Rob Elliot, jochenberger, Jan Papenbrock, Paul King, Marcin Zajączkowski, mrb-twx, Alexander Kazakov, Serban Iordache, Xavier Fournet, timothy-long, John Osberg, AlexElin, Benjamin Muschko, Andreas Neumann, geoand, Burk Hufnagel

Known Issues

  • Groovy 2.4.10 introduced a bug that interfered with the way verifyAll works, it has been fixed in 2.4.12

1.1 (2017-05-01)

What’s New In This release

  • Update docs to include info/examples for Spying instantiated objects

  • Fix integer overflow that could occur when the OutOfMemoryError protection while comparing huge strings kicked in

  • Improve rendering for OutOfMemoryError protection

1.1-rc-4 (2017-03-28)

This should be the last rc for 1.1

What’s New In This release

  • 15 merged pull requests

  • Spies can now be created with an already existing target

  • Fix for scoped Spring Beans

  • Fix incompatibility with Spring 2/3 that was introduced in 1.1-rc-1

  • Fix groovy compatibility

  • Fix ByteBuddy compatibility

  • Fix OutOfMemoryError when comparing huge strings

  • Improve default response for java.util.Optional<T>, will now return empty optional

  • Improve detection of Spring Boot tests

  • Improve documentation for global extensions

Thanks to all the contributors to this release: Taylor Wicksell, Rafael Winterhalter, Marcin Zajączkowski, Eduardo Grajeda, Paul King, Andrii, Björn Kautler, Libor Rysavy

Known issues with groovy 2.4.10 which breaks a smoke test, but should have little impact on normal use (#709).

1.1-rc-3 (released 2016-10-17)

Adds compatibility with ByteBuddy as an alternative to cglib for generating mocks and stubs for classes.

1.1-rc-2 (released 2016-08-22)

1.1 should be here soon but in the meantime there’s a new release candidate.

What’s New In This release

  • Support for the new test annotations in Spring Boot 1.4.

  • Fixed the integration of JUnit method rules which now correctly happen "outside" the setup / cleanup methods.

Thanks to all the contributors to this release: Jochen Berger, Leonard Brünings, Mariusz Gilewicz, Tomasz Juchniewicz, Gamal Mateo, Tobias Schulte, Florian Wilhelm, Kevin Wittek

1.1-rc-1 (released 2016-06-30)

A number of excellent pull requests have been integrated into the 1.1 stream. Currently some features are incubating. We encourage users to try out these new features and provide feedback so we can finalize the content for a 1.1 release.

What’s New In This release

  • 44 merged pull requests

  • The verifyAll method can be used to assert multiple boolean expressions without short-circuiting those after a failure. For example:

then:
verifyAll {
  a == b
  b == c
}
  • Detached mocks via the DetachedMockFactory and SpockMockFactoryBean classes see the Spring Module Docs.

  • Cells in a data table can refer to the current value for a column to the left.

  • Spy can be used to create partial mocks for Java 8 interfaces with default methods just as it can for abstract classes.

  • Improved power assert output when an exception occurs evaluating an assertion.

  • A new @PendingFeature annotation to distinguish incomplete functionality from features with @Ignore.

Special thanks to all the contributors to this release: Dmitry Andreychuk, Aseem Bansal, Daniel Bechler, Fedor Bobin, Leonard Brünings, Leonard Daume, Marcin Erdmann, Jarl Friis, Søren Berg Glasius, Serban Iordache, Michal Kordas, Pap Lőrinc, Vlad Muresan, Etienne Neveu, Glyn Normington, David Norton, Magnus Palmér, Gus Power, Oliver Reissig, Kevin Wittek and Marcin Zajączkowski

1.0 (released 2015-03-02)

1.0 has arrived! Finally (and some years late) the version number communicates what Spock users have known for ages - that Spock isn’t only useful and fun, but also reliable, mature, and here to stay. So please, go out and tell everyone who hasn’t been assimilated that now is the time to join the party!

A special thanks goes to all our tireless speakers and supporters, only a few of which are listed here: Andres Almiray, Cédric Champeau, David Dawson, Rob Fletcher, Sean Gilligan, Ken Kousen, Guillaume Laforge, NFJS Tour, Graeme Rocher, Baruch Sadogursky, Odin Hole Standal, Howard M. Lewis Ship, Ken Sipe, Venkat Subramaniam, Russel Winder.

What’s New In This Release

  • 17 contributors, 21 resolved issues, 18 merged pull requests, some ongoing work. No ground-breaking new features, but significant improvements and fixes across the board.

  • Minimum runtime requirements raised to JRE 1.6 and Groovy 2.0.

  • Improved and restyled reference documentation at http://docs.spockframework.org. Generated with Asciidoctor (what else?).

  • Maven plugin removed. Just let Maven Surefire run your Spock specs like your JUnit tests (see spock-example project).

  • Official support for Java 1.8, Groovy 2.3 and Groovy 2.4. Make sure to pick the groovy-2.0 binaries for Groovy 2.0/2.1/2.2, groovy-2.3 binaries for Groovy 2.3, and groovy-2.4 binaries for Groovy 2.4 and higher.

  • Improved infrastructure to allow for easier community involvement: Switch to GitHub issue tracker, Windows and Linux CI builds, pull requests automatically tested, all development on master branch (bye-bye groovy-x.y branches!).

Other News

What’s Up Next?

With a revamped build/release process and a reforming core team, we hope to release much more frequently from now on. Another big focus will be to better involve the community and their valuable contributions. Last but not least, we are finally shooting for a professional logo and website. Stay tuned for announcements!

Test Long And Prosper,

The Spock Team


New Third Party Extensions

These awesome extensions have been published or updated:

0.7 (released 2012-10-08)

New Reference Documentation

The new Spock reference documentation is available at http://docs.spockframework.org. It will gradually replace the documentation at http://wiki.spockframework.org. Each Spock version is documented separately (e.g. http://docs.spockframework.org/en/spock-0.7-groovy-1.8). Documentation for the latest Spock snapshot is at http://docs.spockframework.org/en/latest. As of Spock 0.7, the chapters on Data Driven Testing and Interaction Based Testing are complete.

Improved Mocking Failure Message for TooManyInvocationsError

The diagnostic message accompanying a TooManyInvocationsError has been greatly improved. Here is an example:

Too many invocations for:

3 * person.sing(_)   (4 invocations)

Matching invocations (ordered by last occurrence):

2 * person.sing("do")   <-- this triggered the error
1 * person.sing("re")
1 * person.sing("mi")

Improved Mocking Failure Message for TooFewInvocationsError

The diagnostic message accompanying a TooFewInvocationsError has been greatly improved. Here is an example:

Too few invocations for:

1 * person.sing("fa")   (0 invocations)

Unmatched invocations (ordered by similarity):

1 * person.sing("re")
1 * person.say("fa")
1 * person2.shout("mi")

Stubs

Besides mocks, Spock now has explicit support for stubs:

def person = Stub(Person)

A stub is a restricted form of mock object that responds to invocations without ever demanding them. Other than not having a cardinality, a stub’s interactions look just like a mock’s interactions. Using a stub over a mock is an effective way to communicate its role to readers of the specification.

Spies

Besides mocks, Spock now has support for spies:

def person = Spy(Person, constructorArgs: ["Fred"])

A spy sits atop a real object, in this example an instance of class Person. All invocations on the spy that don’t match an interaction are delegated to that object. This allows to listen in on and selectively change the behavior of the real object. Furthermore, spies can be used as partial mocks.

Declaring Interactions at Mock Creation Time

Interactions can now be declared at mock creation time:

def person = Mock(Person) {
    sing() >> "tra-la-la"
    3 * eat()
}

This feature is particularly attractive for Stubs.

Groovy Mocks

Spock now offers specialized mock objects for spec’ing Groovy code:

def mock = GroovyMock(Person)
def stub = GroovyStub(Person)
def spy = GroovySpy(Person)

A Groovy mock automatically implements groovy.lang.GroovyObject. It allows stubbing and mocking of dynamic methods just like for statically declared methods. When a Groovy mock is called from Java rather than Groovy code, it behaves like a regular mock.

Global Mocks

A Groovy mock can be made global:

GroovySpy(Person, global: true)

A global mock can only be created for a class type. It effectively replaces all instances of that type and makes them amenable to stubbing and mocking. (You may know this behavior from Groovy’s MockFor and StubFor facilities.) Furthermore, a global mock allows mocking of the type’s constructors and static methods.

Grouping Conditions with Same Target Object

Inspired from Groovy’s Object.with method, the Specification.with method allows to group conditions involving the same target object:

def person = new Person(name: "Fred", age: 33, sex: "male")

expect:
with(person) {
    name == "Fred"
    age == 33
    sex == "male"
}

Grouping Interactions with Same Target Object

The with method can also be used for grouping interactions:

def service = Mock(Service)
app.service = service

when:
app.run()

then:
with(service) {
    1 * start()
    1 * act()
    1 * stop()
}

Polling Conditions

spock.util.concurrent.PollingConditions joins AsyncConditions and BlockingVariable(s) as another utility for testing asynchronous code:

def person = new Person(name: "Fred", age: 22)
def conditions = new PollingConditions(timeout: 10)

when:
Thread.start {
    sleep(1000)
    person.age = 42
    sleep(5000)
    person.name = "Barney"
}

then:
conditions.within(2) {
    assert person.age == 42
}

conditions.eventually {
    assert person.name == "Barney"
}

Experimental DSL Support for Eclipse

Spock now ships with a DSL descriptor that lets Groovy Eclipse better understand certain parts of Spock’s DSL. The descriptor is automatically detected and activated by the IDE. Here is an example:

// currently need to type variable for the following to work
Person person = new Person(name: "Fred", age: 42)

expect:
with(person) {
    name == "Fred" // editor understands and auto-completes 'name'
    age == 42      // editor understands and auto-completes 'age'
}

Another example:

def person = Stub(Person) {
    getName() >> "Fred" // editor understands and auto-completes 'getName()'
    getAge() >> 42      // editor understands and auto-completes 'getAge()'
}

DSL support is activated for Groovy Eclipse 2.7.1 and higher. If necessary, it can be deactivated in the Groovy Eclipse preferences.

Experimental DSL Support for IntelliJ IDEA

Spock now ships with a DSL descriptor that lets Intellij IDEA better understand certain parts of Spock’s DSL. The descriptor is automatically detected and activated by the IDE. Here is an example:

def person = new Person(name: "Fred", age: 42)

expect:
with(person) {
    name == "Fred" // editor understands and auto-completes 'name'
    age == 42      // editor understands and auto-completes 'age'
}

Another example:

def person = Stub(Person) {
    getName() >> "Fred" // editor understands and auto-completes 'getName()'
    getAge() >> 42      // editor understands and auto-completes 'getAge()'
}

DSL support is activated for IntelliJ IDEA 11.1 and higher.

Splitting up Class Specification

Parts of class spock.lang.Specification were pulled up into two new super classes: spock.lang.MockingApi now contains all mocking-related methods, and org.spockframework.lang.SpecInternals contains internal methods which aren’t meant to be used directly.

Improved Failure Messages for notThrown and noExceptionThrown

Instead of just passing through exceptions, Specification.notThrown and Specification.noExceptionThrown now fail with messages like:

Expected no exception to be thrown, but got 'java.io.FileNotFoundException'

Caused by: java.io.FileNotFoundException: ...

HamcrestSupport.expect

Class spock.util.matcher.HamcrestSupport has a new expect method that makes Hamcrest assertions read better in then-blocks:

when:
def x = computeValue()

then:
expect x, closeTo(42, 0.01)

@Beta

Recently introduced classes and methods may be annotated with @Beta, as a sign that they may still undergo incompatible changes. This gives us a chance to incorporate valuable feedback from our users. (Yes, we need your feedback!) Typically, a @Beta annotation is removed within one or two releases.

Fixed Issues

See the issue tracker for a list of fixed issues.

0.6 (released 2012-05-02)

Mocking Improvements

The mocking framework now provides better diagnostic messages in some cases.

Multiple result declarations can be chained. The following causes method bar to throw an IOException when first called, return the numbers one, two, and three on the next calls, and throw a RuntimeException for all subsequent calls:

foo.bar() >> { throw new IOException() } >>> [1, 2, 3] >> { throw new RuntimeException() }

It’s now possible to match any argument list (including the empty list) with foo.bar(*_).

Method arguments can now be constrained with Hamcrest matchers:

import static spock.util.matcher.HamcrestMatchers.closeTo

...

1 * foo.bar(closeTo(42, 0.001))

Extended JUnit Rules Support

In addition to rules implementing org.junit.rules.MethodRule (which has been deprecated in JUnit 4.9), Spock now also supports rules implementing the new org.junit.rules.TestRule interface. Also supported is the new @ClassRule annotation. Rule declarations are now verified and can leave off the initialization part. I that case Spock will automatically initialize the rule by calling the default constructor. The @TestName rule, and rules in general, now honor the @Unroll annotation and any defined naming pattern.

See Issue 240 for a known limitation with Spock’s TestRule support.

Condition Rendering Improvements

When two objects are compared with the == operator, they are unequal, but their string representations are the same, Spock will now print the objects' types:

enteredNumber == 42
|             |
|             false
42 (java.lang.String)

JUnit Fixture Annotations

Fixture methods can now be declared with JUnit’s @Before, @After, @BeforeClass, and @AfterClass annotations, as an addition or alternative to Spock’s own fixture methods. This was particularly needed for Grails 2.0 support.

Tapestry 5.3 Support

Thanks to a contribution from Howard Lewis Ship, the Tapestry module is now compatible with Tapestry 5.3. Older 5.x versions are still supported.

IBM JDK Support

Spock now runs fine on IBM JDKs, working around a bug in the IBM JDK’s verifier.

Improved JUnit Compatibility

org.junit.internal.AssumptionViolatedException is now recognized and handled as known from JUnit. @Unrolled methods no longer cause "yellow" nodes in IDEs.

Improved @Unroll

The @Unroll naming pattern can now be provided in the method name, instead of as an argument to the annotation:

@Unroll
def "maximum of #a and #b is #c"() {
    expect:
    Math.max(a, b) == c

    where:
    a | b | c
    1 | 2 | 2
}

The naming pattern now supports property access and zero-arg method calls:

@Unroll
def "#person.name.toUpperCase() is #person.age years old"() { ... }

The @Unroll annotation can now be applied to a spec class. In this case, all data-driven feature methods in the class will be unrolled.

Improved @Timeout

The @Timeout annotation can now be applied to a spec class. In this case, the timeout applies to all feature methods (individually) that aren’t already annotated with @Timeout. Timed methods are now executed on the regular test framework thread. This can be important for tests that rely on thread-local state (like Grails integration tests). Also the interruption behavior has been improved, to increase the chance that a timeout can be enforced.

The failure exception that is thrown when a timeout occurs now contains the stacktrace of test execution, allowing you to see where the test was “stuck” or how far it got in the allocated time.

Improved Data Table Syntax

Table cells can now be separated with double pipes. This can be used to visually set apart expected outputs from provided inputs:

...
where:
a | b || sum
1 | 2 || 3
3 | 1 || 4

Groovy 1.8/2.0 Support

Spock 0.6 ships in three variants for Groovy 1.7, 1.8, and 2.0. Make sure to pick the right version - for example, for Groovy 1.8 you need to use spock-core-0.6-groovy-1.8 (likewise for all other modules). The Groovy 2.0 variant is based on Groovy 2.0-beta-3-SNAPSHOT and only available from http://m2repo.spockframework.org. The Groovy 1.7 and 1.8 variants are also available from Maven Central. The next version of Spock will no longer support Groovy 1.7.

Grails 2.0 Support

Spock’s Grails plugin was split off into a separate project and now lives at http://github.spockframework.org/spock-grails. The plugin supports both Grails 1.3 and 2.0.

The Spock Grails plugin supports all of the new Grails 2.0 test mixins, effectively deprecating the existing unit testing classes (e.g. UnitSpec). For integration testing, IntegrationSpec must still be used.

IntelliJ IDEA Integration

The folks from JetBrains have added a few handy features around data tables. Data tables will now be layed out automatically when reformatting code. Data variables are no longer shown as "unknown" and have their types inferred from the values in the table (!).

GitHub Repository

All source code has moved to http://github.spockframework.org/. The Grails Spock plugin, Spock Example project, and Spock Web Console now have their own GitHub projects. Also available are slides and code for various Spock presentations (such as this one).

Gradle Build

Spock is now exclusively built with Gradle. Building Spock yourself is as easy as cloning the Github repo and executing gradlew build. No build tool installation is required; the only prerequisite for building Spock is a JDK installation (1.5 or higher).

Fixed Issues

See the issue tracker for a list of fixed issues.