When you have a fully up to date build and you clean and re-run the code generator task on the same code base it should generate _exactly the same output_, so anything that depends on that output will stay up-to-date.

It might also be that your code generator adds some extra information to its output that doesn't depend on its declared inputs, like a timestamp.

        "both have (but zip provider missing)" | Integer.MAX_VALUE | 0        | 88         | 0         | null
        "only left has"                        | 23                | 0        | null       | 0         | null
        "only right has"                       | null              | 0        | 88         | 0         | null
        "none have"                            | null              | 0        | null       | 0         | null
    }

    }

    @Nullable
    ExcludeSpec tryIntersect(ExcludeSpec left, ExcludeSpec right) {
        if (left.equals(right)) {
            return left;
        } else {
            return intersections.stream()
                .filter(i -> i.applies(left, right))
                .findFirst()
                .map(i -> i.intersect(left, right, factory))
                .orElse(null);
        }
    }

    @NonNullApi

 * 2. UPDATE build-logic/lifecycle/src/main/kotlin/PrintStackTracesOnTimeoutBuildService.kt if this class is moved to another package.
 *
 * Used to print all JVMs' thread dumps on the machine. When it starts working, the process/machine might be in a bad state (e.g. deadlock),
 * So we don't want it to depend on any third-party libraries. It's executed directly via `java JavaProcessStackTracesMonitor` in `PrintStackTracesOnTimeoutBuildService`.
 *

    }

    private void recordTopLevelModules(FileCopyDetails details) {
        DeploymentDescriptor deploymentDescriptor = getDeploymentDescriptor();
        // since we might generate the deployment descriptor, record each top-level module
        if (deploymentDescriptor != null && details.getPath().lastIndexOf("/") <= 0) {
            EarModule module;

  virtual const string& GetTestCaseName() const = 0;
  // Test case id to verify identity.
  virtual TypeId GetTestCaseTypeId() const = 0;
  // UnitTest class invokes this method to register tests in this
  // test case right before running them in RUN_ALL_TESTS macro.
  // This method should not be called more then once on any single
  // instance of a ParameterizedTestCaseInfoBase derived class.
  virtual void RegisterTests() = 0;

        project.plugins.hasPlugin(ReportingBasePlugin)
    }

    def "pmd 7.0.0 is loading the right dependencies"() {
        expect:
        PmdPlugin.calculateDefaultDependencyNotation("7.0.0") == ImmutableSet.of("net.sourceforge.pmd:pmd-java:7.0.0", "net.sourceforge.pmd:pmd-ant:7.0.0")
    }

    def "pmd 7.0.0-rc4 is loading the right dependencies"() {
        expect:

     *
     * We currently can't filter single unrolled methods due to a limitation with Spock and JUnit 4/5
     * So we handle selecting the right methods by some custom code in our performance test infrastructure (`TestScenarioSelector.shouldRun()`).
     * The classes filters are still handled by Gradle's/JUnit's infrastructure.
     */

        then:
        def messages = models.left
        messages.size() == 2
        messages[0] == "It works from project :"
        messages[1] == "It works from project :a"
        def model = models.right
        model.size() == 2
        model[0].message == "It works from project :"
        model[1].message == "It works from project :a"

        and:
        fixture.assertStateStored {

        add {
            failOnText {
                conditionType = BuildFailureOnText.ConditionType.CONTAINS
                pattern = "%unmaskedFakeCredentials%"
                failureMessage = "This build might be leaking credentials"
                reverse = false
                stopBuildOnFailure = true
            }
        }
    }
}