Avoid DependsOn

The task dependsOn method should only be used for lifecycle tasks (tasks without task actions).

Explanation

Tasks with actions should declare their inputs and outputs so that Gradle’s up-to-date checking can automatically determine when these tasks need to be run or rerun.

Using dependsOn to link tasks is a much coarser-grained mechanism that does not allow Gradle to understand why a task requires a prerequisite task to run, or which specific files from a prerequisite task are needed. dependsOn forces Gradle to assume that every file produced by a prerequisite task is needed by this task. This can lead to unnecessary task execution and decreased build performance.

Example

Here is a task that writes output to two separate files:

build.gradle.kts
abstract class SimplePrintingTask : DefaultTask() {
    @get:OutputFile
    abstract val messageFile: RegularFileProperty

    @get:OutputFile
    abstract val audienceFile: RegularFileProperty

    @TaskAction (1)
    fun run() {
        messageFile.get().asFile.writeText("Hello")
        audienceFile.get().asFile.writeText("World")
    }
}

tasks.register<SimplePrintingTask>("helloWorld") { (2)
    messageFile.set(layout.buildDirectory.file("message.txt"))
    audienceFile.set(layout.buildDirectory.file("audience.txt"))
}
build.gradle
abstract class SimplePrintingTask extends DefaultTask {
    @OutputFile
    abstract RegularFileProperty getMessageFile()

    @OutputFile
    abstract RegularFileProperty getAudienceFile()

    @TaskAction (1)
    void run() {
        messageFile.get().asFile.write("Hello")
        audienceFile.get().asFile.write("World")
    }
}

tasks.register("helloWorld", SimplePrintingTask) { (2)
    messageFile = layout.buildDirectory.file("message.txt")
    audienceFile = layout.buildDirectory.file("audience.txt")
}
1 Task With Multiple Outputs: helloWorld task prints "Hello" to its messageFile and "World" to its audienceFile.
2 Registering the Task: helloWorld produces "message.txt" and "audience.txt" outputs.

Don’t Do This

If you want to translate the greeting in the message.txt file using another task, you could do this:

build.gradle.kts
abstract class SimpleTranslationTask : DefaultTask() {
    @get:InputFile
    abstract val messageFile: RegularFileProperty

    @get:OutputFile
    abstract val translatedFile: RegularFileProperty

    init {
        messageFile.convention(project.layout.buildDirectory.file("message.txt"))
        translatedFile.convention(project.layout.buildDirectory.file("translated.txt"))
    }

    @TaskAction (1)
    fun run() {
        val message = messageFile.get().asFile.readText(Charsets.UTF_8)
        val translatedMessage = if (message == "Hello") "Bonjour" else "Unknown"

        logger.lifecycle("Translation: " + translatedMessage)
        translatedFile.get().asFile.writeText(translatedMessage)
    }
}

tasks.register<SimpleTranslationTask>("translateBad") {
    dependsOn(tasks.named("helloWorld")) (2)
}
build.gradle
abstract class SimpleTranslationTask extends DefaultTask {
    @InputFile
    abstract RegularFileProperty getMessageFile()

    @OutputFile
    abstract RegularFileProperty getTranslatedFile()

    SimpleTranslationTask() {
        messageFile.convention(project.layout.buildDirectory.file("message.txt"))
        translatedFile.convention(project.layout.buildDirectory.file("translated.txt"))
    }

    @TaskAction (1)
    void run() {
        def message = messageFile.get().asFile.text
        def translatedMessage = message == "Hello" ? "Bonjour" : "Unknown"

        logger.lifecycle("Translation: " + translatedMessage)
        translatedFile.get().asFile.write(translatedMessage)
    }
}

tasks.register("translateBad", SimpleTranslationTask) {
    dependsOn(tasks.named("helloWorld")) (2)
}
1 Translation Task Setup: translateBad requires helloWorld to run first to produce the message file otherwise it will fail with an error as the file does not exist.
2 Explicit Task Dependency: Running translateBad will cause helloWorld to run first, but Gradle does not understand why.

Do This Instead

Instead, you should explicitly wire task inputs and outputs like this:

build.gradle.kts
abstract class SimpleTranslationTask : DefaultTask() {
    @get:InputFile
    abstract val messageFile: RegularFileProperty

    @get:OutputFile
    abstract val translatedFile: RegularFileProperty

    init {
        messageFile.convention(project.layout.buildDirectory.file("message.txt"))
        translatedFile.convention(project.layout.buildDirectory.file("translated.txt"))
    }

    @TaskAction (1)
    fun run() {
        val message = messageFile.get().asFile.readText(Charsets.UTF_8)
        val translatedMessage = if (message == "Hello") "Bonjour" else "Unknown"

        logger.lifecycle("Translation: " + translatedMessage)
        translatedFile.get().asFile.writeText(translatedMessage)
    }
}

tasks.register<SimpleTranslationTask>("translateGood") {
    inputs.file(tasks.named<SimplePrintingTask>("helloWorld").map { messageFile }) (1)
}
build.gradle
abstract class SimpleTranslationTask extends DefaultTask {
    @InputFile
    abstract RegularFileProperty getMessageFile()

    @OutputFile
    abstract RegularFileProperty getTranslatedFile()

    SimpleTranslationTask() {
        messageFile.convention(project.layout.buildDirectory.file("message.txt"))
        translatedFile.convention(project.layout.buildDirectory.file("translated.txt"))
    }

    @TaskAction (1)
    void run() {
        def message = messageFile.get().asFile.text
        def translatedMessage = message == "Hello" ? "Bonjour" : "Unknown"

        logger.lifecycle("Translation: " + translatedMessage)
        translatedFile.get().asFile.write(translatedMessage)
    }
}

tasks.register("translateGood", SimpleTranslationTask) {
    inputs.file(tasks.named("helloWorld", SimplePrintingTask).map { messageFile }) (1)
}
1 Register Implicit Task Dependency: translateGood requires only one of the files that is produced by helloWorld.

Gradle now understands that translateGood requires helloWorld to have run successfully first because it needs to create the message.txt file which is then used by the translation task. Gradle can use this information to optimize task scheduling. Using the map method avoids eagerly retrieving the helloWorld task until the output is needed to determine if translateGood should run.

References

Tags

#tasks, #inputs-and-outputs, #up-to-date-checking

Favor @CacheableTask and @DisableCachingByDefault over cacheIf(Spec) and doNotCacheIf(String, Spec)

The cacheIf and doNotCacheIf methods should only be used in situations where the cacheability of a task varies between different task instances or cannot be determined until the task is executed by Gradle. You should instead favor annotating the task class itself with @CacheableTask annotation for any task that is always cacheable. Likewise, the @DisableCachingByDefault should be used to always disable caching for all instances of a task type.

Explanation

Annotating a task type will ensure that each task instance of that type is properly understood by Gradle to be cacheable (or not cacheable). This removes the need to remember to configure each of the task instances separately in build scripts.

Using the annotations also documents the intended cacheability of the task type within its own source, appearing in Javadoc and making the task’s behavior clear to other developers without requiring them to inspect each task instance’s configuration. It is also slightly more efficient than running a test to determine cacheability.

Remember that only tasks that produce reproducible and relocatable output should be marked as @CacheableTask.

Example

Don’t Do This

If you want to reuse the output of a task, you shouldn’t do this:

build.gradle.kts
abstract class BadCalculatorTask : DefaultTask() { (1)
    @get:Input
    abstract val first: Property<Int>

    @get:Input
    abstract val second: Property<Int>

    @get:OutputFile
    abstract val outputFile: RegularFileProperty

    @TaskAction
    fun run() {
        val result = first.get() + second.get()
        logger.lifecycle("Result: $result")
        outputFile.get().asFile.writeText(result.toString())
    }
}

tasks.register<Delete>("clean") {
    delete(layout.buildDirectory)
}

tasks.register<BadCalculatorTask>("addBad1") {
    first = 10
    second = 25
    outputFile = layout.buildDirectory.file("badOutput.txt")
    outputs.cacheIf { true } (2)
}

tasks.register<BadCalculatorTask>("addBad2") { (3)
    first = 3
    second = 7
    outputFile = layout.buildDirectory.file("badOutput2.txt")
}
build.gradle
abstract class BadCalculatorTask extends DefaultTask {
    @Input
    abstract Property<Integer> getFirst()

    @Input
    abstract Property<Integer> getSecond()

    @OutputFile
    abstract RegularFileProperty getOutputFile()

    @TaskAction
    void run() {
        def result = first.get() + second.get()
        logger.lifecycle("Result: " + result)
        outputFile.get().asFile.write(result.toString())
    }
}

tasks.register("clean", Delete) {
    delete layout.buildDirectory
}

tasks.register("addBad1", BadCalculatorTask) {
    first = 10
    second = 25
    outputFile = layout.buildDirectory.file("badOutput.txt")
    outputs.cacheIf { true }
}

tasks.register("addBad2", BadCalculatorTask) {
    first = 3
    second = 7
    outputFile = layout.buildDirectory.file("badOutput2.txt")
}
1 Define a Task: The BadCalculatorTask type is deterministic and produces relocatable output, but is not annotated.
2 Mark the Task Instance as Cacheable: This example shows how to mark a specific task instance as cacheable.
3 Forget to Mark a Task Instance as Cacheable: Unfortunately, the addBad2 instance of the BadCalculatorTask type is not marked as cacheable, so it will not be cached, despite behaving the same as addBad1.

Do This Instead

As this task meets the criteria for cacheability (we can imagine a more complex calculation in the @TaskAction that would benefit from automatic work avoidance via caching), you should mark the task type itself as cacheable like this:

build.gradle.kts
@CacheableTask (1)
abstract class GoodCalculatorTask : DefaultTask() {
    @get:Input
    abstract val first: Property<Int>

    @get:Input
    abstract val second: Property<Int>

    @get:OutputFile
    abstract val outputFile: RegularFileProperty

    @TaskAction
    fun run() {
        val result = first.get() + second.get()
        logger.lifecycle("Result: $result")
        outputFile.get().asFile.writeText(result.toString())
    }
}

tasks.register<Delete>("clean") {
    delete(layout.buildDirectory)
}

tasks.register<GoodCalculatorTask>("addGood1") { (2)
    first = 10
    second = 25
    outputFile = layout.buildDirectory.file("goodOutput.txt")
}

tasks.register<GoodCalculatorTask>("addGood2") {
    first = 3
    second = 7
    outputFile = layout.buildDirectory.file("goodOutput2.txt")
}
build.gradle
@CacheableTask (1)
abstract class GoodCalculatorTask extends DefaultTask {
    @Input
    abstract Property<Integer> getFirst()

    @Input
    abstract Property<Integer> getSecond()

    @OutputFile
    abstract RegularFileProperty getOutputFile()

    @TaskAction
    void run() {
        def result = first.get() + second.get()
        logger.lifecycle("Result: " + result)
        outputFile.get().asFile.write(result.toString())
    }
}

tasks.register("clean", Delete) {
    delete layout.buildDirectory
}

tasks.register("addGood1", GoodCalculatorTask) {
    first = 10
    second = 25
    outputFile = layout.buildDirectory.file("goodOutput.txt")
}

tasks.register("addGood2", GoodCalculatorTask) { (2)
    first = 3
    second = 7
    outputFile = layout.buildDirectory.file("goodOutput2.txt")
}
1 Annotate the Task Type: Applying the @CacheableTask to a task type informs Gradle that instances of this task should always be cached.
2 Nothing Else Needs To Be Done: When we register task instances, nothing else needs to be done - Gradle knows to cache them.

References

Tags

#tasks, #caching

Do not call get() on a Provider outside a Task action

When configuring tasks and extensions do not call get() on a provider, use map(), or flatMap() instead.

Explanation

A provider should be evaluated as late as possible. Calling get() forces immediate evaluation, which can trigger unintended side effects, such as:

  • The value of the provider becomes an input to configuration, causing potential configuration cache misses.

  • The value may be evaluated too early, meaning you might not be using the final or correct value of the property. This may lead to painful and hard to debug ordering issues.

  • It breaks Gradle’s ability to build dependencies and to track task inputs and outputs, making automatic task dependency wiring impossible. See Working with task inputs and outputs

It is preferable to avoid explicitly evaluating a Provider at all, and deferring to map/flatMap to connect Providers to Providers implicitly.

Example

Here is a task that writes an input String to a file:

build.gradle.kts
abstract class MyTask : DefaultTask() {
    @get:Input
    abstract val myInput: Property<String>

    @get:OutputFile
    abstract val myOutput: RegularFileProperty

    @TaskAction
    fun doAction() {
        val outputFile = myOutput.get().asFile
        val outputText = myInput.get() (1)
        println(outputText)
        outputFile.writeText(outputText)
    }
}

val currentEnvironment: Provider<String> = providers.gradleProperty("currentEnvironment").orElse("234") (2)
build.gradle
abstract class MyTask extends DefaultTask {
    @Input
    abstract Property<String> getMyInput()

    @OutputFile
    abstract RegularFileProperty getMyOutput()

    @TaskAction
    void doAction() {
        def outputFile = myOutput.get().asFile
        def outputText = myInput.get() (1)
        println(outputText)
        outputFile.write(outputText)
    }
}

Provider<String> currentEnvironment = providers.gradleProperty("currentEnvironment").orElse("234") (2)
1 Using Provider.get() in the task action
2 Gradle property that we wish to use as input

Don’t Do This

You could call get() at configuration time to set up this task:

build.gradle.kts
tasks.register<MyTask>("avoidThis") {
    myInput = "currentEnvironment=${currentEnvironment.get()}"  (1)
    myOutput = layout.buildDirectory.get().asFile.resolve("output-avoid.txt")  (2)
}
build.gradle
tasks.register("avoidThis", MyTask) {
    myInput = "currentEnvironment=${currentEnvironment.get()}"  (1)
    myOutput = new File(layout.buildDirectory.get().asFile, "output-avoid.txt")  (2)
}
1 Reading the value of currentEnvironment at configuration time: This value might change by the time the task start executing.
2 Reading the value of buildDirectory at configuration time: This value might change by the time the task start executing.

Do This Instead

Instead, you should explicitly wire task inputs and outputs like this:

build.gradle.kts
tasks.register<MyTask>("doThis") {
    myInput = currentEnvironment.map { "currentEnvironment=$it" }  (1)
    myOutput = layout.buildDirectory.file("output-do.txt")  (2)
}
build.gradle
tasks.register("doThis", MyTask) {
    myInput = currentEnvironment.map { "currentEnvironment=$it" }  (1)
    myOutput = layout.buildDirectory.file("output-do.txt")  (2)
}
1 Using map() to transform currentEnvironment: map transform runs only when the value is read.
2 Using file() to create a new Provider<RegularFile>: the value of the buildDirectory is only checked when the value of the provider is read.

References

Tags

#tasks, #inputs-and-outputs

Group and Describe custom Tasks

When defining custom task types or registering ad-hoc tasks, always set a clear group and description.

Explanation

A good group name is short, lowercase, and reflects the purpose or domain of the task. For example: documentation, verification, release, or publishing.

Before creating a new group, look for an existing group name that aligns with your task’s intent. It’s often better to reuse an established category to keep the task output organized and familiar to users.

This information is used in the Tasks Report (shown via ./gradlew tasks) to group and describe available tasks in a readable format.

Providing a group and description ensures that your tasks are:

  • Displayed clearly in the report

  • Categorized appropriately

  • Understandable to other users (and to your future self)

Tasks with no group are hidden from the Tasks Report unless --all is specified.

Example

Don’t Do This

Tasks without a group appear under the "other" category in ./gradlew tasks --all output, making them harder to locate:

app/build.gradle.kts
tasks.register("generateDocs") {
    // Build logic to generate documentation
}
app/build.gradle
tasks.register('generateDocs') {
    // Build logic to generate documentation
}
$ gradlew :app:tasks --all

> Task :app:tasks

------------------------------------------------------------
Tasks runnable from project ':app'
------------------------------------------------------------

Other tasks
-----------
compileJava - Compiles main Java source.
compileTestJava - Compiles test Java source.
generateDocs
processResources - Processes main resources.
processTestResources - Processes test resources.
startScripts - Creates OS specific scripts to run the project as a JVM application.

Do this Instead

When defining custom tasks, always assign a clear group and description:

app/build.gradle.kts
tasks.register("generateDocs") {
    group = "documentation"
    description = "Generates project documentation from source files."
    // Build logic to generate documentation
}
app/build.gradle
tasks.register('generateDocs') {
    group = 'documentation'
    description = 'Generates project documentation from source files.'
    // Build logic to generate documentation
}
$ gradlew :app:tasks --all

> Task :app:tasks

------------------------------------------------------------
Tasks runnable from project ':app'
------------------------------------------------------------

Documentation tasks
-------------------
generateDocs - Generates project documentation from source files.
javadoc - Generates Javadoc API documentation for the 'main' feature.

References

Tags

#tasks

Avoid using eager APIs on File Collections

When working with Gradle’s file collection types, be careful to avoid triggering dependency resolution during the configuration phase.

Explanation

Gradle’s Configuration and FileCollection types extend the JDK’s Collection<File> interface.

However, calling some available methods from this interface—such as .size(), .isEmpty(), getFiles(), asPath(), or .toList()—on these Gradle types will implicitly trigger resolution of their dependencies. The same is possible using Kotlin stdlib collection extension methods or Groovy GDK collection extensions. Converting a Configuration to a Set<File> also discards any implicit task dependencies it carries.

You should avoid using these methods when configuring your build. Instead, use the methods defined directly on the Gradle interfaces - this is a necessary first step towards preventing eager resolutions. Be sure to use lazy types and APIs that defer resolution to wire task dependencies and inputs correctly. Some methods that cause resolution are not obvious. Be sure to check the actual behavior when using configurations in an atypical way.

Example

Don’t Do This

build.gradle.kts
abstract class FileCounterTask: DefaultTask() {
    @get:InputFiles
    abstract val countMe: ConfigurableFileCollection

    @TaskAction
    fun countFiles() {
        logger.lifecycle("Count: " + countMe.files.size)
    }
}

tasks.register<FileCounterTask>("badCountingTask") {
    if (!configurations.runtimeClasspath.get().isEmpty()) { (1)
        logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.get().state == RESOLVED))
        countMe.from(configurations.runtimeClasspath)
    }
}

tasks.register<FileCounterTask>("badCountingTask2") {
    val files = configurations.runtimeClasspath.get().files (2)
    countMe.from(files)
    logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.get().state == RESOLVED))
}

tasks.register<FileCounterTask>("badCountingTask3") {
    val files = configurations.runtimeClasspath.get() + layout.projectDirectory.file("extra.txt") (3)
    countMe.from(files)
    logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.get().state == RESOLVED))
}

tasks.register<Zip>("badZippingTask") { (4)
    if (!configurations.runtimeClasspath.get().isEmpty()) {
        logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.get().state == RESOLVED))
        from(configurations.runtimeClasspath)
    }
}
build.gradle
abstract class FileCounterTask extends DefaultTask {
    @InputFiles
    abstract ConfigurableFileCollection getCountMe();

    @TaskAction
    void countFiles() {
        logger.lifecycle("Count: " + countMe.files.size())
    }
}

tasks.register("badCountingTask", FileCounterTask) {
    if (!configurations.runtimeClasspath.isEmpty()) { (1)
        logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.state == RESOLVED))
        countMe.from(configurations.runtimeClasspath)
    }
}

tasks.register("badCountingTask2", FileCounterTask) {
    def files = configurations.runtimeClasspath.files (2)
    countMe.from(files)
    logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.state == RESOLVED))
}

tasks.register("badCountingTask3", FileCounterTask) {
    def files = configurations.runtimeClasspath + layout.projectDirectory.file("extra.txt") (3)
    countMe.from(files)
    logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.state == RESOLVED))
}

tasks.register("badZippingTask", Zip) { (4)
    if (!configurations.runtimeClasspath.isEmpty()) {
        logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.state == RESOLVED))
        from(configurations.runtimeClasspath)
    }
}
1 isEmpty() causes resolution: Many seemingly harmless Collection API methods like isEmpty() cause Gradle to resolve dependencies.
2 Accessing files directly: Using getFiles() to access the files in a Configuration will also cause Gradle to resolve the file collection.
3 Adding a file via plus operator: Using the plus operator will force the runtimeClasspath configuration to be resolved implicitly. The implementation of Configuration doesn’t override the plus operator for regular files, therefore it falls back to using the eager API, which causes resolution.
4 Be careful with indirect inputs: Some built-in tasks, for example subtypes of AbstractCopyTask like Zip, allow adding inputs indirectly and can have the same problems.

Do This Instead

To avoid issues, always defer resolution until the execution phase. Use APIs that support lazy evaluation.

build.gradle.kts
abstract class FileCounterTask: DefaultTask() {
    @get:InputFiles
    abstract val countMe: ConfigurableFileCollection

    @TaskAction
    fun countFiles() {
        logger.lifecycle("Count: " + countMe.files.size)
    }
}

tasks.register<FileCounterTask>("goodCountingTask") {
    countMe.from(configurations.runtimeClasspath) (1)
    countMe.from(layout.projectDirectory.file("extra.txt"))
    logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.get().state == RESOLVED))
}
build.gradle
abstract class FileCounterTask extends DefaultTask {
    @InputFiles
    abstract ConfigurableFileCollection getCountMe();

    @TaskAction
    void countFiles() {
        logger.lifecycle("Count: " + countMe.files.size())
    }
}

tasks.register("goodCountingTask", FileCounterTask) {
    countMe.from(configurations.runtimeClasspath) (1)
    countMe.from(layout.projectDirectory.file("extra.txt")) (2)
    logger.lifecycle("Resolved: " + (configurations.runtimeClasspath.state == RESOLVED))
}
1 Add configurations to Task properties or Specs directly: This will defer resolution until the task is executed.
2 Add files to Specs separately: This allows combining files with file collections without triggering implicit resolutions.

Tags

#tasks, #inputs-and-outputs, #configurations

Don’t resolve Configurations before Task Execution

Resolving configurations before the task execution phase can lead to incorrect results and slower builds.

Explanation

Resolving a configuration - either directly via calling its resolve() method or indirectly via accessing its set of artifacts - returns a set of files that does not preserve references to the tasks that produced those files.

Configurations are file collections and can be added to @InputFiles properties on other tasks. It is important to do this correctly to avoid breaking automatic task dependency wiring between a consumer task and any tasks that are implicitly required to produce the artifacts being consumed. For example, if a configuration contains a project dependency, Gradle knows that consumers of the configuration must first run any tasks that produce that project’s artifacts.

In addition to correctness concerns, resolving configurations during the configuration phase can slow down the build, even when running unrelated tasks (e.g., help) that don’t require the resolved dependencies.

Example

Don’t Do This

build.gradle.kts
dependencies {
    runtimeOnly(project(":lib")) (1)
}

abstract class BadClasspathPrinter : DefaultTask() {
    @get:InputFiles
    var classpath: Set<File> = emptySet() (2)

    private fun calculateDigest(fileOrDirectory: File): Int {
        require(fileOrDirectory.exists()) { "File or directory $fileOrDirectory doesn't exist" }
        return 0 // actual implementation is stripped
    }

    @TaskAction
    fun run() {
        logger.lifecycle(
            classpath.joinToString("\n") {
                val digest = calculateDigest(it) (3)
                "$it#$digest"
            }
        )
    }
}

tasks.register("badClasspathPrinter", BadClasspathPrinter::class) {
    classpath = configurations.named("runtimeClasspath").get().resolve() (4)
}
build.gradle
dependencies {
    runtimeOnly(project(":lib")) (1)
}

abstract class BadClasspathPrinter extends DefaultTask {
    @InputFiles
    Set<File> classpath = [] as Set (2)

    protected int calculateDigest(File fileOrDirectory) {
        if (!fileOrDirectory.exists()) {
            throw new IllegalArgumentException("File or directory $fileOrDirectory doesn't exist")
        }
        return 0 // actual implementation is stripped
    }

    @TaskAction
    void run() {
        logger.lifecycle(
            classpath.collect { file ->
                def digest = calculateDigest(file) (3)
                "$file#$digest"
            }.join("\n")
        )
    }
}

tasks.register("badClasspathPrinter", BadClasspathPrinter) {
    classpath = configurations.named("runtimeClasspath").get().resolve() (4)
}
1 Add project dependency: The :lib project must be built in order to resolve the runtime classpath successfully.
2 Declare input property as Set of files: A simple Set input doesn’t track task dependencies.
3 Dependency artifacts are used to calculate digest: Artifacts from the already resolved classpath are used to calculate the digest.
4 Resolve runtimeClasspath: The implicit task dependency on :library:jar task is lost here when the configuration is resolved prior to task execution. The lib project will not be built when the :app:badClasspathPrinter task is run, leading to a failure in calculateDigest because the lib.jar file will not exist.

Do This Instead

To avoid issues, always defer resolution to the execution phase by using lazy APIs like FileCollection.

build.gradle.kts
dependencies {
    runtimeOnly(project(":lib")) (1)
}

abstract class GoodClasspathPrinter : DefaultTask() {
    @get:InputFiles
    abstract val classpath: ConfigurableFileCollection (2)

    private fun calculateDigest(fileOrDirectory: File): Int {
        require(fileOrDirectory.exists()) { "File or directory $fileOrDirectory doesn't exist" }
        return 0 // actual implementation is stripped
    }

    @TaskAction
    fun run() {
        logger.lifecycle(
            classpath.joinToString("\n") {
                val digest = calculateDigest(it) (3)
                "$it#$digest"
            }
        )
    }
}

tasks.register("goodClasspathPrinter", GoodClasspathPrinter::class.java) {
    classpath.from(configurations.named("runtimeClasspath")) (4)
}
build.gradle
dependencies {
    runtimeOnly(project(":lib")) (1)
}

abstract class GoodClasspathPrinter extends DefaultTask {

    @InputFiles
    abstract ConfigurableFileCollection getClasspath() (2)

    protected int calculateDigest(File fileOrDirectory) {
        if (!fileOrDirectory.exists()) {
            throw new IllegalArgumentException("File or directory $fileOrDirectory doesn't exist")
        }
        return 0 // actual implementation is stripped
    }

    @TaskAction
    void run() {
        logger.lifecycle(
            classpath.collect { file ->
                def digest = calculateDigest(file) (3)
                "$file#$digest"
            }.join("\n")
        )
    }
}

tasks.register("goodClasspathPrinter", GoodClasspathPrinter) {
    classpath.from(configurations.named("runtimeClasspath")) (4)
}
1 Add project dependency: This is the same.
2 Declare input files property as ConfigurableFileCollection: This lazy collection type will track task dependencies.
3 Dependency artifacts are resolved to calculate digest: The classpath will be resolved at execution time to calculate the digest.
4 Configuration is passed to input property directly: Using from causes the configuration to be lazily wired to the input proeprty. The configuration will be resolved when necessary, preserving task dependencies. The output reveals that the lib project is now built when the :app:goodClasspathPrinter task is run because of the implicit task dependency, and the lib.jar file is found when calculating the digest.

Tags

#tasks, #inputs-and-outputs, #configurations