Gradle Task to check if configured dependencies are available - gradle

I have a project that builds into an SDK library.
Before building a release I would like to verify that all configured dependencies are available using a gradle task. Somehow current build operations pass, because it manages to resolve a different version of a misconfigured dependency as it gets used by some other library. However, this is not a guaranteed situation and therefor I'd like a task that verifies if all configured dependencies are actually available or not and otherwise, fail.
I started a Task like this:
abstract class CheckDependenciesTask : DefaultTask() {
#TaskAction
fun checkDependencies() {
project.configurations.forEach { config ->
if (config.isCanBeResolved) {
println("Resolving ${config.name}")
config.resolve()
} else {
println("Not resolving ${config.name}")
}
}
}
}
But that really aims to resolve configs, rather than dependencies. Somehow I'm not able to figure out how to check if dependencies are available. When running lintAnalyzeDebug --info it neatly prints that they are missing, so there must be some command that checks this, but then just fails instead of try to resolve it.
Any idea or pointers on how to achieve this would be appreciated. Thank you.

Related

How can I generate an entire gradle subproject and have it as a dependency in another project?

So let's say I have the following settings.gradle:
include 'math-server'
project(':math-server').projectDir = file('math/server')
include 'math-client-gen'
project(':math-client-gen').projectDir = file('math/client')
include 'ai'
Now I'd like to not commit the any of the files in math-client-gen (including the build.gradle) since those are generated by a build job in math-server:
// math/server/build.gradle
task generateClient(type:Exec) {
workingDir '.'
inputs.dir('./app')
inputs.dir('.')
outputs.dir('../client')
commandLine './client-generator/generate.sh'
}
The generate.sh leverages the openapi client generator for kotlin.
Now the ai project relies on the math-client-gen:
// ai/build.gradle
dependencies {
compile project(':math-client-gen')
}
Now I have currently found two suboptimal ways to make this work.
Option 1 is to run ./gradlew :math-server:generateClient before I'm able to run ./gradlew :ai:build. This sucks, since you cannot build ai on its own anymore.
Option 2 is to commit the files, which of course also isn't the way it should be.
I'm sure there is a better way to do it with gradle, but I just didn't manage to find it yet. As a compromise, I'd be willing to commit the generated math-client-gen/build.gradle if it doesn't work without that.
What's the best solution to this problem?
Note: I also saw something like:
implementation files(layout.buildDirectory.dir('classes')) {
builtBy 'compile'
}
in the docs, that looks promising, but i'd like to have it for an entire subproject and not just some source files if possible.
// ai/build.gradle
afterEvaluate {
tasks.build.dependsOn(project(":math-client").tasks["generateClient"])
}
To automate your first option.
I ended up committing the build.gradle of the math-client-gen and have this line there:
// math/client/build.gradle
compileKotlin {
dependsOn ':math-server:buildMathClient'
}
This ensures that the client is always generated when this project is listed as a dependency somewhere.
Then you can simply add the math-client-gen as a dependency in other projects, no additional steps required:
// ai/build.gradle
dependencies {
compile project(':math-client-gen')
}

Does Gradle automatically infer dependency between tasks? If so, when?

In my build script, when I configure the downloadAndUnzipFile task, I am explicitly querying output of downloadZipFile task. I expected this is sufficient for Gradle to infer a dependency between the tasks, but it apparently is not, because I get an error when invoking downloadAndUnzipFile`.
Execution failed for task ':downloadAndUnzipFile'.
> Cannot expand ZIP '/home/jdanek/repos/testing/gradle-infer-deps/build/1.0.zip' as it does not exist.
My build script build.gradle.kts is
import de.undercouch.gradle.tasks.download.Download
group = "org.example"
version = "1.0-SNAPSHOT"
plugins {
id("de.undercouch.download").version("4.0.4")
}
tasks {
val downloadZipFile by registering(Download::class) {
src("https://github.com/michel-kraemer/gradle-download-task/archive/1.0.zip")
dest(File(buildDir, "1.0.zip"))
}
val downloadAndUnzipFile by registering(Copy::class) {
from(zipTree(downloadZipFile.get().outputFiles.first()))
into(buildDir)
}
}
I also tried
from(zipTree(downloadZipFile.get().outputFiles.first()))
and that does not define a dependency either.
My Gradle is the latest 6.2.2.
In order for Gradle to discover task dependencies, they have to use specific types for their inputs and outputs so that Gradle can track the dependencies for you. See this documentation page on the topic.
In your use case, the de.undercouch.download plugin seems to expose a simple List<File> which is not a rich type, so Gradle cannot figure out the link. In that case you have be explicit about the task dependency, using dependsOn(downloadZipFile)

WIthin nebula/gradle, how can I inject the version being released into the jar being published?

We have a tool that runs from the command line. One of the commands is -version.
Before we converted to the nebula release plugin, the version was in the gradle.properties file, and as part of the build we copied it from there to a src/main/resources/version.txt file, that was later read by the tool to output the version.
But now the version is never in a file that's checked into git. Instead, it is only known during the nebula release process.
We want to obtain that version during the nebula release process and inject it into the jar that nebula is about to publish. For example, it could be added to the manifest.
We've tried to figure out how to do this, but don't see any examples online, and nothing about it in the documentation.
Simply create a task that caches the version that is dynamically inferred by Nebula.
Since you originally copied/created src/main/resources/version.txt, we'll use that that model our task.
Assuming a simple/standard Java project, using the Kotlin DSL:
val cacheNebulaVersion by tasks.registering {
mustRunAfter(tasks.named("release"))
doLast {
val sourceSets = project.extensions.getByName("sourceSets") as SourceSetContainer
sourceSets.getByName(SourceSet.MAIN_SOURCE_SET_NAME).output.resourcesDir?.let {
// If there are not existing resources in your project then you must create
// the resources dir otherwise a FileNotFoundException will be thrown.
if (!it.exists()) {
it.mkdirs()
}
File(it, "version.txt").printWriter().use { out ->
out.println(project.version)
}
}
}
}
When I invoke ./gradlew clean build snapshot cacheNebulaVersion, the version produced by Nebula is cached/created at src/main/resources/version.txt in the build output. The task above does not bundle it with the jar.
Hopefully that gives you an idea what to do.

Changing configuration with the gretty plugin?

I haven't done anything with Gradle for a while, so it appears I've forgotten how configuration resolution works.
I'm trying to use the gretty plugin (instead of core, deprecated jetty), but I cannot seem to create a custom configuration.
I've boiled it down to a very short, simple script (using Gradle 3.4):
buildscript {
repositories {
maven {
url 'https://plugins.gradle.org/m2/'
}
}
dependencies {
classpath 'org.akhikhl.gretty:gretty:1.4.0'
}
}
plugins {
id 'org.akhikhl.gretty' version '1.4.0'
}
configurations {
fooTest
}
configurations.fooTest.each {
println it.toString()
}
It seems to not like me iterating over the fooTest configuration.
Assuming I need to know the dependencies for that configuration (I stripped that part from the code above)
What am I doing wrong here?
The script above gives me this:
org.gradle.api.InvalidUserDataException: Cannot change strategy of configuration ':fooTest' after it has been resolved.
The key point here was that I needed an unresolved configuration to loop over. Admittedly this information was neglected in the initial description as I didn't know it was critical information. We needed to loop over the files in the dependency and copy/unzip them into certain locations.
However, we cannot do that with a resolved configuration. That said, we can copy the configuration into a unresolved one, and loop over that instead:
configurations.fooTest.copy().each {
println it.toString()
}
This will successfully print out the files involved in the dependency (or unzip them, as my case needs).

Why does gradle resolve dependencies at configure time?

We have some "heavy weight" artifacts, and we don't necessarily always need them. If we have some tasks with dependencies to them, they will get downloaded, regardless of the task which we invoke.
It would be nice if gradle only downloaded dependencies when it actually needs them.
Example
configurations {
webresources
}
dependencies {
webresources my:hugeIconCollection:2.0#zip
}
def unpackIcons = tasks.register('unpackIcons', Copy) {
configurations.webresources.resolveConfiguration.resolvedArtifacts.each {
from zipTree(it.file)
into 'resources/path'
}
}
Now simply running gradle help will already download the hugeIconCollection.zip.
The code uses the "Configuration Avoidance API", but it doesn't help. Still it would be desirable that the resource only be downloaded when the unpackIcons task is run (or a task that dependsOn it), and then the file be unzipped into the destination path.
Gradle downloads dependencies just-in-time on their first use. If your build downloads dependencies at configuration time, then because your build scripts/plugins are using them at configuration time. In most cases this indicates a problem with the build scripts/plugins.

Resources