apache-spark - 使用 Gradle 构建 Spark fat jar : shadow plugin yields corrupted JAR file

Question

我正在尝试使用 Gradle 构建 Spark fat jar。构建成功，但文件以一种微妙的方式损坏:尝试运行它会产生:

Error: Could not find or load main class shadow_test.Main
Caused by: java.lang.ClassNotFoundException: shadow_test.Main

JAR 本身看起来不错:缺少的类就在那里，当我解压缩它时，我可以正常运行该项目。

这是gradle.build文件:

plugins {
    id "scala"
    id 'com.github.johnrengelman.shadow' version '7.1.2'
}

ext {
    ver = [
            scala   : '2.11.12',
            scala_rt: '2.11',
            spark   : '2.4.4'
    ]
}

configurations {
    // Dependencies that will be provided at runtime in the cloud execution
    provided

    compileOnly.extendsFrom(provided)
    testImplementation.extendsFrom provided
}

repositories {
    mavenCentral()
}

dependencies {
    implementation "org.scala-lang:scala-library:$ver.scala"

    provided "org.apache.xbean:xbean-asm6-shaded:4.10"
    provided "org.apache.spark:spark-sql_$ver.scala_rt:$ver.spark"
    provided "org.apache.spark:spark-hive_$ver.scala_rt:$ver.spark"

    testImplementation "org.testng:testng:6.14.3"
}

tasks.register("allJar", com.github.jengelman.gradle.plugins.shadow.tasks.ShadowJar) {
    manifest {
        attributes "Main-Class": "shadow_test.Main"
    }

    from sourceSets.main.output
    configurations = [project.configurations.runtimeClasspath, project.configurations.provided]

    zip64 true

    mergeServiceFiles()

    with jar
}

test {
    useTestNG()
}

Gradle 版本为 7.3.3

重现此问题的最小项目的完整代码可以在 https://github.com/SashaOv/shadow-jar-repro 找到。

感谢您的任何线索

Answer 1

该问题已由 Gradle 社区解决:https://discuss.gradle.org/t/possible-to-build-spark-fat-jar-with-gradle/42235/2?u=sashao

我缺少签名文件的排除:

exclude 'META-INF/*.DSA'
exclude 'META-INF/*.SF'