Forking

Introduction

Version 0.5 of sbt adds the ability to fork the Scala compiler and runner for the compile, test-compile, and run tasks. Forking is required when you want to use a version of Scala that sbt does not support (such as Scala trunk). When forking the compiler is enabled, not all sbt features are available. In particular, partial recompilation, testing, and main class detection are not available. This usage is described in the first section.

You might additionally use the forking API for spawning Java processes in general. This usage is described in the second section.

Basic Usage

You can choose to enable forking the compiler, runner, or both. You can change the Java and Scala locations used to build the project and the working directory for the runner.

The method to override in your project definition is fork, which is of type Option[ForkScala]. By default it is None, which indicates forking is disabled. To enable forking, create a new ForkScalaRun, ForkScalaCompiler, or ForkScalaCompiler with ForkScalaRun. For example, to override the working directory used when running, use:

  override def fork = Some(new ForkScalaRun { override def workingDirectory = Some(info.projectPath.asFile) } )

ForkScalaRun and ForkScalaCompiler inherit from ForkScala, which defines the following two methods:

    def javaHome: Option[File] = None
    def scalaJars: Iterable[File] = None

By default, the forked process uses the same Java and Scala that sbt is using. Override the above methods to use a different version. scalaJars needs to include scala-compiler.jar and scala-library.jar.

ForkScalaRun also defines these methods:

    def workingDirectory: Option[File] = None
    def runJVMOptions: Seq[String] = Nil

These allow you to change the working directory when using the run action and to provide options to the forked Java Virtual Machine.

ForkScalaCompiler defines:

    def compileJVMOptions: Seq[String] = Nil

for providing options to the forked Java process for the Scala compiler.

Direct Usage

To fork a new process, use the Fork API. The methods of interest are Fork.java, Fork.javac, Fork.scala, and Fork.scalac. See the ForkJava and ForkScala classes for the arguments and types.

The last argument in all cases is either the Logger to which output should go or the OutputStrategy to use (such as standard output or another OutputStream).