Skip to content

Latest commit

 

History

History
66 lines (42 loc) · 3.5 KB

README.md

File metadata and controls

66 lines (42 loc) · 3.5 KB
Raw

IteRace

IteRace is a static race detection tool developed at University of Illinois. Static race detectors suffer from imprecision (due to conservative assumptions) which usually manifests itself in an unmanageable number of warnings that the programmer needs to inspect. IteRace tackles this problem by specialization:

  • it is aware of and uses to its advantage the thread and data-flow structure of loop-parallel operations. As parallel collections are not yet available in Java (they will be in Java8 next year), IteRace analyzes a collection which mostly follows ParallelArray's API
  • reports races in application, not library, code. For example, you don't have to track down races occurring in HashSet - all it will tell you is that you have inadvertently shared a particular HashSet object.
  • filters races based on a thread-safety model of classes. It is slightly more involved then that, but, for example, you won't get race reports on an AtomicInteger.

You can find more details in our ISSTA '13 paper.

Getting started

IteRace is implemented in Scala, relies on WALA as the underlying analysis engine, and uses SBT for building.

Steps:

  1. Make sure you have Scala 2.10, Maven, Apache Ivy, and SBT 0.13

  2. Clone WALA and WALAFacade

  3. Clone cos/Util

    • git clone https://github.com/cos/Util.git
    • cd Util
    • sbt publishLocal

  4. Clone IteRace and compile

    • git clone https://github.com/cos/IteRace.git
    • cd IteRace
    • sbt compile

  5. Edit /src/main/resources/local.config to point to your Java library jar

If you use Eclipse, you can sbt eclipse and import the project into your workspace (dependencies will be linked to the Ivy repo).

Using IteRace

As a library

See this example project

As a standalone tool

SBT puts the compiled binary in the /target directory. You can play with IteRace using its Interactive tool (iterace.Interactive main class). It accepts configuration options as arguments, or you can provide it a configuration file via the configFile=... argument.

Configuration options

Examples:

  • wala.entry.class="NBodySimulation"
  • wala.entry.method = "main([Ljava/lang/String;)V" (also the default)
  • iterace.races-file="test.races"
  • iterace.log-file="verbose.log"
  • wala.dependencies.binary+="path/to/class/folder"
  • wala.dependencies.jar+="path/to/a/jar"
  • wala.dependencies.source+="path/to/sources" (WALA is sometimes fickle about loading sources)
  • wala.exclussions="javax\/.*"

See src/main/resources for the default option values.

Also

We haven't yet made the tool as user-friendly as we would like to. If you encounter any problem with the setup or use of the tool, report it here and I'll try to solve it quickly.

This git repo contains the workspace we used for running the benchmarks and gathering the results presented in the paper. It is not as user friendly as we would like but we'll improve it shortly.