D4: Fast Concurrency Debugging with Parallel Differential Analysis

D4 is a tool that detects data races and deadlocks interactively in the Eclipse IDE for Java. For most code changes, it finishes in less than 0.1s after the change is introduced. A video demo of D4 is available below:

Software Dependencies

• Java 1.8 to compile
• Eclipse Mars, Photon
• Eclipse PDE
• WALA 1.5.0
• Incremental Pointer Analysis(submodule)
• Akka (included)

Build and Run D4

git clone https://github.com/wala/WALA

git clone --recurse-submodules https://github.com/parasol-aser/D4

Import all the projects into Eclipse, and launch the plugin: /edu.tamu.cse.aser.d4 -> MANIFEST.MF -> Testing -> Launch an Eclipse application.

In the newly launched Eclipse application, right-click a main class in Package Explorer, select ASER -> D4.

The incremental detection can be triggered by saving the changed files.

** Add the following method to the com.ibm.wala.cast.loader.AstClass:

  public void updateMethod(Selector selector, IMethod method) {
    if (declaredMethods.containsKey(selector)) {
      declaredMethods.put(selector, method);
    } else if (getSuperclass() != null) {
      if(getSuperclass() instanceof AstClass){
        ((AstClass) getSuperclass()).updateMethod(selector,method);
      }
    }
  }

D4 views

Go to Window -> Show View -> Others to display D4 views (i.e., D4 Concurrent Relations, D4 Race List, D4 Deadlock List).

Ignore certain variable or method

To exclude variables or methods from being detected by D4, right-click the variables/methods shown in the Outline view or the bugs/relations, select D4 -> Ignore This Variable/Method. To add them back, select D4 -> Consider This Variable/Method.

D4 Docker Image

A docker image running D4 on a local machine is available, run docker pull aprildocker/d4_ubuntu_java8:firsttry

Authors

Bozhen Liu, Texas A&M University

Jeff Huang, Texas A&M University

Paper

D4 is powered by a set of parallel incremental algorithms for pointer analysis and happens-before analysis. You can find the technical details in our PLDI'18 paper. Our incremental pointer analysis code has also been integrated into the popular WALA framework.

If you use this resource, please cite our paper:

[PLDI'18] "D4: Fast Concurrency Debugging with Parallel Differential Analysis"