Checker Framework Inference

This project is a general type inference framework, built upon the EISOP Checker Framework.

Given a program with no type annotations, Checker Framework Inference produces a program with type annotations.

By contrast, given a program with type annotations, the Checker Framework determines verifies the program's correctness or reveals errors in it.

Developer Notes

If you want to extend the framework for your own type system or add additional constraint solvers, please send us mail.

The checker-framework-inference Google Drive folder contains additional documents for developers:

https://drive.google.com/drive/u/1/folders/0B7vOZvme6aAOfjQ0bWlFU1VoeVZCVjExVmJLM0lGY3NBV0VLcENYdm03c0RCNGFzZURHX2c

That information is being moved to here in the repository.

Configure Eclipse to edit Checker Framework Inference

The instructions here assumes you have cloned this project into a folder called checker-framework-inference.

1. Follow the instructions in the EISOP Checker Framework Manual to download, build, and configure Eclipse to edit the Checker Framework. The Checker Framework Inference Eclipse project depends on the eclipse projects from Checker Framework.

2. Follow the instructions below to build Checker Framework Inference

3. Build the dependencies jar file:

./gradlew dependenciesJar

4. Enter the main Eclipse working screen and in the “File” menu, select “Import” -> “General” -> “Existing Projects into workspace”. After the “Import Projects” window appears, select “Select Root Directory”, and select the checker-framework-inference directory. Eclipse should successfully build all the imported projects.

Requirements

You will need a JDK (version 8) and gradle.

Following the instructions in the Checker Framework manual to install the Checker Framework from source.

NOTE: gradle on Ubuntu 14.10 hard-codes JAVA_HOME. To change this, edit /usr/share/gradle/bin/gradle and replace

    export JAVA_HOME=/usr/lib/jvm/default-java

with

    [ -n "$JAVA_HOME" ] || export JAVA_HOME=/usr/lib/jvm/default-java

Building

To clone and build all dependencies:

./gradlew cloneAndBuildDependencies

To build:

./gradlew dist

To test the build:

./gradlew testCheckerInferenceScript
./gradlew testCheckerInferenceDevScript
./gradlew testDataflowExternalSolvers

Execution

Verify you have all the requirements.

./scripts/inference

is the script used to run inference.

Example:

./scripts/inference --logLevel FINE --mode ROUNDTRIP --checker ostrusted.OsTrustedChecker --solver checkers.inference.solver.PropagationSolver -afud /path/to/Annotation/File/Utilities/output/directory [List of files]

There are a couple of required options:

• --mode Specifies what the tools should do. Available options are [INFER, TYPECHECK, ROUNDTRIP, ROUNDTRIP_TYPECHECK]

  • INFER: Generates and solves the constraints and writes the results to default.jaif file

  • TYPECHECK: Typechecks the existing code

  • ROUNDTRIP: Generates and solves the constraints and then inserts the results back into the original source code

  • ROUNDTRIP_TYPECHECK: Executes roundtrip and then typechecks the result

• --checker Specifies which checker to run. The three most supported checkers at the moment are ostrusted.OsTrustedChecker, checkers.tainting.TaintingChecker and dataflow.DataflowChecker.

  You can find details of dataflow.DataflowChecker in README.dataflow

• --solver Which solver to use on the constraints.

• --targetclasspath The classpath that is required by target program.

checkers.inference.solver.PropagationSolver and checkers.inference.solver.SolverEngine are real solvers at the moment.

Omitting the solver will create an output that numbers all the annotation positions in the program.

checkers.inference.solver.DebugSolver will output all the constraints generated.

Other options can be found by ./scripts/inference --help.

Use of SolverEngine

Generic solver is designed for solving type constraints from arbitrary type system.

You can invoke generic solver through:

--solver checkers.inference.solver.SolverEngine

There are a couple of arguments that generic solver can accept:

• solver Specifies what concrete solver is going to use.

  At this moment, we have below available back ends:

  • MaxSat: Encodes constraints as Max-SAT problem and use Sat4j library to solve.

  • Lingeling: Encodes constraints as SAT problem and use Lingeling solver to solve.

  • LogiQL: Encodes constraints as statements of LogiQL language and use LogicBlox to solve.

  • Z3 with bit vector theory: Encodes constraints as Max-SMT problem with bit vector theory, and use Z3 library to solve.

  MaxSat solver is used by default.

• useGraph Specifies whether to separate constraints into multiple components through constraint graph and solve them respectively. The default value is true.

• solveInParallel If constraints are separated by constraint graph, this argument indicates whether to solve the components in parallel (multithreading). The default value is true.

• collectStatistics Specifies whether to collect statistic with respect to timing, size of constraints, size of encoding, etc. The default value is false.

For example, generic solver can be invoked through following command:

./scripts/inference --mode INFER --checker ostrusted.OsTrustedChecker --solver checkers.inference.solver.SolverEngine --solverArgs solver=MaxSat,useGraph=true,collectStatistics=true,solveInParallel=false [List of files]