[Java Integration]: Add in WALA support

[arthurscchan]

Add in WALA support as an alternative to generate more static analysing data for java code.
This PR is targeting step 2 and 3 from Issue #536

Signed-off-by: Arthur Chan [email protected]

[arthurscchan]

Node: synthetic < Primordial, Lcom/ibm/wala/FakeRootClass, fakeRootMethod()V > Context: Everywhere
 - invokestatic < Primordial, Lcom/ibm/wala/FakeRootClass, fakeWorldClinit()V >@0
   -> Node: synthetic < Primordial, Lcom/ibm/wala/FakeRootClass, fakeWorldClinit()V > Context: Everywhere
 - invokespecial < Primordial, Ljava/lang/Object, <init>()V >@4
   -> Node: < Primordial, Ljava/lang/Object, <init>()V > Context: Everywhere
 - invokestatic < Application, LTestFuzzer, main([Ljava/lang/String;)V >@5
   -> Node: < Application, LTestFuzzer, main([Ljava/lang/String;)V > Context: Everywhere
Node: synthetic < Primordial, Lcom/ibm/wala/FakeRootClass, fakeWorldClinit()V > Context: Everywhere
 - invokestatic < Primordial, Ljava/lang/Object, <clinit>()V >@0
   -> Node: < Primordial, Ljava/lang/Object, <clinit>()V > Context: Everywhere
 - invokestatic < Primordial, Ljava/lang/String, <clinit>()V >@1
   -> Node: < Primordial, Ljava/lang/String, <clinit>()V > Context: Everywhere
Node: < Primordial, Ljava/lang/Object, <clinit>()V > Context: Everywhere
 - invokestatic < Primordial, Ljava/lang/Object, registerNatives()V >@0
   -> Node: < Primordial, Ljava/lang/Object, registerNatives()V > Context: Everywhere
Node: < Primordial, Ljava/lang/Object, registerNatives()V > Context: Everywhere
Node: < Primordial, Ljava/lang/String, <clinit>()V > Context: Everywhere
 - invokespecial < Primordial, Ljava/lang/String$CaseInsensitiveComparator, <init>(Ljava/lang/String$1;)V >@12
   -> Node: < Primordial, Ljava/lang/String$CaseInsensitiveComparator, <init>(Ljava/lang/String$1;)V > Context: Everywhere
Node: < Primordial, Ljava/lang/String$CaseInsensitiveComparator, <init>(Ljava/lang/String$1;)V > Context: Everywhere
 - invokespecial < Primordial, Ljava/lang/String$CaseInsensitiveComparator, <init>()V >@1
   -> Node: < Primordial, Ljava/lang/String$CaseInsensitiveComparator, <init>()V > Context: Everywhere
Node: < Primordial, Ljava/lang/String$CaseInsensitiveComparator, <init>()V > Context: Everywhere
 - invokespecial < Primordial, Ljava/lang/Object, <init>()V >@1
   -> Node: < Primordial, Ljava/lang/Object, <init>()V > Context: Everywhere
Node: < Primordial, Ljava/lang/Object, <init>()V > Context: Everywhere
Node: < Application, LTestFuzzer, main([Ljava/lang/String;)V > Context: Everywhere
 - invokestatic < Application, LTestFuzzer, fuzzerTestOneInput(Lcom/code_intelligence/jazzer/api/FuzzedDataProvider;)V >@9
   -> Node: < Application, LTestFuzzer, fuzzerTestOneInput(Lcom/code_intelligence/jazzer/api/FuzzedDataProvider;)V > Context: Everywhere
Node: < Application, LTestFuzzer, fuzzerTestOneInput(Lcom/code_intelligence/jazzer/api/FuzzedDataProvider;)V > Context: Everywhere

The above is a dump of the CallGraph object generated when statically analyse the following program.

import com.code_intelligence.jazzer.api.FuzzedDataProvider;
import com.code_intelligence.jazzer.api.CannedFuzzedDataProvider;


public class TestFuzzer {
        public static void fuzzerTestOneInput(FuzzedDataProvider data) {
                return;
        }

        public static void main(String[] args) {
                TestFuzzer.fuzzerTestOneInput(new CannedFuzzedDataProvider("RANDOM"));
        }
}

[arthurscchan]

Some findings right now

1. The call graph will shows everything it goes through, including the WALA logic itself, it is possible to use exclude options to exclude result from files in certain java package. This has been confirmed.
2. By default, it will look for a public class with main method to act as starting point for the analysis. This could be specified by commands. Investigating different settings to allow this.
3. The call graph object store much more information, including method signature and line number in source code, that could be retrieved through different method. Will read the javadoc and investigate on this after lunch.

Javadoc of the WALA is at follows
https://wala.github.io/javadoc/

[DavidKorczynski]

3. The call graph object store much more information, including method signature and line number in source code, that could be retrieved through different method.

This is promising

[DavidKorczynski]

Am not sure about legalities with license, but I think it might be best to keep all that is WALA licensed (eclipse license) in a separate repository, similar to what we did with PyCG for the Python frontend. Can we move some of this to Ada Logics for now?