NPEX

NPEX is an automated program repair system for Null Pointer Exceptions (NPE) in Java. The key feature of NPEX is that it automatically inferences repair specification of the buggy program and uses the inferred specification to validate patches. This feature enables NPEX to repair an NPE bug with only the buggy program and the corresponding NPE stack trace. In contrast to NPEX, the state-of-the-art NPE repair techniques requires test case written by developers as additional inputs for patch validation. The key idea of NPEX is to learn a statistical model that predicts how developers would handle NPEs by mining null-handling patterns from existing codebases, and to use a variant of symbolic execution that can infer the repair specification from the buggy program using the model.

Installation Guides

Requirements

• Java 15+
• Maven 3.6+
• Ocaml 4.10+
• Python 3.8+

Building Instructions

1. Clone the source codes from GitHub and go into the cloned directory:

$ git clone https://github.com/kupl/NPEX --recursive && cd ./NPEX

Please don't forget the --recursive option because we manage our specification inference & validation module as git submodules. The option will automatically checkouts its latest version.

2. Build NPEX Java modules. We implemented those modules using the maven build system. Please run the following command:

$ mvn install

3. Build the specification inference & validation module, which is implemented on top of the Infer framework. Please run the following command:

$ cd npex-analyzer && ./build-infer.sh java

Getting Started

We explain how use NPEX to fix a given NPE. Let's consider fixing NPE in aries-jpa_7712046. If you type mvn test at the project, the test will fail with NPE. To fix this NPE, we need to first prepare inputs of NPEX.

Step 1. Preparing Inputs

NPEX takes the following three inputs:

• npe.json: Information of the given NPE.
  • filepath, line, npe_class, npe_method of the given NPE
  • deref_field: last field name of null pointer (e.g., "x" for x, "f" for x.f())
• traces.json: NPE Stack trace.
• Buggy program builded by Infer and Spoon.

For example, NPEX takes the following two inputs for aries-jpa_7712046.

npe.json = {
    "filepath": "jpa-blueprint/src/main/java/org/apache/aries/jpa/blueprint/impl/AnnotationScanner.java",
    "line": 39,
    "npe_class": "AnnotationScanner",
    "npe_method": "parseClass",
    "deref_field": "cl",
}

traces.json =
[
    {
        "filepath": "jpa-blueprint/src/main/java/org/apache/aries/jpa/blueprint/impl/AnnotationScanner.java",
        "line": 39,
        "method_name": "parseClass"
    },
    {
        "filepath": "jpa-blueprint/src/main/java/org/apache/aries/jpa/blueprint/impl/AnnotationScanner.java",
        "line": 33,
        "method_name": "getJpaAnnotatedMembers"
    },
    {
        "filepath": "jpa-blueprint/src/test/java/org/apache/aries/jpa/blueprint/impl/AnnotationScannerTest.java",
        "line": 96,
        "method_name": "getFactoryTest"
    }
]

For npe.json, we need to manually write the information of NPE. For traces.json, we provide a script for parsing the stack trace of NPE. For example, we can get traces.json by the following stack trace and the command.

stack_trace.txt = 
	java.lang.NullPointerException: null
        at org.apache.aries.jpa.blueprint.impl.AnnotationScanner.parseClass(AnnotationScanner.java:39)
        at org.apache.aries.jpa.blueprint.impl.AnnotationScanner.getJpaAnnotatedMembers(AnnotationScanner.java:33)
        at org.apache.aries.jpa.blueprint.impl.AnnotationScannerTest.getFactoryTest(AnnotationScannerTest.java:96)

npex prepare --parse_trace stack_trace.txt

Then, we also need to build the buggy program with Infer and Spoon.

npex prepare --build

Currently, we only support a java program with Maven2 build system.

Step 2. Fault Localization

Then we perform fault localization by the given npe.json and the stack trace traces.json.

npex run --localize

Then, localized NPEs (JSON file with prefix npe_) will be generated. For the example, the following JSONs are generated:

npe_AnnotationScanner.java_30.json
npe_AnnotationScanner.java_32.json 
npe_AnnotationScanner.java_33.json 
npe_AnnotationScanner.java_38.json 
npe_AnnotationScanner.java_39.json

Step 3. Patch Enumeration

Then, we enumerate patches for all the localized NPEs. The following command will takes all npe*.json and enumerate patches.

npex run --enumerate

The generated patches are located at patches directory.

Step 4. Instantiate Null Handling Model

To infer a specification, we need a null handling model which is a map from NPE expression to alternative expression.

npex run --predict

Then, a null handling model is generated at model.json. This procedure uses a learned classifier to instantiate a null handling model for a given program. The learned classifier is located at tools/npex/example/classifier-example

Step 5. Specification Inference & Patch Validation

The last step of NPEX is to infer a specification and validate patches by the inferred specification.

npex run --validate

Then the validation result is stored at result.json. In this example, the results is

{
    "number_of_patches": 27,
    "time_to_verify": 172.13991022109985,
    "time_to_inference": 0.6845083236694336,
    "time_to_capture_original": 0.0,
    "time_to_capture_patches": 162.69369459152222,
    "verified_patches": [
        "SkipBreakStrategy_33-32_1"
    ],
    "rejected_patches": [
        "ReplacePointerStrategy_32_0",
        ...
    ]
}

License

This project is licensed under the terms of MIT license.

Issues

If you encounter any issues, please refer to the Issue page as the solutions already exists.
If you don't get any solutions from the page, please report it as a new issue for public interest.
The email address you can contact is as follows:

• Junhee Lee : junhee_lee@korea.ac.kr
• Seongjoon Hong : seongjoon@korea.ac.kr

We are welcome to any questions or comments for our lovely project!