README.md

DEVLoRe

Our experimental data is available at release page

Useful links: MethodRecorder, DebugRecorder

How to use

• Step 1. Clone project

git clone https://github.com/XYZboom/DEVLoRe.git

• Step 2. Install python requirements

pip install -r requirements.txt

• Step 3. Prepare Defects4j environment

Just follow the readme in Defects4j (https://github.com/rjust/defects4j)

Note that we currently support Defects4j v2.0.0 only.

• Step 4. Download tools

Download tools from release page.

• Step 5. Setup environment variables
• Create a ".env" file at the root of this project.
• Set environment variables like this:

OPENAI_API_KEY=your_openai_api_key
DEFECTS4J_EXEC=/path/to/defects4j/framework/bin/defects4j
OUTPUT_PATH=where_to_output
D4J_JSON_PATH=where_to_find_baseline_buggy_locations
EXTRACT_JAR_PATH=/path/to/RelatedMethodExtractor.jar

• Step 6. Run data preparation scripts

After you finish the setup of Defects4J, you can see an "ant" file at {Defects4J_HOME}/major/bin/ant. Add MethodRecorder as a java agent commandline argument in this file. Remember to back up this file!

#!/bin/sh
# This is not a script to directly run! 
# This is a example how to add MethodRecorder as a java agent in Defects4J
BASE="`dirname $0`/.."

if [ -z "$JAVA_HOME" ]; then
    CMD="java"
else
    CMD="$JAVA_HOME/bin/java"
fi

# Add this "-javaagent:..." line
$CMD \
    -javaagent:/path/to/MethodRecorder.jar=./temp.properties -Xverify:none \ 
    -XX:ReservedCodeCacheSize=256M \
    -XX:MaxPermSize=1G \
    -Djava.awt.headless=true \
    -Xbootclasspath/a:$BASE/config/config.jar \
    -jar $BASE/lib/ant-launcher.jar $*

After this, run ExtractSkeleton.py to get method skeleton.

• Step 7. Run modules

• Locate buggy method

python LocateMethod.py --add-issue-info=True --add-stack-info=True
# --add-issue-info=True add issue info in prompt
# --add-stack-info=True add stack info in prompt

Issue information should put in {OUTPUT_PATH}/issue_content. You can find issue content extract by us in our data zip file. You can also extract issue content by your own, for example, issue url of project "Lang" in Defects4J is located at here.

• Extract buggy method content

Run ExtractMethodAndField.py to extract buggy method content.

The commandline arguments must be same as the ones when running LocateMethod.py.

That is: if you run LocateMethod.py with "--add-issue-info=True" and without "--add-issue-info=True", then you must run ExtractMethodAndField.py with "--add-issue-info=True" and without " --add-issue-info=True".

• Extract debugging information

If you don't want provide debugging information, just see the next step "Locate buggy line".

After you finish the setup of Defects4J, you can see an "ant" file at {Defects4J_HOME}/major/bin/ant. Add DebugRecorder as a java agent commandline argument in this file. Remember to back up this file!

#!/bin/sh
# This is not a script to directly run! 
# This is a example how to add DebugRecord as a java agent in Defects4J
BASE="`dirname $0`/.."

if [ -z "$JAVA_HOME" ]; then
    CMD="java"
else
    CMD="$JAVA_HOME/bin/java"
fi

# Add this "-javaagent:..." line
$CMD \
    -javaagent:/path/to/DebugRecorder.jar=./temp.properties -Xverify:none \ 
    -XX:ReservedCodeCacheSize=256M \
    -XX:MaxPermSize=1G \
    -Djava.awt.headless=true \
    -Xbootclasspath/a:$BASE/config/config.jar \
    -jar $BASE/lib/ant-launcher.jar $*

Run ExtractDebugInfo.py with same commandline arguments above.

• Locate buggy line

python LocateLine.py --add-issue-info=True --add-stack-info=True --add-debug-info=True
# --add-debug-info=True add debugging info in prompt

• Repair

python Repair.py --add-issue-info=True --add-stack-info=True --add-debug-info=True

• Evaluate RESTORE "ant" file to its ORIGINAL STATE before evaluate!

python Evaluate.py --add-debug-info=True
  --patch-valid=False
  --add-issue-info=True
  --add-stack-info=True
# set --patch-valid=True to extract patches from evaluate result.