Impurity annotation

scripts/Makefile

@@ -1,6 +1,7 @@
 MAKEFILE_DIR := $(dir $(realpath $(lastword $(MAKEFILE_LIST))))
 
 all: \
+    build/checker-framework/checker/dist/checker-qual.jar \
     build/evosuite-1.2.0.jar \
     build/evosuite-standalone-runtime-1.2.0.jar \
     build/hamcrest-core-1.3.jar \
@@ -64,13 +65,21 @@ build/junit-4.12.jar:
 
 build/hamcrest-core-1.3.jar:
 	mkdir -p build
-	cp ../subject-programs/hamcrest-core-1.3.jar build/
+	cp ../subject-programs/jars/hamcrest-core-1.3.jar build/
 
 build/major/bin/ant-replacecall:
 	LATEST_RELEASE=$$(curl -s https://api.github.com/repos/randoop/grt-replacecall/releases/latest | grep -o '"tag_name": "[^"]*' | cut -d'"' -f4) && \
 	wget "https://github.com/randoop/grt-replacecall/releases/download/$${LATEST_RELEASE}/ant-replacecall" && \
 	mv ant-replacecall build/major/bin/
 
+build/checker-framework/checker/dist/checker-qual.jar:
+	mkdir -p build
+	CFVER=3.50.0 && \
+	cd build && wget "https://github.com/typetools/checker-framework/releases/download/checker-framework-$${CFVER}/checker-framework-$${CFVER}.zip" && \
+	unzip -o checker-framework-$${CFVER}.zip && \
+	rm -rf checker-framework && \
+	mv checker-framework-$${CFVER} checker-framework
+
 build/defects4j:
 	mkdir -p build
 	cd build && git clone https://github.com/rjust/defects4j && cd defects4j && cpanm --installdeps . && ./init.sh

scripts/mutation-evosuite.sh

@@ -172,7 +172,7 @@ echo
 SRC_BASE_DIR="$(realpath "$SCRIPT_DIR/../subject-programs/src/$SUBJECT_PROGRAM")"
 
 # Path to the jar file of the subject program.
-SRC_JAR=$(realpath "$SCRIPT_DIR/../subject-programs/$SUBJECT_PROGRAM.jar")
+SRC_JAR=$(realpath "$SCRIPT_DIR/../subject-programs/jars/$SUBJECT_PROGRAM.jar")
 
 # Number of classes in given jar file.
 NUM_CLASSES=$(jar -tf "$SRC_JAR" | grep -c '.class')
@@ -318,7 +318,7 @@ case "$SUBJECT_PROGRAM" in
       "$SRC_BASE_DIR/lib/jaxen-core.jar" \
       "$SRC_BASE_DIR/lib/jaxen-jdom.jar" \
       "$SRC_BASE_DIR/lib/saxpath.jar" \
-      "$SCRIPT_DIR/../subject-programs/jaxen-1.1.6.jar"
+      "$SCRIPT_DIR/../subject-programs/jars/jaxen-1.1.6.jar"
     ;;
 
   "joda-time-2.3")

scripts/mutation-randoop.sh

@@ -56,12 +56,12 @@ set -o pipefail
 Generator=Randoop
 generator=randoop
 SCRIPT_DIR="$(CDPATH='' cd -- "$(dirname -- "$0")" && pwd -P)"
-SCRIPT_NAME=$(basename -- "$0")
-MAJOR_HOME=$(realpath "${SCRIPT_DIR}/build/major/")                     # Major home directory, for mutation testing
-RANDOOP_JAR=$(realpath "${SCRIPT_DIR}/build/randoop-all-4.3.4.jar")     # Randoop jar file
-JACOCO_AGENT_JAR=$(realpath "${SCRIPT_DIR}/build/jacocoagent.jar")      # For Bloodhound
-JACOCO_CLI_JAR=$(realpath "${SCRIPT_DIR}/build/jacococli.jar")          # For coverage report generation
-REPLACECALL_JAR=$(realpath "${SCRIPT_DIR}/build/replacecall-4.3.4.jar") # For replacing undesired method calls
+MAJOR_HOME=$(realpath "${SCRIPT_DIR}/build/major/")                                                # Major home directory, for mutation testing
+RANDOOP_JAR=$(realpath "${SCRIPT_DIR}/build/randoop-all-4.3.4.jar")                                # Randoop jar file
+JACOCO_AGENT_JAR=$(realpath "${SCRIPT_DIR}/build/jacocoagent.jar")                                 # For Bloodhound
+JACOCO_CLI_JAR=$(realpath "${SCRIPT_DIR}/build/jacococli.jar")                                     # For coverage report generation
+REPLACECALL_JAR=$(realpath "${SCRIPT_DIR}/build/replacecall-4.3.4.jar")                            # For replacing undesired method calls
+CHECKER_QUAL_JAR=$(realpath "${SCRIPT_DIR}/build/checker-framework/checker/dist/checker-qual.jar") # For Randoop Impurity
 
 . "$SCRIPT_DIR/defs.sh" # Define shell functions.
 
@@ -70,6 +70,7 @@ require_file "$RANDOOP_JAR"
 require_file "$JACOCO_AGENT_JAR"
 require_file "$JACOCO_CLI_JAR"
 require_file "$REPLACECALL_JAR"
+require_file "$CHECKER_QUAL_JAR"
 
 usejdk8
 JAVA_VER=$(java -version 2>&1 | awk -F '"' '/version/ {print $2}' | awk -F '.' '{sub("^$", "0", $2); print ($1=="1")?$2:$1}')
@@ -220,7 +221,13 @@ echo
 SRC_BASE_DIR="$(realpath "$SCRIPT_DIR/../subject-programs/src/$SUBJECT_PROGRAM")"
 
 # Path to the jar file of the subject program.
-SRC_JAR=$(realpath "$SCRIPT_DIR/../subject-programs/$SUBJECT_PROGRAM.jar")
+if [[ -n "${FEATURE_FLAGS[GRT_FUZZING]}" ]]; then
+  # If randoop features contain "GRT_FUZZING", use annotated subject program jar.
+  SRC_JAR=$(realpath "$SCRIPT_DIR/../subject-programs/annotated-jars/$SUBJECT_PROGRAM.jar")
+else
+  # Else, use the original subject program jar.
+  SRC_JAR=$(realpath "$SCRIPT_DIR/../subject-programs/jars/$SUBJECT_PROGRAM.jar")
+fi
 
 # Number of classes in given jar file.
 NUM_CLASSES=$(jar -tf "$SRC_JAR" | grep -c '.class')
@@ -353,7 +360,7 @@ case "$SUBJECT_PROGRAM" in
       "$SRC_BASE_DIR/lib/jaxen-core.jar" \
       "$SRC_BASE_DIR/lib/jaxen-jdom.jar" \
       "$SRC_BASE_DIR/lib/saxpath.jar" \
-      "$SCRIPT_DIR/../subject-programs/jaxen-1.1.6.jar"
+      "$SCRIPT_DIR/../subject-programs/jars/jaxen-1.1.6.jar"
     ;;
 
   "joda-time-2.3")
@@ -456,7 +463,7 @@ RANDOOP_BASE_COMMAND=(
   -Xbootclasspath/a:"$JACOCO_AGENT_JAR:$REPLACECALL_JAR"
   -javaagent:"$JACOCO_AGENT_JAR"
   -javaagent:"$REPLACECALL_COMMAND"
-  -classpath "$RANDOOP_CLASSPATH:$RANDOOP_JAR"
+  -classpath "$RANDOOP_CLASSPATH:$RANDOOP_JAR:$CHECKER_QUAL_JAR"
   randoop.main.Main
   gentests
   --testjar="$TARGET_JAR"

subject-programs/annotated-jars/README.md

@@ -0,0 +1,62 @@
+# Annotated Subject Program JARs
+
+This folder holds the subject JARs after we run Checker Framework whole-program
+inference and rewrite the sources with purity annotations. Compared to
+`subject-programs/jars/`, the compiled bytecode here is the same, but the
+underlying Java files now carry `@Pure`, `@SideEffectFree`, `@Impure`, and
+related qualifiers from `org.checkerframework.framework.qual`. No other code or
+resources are altered.
+
+## Why the annotations matter
+
+The `GRT_FUZZING` feature in `scripts/mutation-randoop.sh` loads these annotated artifacts and, for each method, determines whether it is pure (no state change) or impure (mutates the receiver, its arguments, or global state). During fuzzing, we prioritize impure calls to mutate an object’s state before exercising additional API entry points. This produces many more distinct runtime states than repeatedly calling pure methods, which in turn helps Randoop uncover deeper behaviors and tricky branch conditions.
+
+## Rebuilding an annotated JAR
+
+1. **Fetch sources** – Run `scripts/get-all-subject-src.sh` to populate
+   `subject-programs/src/<subject-program>/`.
+2. **Build the plain JAR** – In the subject directory, run the build command
+   listed later (also see `subject-programs/README.build`). The result lands in
+   `subject-programs/jars/` or the subject’s usual build folder.
+3. **Set up inference tooling** – Point your environment at the local Checker
+   Framework build:
+
+   ```bash
+   export CHECKERFRAMEWORK=/path/to/grt-testing/scripts/build/checker-framework/checker
+   export PATH="$CHECKERFRAMEWORK/../annotation-file-utilities/bin:$PATH"
+   export JAVAC_JAR="$CHECKERFRAMEWORK/dist/javac.jar"
+   ```
+
+4. **Assemble the runtime classpath** – Start with the JAR you just built, then
+    append any dependencies:
+    - Ant projects usually ship extra JARs in a local `jars/` or `lib/` folder.
+    - Maven projects can generate a classpath with
+       `mvn -q dependency:build-classpath -Dmdep.outputFile=target/wpi-classpath.txt`.
+5. **Run inference** – From the subject directory, execute:
+
+   ```bash
+   $CHECKERFRAMEWORK/bin/infer-and-annotate.sh \
+     "org.checkerframework.framework.util.PurityChecker" \
+     "$RUNTIME_CLASSPATH" \
+     $(find src -name "*.java")
+   ```
+
+   The script rewrites the sources in-place with the inferred annotations.
+6. **Rebuild** – Repeat the command from step 2 to produce an annotated JAR.
+
+### Worked example (a4j-1.0b)
+
+```bash
+cd subject-programs/src/a4j-1.0b
+ant createJar                              # build the baseline JAR
+export CHECKERFRAMEWORK=...                # reuse the env vars above
+export PATH="$CHECKERFRAMEWORK/../annotation-file-utilities/bin:$PATH"
+export JAVAC_JAR="$CHECKERFRAMEWORK/dist/javac.jar"
+RUNTIME_CLASSPATH="../../jars/a4j-1.0b.jar:jars/jox116.jar:jars/log4j-1.2.4.jar"
+$CHECKERFRAMEWORK/bin/infer-and-annotate.sh \
+  "org.checkerframework.framework.util.PurityChecker" \
+  "$RUNTIME_CLASSPATH" \
+  $(find src -name "*.java")
+ant createJar                              # rebuild with annotations
+cp dist/a4j.jar ../../annotated-jars/a4j-1.0b.jar
+```