Decode bytecode into IR expressions/statements

Adds an implementation of an interpreter that converts JVM bytecode to
expression trees and statements as it goes, handling stack manipiulation
instructions on the way.

Signed-off-by: Gary Tierney <gary.tierney@fastmail.com>

deob-cli/src/main/java/dev/openrs2/deob/cli/DeobfuscatorClassLoader.kt

@@ -33,7 +33,7 @@ object SystemClassLoader : DeobfuscatorClassLoader {
 
 class ClasspathClassLoader(val classPath: List<Path>) : DeobfuscatorClassLoader {
     override fun load(name: String): ClassNode {
-        val relativePath = Paths.get("/${name.replace('.', File.separatorChar)}.class")
+        val relativePath = Paths.get("${name.replace('.', File.separatorChar)}.class")
 
         for (entry in classPath) {
             val classFilePath = entry.resolve(relativePath)

deob-cli/src/main/java/dev/openrs2/deob/cli/ir/MethodScopedCommand.kt

@@ -5,7 +5,7 @@ import com.github.ajalt.clikt.core.requireObject
 import com.github.ajalt.clikt.parameters.arguments.argument
 import dev.openrs2.deob.cli.DeobfuscatorOptions
 import dev.openrs2.deob.ir.Method
-import dev.openrs2.deob.ir.translation.IrDecompiler
+import dev.openrs2.deob.ir.translation.BytecodeToIrTranlator
 
 abstract class MethodScopedCommand(command: String) : CliktCommand(name = command) {
     val className: String by argument(help = "Fully qualified name of the class name to disassemble")
@@ -16,8 +16,8 @@ abstract class MethodScopedCommand(command: String) : CliktCommand(name = comman
         val clazz = PrintCfgCommand.options.classLoader.load(PrintCfgCommand.className)
         val method = clazz.methods.find { it.name == PrintCfgCommand.methodName }!!
 
-        val decompiler = IrDecompiler(clazz, method)
-        val ir = decompiler.decompile()
+        val decompiler = BytecodeToIrTranlator()
+        val ir = decompiler.decompile(clazz, method)
 
         run(ir)
     }

deob-cli/src/main/java/dev/openrs2/deob/cli/ir/PrintCfgCommand.kt

@@ -1,10 +1,9 @@
 package dev.openrs2.deob.cli.ir
 
 import com.google.common.graph.EndpointPair
+import dev.openrs2.deob.ir.BasicBlock
 import dev.openrs2.deob.ir.Method
-import dev.openrs2.deob.ir.flow.BasicBlock
 import org.jgrapht.Graph
-import org.jgrapht.graph.guava.MutableGraphAdapter
 import org.jgrapht.nio.DefaultAttribute
 import org.jgrapht.nio.GraphExporter
 import org.jgrapht.nio.dot.DOTExporter
@@ -28,9 +27,6 @@ fun dotExporter(): BlockGraphExporter {
 
 object PrintCfgCommand : MethodScopedCommand("ir-print-cfg") {
     override fun run(method: Method) {
-        val graph: BlockGraph = MutableGraphAdapter(method.cfg)
-        val exporter: BlockGraphExporter = dotExporter()
-
-        exporter.exportGraph(graph, System.out)
+        TODO("Make this output expressions")
     }
 }

deob-ir/build.gradle.kts

@@ -6,8 +6,6 @@ plugins {
 
 dependencies {
     implementation(project(":asm"))
-    implementation(project(":deob"))
-    implementation(project(":deob-annotations"))
     implementation("com.google.guava:guava:${Versions.guava}")
 }
 

deob-ir/src/main/java/dev/openrs2/deob/ir/BasicBlock.kt

@@ -0,0 +1,27 @@
+package dev.openrs2.deob.ir
+
+import com.google.common.graph.MutableValueGraph
+import dev.openrs2.deob.ir.flow.ControlFlowTransfer
+
+typealias BasicBlockGraph = MutableValueGraph<BasicBlock, ControlFlowTransfer>
+
+class BasicBlock(
+    internal val graph: BasicBlockGraph
+) {
+    val statements = mutableListOf<Stmt>()
+
+    fun successors(): Set<BasicBlock> = graph.successors(this)
+    fun predecessors(): Set<BasicBlock> = graph.predecessors(this)
+
+    override fun equals(other: Any?): Boolean {
+        return this === other
+    }
+
+    override fun hashCode(): Int {
+        return System.identityHashCode(this)
+    }
+
+    override fun toString(): String {
+        return "BasicBlock(expressions=\n${statements.joinToString("\n")}\n)"
+    }
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/Expr.kt

@@ -0,0 +1,52 @@
+package dev.openrs2.deob.ir
+
+import dev.openrs2.asm.MemberRef
+import org.objectweb.asm.Type
+
+sealed class Storage
+
+class LocalVarStorage(val slot: Int = -1) : Storage()
+class FieldStorage(val instance: Expr, val member: MemberRef) : Storage()
+class StaticStorage(val member: MemberRef) : Storage()
+
+sealed class UnaryOp {
+    object Negate : UnaryOp()
+    data class InstanceOf(val type: Type) : UnaryOp()
+    data class Cast(val type: Type) : UnaryOp()
+}
+
+enum class BinOp {
+    Equals,
+    NotEquals,
+    LessThan,
+    GreaterThanOrEquals,
+    GreaterThan,
+    LessThanOrEquals,
+    Remainder,
+    Divide,
+    Multiply,
+    Add,
+    Subtract,
+    ShiftLeft,
+    ShiftRight,
+    UnsignedShiftRight,
+    ExclusiveOr
+}
+
+data class BinaryExpr(val lhs: Expr, val rhs: Expr, val op: dev.openrs2.deob.ir.BinOp) : Expr()
+
+sealed class Expr
+
+data class IndexExpr(val array: Expr, val index: Expr) : Expr()
+
+data class UnaryExpr(val operand: Expr, val operator: UnaryOp) : Expr()
+
+data class ConstExpr(val value: Any?) : Expr()
+
+data class CallExpr(val instance: Expr?, val method: MemberRef, val arguments: List<Expr>) : Expr()
+
+data class VarExpr(val storage: Storage) : Expr() {
+    companion object {
+        fun local(slot: Int) = VarExpr(LocalVarStorage(slot))
+    }
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/Method.kt

@@ -1,11 +1,11 @@
 package dev.openrs2.deob.ir
 
-import com.google.common.graph.MutableGraph
-import dev.openrs2.deob.ir.flow.BasicBlock
+import com.google.common.graph.MutableValueGraph
+import dev.openrs2.deob.ir.flow.ControlFlowTransfer
 import org.objectweb.asm.tree.ClassNode
 import org.objectweb.asm.tree.MethodNode
 
 class Method(val owner: ClassNode, val method: MethodNode, val entry: BasicBlock) {
-    val cfg: MutableGraph<BasicBlock>
+    val cfg: MutableValueGraph<BasicBlock, ControlFlowTransfer>
         get() = entry.graph
 }

deob-ir/src/main/java/dev/openrs2/deob/ir/Stmt.kt

@@ -0,0 +1,22 @@
+package dev.openrs2.deob.ir
+
+sealed class Stmt {
+    fun accept(visitor: StmtVisitor) = when (this) {
+        is AssignmentStmt -> visitor.visitAssignmen(slot, expr)
+        is CallStmt -> visitor.visitCall(expr)
+        is IfStmt -> visitor.visitIf(expr)
+        is ReturnStmt -> visitor.visitReturn(expr)
+    }
+}
+
+data class AssignmentStmt(val slot: VarExpr, val expr: Expr) : Stmt()
+data class CallStmt(val expr: CallExpr) : Stmt()
+data class IfStmt(val expr: BinaryExpr) : Stmt()
+data class ReturnStmt(val expr: Expr?) : Stmt()
+
+interface StmtVisitor {
+    fun visitAssignmen(variable: VarExpr, value: Expr)
+    fun visitCall(expr: CallExpr)
+    fun visitIf(conditional: BinaryExpr)
+    fun visitReturn(value: Expr?)
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/flow/BasicBlock.kt

@@ -1,48 +0,0 @@
-package dev.openrs2.deob.ir.flow
-
-import com.google.common.graph.MutableGraph
-import org.objectweb.asm.tree.InsnList
-import org.objectweb.asm.tree.LabelNode
-import org.objectweb.asm.util.Textifier
-import org.objectweb.asm.util.TraceMethodVisitor
-
-class BasicBlock(
-    internal val graph: MutableGraph<BasicBlock>,
-    private val code: InsnList,
-    private var start: Int,
-    private var end: Int
-) {
-    fun successors() : Set<BasicBlock> = graph.successors(this)
-    fun predecessors() : Set<BasicBlock> = graph.predecessors(this)
-
-    var label: LabelNode? = null
-
-    fun merge(other: BasicBlock) {
-        for (succ in other.successors()) {
-            graph.putEdge(this, succ)
-        }
-
-        graph.removeEdge(this, other)
-        graph.removeNode(other)
-
-        end = other.end
-    }
-
-    override fun equals(other: Any?): Boolean {
-        return this === other
-    }
-
-    override fun hashCode(): Int {
-        return System.identityHashCode(this)
-    }
-
-    override fun toString(): String {
-        val textifier = Textifier()
-        val methodTracer = TraceMethodVisitor(textifier)
-
-        code.accept(methodTracer)
-
-        val text = textifier.text.subList(start, end + 1) as List<String>
-        return text.reduce { l, r -> l + r }
-    }
-}

deob-ir/src/main/java/dev/openrs2/deob/ir/flow/ControlFlowAnalyzer.kt

@@ -0,0 +1,80 @@
+package dev.openrs2.deob.ir.flow
+
+import com.google.common.graph.Graph
+import com.google.common.graph.GraphBuilder
+import org.objectweb.asm.tree.MethodNode
+import org.objectweb.asm.tree.analysis.Analyzer
+import org.objectweb.asm.tree.analysis.BasicInterpreter
+import org.objectweb.asm.tree.analysis.BasicValue
+import java.util.Stack
+
+/**
+ * A bytecode analyzer that produces a graph representing the basic control structure of a method.
+ *
+ * @param reduce if the graph should be reduced by coalescing all nodes with an immediate dominator
+ * with only a single successor.
+ */
+class ControlFlowAnalyzer(private val reduce: Boolean = false) : Analyzer<BasicValue>(BasicInterpreter()) {
+    val instructionsToBlocks = mutableMapOf<Int, Int>()
+
+    private val graph = GraphBuilder
+        .directed()
+        .allowsSelfLoops(true)
+        .build<Int>()
+
+    override fun newControlFlowEdge(insnIndex: Int, successorIndex: Int) {
+        graph.putEdge(insnIndex, successorIndex)
+    }
+
+    override fun newControlFlowExceptionEdge(insnIndex: Int, successorIndex: Int): Boolean {
+        graph.putEdge(insnIndex, successorIndex)
+        return true
+    }
+
+    private fun reduceGraph() {
+        val nodeQueue = Stack<Int>()
+        val nodesVisited = mutableSetOf<Int>()
+
+        nodeQueue.push(0)
+
+        while (nodeQueue.isNotEmpty()) {
+            // Coalesce all nodes that are immediately dominated and are the sole successor
+            // of their dominator
+            val current = nodeQueue.pop()
+            val nextSuccessors = generateSequence { graph.successors(current).singleOrNull() }.iterator()
+
+            for (successor in nextSuccessors) {
+                val isImmediateDominator = current == graph.predecessors(successor).singleOrNull()
+
+                if (!isImmediateDominator || successor == current) {
+                    break
+                }
+
+                for (domSuccessor in graph.successors(successor)) {
+                    graph.putEdge(current, domSuccessor)
+                }
+
+                graph.removeEdge(current, successor)
+                graph.removeNode(successor)
+            }
+
+            for (successor in graph.successors(current)) {
+                if (!nodesVisited.contains(successor)) {
+                    nodeQueue.push(successor)
+                }
+            }
+
+            nodesVisited += current
+        }
+    }
+
+    fun createGraph(owner: String, method: MethodNode): Graph<Int> {
+        analyze(owner, method)
+
+        if (reduce) {
+            reduceGraph()
+        }
+
+        return graph
+    }
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/flow/ControlFlowTransfer.kt

@@ -0,0 +1,8 @@
+package dev.openrs2.deob.ir.flow
+
+sealed class ControlFlowTransfer {
+
+    data class ConditionalJump(val successful: Boolean) : ControlFlowTransfer()
+
+    object Goto : ControlFlowTransfer()
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/flow/data/DataFlowAnalyzer.kt

@@ -1,7 +1,8 @@
-package dev.openrs2.deob.analysis
+package dev.openrs2.deob.ir.flow.data
 
 import com.google.common.graph.Graph
 import com.google.common.graph.Graphs
+import dev.openrs2.deob.ir.flow.ControlFlowAnalyzer
 import org.objectweb.asm.tree.AbstractInsnNode
 import org.objectweb.asm.tree.MethodNode
 

deob-ir/src/main/java/dev/openrs2/deob/ir/flow/data/LiveVariableAnalyzer.kt

@@ -1,4 +1,4 @@
-package dev.openrs2.deob.analysis
+package dev.openrs2.deob.ir.flow.data
 
 import org.objectweb.asm.Opcodes
 import org.objectweb.asm.tree.AbstractInsnNode

deob-ir/src/main/java/dev/openrs2/deob/ir/translation/BytecodeToIrTranlator.kt

@@ -0,0 +1,16 @@
+package dev.openrs2.deob.ir.translation
+
+import dev.openrs2.deob.ir.Method
+import dev.openrs2.deob.ir.translation.decompiler.IrAnalyzer
+import dev.openrs2.deob.ir.translation.decompiler.IrInterpreter
+import org.objectweb.asm.tree.ClassNode
+import org.objectweb.asm.tree.MethodNode
+
+class BytecodeToIrTranlator {
+    fun decompile(owner: ClassNode, method: MethodNode): Method {
+        val irAnalyzer = IrAnalyzer(IrInterpreter())
+        val entryBlock = irAnalyzer.decode(owner.name, method)
+
+        return Method(owner, method, entryBlock)
+    }
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/translation/IrDecompiler.kt

@@ -1,89 +0,0 @@
-package dev.openrs2.deob.ir.translation
-
-import com.google.common.graph.GraphBuilder
-import dev.openrs2.deob.analysis.ControlFlowAnalyzer
-import dev.openrs2.deob.ir.Method
-import dev.openrs2.deob.ir.flow.BasicBlock
-import org.objectweb.asm.tree.ClassNode
-import org.objectweb.asm.tree.LabelNode
-import org.objectweb.asm.tree.MethodNode
-import org.objectweb.asm.tree.analysis.Analyzer
-import org.objectweb.asm.tree.analysis.BasicInterpreter
-import org.objectweb.asm.tree.analysis.BasicValue
-import java.util.Stack
-
-class IrDecompiler(private val owner: ClassNode, private val method: MethodNode) :
-    Analyzer<BasicValue>(BasicInterpreter()) {
-
-    private val graph = GraphBuilder
-        .directed()
-        .allowsSelfLoops(true)
-        .build<BasicBlock>()
-
-    private val blocks = mutableMapOf<Int, BasicBlock>()
-    private fun bb(index: Int) = blocks.computeIfAbsent(index) {
-        val bb = BasicBlock(graph, method.instructions, it, it)
-        val leader = method.instructions[index]
-
-        if (leader is LabelNode) {
-            bb.label = leader
-        }
-
-        bb
-    }
-
-    override fun newControlFlowEdge(insnIndex: Int, successorIndex: Int) {
-        graph.putEdge(bb(insnIndex), bb(successorIndex))
-    }
-
-    override fun newControlFlowExceptionEdge(insnIndex: Int, successorIndex: Int): Boolean {
-        // @TODO: Attach basic blocks with exception types.
-        return true
-    }
-
-    fun decompile(): Method {
-        analyze(owner.name, method)
-
-        val entryBlock = blocks[0] ?: throw IllegalStateException("No method entry block found")
-
-        val remainingBlocks = Stack<BasicBlock>()
-        remainingBlocks.push(entryBlock)
-
-        val visited = mutableSetOf<BasicBlock>()
-
-        while (remainingBlocks.isNotEmpty()) {
-            val bb = remainingBlocks.pop()
-            var next = bb.findTrivialSuccessor()
-
-            while (next != null && next != bb) {
-                bb.merge(next)
-
-                next = bb.findTrivialSuccessor()
-            }
-
-            for (succ in bb.successors()) {
-                if (!visited.contains(succ)) {
-                    remainingBlocks.push(succ)
-                }
-            }
-
-            visited.add(bb)
-        }
-
-        return Method(owner, method, entryBlock)
-    }
-}
-
-private fun BasicBlock.findTrivialSuccessor(): BasicBlock? {
-    val successors = successors()
-    if (successors.size != 1) return null
-
-    val successor = successors.first()!!
-    val nextPredecessors = successor.predecessors()
-
-    if (nextPredecessors.size == 1) {
-        return successor
-    } else {
-        return null
-    }
-}

deob-ir/src/main/java/dev/openrs2/deob/ir/translation/decompiler/IrAnalyzer.kt

@@ -0,0 +1,106 @@
+package dev.openrs2.deob.ir.translation.decompiler
+
+import com.google.common.graph.ValueGraphBuilder
+import dev.openrs2.deob.ir.AssignmentStmt
+import dev.openrs2.deob.ir.BasicBlock
+import dev.openrs2.deob.ir.BasicBlockGraph
+import dev.openrs2.deob.ir.BinaryExpr
+import dev.openrs2.deob.ir.CallExpr
+import dev.openrs2.deob.ir.CallStmt
+import dev.openrs2.deob.ir.Expr
+import dev.openrs2.deob.ir.IfStmt
+import dev.openrs2.deob.ir.ReturnStmt
+import dev.openrs2.deob.ir.StmtVisitor
+import dev.openrs2.deob.ir.VarExpr
+import dev.openrs2.deob.ir.flow.ControlFlowTransfer
+import org.objectweb.asm.Opcodes
+import org.objectweb.asm.tree.InsnList
+import org.objectweb.asm.tree.JumpInsnNode
+import org.objectweb.asm.tree.MethodNode
+import org.objectweb.asm.tree.analysis.Analyzer
+
+class IrAnalyzer(private val interpreter: IrInterpreter) : StmtVisitor, Analyzer<IrValue>(interpreter) {
+    init {
+        interpreter.visitor = this
+    }
+
+    /**
+     * A mapping of program counters (instruction offsets) to the [BasicBlock]s
+     * that define them.
+     */
+    private lateinit var blocks: MutableMap<Int, BasicBlock>
+
+    /**
+     * The control flow graph defining the method.
+     */
+    private lateinit var graph: BasicBlockGraph
+
+    /**
+     * The current [BasicBlock] of the method that is being analyzed.
+     */
+    private lateinit var bb: BasicBlock
+
+    /**
+     * The code belonging to the current method being analyzed.
+     */
+    private lateinit var code: InsnList
+
+    /**
+     * Get or create a new basic block that begins at the instruction with the given [index].
+     */
+    private fun bb(index: Int) = blocks.computeIfAbsent(index) { BasicBlock(graph) }
+
+    /**
+     * Run the analyzer on the [method] owned by the class with name [owner] and return the [BasicBlock]
+     * that defines method entry.
+     */
+    fun decode(owner: String, method: MethodNode): BasicBlock {
+        analyze(owner, method)
+        return blocks[0] ?: error("No entry BasicBlock found")
+    }
+
+    override fun init(owner: String, method: MethodNode) {
+        graph = ValueGraphBuilder.directed()
+            .allowsSelfLoops(true)
+            .build()
+
+        bb = BasicBlock(graph)
+        code = method.instructions
+        blocks = mutableMapOf(0 to bb)
+    }
+
+    override fun newControlFlowEdge(insnIndex: Int, successorIndex: Int) {
+        when (val insn = code[insnIndex]) {
+            is JumpInsnNode -> {
+                val transfer = if (insn.opcode == Opcodes.GOTO) {
+                    ControlFlowTransfer.ConditionalJump(code[successorIndex] == insn.label)
+                } else {
+                    ControlFlowTransfer.Goto
+                }
+
+                graph.putEdgeValue(bb, bb(successorIndex), transfer)
+            }
+        }
+
+        val startsBlock = blocks[insnIndex]
+        if (startsBlock != null) {
+            bb = startsBlock
+        }
+    }
+
+    override fun visitAssignmen(variable: VarExpr, value: Expr) {
+        bb.statements.add(AssignmentStmt(variable, value))
+    }
+
+    override fun visitCall(expr: CallExpr) {
+        bb.statements.add(CallStmt(expr))
+    }
+
+    override fun visitIf(conditional: BinaryExpr) {
+        bb.statements.add(IfStmt(conditional))
+    }
+
+    override fun visitReturn(value: Expr?) {
+        bb.statements.add(ReturnStmt(value))
+    }
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/translation/decompiler/IrInterpreter.kt

@@ -0,0 +1,334 @@
+package dev.openrs2.deob.ir.translation.decompiler
+
+import dev.openrs2.asm.MemberRef
+import dev.openrs2.asm.intConstant
+import dev.openrs2.asm.toPrettyString
+import dev.openrs2.deob.ir.BinOp
+import dev.openrs2.deob.ir.BinaryExpr
+import dev.openrs2.deob.ir.CallExpr
+import dev.openrs2.deob.ir.ConstExpr
+import dev.openrs2.deob.ir.FieldStorage
+import dev.openrs2.deob.ir.IndexExpr
+import dev.openrs2.deob.ir.LocalVarStorage
+import dev.openrs2.deob.ir.StaticStorage
+import dev.openrs2.deob.ir.StmtVisitor
+import dev.openrs2.deob.ir.Storage
+import dev.openrs2.deob.ir.UnaryExpr
+import dev.openrs2.deob.ir.UnaryOp
+import dev.openrs2.deob.ir.VarExpr
+import org.objectweb.asm.Opcodes
+import org.objectweb.asm.Opcodes.ACONST_NULL
+import org.objectweb.asm.Opcodes.ALOAD
+import org.objectweb.asm.Opcodes.ARETURN
+import org.objectweb.asm.Opcodes.ASM7
+import org.objectweb.asm.Opcodes.ASTORE
+import org.objectweb.asm.Opcodes.CHECKCAST
+import org.objectweb.asm.Opcodes.D2F
+import org.objectweb.asm.Opcodes.D2L
+import org.objectweb.asm.Opcodes.DADD
+import org.objectweb.asm.Opcodes.DCMPG
+import org.objectweb.asm.Opcodes.DDIV
+import org.objectweb.asm.Opcodes.DMUL
+import org.objectweb.asm.Opcodes.DNEG
+import org.objectweb.asm.Opcodes.DREM
+import org.objectweb.asm.Opcodes.DSUB
+import org.objectweb.asm.Opcodes.DUP
+import org.objectweb.asm.Opcodes.F2L
+import org.objectweb.asm.Opcodes.GETFIELD
+import org.objectweb.asm.Opcodes.GETSTATIC
+import org.objectweb.asm.Opcodes.I2B
+import org.objectweb.asm.Opcodes.I2C
+import org.objectweb.asm.Opcodes.I2D
+import org.objectweb.asm.Opcodes.I2F
+import org.objectweb.asm.Opcodes.I2L
+import org.objectweb.asm.Opcodes.I2S
+import org.objectweb.asm.Opcodes.IADD
+import org.objectweb.asm.Opcodes.IALOAD
+import org.objectweb.asm.Opcodes.ICONST_M1
+import org.objectweb.asm.Opcodes.IDIV
+import org.objectweb.asm.Opcodes.IFEQ
+import org.objectweb.asm.Opcodes.IFGE
+import org.objectweb.asm.Opcodes.IFGT
+import org.objectweb.asm.Opcodes.IFLE
+import org.objectweb.asm.Opcodes.IFLT
+import org.objectweb.asm.Opcodes.IFNE
+import org.objectweb.asm.Opcodes.IFNONNULL
+import org.objectweb.asm.Opcodes.IFNULL
+import org.objectweb.asm.Opcodes.IF_ACMPEQ
+import org.objectweb.asm.Opcodes.IF_ACMPNE
+import org.objectweb.asm.Opcodes.IF_ICMPEQ
+import org.objectweb.asm.Opcodes.IF_ICMPGE
+import org.objectweb.asm.Opcodes.IF_ICMPGT
+import org.objectweb.asm.Opcodes.IF_ICMPLE
+import org.objectweb.asm.Opcodes.IF_ICMPLT
+import org.objectweb.asm.Opcodes.IF_ICMPNE
+import org.objectweb.asm.Opcodes.ILOAD
+import org.objectweb.asm.Opcodes.IMUL
+import org.objectweb.asm.Opcodes.INEG
+import org.objectweb.asm.Opcodes.INSTANCEOF
+import org.objectweb.asm.Opcodes.IREM
+import org.objectweb.asm.Opcodes.IRETURN
+import org.objectweb.asm.Opcodes.ISHL
+import org.objectweb.asm.Opcodes.ISHR
+import org.objectweb.asm.Opcodes.ISTORE
+import org.objectweb.asm.Opcodes.ISUB
+import org.objectweb.asm.Opcodes.IUSHR
+import org.objectweb.asm.Opcodes.IXOR
+import org.objectweb.asm.Opcodes.L2D
+import org.objectweb.asm.Opcodes.L2F
+import org.objectweb.asm.Opcodes.LALOAD
+import org.objectweb.asm.Opcodes.LDC
+import org.objectweb.asm.Opcodes.LOOKUPSWITCH
+import org.objectweb.asm.Opcodes.LSHL
+import org.objectweb.asm.Opcodes.LSHR
+import org.objectweb.asm.Opcodes.LUSHR
+import org.objectweb.asm.Opcodes.LXOR
+import org.objectweb.asm.Opcodes.NEW
+import org.objectweb.asm.Opcodes.PUTFIELD
+import org.objectweb.asm.Opcodes.SIPUSH
+import org.objectweb.asm.Opcodes.SWAP
+import org.objectweb.asm.Opcodes.TABLESWITCH
+import org.objectweb.asm.Type
+import org.objectweb.asm.tree.AbstractInsnNode
+import org.objectweb.asm.tree.FieldInsnNode
+import org.objectweb.asm.tree.JumpInsnNode
+import org.objectweb.asm.tree.LdcInsnNode
+import org.objectweb.asm.tree.LookupSwitchInsnNode
+import org.objectweb.asm.tree.MethodInsnNode
+import org.objectweb.asm.tree.MultiANewArrayInsnNode
+import org.objectweb.asm.tree.TableSwitchInsnNode
+import org.objectweb.asm.tree.TypeInsnNode
+import org.objectweb.asm.tree.VarInsnNode
+import org.objectweb.asm.tree.analysis.Interpreter
+
+class IrInterpreter : Interpreter<IrValue>(ASM7) {
+    lateinit var visitor: StmtVisitor
+
+    override fun newValue(type: Type?) = IrValue(null)
+
+    override fun newEmptyValue(local: Int) = IrValue(VarExpr.local(local))
+
+    override fun newParameterValue(isInstanceMethod: Boolean, local: Int, type: Type) =
+        IrValue(VarExpr.local(local), type)
+
+    override fun naryOperation(insn: AbstractInsnNode?, values: MutableList<out IrValue>): IrValue? = when (insn) {
+        is MethodInsnNode -> handleMethodOperation(insn, values)
+        is MultiANewArrayInsnNode -> handleNewMultiArrayOperation(insn, values)
+        else -> error("Unrecognized nary operation instruction: $insn")
+    }
+
+    override fun ternaryOperation(
+        insn: AbstractInsnNode?,
+        value1: IrValue?,
+        value2: IrValue?,
+        value3: IrValue?
+    ): IrValue {
+        TODO("Not yet implemented")
+    }
+
+    override fun merge(value1: IrValue?, value2: IrValue?): IrValue? {
+        return value1
+    }
+
+    override fun newReturnTypeValue(type: Type?): IrValue? {
+        if (type == Type.VOID_TYPE) {
+            return null
+        }
+
+        return IrValue(null, type)
+    }
+
+    /**
+     * We ignore this, since we catch all of the return instructions (outwith RETURN) in [unaryOperation].
+     */
+    override fun returnOperation(insn: AbstractInsnNode, value: IrValue, expected: IrValue) {}
+
+    /**
+     * Interprets a bytecode instruction with a single argument. This method is called for the
+     * following opcodes:
+     *
+     * <p>INEG, LNEG, FNEG, DNEG, IINC, I2L, I2F, I2D, L2I, L2F, L2D, F2I, F2L, F2D, D2I, D2L, D2F,
+     * I2B, I2C, I2S, IFEQ, IFNE, IFLT, IFGE, IFGT, IFLE, TABLESWITCH, LOOKUPSWITCH, IRETURN, LRETURN,
+     * FRETURN, DRETURN, ARETURN, PUTSTATIC, GETFIELD, NEWARRAY, ANEWARRAY, ARRAYLENGTH, ATHROW,
+     * CHECKCAST, INSTANCEOF, MONITORENTER, MONITOREXIT, IFNULL, IFNONNULL
+     *
+     * @param insn the bytecode instruction to be interpreted.
+     * @param value the argument of the instruction to be interpreted.
+     * @return the result of the interpretation of the given instruction.
+     * @throws AnalyzerException if an error occurred during the interpretation.
+     */
+    override fun unaryOperation(insn: AbstractInsnNode, value: IrValue) = when (insn.opcode) {
+        I2B -> handleCastOperation(insn, value, Type.BYTE_TYPE)
+        I2C -> handleCastOperation(insn, value, Type.CHAR_TYPE)
+        I2S -> handleCastOperation(insn, value, Type.SHORT_TYPE)
+        I2D, L2D -> handleCastOperation(insn, value, Type.DOUBLE_TYPE)
+        I2L, D2L, F2L -> handleCastOperation(insn, value, Type.LONG_TYPE)
+        I2F, D2F, L2F -> handleCastOperation(insn, value, Type.FLOAT_TYPE)
+        CHECKCAST -> handleCastOperation(insn, value, (insn as TypeInsnNode).desc.let(Type::getType))
+        INSTANCEOF -> handleCastOperation(
+            insn,
+            value,
+            (insn as TypeInsnNode).desc.let(Type::getType),
+            instanceOf = true
+        )
+        GETFIELD -> handleVarLoadOperation(
+            insn,
+            FieldStorage(value.expr!!, MemberRef(insn as FieldInsnNode))
+        )
+        IFNULL, IFNONNULL -> handleConditionalOperation(
+            insn as JumpInsnNode,
+            value,
+            IrValue(ConstExpr(null), value.type)
+        )
+        TABLESWITCH -> handleTableSwitchOperation(insn as TableSwitchInsnNode, value)
+        LOOKUPSWITCH -> handleLookupSwitchOperation(insn as LookupSwitchInsnNode, value)
+        in INEG..DNEG -> handleUnaryOperation(insn,value, UnaryOp.Negate)
+        in IFEQ..IFLE -> handleConditionalOperation(insn as JumpInsnNode, value, IrValue(ConstExpr(0), Type.INT_TYPE))
+        in IRETURN..ARETURN -> handleReturnOperation(insn, value)
+        else -> error("Invalid instruction for unary expression: ${insn.toPrettyString()}")
+    }
+
+    private fun handleTableSwitchOperation(tableSwitchInsnNode: TableSwitchInsnNode, value: IrValue): IrValue? {
+        TODO("Not yet implemented")
+    }
+
+    private fun handleLookupSwitchOperation(lookupSwitchInsnNode: LookupSwitchInsnNode, value: IrValue): IrValue? {
+        TODO("Not yet implemented")
+    }
+
+    /**
+     * Interprets a bytecode instruction with two arguments. This method is called for the following
+     * opcodes:
+     *
+     * <p>IALOAD, LALOAD, FALOAD, DALOAD, AALOAD, BALOAD, CALOAD, SALOAD, IADD, LADD, FADD, DADD,
+     * ISUB, LSUB, FSUB, DSUB, IMUL, LMUL, FMUL, DMUL, IDIV, LDIV, FDIV, DDIV, IREM, LREM, FREM, DREM,
+     * ISHL, LSHL, ISHR, LSHR, IUSHR, LUSHR, IAND, LAND, IOR, LOR, IXOR, LXOR, LCMP, FCMPL, FCMPG,
+     * DCMPL, DCMPG, IF_ICMPEQ, IF_ICMPNE, IF_ICMPLT, IF_ICMPGE, IF_ICMPGT, IF_ICMPLE, IF_ACMPEQ,
+     * IF_ACMPNE, PUTFIELD
+     *
+     * @param insn the bytecode instruction to be interpreted.
+     * @param value1 the first argument of the instruction to be interpreted.
+     * @param value2 the second argument of the instruction to be interpreted.
+     * @return the result of the interpretation of the given instruction.
+     * @throws AnalyzerException if an error occurred during the interpretation.
+     */
+    override fun binaryOperation(insn: AbstractInsnNode, lhs: IrValue, rhs: IrValue) = when (insn.opcode) {
+        PUTFIELD -> handleVarStoreOperation(insn, FieldStorage(lhs.expr!!, MemberRef(insn as FieldInsnNode)), rhs)
+        in IALOAD..LALOAD -> handleArrayLoadOperation(insn, lhs, rhs)
+        in IADD..DCMPG -> handleBinaryOperation(insn, lhs, rhs)
+        in IF_ICMPEQ..IF_ACMPNE -> handleConditionalOperation(insn as JumpInsnNode, lhs, rhs)
+        else -> error("Unhandled binary operation instruction: $insn")
+    }
+
+    private fun handleArrayLoadOperation(insn: AbstractInsnNode, index: IrValue, array: IrValue): IrValue? =
+        IrValue(IndexExpr(array.expr!!, index.expr!!))
+
+    override fun copyOperation(insn: AbstractInsnNode, value: IrValue) = when (insn.opcode) {
+        in ISTORE..ASTORE -> {
+            handleVarStoreOperation(insn, LocalVarStorage((insn as VarInsnNode).`var`), value)
+            value
+        }
+        in ILOAD..ALOAD -> handleVarLoadOperation(insn, LocalVarStorage((insn as VarInsnNode).`var`))
+        in DUP..SWAP -> {
+            val local = LocalVarStorage()
+
+            handleVarStoreOperation(insn, local, value)
+            handleVarLoadOperation(insn, local)
+        }
+        else -> error("Invalid copy instruction: $insn")
+    }
+
+    override fun newOperation(insn: AbstractInsnNode) = when (insn.opcode) {
+        NEW -> IrValue(null, (insn as TypeInsnNode).desc.let(Type::getType))
+        ACONST_NULL -> handleConstantOperation(null)
+        LDC -> handleConstantOperation((insn as LdcInsnNode).cst)
+        GETSTATIC -> handleVarLoadOperation(insn, StaticStorage(MemberRef((insn as FieldInsnNode))))
+        in ICONST_M1..SIPUSH -> handleConstantOperation(insn.intConstant!!)
+        else -> error("Unhandled new operation instruction: $insn")
+    }
+
+    private fun handleConstantOperation(value: Any?) = IrValue(ConstExpr(value))
+
+    fun handleBinaryOperation(insn: AbstractInsnNode, lhs: IrValue, rhs: IrValue): IrValue? {
+        val op = when (insn.opcode) {
+            in ISUB..DSUB -> BinOp.Subtract
+            in IADD..DADD -> BinOp.Add
+            in IMUL..DMUL -> BinOp.Multiply
+            in IDIV..DDIV -> BinOp.Divide
+            in IREM..DREM -> BinOp.Remainder
+            in ISHL..LSHL -> BinOp.ShiftLeft
+            in ISHR..LSHR -> BinOp.ShiftRight
+            in IUSHR..LUSHR -> BinOp.UnsignedShiftRight
+            in IXOR..LXOR -> BinOp.ExclusiveOr
+            else -> error("Invalid instruction for binary expression: ${insn.toPrettyString()}")
+        }
+
+        return IrValue(BinaryExpr(lhs.expr!!, rhs.expr!!, op))
+    }
+
+    fun handleCastOperation(insn: AbstractInsnNode, value: IrValue, type: Type, instanceOf: Boolean = false): IrValue {
+        val op = if (instanceOf) {
+            UnaryOp.InstanceOf(type)
+        } else {
+            UnaryOp.Cast(type)
+        }
+
+        return IrValue(UnaryExpr(value.expr!!, op))
+    }
+
+    private fun handleConditionalOperation(insn: JumpInsnNode, lhs: IrValue, rhs: IrValue): IrValue? {
+        val op = when (insn.opcode) {
+            IFEQ, IF_ACMPEQ, IF_ICMPEQ, IFNULL -> BinOp.Equals
+            IFNE, IF_ACMPNE, IF_ICMPNE, IFNONNULL -> BinOp.NotEquals
+            IFLE, IF_ICMPLE -> BinOp.LessThanOrEquals
+            IFLT, IF_ICMPLT -> BinOp.LessThan
+            IFGE, IF_ICMPGE -> BinOp.GreaterThanOrEquals
+            IFGT, IF_ICMPGT -> BinOp.GreaterThan
+            else -> error("Not a valid conditional instruction: $insn")
+        }
+
+        visitor.visitIf(BinaryExpr(lhs.expr!!, rhs.expr!!, op))
+        return null
+    }
+
+    fun handleMethodOperation(insn: MethodInsnNode, values: MutableList<out IrValue>): IrValue? {
+        val method = MemberRef(insn)
+        val returnType = Type.getReturnType(insn.desc)
+        val instance = when (insn.opcode) {
+            Opcodes.INVOKESTATIC -> null
+            else -> values.removeAt(0).expr ?: error("No expression found for isntance slot")
+        }
+
+        val call = CallExpr(instance, method, values.map { it.expr!! })
+
+        return when (returnType) {
+            Type.VOID_TYPE -> {
+                visitor.visitCall(call)
+                null
+            }
+            else -> IrValue(call, returnType)
+        }
+    }
+
+    fun handleNewMultiArrayOperation(insn: MultiANewArrayInsnNode, values: MutableList<out IrValue>): IrValue? {
+        return null
+    }
+
+    fun handleReturnOperation(insn: AbstractInsnNode, value: IrValue?): IrValue? {
+        visitor.visitReturn(value?.expr)
+        return null
+    }
+
+    fun handleUnaryOperation(insn: AbstractInsnNode, value: IrValue, operator: UnaryOp): IrValue {
+        return IrValue(UnaryExpr(value.expr!!, operator))
+    }
+
+    fun handleVarLoadOperation(insn: AbstractInsnNode, storage: Storage): IrValue? {
+        return IrValue(VarExpr(storage))
+    }
+
+    fun handleVarStoreOperation(insn: AbstractInsnNode, storage: Storage, value: IrValue): IrValue? {
+        visitor.visitAssignmen(VarExpr(storage), value.expr!!)
+        return null
+    }
+}

deob-ir/src/main/java/dev/openrs2/deob/ir/translation/decompiler/IrValue.kt

@@ -0,0 +1,9 @@
+package dev.openrs2.deob.ir.translation.decompiler
+
+import dev.openrs2.deob.ir.Expr
+import org.objectweb.asm.Type
+import org.objectweb.asm.tree.analysis.Value
+
+class IrValue(val expr: Expr?, val type: Type? = null) : Value {
+    override fun getSize() = type?.size ?: 1
+}

deob-ir/src/test/java/dev/openrs2/deob/ir/translation/BytecodeToIrTranlatorTests.kt

@@ -4,12 +4,14 @@ import dev.openrs2.deob.ir.translation.fixture.Fixture
 import dev.openrs2.deob.ir.translation.fixture.FixtureMethod
 import org.junit.jupiter.api.Test
 
-class IrDecompilerTests {
+class BytecodeToIrTranlatorTests {
 
     @Test
     fun `Creates entry basic block`() {
-        class CfgSample(val a: Boolean, val b: Boolean) :
-            Fixture {
+        class CfgSample() : Fixture {
+            val a = true
+            val b = false
+
             override fun test() {
                 if (a) {
                     println("a")
@@ -20,8 +22,10 @@ class IrDecompilerTests {
         }
 
         val fixture = FixtureMethod.from(CfgSample::class)
-        val decompiler = IrDecompiler(fixture.owner, fixture.method)
-        val irMethod = decompiler.decompile()
+        val decompiler = BytecodeToIrTranlator()
+        val irMethod = decompiler.decompile(fixture.owner, fixture.method)
+
+        // @TODO - some way of asserting on the output cfg?
     }
 }
 

deob-ir/src/test/java/dev/openrs2/deob/ir/translation/fixture/Fixture.java

@@ -1,9 +1,9 @@
 package dev.openrs2.deob.ir.translation.fixture;
 
-import dev.openrs2.deob.ir.translation.IrDecompilerTests;
+import dev.openrs2.deob.ir.translation.BytecodeToIrTranlatorTests;
 
 /**
- * Marker interface for {@link IrDecompilerTests}.
+ * Marker interface for {@link BytecodeToIrTranlatorTests}.
  */
 public interface Fixture {
 	void test();

deob/build.gradle.kts

@@ -11,6 +11,7 @@ application {
 dependencies {
     implementation(project(":bundler"))
     implementation(project(":deob-annotations"))
+    implementation(project(":deob-ir"))
     implementation("com.google.guava:guava:${Versions.guava}")
 }
 

deob/src/main/java/dev/openrs2/deob/analysis/ControlFlowAnalyzer.kt

@@ -1,29 +0,0 @@
-package dev.openrs2.deob.analysis
-
-import com.google.common.graph.Graph
-import com.google.common.graph.GraphBuilder
-import org.objectweb.asm.tree.MethodNode
-import org.objectweb.asm.tree.analysis.Analyzer
-import org.objectweb.asm.tree.analysis.BasicInterpreter
-import org.objectweb.asm.tree.analysis.BasicValue
-
-class ControlFlowAnalyzer : Analyzer<BasicValue>(BasicInterpreter()) {
-    private val graph = GraphBuilder
-        .directed()
-        .allowsSelfLoops(true)
-        .immutable<Int>()
-
-    override fun newControlFlowEdge(insnIndex: Int, successorIndex: Int) {
-        graph.putEdge(insnIndex, successorIndex)
-    }
-
-    override fun newControlFlowExceptionEdge(insnIndex: Int, successorIndex: Int): Boolean {
-        graph.putEdge(insnIndex, successorIndex)
-        return true
-    }
-
-    fun createGraph(owner: String, method: MethodNode): Graph<Int> {
-        analyze(owner, method)
-        return graph.build()
-    }
-}

deob/src/main/java/dev/openrs2/deob/transform/DummyLocalTransformer.kt

@@ -6,7 +6,7 @@ import dev.openrs2.asm.classpath.Library
 import dev.openrs2.asm.deleteExpression
 import dev.openrs2.asm.pure
 import dev.openrs2.asm.transform.Transformer
-import dev.openrs2.deob.analysis.LiveVariableAnalyzer
+import dev.openrs2.deob.ir.flow.data.LiveVariableAnalyzer
 import org.objectweb.asm.Opcodes
 import org.objectweb.asm.tree.AbstractInsnNode
 import org.objectweb.asm.tree.ClassNode

deob/src/test/java/dev/openrs2/deob/test/BytecodeTest.kt

@@ -0,0 +1,35 @@
+package dev.openrs2.deob.test
+
+import org.objectweb.asm.Opcodes
+import org.objectweb.asm.commons.InstructionAdapter
+import org.objectweb.asm.tree.MethodNode
+
+@DslMarker
+annotation class BytecodeDslMarker
+
+enum class Expectation {
+    Removed,
+    Added,
+    Present
+}
+
+data class BytecodeTest(val code: MethodNode, val expectations: List<Expectation>)
+
+@BytecodeDslMarker
+class BytecodeTestBuilder : InstructionAdapter(Opcodes.ASM7, MethodNode()) {
+    val method = mv as MethodNode
+    val expectations = mutableListOf<Expectation>()
+
+    override fun visitInsn(opcode: Int) {
+        super.visitInsn(opcode)
+        expectations.add(Expectation.Present)
+    }
+
+    operator fun Unit.unaryMinus() {
+        expectations[expectations.size - 1] = Expectation.Removed
+    }
+
+    operator fun Unit.unaryPlus() {
+        expectations[expectations.size - 1] = Expectation.Added
+    }
+}