/* NSC -- new Scala compiler * Copyright 2005-2014 LAMP/EPFL * @author Martin Odersky */ package scala.tools.nsc package backend.jvm import scala.tools.asm import scala.tools.nsc.backend.jvm.opt._ import scala.tools.nsc.backend.jvm.BTypes.{InlineInfo, MethodInlineInfo, InternalName} import BackendReporting._ import scala.tools.nsc.settings.ScalaSettings /** * This class mainly contains the method classBTypeFromSymbol, which extracts the necessary * information from a symbol and its type to create the corresponding ClassBType. It requires * access to the compiler (global parameter). * * The mixin CoreBTypes defines core BTypes that are used in the backend. Building these BTypes * uses classBTypeFromSymbol, hence requires access to the compiler (global). * * BTypesFromSymbols extends BTypes because the implementation of BTypes requires access to some * of the core btypes. They are declared in BTypes as abstract members. Note that BTypes does * not have access to the compiler instance. */ class BTypesFromSymbols[G <: Global](val global: G) extends BTypes { import global._ import definitions._ val bCodeICodeCommon: BCodeICodeCommon[global.type] = new BCodeICodeCommon(global) val bCodeAsmCommon: BCodeAsmCommon[global.type] = new BCodeAsmCommon(global) import bCodeAsmCommon._ // Why the proxy, see documentation of class [[CoreBTypes]]. val coreBTypes = new CoreBTypesProxy[this.type](this) import coreBTypes._ val byteCodeRepository = new ByteCodeRepository(global.classPath, javaDefinedClasses, recordPerRunCache(collection.concurrent.TrieMap.empty)) val localOpt: LocalOpt[this.type] = new LocalOpt(this) val inliner: Inliner[this.type] = new Inliner(this) val closureOptimizer: ClosureOptimizer[this.type] = new ClosureOptimizer(this) val callGraph: CallGraph[this.type] = new CallGraph(this) val backendReporting: BackendReporting = new BackendReportingImpl(global) final def initializeCoreBTypes(): Unit = { coreBTypes.setBTypes(new CoreBTypes[this.type](this)) } def recordPerRunCache[T <: collection.generic.Clearable](cache: T): T = perRunCaches.recordCache(cache) def compilerSettings: ScalaSettings = settings // helpers that need access to global. // TODO @lry create a separate component, they don't belong to BTypesFromSymbols final val strMODULE_INSTANCE_FIELD = nme.MODULE_INSTANCE_FIELD.toString private val primitiveCompilationUnits = Set( "Unit.scala", "Boolean.scala", "Char.scala", "Byte.scala", "Short.scala", "Int.scala", "Float.scala", "Long.scala", "Double.scala" ) /** * True if the current compilation unit is of a primitive class (scala.Boolean et al). * Used only in assertions. */ def isCompilingPrimitive = { primitiveCompilationUnits(currentUnit.source.file.name) } def isCompilingArray = { currentUnit.source.file.name == "Array.scala" } // end helpers /** * The ClassBType for a class symbol `classSym`. * * The class symbol scala.Nothing is mapped to the class scala.runtime.Nothing$. Similarly, * scala.Null is mapped to scala.runtime.Null$. This is because there exist no class files * for the Nothing / Null. If used for example as a parameter type, we use the runtime classes * in the classfile method signature. * * Note that the referenced class symbol may be an implementation class. For example when * compiling a mixed-in method that forwards to the static method in the implementation class, * the class descriptor of the receiver (the implementation class) is obtained by creating the * ClassBType. */ final def classBTypeFromSymbol(classSym: Symbol): ClassBType = { assert(classSym != NoSymbol, "Cannot create ClassBType from NoSymbol") assert(classSym.isClass, s"Cannot create ClassBType from non-class symbol $classSym") assertClassNotArrayNotPrimitive(classSym) assert(!primitiveTypeMap.contains(classSym) || isCompilingPrimitive, s"Cannot create ClassBType for primitive class symbol $classSym") if (classSym == NothingClass) RT_NOTHING else if (classSym == NullClass) RT_NULL else { val internalName = classSym.javaBinaryName.toString classBTypeFromInternalName.getOrElse(internalName, { // The new ClassBType is added to the map in its constructor, before we set its info. This // allows initializing cyclic dependencies, see the comment on variable ClassBType._info. val res = ClassBType(internalName) if (completeSilentlyAndCheckErroneous(classSym)) { res.info = Left(NoClassBTypeInfoClassSymbolInfoFailedSI9111(classSym.fullName)) res } else { setClassInfo(classSym, res) } }) } } /** * Builds a [[MethodBType]] for a method symbol. */ final def methodBTypeFromSymbol(methodSymbol: Symbol): MethodBType = { assert(methodSymbol.isMethod, s"not a method-symbol: $methodSymbol") val resultType: BType = if (methodSymbol.isClassConstructor || methodSymbol.isConstructor) UNIT else typeToBType(methodSymbol.tpe.resultType) MethodBType(methodSymbol.tpe.paramTypes map typeToBType, resultType) } /** * This method returns the BType for a type reference, for example a parameter type. * * If `t` references a class, typeToBType ensures that the class is not an implementation class. * See also comment on classBTypeFromSymbol, which is invoked for implementation classes. */ final def typeToBType(t: Type): BType = { import definitions.ArrayClass /** * Primitive types are represented as TypeRefs to the class symbol of, for example, scala.Int. * The `primitiveTypeMap` maps those class symbols to the corresponding PrimitiveBType. */ def primitiveOrClassToBType(sym: Symbol): BType = { assertClassNotArray(sym) assert(!sym.isImplClass, sym) primitiveTypeMap.getOrElse(sym, classBTypeFromSymbol(sym)) } /** * When compiling Array.scala, the type parameter T is not erased and shows up in method * signatures, e.g. `def apply(i: Int): T`. A TyperRef to T is replaced by ObjectReference. */ def nonClassTypeRefToBType(sym: Symbol): ClassBType = { assert(sym.isType && isCompilingArray, sym) ObjectReference } t.dealiasWiden match { case TypeRef(_, ArrayClass, List(arg)) => ArrayBType(typeToBType(arg)) // Array type such as Array[Int] (kept by erasure) case TypeRef(_, sym, _) if !sym.isClass => nonClassTypeRefToBType(sym) // See comment on nonClassTypeRefToBType case TypeRef(_, sym, _) => primitiveOrClassToBType(sym) // Common reference to a type such as scala.Int or java.lang.String case ClassInfoType(_, _, sym) => primitiveOrClassToBType(sym) // We get here, for example, for genLoadModule, which invokes typeToBType(moduleClassSymbol.info) /* AnnotatedType should (probably) be eliminated by erasure. However we know it happens for * meta-annotated annotations (@(ann @getter) val x = 0), so we don't emit a warning. * The type in the AnnotationInfo is an AnnotatedTpe. Tested in jvm/annotations.scala. */ case a @ AnnotatedType(_, t) => debuglog(s"typeKind of annotated type $a") typeToBType(t) /* ExistentialType should (probably) be eliminated by erasure. We know they get here for * classOf constants: * class C[T] * class T { final val k = classOf[C[_]] } */ case e @ ExistentialType(_, t) => debuglog(s"typeKind of existential type $e") typeToBType(t) /* The cases below should probably never occur. They are kept for now to avoid introducing * new compiler crashes, but we added a warning. The compiler / library bootstrap and the * test suite don't produce any warning. */ case tp => currentUnit.warning(tp.typeSymbol.pos, s"an unexpected type representation reached the compiler backend while compiling $currentUnit: $tp. " + "If possible, please file a bug on issues.scala-lang.org.") tp match { case ThisType(ArrayClass) => ObjectReference // was introduced in 9b17332f11 to fix SI-999, but this code is not reached in its test, or any other test case ThisType(sym) => classBTypeFromSymbol(sym) case SingleType(_, sym) => primitiveOrClassToBType(sym) case ConstantType(_) => typeToBType(t.underlying) case RefinedType(parents, _) => parents.map(typeToBType(_).asClassBType).reduceLeft((a, b) => a.jvmWiseLUB(b).get) } } } def assertClassNotArray(sym: Symbol): Unit = { assert(sym.isClass, sym) assert(sym != definitions.ArrayClass || isCompilingArray, sym) } def assertClassNotArrayNotPrimitive(sym: Symbol): Unit = { assertClassNotArray(sym) assert(!primitiveTypeMap.contains(sym) || isCompilingPrimitive, sym) } private def setClassInfo(classSym: Symbol, classBType: ClassBType): ClassBType = { // Check for isImplClass: trait implementation classes have NoSymbol as superClass // Check for hasAnnotationFlag for SI-9393: the classfile / java source parsers add // scala.annotation.Annotation as superclass to java annotations. In reality, java // annotation classfiles have superclass Object (like any interface classfile). val superClassSym = if (classSym.isImplClass || classSym.hasJavaAnnotationFlag) ObjectClass else { val sc = classSym.superClass // SI-9393: Java annotation classes don't have the ABSTRACT/INTERFACE flag, so they appear // (wrongly) as superclasses. Fix this for BTypes: the java annotation will appear as interface // (handled by method implementedInterfaces), the superclass is set to Object. if (sc.hasJavaAnnotationFlag) ObjectClass else sc } assert( if (classSym == ObjectClass) superClassSym == NoSymbol else if (classSym.isInterface) superClassSym == ObjectClass else // A ClassBType for a primitive class (scala.Boolean et al) is only created when compiling these classes. ((superClassSym != NoSymbol) && !superClassSym.isInterface) || (isCompilingPrimitive && primitiveTypeMap.contains(classSym)), s"Bad superClass for $classSym: $superClassSym" ) val superClass = if (superClassSym == NoSymbol) None else Some(classBTypeFromSymbol(superClassSym)) val interfaces = implementedInterfaces(classSym).map(classBTypeFromSymbol) val flags = { if (classSym.isJava) javaClassfileFlags(classSym) // see comment on javaClassfileFlags else javaFlags(classSym) } /* The InnerClass table of a class C must contain all nested classes of C, even if they are only * declared but not otherwise referenced in C (from the bytecode or a method / field signature). * We collect them here. * * Nested classes that are also referenced in C will be added to the innerClassBufferASM during * code generation, but those duplicates will be eliminated when emitting the InnerClass * attribute. * * Why do we need to collect classes into innerClassBufferASM at all? To collect references to * nested classes, but NOT nested in C, that are used within C. */ val nestedClassSymbols = { val linkedClass = exitingPickler(classSym.linkedClassOfClass) // linkedCoC does not work properly in late phases // The lambdalift phase lifts all nested classes to the enclosing class, so if we collect // member classes right after lambdalift, we obtain all nested classes, including local and // anonymous ones. val nestedClasses = { val allNested = exitingPhase(currentRun.lambdaliftPhase)(memberClassesForInnerClassTable(classSym)) val nested = { // Classes nested in value classes are nested in the companion at this point. For InnerClass / // EnclosingMethod, we use the value class as the outer class. So we remove nested classes // from the companion that were originally nested in the value class. if (exitingPickler(linkedClass.isDerivedValueClass)) allNested.filterNot(classOriginallyNestedInClass(_, linkedClass)) else allNested } if (isTopLevelModuleClass(classSym)) { // For Java compatibility, member classes of top-level objects are treated as members of // the top-level companion class, see comment below. val members = exitingPickler(memberClassesForInnerClassTable(classSym)) nested diff members } else { nested } } val companionModuleMembers = if (considerAsTopLevelImplementationArtifact(classSym)) Nil else { // If this is a top-level non-impl (*) class, the member classes of the companion object are // added as members of the class. For example: // class C { } // object C { // class D // def f = { class E } // } // The class D is added as a member of class C. The reason is: for Java compatibility, the // InnerClass attribute for D has "C" (NOT the module class "C$") as the outer class of D // (done by buildNestedInfo). See comment in BTypes. // For consistency, the InnerClass entry for D needs to be present in C - to Java it looks // like D is a member of C, not C$. // // (*) We exclude impl classes: if the classfile for the impl class exists on the classpath, // a linkedClass symbol is found for which isTopLevelModule is true, so we end up searching // members of that weird impl-class-module-class-symbol. that search probably cannot return // any classes, but it's better to exclude it. val javaCompatMembers = { if (linkedClass != NoSymbol && isTopLevelModuleClass(linkedClass)) // phase travel to exitingPickler: this makes sure that memberClassesForInnerClassTable only sees member // classes, not local classes of the companion module (E in the example) that were lifted by lambdalift. exitingPickler(memberClassesForInnerClassTable(linkedClass)) else Nil } // Classes nested in value classes are nested in the companion at this point. For InnerClass / // EnclosingMethod we use the value class as enclosing class. Here we search nested classes // in the companion that were originally nested in the value class, and we add them as nested // in the value class. val valueClassCompanionMembers = { if (linkedClass != NoSymbol && exitingPickler(classSym.isDerivedValueClass)) { val moduleMemberClasses = exitingPhase(currentRun.lambdaliftPhase)(memberClassesForInnerClassTable(linkedClass)) moduleMemberClasses.filter(classOriginallyNestedInClass(_, classSym)) } else Nil } javaCompatMembers ++ valueClassCompanionMembers } nestedClasses ++ companionModuleMembers } /** * For nested java classes, the scala compiler creates both a class and a module (and therefore * a module class) symbol. For example, in `class A { class B {} }`, the nestedClassSymbols * for A contain both the class B and the module class B. * Here we get rid of the module class B, making sure that the class B is present. */ val nestedClassSymbolsNoJavaModuleClasses = nestedClassSymbols.filter(s => { if (s.isJavaDefined && s.isModuleClass) { // We could also search in nestedClassSymbols for s.linkedClassOfClass, but sometimes that // returns NoSymbol, so it doesn't work. val nb = nestedClassSymbols.count(mc => mc.name == s.name && mc.owner == s.owner) assert(nb == 2, s"Java member module without member class: $s - $nestedClassSymbols") false } else true }) val nestedClasses = nestedClassSymbolsNoJavaModuleClasses.map(classBTypeFromSymbol) val nestedInfo = buildNestedInfo(classSym) val inlineInfo = buildInlineInfo(classSym, classBType.internalName) classBType.info = Right(ClassInfo(superClass, interfaces, flags, nestedClasses, nestedInfo, inlineInfo)) classBType } private def buildNestedInfo(innerClassSym: Symbol): Option[NestedInfo] = { assert(innerClassSym.isClass, s"Cannot build NestedInfo for non-class symbol $innerClassSym") val isTopLevel = innerClassSym.rawowner.isPackageClass // impl classes are considered top-level, see comment in BTypes if (isTopLevel || considerAsTopLevelImplementationArtifact(innerClassSym)) None else if (innerClassSym.rawowner.isTerm) { // This case should never be reached: the lambdalift phase mutates the rawowner field of all // classes to be the enclosing class. SI-9392 shows an errant macro that leaves a reference // to a local class symbol that no longer exists, which is not updated by lambdalift. devWarning(innerClassSym.pos, s"""The class symbol $innerClassSym with the term symbol ${innerClassSym.rawowner} as `rawowner` reached the backend. |Most likely this indicates a stale reference to a non-existing class introduced by a macro, see SI-9392.""".stripMargin) None } else { // See comment in BTypes, when is a class marked static in the InnerClass table. val isStaticNestedClass = isOriginallyStaticOwner(innerClassSym.originalOwner) // After lambdalift (which is where we are), the rawowner field contains the enclosing class. val enclosingClass = { // (1) Example java source: class C { static class D { } } // The Scala compiler creates a class and a module symbol for C. Because D is a static // nested class, the symbol for D is nested in the module class C (not in the class C). // For the InnerClass attribute, we use the class symbol C, which represents the situation // in the source code. // (2) Java compatibility. See the big comment in BTypes that summarizes the InnerClass spec. if ((innerClassSym.isJavaDefined && innerClassSym.rawowner.isModuleClass) || // (1) (!isAnonymousOrLocalClass(innerClassSym) && isTopLevelModuleClass(innerClassSym.rawowner))) { // (2) // phase travel for linkedCoC - does not always work in late phases exitingPickler(innerClassSym.rawowner.linkedClassOfClass) match { case NoSymbol => // For top-level modules without a companion class, see doc of mirrorClassClassBType. mirrorClassClassBType(exitingPickler(innerClassSym.rawowner)) case companionClass => classBTypeFromSymbol(companionClass) } } else { classBTypeFromSymbol(innerClassSym.rawowner) } } val outerName: Option[String] = { if (isAnonymousOrLocalClass(innerClassSym)) None else Some(enclosingClass.internalName) } val innerName: Option[String] = { // phase travel necessary: after flatten, the name includes the name of outer classes. // if some outer name contains $anon, a non-anon class is considered anon. if (exitingPickler(innerClassSym.isAnonymousClass || innerClassSym.isAnonymousFunction)) None else Some(innerClassSym.rawname + innerClassSym.moduleSuffix) // moduleSuffix for module classes } Some(NestedInfo(enclosingClass, outerName, innerName, isStaticNestedClass)) } } /** * Build the InlineInfo for a ClassBType from the class symbol. * * Note that the InlineInfo is only built from the symbolic information for classes that are being * compiled. For all other classes we delegate to inlineInfoFromClassfile. The reason is that * mixed-in methods are only added to class symbols being compiled, but not to other classes * extending traits. Creating the InlineInfo from the symbol would prevent these mixins from being * inlined. * * So for classes being compiled, the InlineInfo is created here and stored in the ScalaInlineInfo * classfile attribute. */ private def buildInlineInfo(classSym: Symbol, internalName: InternalName): InlineInfo = { def buildFromSymbol = buildInlineInfoFromClassSymbol(classSym, classBTypeFromSymbol(_).internalName, methodBTypeFromSymbol(_).descriptor) // phase travel required, see implementation of `compiles`. for nested classes, it checks if the // enclosingTopLevelClass is being compiled. after flatten, all classes are considered top-level, // so `compiles` would return `false`. if (exitingPickler(currentRun.compiles(classSym))) buildFromSymbol // InlineInfo required for classes being compiled, we have to create the classfile attribute else if (!compilerSettings.YoptInlinerEnabled) BTypes.EmptyInlineInfo // For other classes, we need the InlineInfo only inf the inliner is enabled. else { // For classes not being compiled, the InlineInfo is read from the classfile attribute. This // fixes an issue with mixed-in methods: the mixin phase enters mixin methods only to class // symbols being compiled. For non-compiled classes, we could not build MethodInlineInfos // for those mixin members, which prevents inlining. byteCodeRepository.classNode(internalName) match { case Right(classNode) => inlineInfoFromClassfile(classNode) case Left(missingClass) => InlineInfo(None, false, Map.empty, Some(ClassNotFoundWhenBuildingInlineInfoFromSymbol(missingClass))) } } } /** * For top-level objects without a companion class, the compilere generates a mirror class with * static forwarders (Java compat). There's no symbol for the mirror class, but we still need a * ClassBType (its info.nestedClasses will hold the InnerClass entries, see comment in BTypes). */ def mirrorClassClassBType(moduleClassSym: Symbol): ClassBType = { assert(isTopLevelModuleClass(moduleClassSym), s"not a top-level module class: $moduleClassSym") val internalName = moduleClassSym.javaBinaryName.dropModule.toString classBTypeFromInternalName.getOrElse(internalName, { val c = ClassBType(internalName) // class info consistent with BCodeHelpers.genMirrorClass val nested = exitingPickler(memberClassesForInnerClassTable(moduleClassSym)) map classBTypeFromSymbol c.info = Right(ClassInfo( superClass = Some(ObjectReference), interfaces = Nil, flags = asm.Opcodes.ACC_SUPER | asm.Opcodes.ACC_PUBLIC | asm.Opcodes.ACC_FINAL, nestedClasses = nested, nestedInfo = None, InlineInfo(None, true, Map.empty, None))) // no InlineInfo needed, scala never invokes methods on the mirror class c }) } /** * True for module classes of package level objects. The backend will generate a mirror class for * such objects. */ final def isTopLevelModuleClass(sym: Symbol): Boolean = exitingPickler { // phase travel to pickler required for isNestedClass (looks at owner) val r = sym.isModuleClass && !sym.isNestedClass // The mixin phase adds the `lateMODULE` flag to trait implementation classes. Since the flag // is late, it should not be visible here inside the time travel. We check this. if (r) assert(!sym.isImplClass, s"isModuleClass should be false for impl class $sym") r } /** * True for module classes of modules that are top-level or owned only by objects. Module classes * for such objects will get a MODULE$ flag and a corresponding static initializer. */ final def isStaticModuleClass(sym: Symbol): Boolean = { /* (1) Phase travel to to pickler is required to exclude implementation classes; they have the * lateMODULEs after mixin, so isModuleClass would be true. * (2) isStaticModuleClass is a source-level property. See comment on isOriginallyStaticOwner. */ exitingPickler { // (1) sym.isModuleClass && isOriginallyStaticOwner(sym.originalOwner) // (2) } } // legacy, to be removed when the @remote annotation gets removed final def isRemote(s: Symbol) = s hasAnnotation definitions.RemoteAttr final def hasPublicBitSet(flags: Int) = (flags & asm.Opcodes.ACC_PUBLIC) != 0 /** * Return the Java modifiers for the given symbol. * Java modifiers for classes: * - public, abstract, final, strictfp (not used) * for interfaces: * - the same as for classes, without 'final' * for fields: * - public, private (*) * - static, final * for methods: * - the same as for fields, plus: * - abstract, synchronized (not used), strictfp (not used), native (not used) * for all: * - deprecated * * (*) protected cannot be used, since inner classes 'see' protected members, * and they would fail verification after lifted. */ final def javaFlags(sym: Symbol): Int = { // constructors of module classes should be private. introduced in b06edbc, probably to prevent // creating module instances from java. for nested modules, the constructor needs to be public // since they are created by the outer class and stored in a field. a java client can create // new instances via outerClassInstance.new InnerModuleClass$(). // TODO: do this early, mark the symbol private. val privateFlag = sym.isPrivate || (sym.isPrimaryConstructor && isTopLevelModuleClass(sym.owner)) // Symbols marked in source as `final` have the FINAL flag. (In the past, the flag was also // added to modules and module classes, not anymore since 296b706). // Note that the presence of the `FINAL` flag on a symbol does not correspond 1:1 to emitting // ACC_FINAL in bytecode. // // Top-level modules are marked ACC_FINAL in bytecode (even without the FINAL flag). Nested // objects don't get the flag to allow overriding (under -Yoverride-objects, SI-5676). // // For fields, only eager val fields can receive ACC_FINAL. vars or lazy vals can't: // Source: http://docs.oracle.com/javase/specs/jls/se7/html/jls-17.html#jls-17.5.3 // "Another problem is that the specification allows aggressive // optimization of final fields. Within a thread, it is permissible to // reorder reads of a final field with those modifications of a final // field that do not take place in the constructor." // // A var or lazy val which is marked final still has meaning to the // scala compiler. The word final is heavily overloaded unfortunately; // for us it means "not overridable". At present you can't override // vars regardless; this may change. // // The logic does not check .isFinal (which checks flags for the FINAL flag, // and includes symbols marked lateFINAL) instead inspecting rawflags so // we can exclude lateFINAL. Such symbols are eligible for inlining, but to // avoid breaking proxy software which depends on subclassing, we do not // emit ACC_FINAL. val finalFlag = ( (((sym.rawflags & symtab.Flags.FINAL) != 0) || isTopLevelModuleClass(sym)) && !sym.enclClass.isInterface && !sym.isClassConstructor && !sym.isMutable // lazy vals and vars both ) // Primitives are "abstract final" to prohibit instantiation // without having to provide any implementations, but that is an // illegal combination of modifiers at the bytecode level so // suppress final if abstract if present. import asm.Opcodes._ GenBCode.mkFlags( if (privateFlag) ACC_PRIVATE else ACC_PUBLIC, if (sym.isDeferred || sym.hasAbstractFlag) ACC_ABSTRACT else 0, if (sym.isInterface) ACC_INTERFACE else 0, if (finalFlag && !sym.hasAbstractFlag) ACC_FINAL else 0, if (sym.isStaticMember) ACC_STATIC else 0, if (sym.isBridge) ACC_BRIDGE | ACC_SYNTHETIC else 0, if (sym.isArtifact) ACC_SYNTHETIC else 0, if (sym.isClass && !sym.isInterface) ACC_SUPER else 0, if (sym.hasJavaEnumFlag) ACC_ENUM else 0, if (sym.isVarargsMethod) ACC_VARARGS else 0, if (sym.hasFlag(symtab.Flags.SYNCHRONIZED)) ACC_SYNCHRONIZED else 0, if (sym.isDeprecated) asm.Opcodes.ACC_DEPRECATED else 0 ) } def javaFieldFlags(sym: Symbol) = { javaFlags(sym) | GenBCode.mkFlags( if (sym hasAnnotation TransientAttr) asm.Opcodes.ACC_TRANSIENT else 0, if (sym hasAnnotation VolatileAttr) asm.Opcodes.ACC_VOLATILE else 0, if (sym.isMutable) 0 else asm.Opcodes.ACC_FINAL ) } }