JDK 11 jdk.scripting.nashorn.jmod - Scripting Nashorn Module

JDK 11 jdk.scripting.nashorn.jmod is the JMOD file for JDK 11 Scripting Nashorn module.

JDK 11 Scripting Nashorn module compiled class files are stored in \fyicenter\jdk-11.0.1\jmods\jdk.scripting.nashorn.jmod.

JDK 11 Scripting Nashorn module compiled class files are also linked and stored in the \fyicenter\jdk-11.0.1\lib\modules JImage file.

JDK 11 Scripting Nashorn module source code files are stored in \fyicenter\jdk-11.0.1\lib\src.zip\jdk.scripting.nashorn.

You can click and view the content of each source code file in the list below.

✍: FYIcenter


 * Copyright (c) 2010, 2013, Oracle and/or its affiliates. All rights reserved.
 * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
package jdk.nashorn.internal.runtime;

import static jdk.nashorn.internal.lookup.Lookup.MH;
import static jdk.nashorn.internal.runtime.UnwarrantedOptimismException.INVALID_PROGRAM_POINT;
import static jdk.nashorn.internal.runtime.UnwarrantedOptimismException.isValid;

import java.lang.invoke.CallSite;
import java.lang.invoke.MethodHandle;
import java.lang.invoke.MethodHandles;
import java.lang.invoke.MethodType;
import java.lang.invoke.MutableCallSite;
import java.lang.invoke.SwitchPoint;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Collections;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.TreeMap;
import java.util.function.Supplier;
import java.util.logging.Level;
import jdk.dynalink.linker.GuardedInvocation;
import jdk.nashorn.internal.codegen.Compiler;
import jdk.nashorn.internal.codegen.Compiler.CompilationPhases;
import jdk.nashorn.internal.codegen.TypeMap;
import jdk.nashorn.internal.codegen.types.ArrayType;
import jdk.nashorn.internal.codegen.types.Type;
import jdk.nashorn.internal.ir.FunctionNode;
import jdk.nashorn.internal.objects.annotations.SpecializedFunction.LinkLogic;
import jdk.nashorn.internal.runtime.events.RecompilationEvent;
import jdk.nashorn.internal.runtime.linker.Bootstrap;
import jdk.nashorn.internal.runtime.logging.DebugLogger;

 * An version of a JavaScript function, native or JavaScript.
 * Supports lazily generating a constructor version of the invocation.
final class CompiledFunction {

    private static final MethodHandle NEWFILTER = findOwnMH("newFilter", Object.class, Object.class, Object.class);
    private static final MethodHandle RELINK_COMPOSABLE_INVOKER = findOwnMH("relinkComposableInvoker", void.class, CallSite.class, CompiledFunction.class, boolean.class);
    private static final MethodHandle HANDLE_REWRITE_EXCEPTION = findOwnMH("handleRewriteException", MethodHandle.class, CompiledFunction.class, OptimismInfo.class, RewriteException.class);
    private static final MethodHandle RESTOF_INVOKER = MethodHandles.exactInvoker(MethodType.methodType(Object.class, RewriteException.class));

    private final DebugLogger log;

    static final Collection<CompiledFunction> NO_FUNCTIONS = Collections.emptySet();

     * The method type may be more specific than the invoker, if. e.g.
     * the invoker is guarded, and a guard with a generic object only
     * fallback, while the target is more specific, we still need the
     * more specific type for sorting
    private MethodHandle invoker;
    private MethodHandle constructor;
    private OptimismInfo optimismInfo;
    private final int flags; // from FunctionNode
    private final MethodType callSiteType;

    private final Specialization specialization;

    CompiledFunction(final MethodHandle invoker) {
        this(invoker, null, null);

    static CompiledFunction createBuiltInConstructor(final MethodHandle invoker, final Specialization specialization) {
        return new CompiledFunction(MH.insertArguments(invoker, 0, false), createConstructorFromInvoker(MH.insertArguments(invoker, 0, true)), specialization);

    CompiledFunction(final MethodHandle invoker, final MethodHandle constructor, final Specialization specialization) {
        this(invoker, constructor, 0, null, specialization, DebugLogger.DISABLED_LOGGER);

    CompiledFunction(final MethodHandle invoker, final MethodHandle constructor, final int flags, final MethodType callSiteType, final Specialization specialization, final DebugLogger log) {
        this.specialization = specialization;
        if (specialization != null && specialization.isOptimistic()) {
             * An optimistic builtin with isOptimistic=true works like any optimistic generated function, i.e. it
             * can throw unwarranted optimism exceptions. As native functions trivially can't have parts of them
             * regenerated as "restOf" methods, this only works if the methods are atomic/functional in their behavior
             * and doesn't modify state before an UOE can be thrown. If they aren't, we can reexecute a wider version
             * of the same builtin in a recompilation handler for FinalScriptFunctionData. There are several
             * candidate methods in Native* that would benefit from this, but I haven't had time to implement any
             * of them currently. In order to fit in with the relinking framework, the current thinking is
             * that the methods still take a program point to fit in with other optimistic functions, but
             * it is set to "first", which is the beginning of the method. The relinker can tell the difference
             * between builtin and JavaScript functions. This might change. TODO
            this.invoker = MH.insertArguments(invoker, invoker.type().parameterCount() - 1, UnwarrantedOptimismException.FIRST_PROGRAM_POINT);
            throw new AssertionError("Optimistic (UnwarrantedOptimismException throwing) builtin functions are currently not in use");
        this.invoker = invoker;
        this.constructor = constructor;
        this.flags = flags;
        this.callSiteType = callSiteType;
        this.log = log;

    CompiledFunction(final MethodHandle invoker, final RecompilableScriptFunctionData functionData,
            final Map<Integer, Type> invalidatedProgramPoints, final MethodType callSiteType, final int flags) {
        this(invoker, null, flags, callSiteType, null, functionData.getLogger());
        if ((flags & FunctionNode.IS_DEOPTIMIZABLE) != 0) {
            optimismInfo = new OptimismInfo(functionData, invalidatedProgramPoints);
        } else {
            optimismInfo = null;

    static CompiledFunction createBuiltInConstructor(final MethodHandle invoker) {
        return new CompiledFunction(MH.insertArguments(invoker, 0, false), createConstructorFromInvoker(MH.insertArguments(invoker, 0, true)), null);

    boolean isSpecialization() {
        return specialization != null;

    boolean hasLinkLogic() {
        return getLinkLogicClass() != null;

    Class<? extends LinkLogic> getLinkLogicClass() {
        if (isSpecialization()) {
            final Class<? extends LinkLogic> linkLogicClass = specialization.getLinkLogicClass();
            assert !LinkLogic.isEmpty(linkLogicClass) : "empty link logic classes should have been removed by nasgen";
            return linkLogicClass;
        return null;

    boolean convertsNumericArgs() {
        return isSpecialization() && specialization.convertsNumericArgs();

    int getFlags() {
        return flags;

     * An optimistic specialization is one that can throw UnwarrantedOptimismException.
     * This is allowed for native methods, as long as they are functional, i.e. don't change
     * any state between entering and throwing the UOE. Then we can re-execute a wider version
     * of the method in the continuation. Rest-of method generation for optimistic builtins is
     * of course not possible, but this approach works and fits into the same relinking
     * framework
     * @return true if optimistic builtin
    boolean isOptimistic() {
        return isSpecialization() ? specialization.isOptimistic() : false;

    boolean isApplyToCall() {
        return (flags & FunctionNode.HAS_APPLY_TO_CALL_SPECIALIZATION) != 0;

    boolean isVarArg() {
        return isVarArgsType(invoker.type());

    public String toString() {
        final StringBuilder sb = new StringBuilder();
        final Class<? extends LinkLogic> linkLogicClass = getLinkLogicClass();

            append(" ctor=").
            append(" weight=").
            append(" linkLogic=").
            append(linkLogicClass != null ? linkLogicClass.getSimpleName() : "none");

        return sb.toString();

    boolean needsCallee() {
        return ScriptFunctionData.needsCallee(invoker);

     * Returns an invoker method handle for this function. Note that the handle is safely composable in
     * the sense that you can compose it with other handles using any combinators even if you can't affect call site
     * invalidation. If this compiled function is non-optimistic, then it returns the same value as
     * {@link #getInvokerOrConstructor(boolean)}. However, if the function is optimistic, then this handle will
     * incur an overhead as it will add an intermediate internal call site that can relink itself when the function
     * needs to regenerate its code to always point at the latest generated code version.
     * @return a guaranteed composable invoker method handle for this function.
    MethodHandle createComposableInvoker() {
        return createComposableInvoker(false);

     * Returns an invoker method handle for this function when invoked as a constructor. Note that the handle should be
     * considered non-composable in the sense that you can only compose it with other handles using any combinators if
     * you can ensure that the composition is guarded by {@link #getOptimisticAssumptionsSwitchPoint()} if it's
     * non-null, and that you can relink the call site it is set into as a target if the switch point is invalidated. In
     * all other cases, use {@link #createComposableConstructor()}.
     * @return a direct constructor method handle for this function.
    private MethodHandle getConstructor() {
        if (constructor == null) {
            constructor = createConstructorFromInvoker(createInvokerForPessimisticCaller());

        return constructor;

     * Creates a version of the invoker intended for a pessimistic caller (return type is Object, no caller optimistic
     * program point available).
     * @return a version of the invoker intended for a pessimistic caller.
    private MethodHandle createInvokerForPessimisticCaller() {
        return createInvoker(Object.class, INVALID_PROGRAM_POINT);

     * Compose a constructor from an invoker.
     * @param invoker         invoker
     * @return the composed constructor
    private static MethodHandle createConstructorFromInvoker(final MethodHandle invoker) {
        final boolean needsCallee = ScriptFunctionData.needsCallee(invoker);
        // If it was (callee, this, args...), permute it to (this, callee, args...). We're doing this because having
        // "this" in the first argument position is what allows the elegant folded composition of
        // (newFilter x constructor x allocator) further down below in the code. Also, ensure the composite constructor
        // always returns Object.
        final MethodHandle swapped = needsCallee ? swapCalleeAndThis(invoker) : invoker;

        final MethodHandle returnsObject = MH.asType(swapped, swapped.type().changeReturnType(Object.class));

        final MethodType ctorType = returnsObject.type();

        // Construct a dropping type list for NEWFILTER, but don't include constructor "this" into it, so it's actually
        // captured as "allocation" parameter of NEWFILTER after we fold the constructor into it.
        // (this, [callee, ]args...) => ([callee, ]args...)
        final Class<?>[] ctorArgs = ctorType.dropParameterTypes(0, 1).parameterArray();

        // Fold constructor into newFilter that replaces the return value from the constructor with the originally
        // allocated value when the originally allocated value is a JS primitive (String, Boolean, Number).
        // (result, this, [callee, ]args...) x (this, [callee, ]args...) => (this, [callee, ]args...)
        final MethodHandle filtered = MH.foldArguments(MH.dropArguments(NEWFILTER, 2, ctorArgs), returnsObject);

        // allocate() takes a ScriptFunction and returns a newly allocated ScriptObject...
        if (needsCallee) {
            // ...we either fold it into the previous composition, if we need both the ScriptFunction callee object and
            // the newly allocated object in the arguments, so (this, callee, args...) x (callee) => (callee, args...),
            // or...
            return MH.foldArguments(filtered, ScriptFunction.ALLOCATE);

        // ...replace the ScriptFunction argument with the newly allocated object, if it doesn't need the callee
        // (this, args...) filter (callee) => (callee, args...)
        return MH.filterArguments(filtered, 0, ScriptFunction.ALLOCATE);

     * Permutes the parameters in the method handle from {@code (callee, this, ...)} to {@code (this, callee, ...)}.
     * Used when creating a constructor handle.
     * @param mh a method handle with order of arguments {@code (callee, this, ...)}
     * @return a method handle with order of arguments {@code (this, callee, ...)}
    private static MethodHandle swapCalleeAndThis(final MethodHandle mh) {
        final MethodType type = mh.type();
        assert type.parameterType(0) == ScriptFunction.class : type;
        assert type.parameterType(1) == Object.class : type;
        final MethodType newType = type.changeParameterType(0, Object.class).changeParameterType(1, ScriptFunction.class);
        final int[] reorder = new int[type.parameterCount()];
        reorder[0] = 1;
        assert reorder[1] == 0;
        for (int i = 2; i < reorder.length; ++i) {
            reorder[i] = i;
        return MethodHandles.permuteArguments(mh, newType, reorder);

     * Returns an invoker method handle for this function when invoked as a constructor. Note that the handle is safely
     * composable in the sense that you can compose it with other handles using any combinators even if you can't affect
     * call site invalidation. If this compiled function is non-optimistic, then it returns the same value as
     * {@link #getConstructor()}. However, if the function is optimistic, then this handle will incur an overhead as it
     * will add an intermediate internal call site that can relink itself when the function needs to regenerate its code
     * to always point at the latest generated code version.
     * @return a guaranteed composable constructor method handle for this function.
    MethodHandle createComposableConstructor() {
        return createComposableInvoker(true);

    boolean hasConstructor() {
        return constructor != null;

    MethodType type() {
        return invoker.type();

    int weight() {
        return weight(type());

    private static int weight(final MethodType type) {
        if (isVarArgsType(type)) {
            return Integer.MAX_VALUE; //if there is a varargs it should be the heavist and last fallback

        int weight = Type.typeFor(type.returnType()).getWeight();
        for (int i = 0 ; i < type.parameterCount() ; i++) {
            final Class<?> paramType = type.parameterType(i);
            final int pweight = Type.typeFor(paramType).getWeight() * 2; //params are more important than call types as return values are always specialized
            weight += pweight;

        weight += type.parameterCount(); //more params outweigh few parameters

        return weight;

    static boolean isVarArgsType(final MethodType type) {
        assert type.parameterCount() >= 1 : type;
        return type.parameterType(type.parameterCount() - 1) == Object[].class;

    static boolean moreGenericThan(final MethodType mt0, final MethodType mt1) {
        return weight(mt0) > weight(mt1);

    boolean betterThanFinal(final CompiledFunction other, final MethodType callSiteMethodType) {
        // Prefer anything over nothing, as we can't compile new versions.
        if (other == null) {
            return true;
        return betterThanFinal(this, other, callSiteMethodType);

    private static boolean betterThanFinal(final CompiledFunction cf, final CompiledFunction other, final MethodType callSiteMethodType) {
        final MethodType thisMethodType  = cf.type();
        final MethodType otherMethodType = other.type();
        final int thisParamCount = getParamCount(thisMethodType);
        final int otherParamCount = getParamCount(otherMethodType);
        final int callSiteRawParamCount = getParamCount(callSiteMethodType);
        final boolean csVarArg = callSiteRawParamCount == Integer.MAX_VALUE;
        // Subtract 1 for callee for non-vararg call sites
        final int callSiteParamCount = csVarArg ? callSiteRawParamCount : callSiteRawParamCount - 1;

        // Prefer the function that discards less parameters
        final int thisDiscardsParams = Math.max(callSiteParamCount - thisParamCount, 0);
        final int otherDiscardsParams = Math.max(callSiteParamCount - otherParamCount, 0);
        if(thisDiscardsParams < otherDiscardsParams) {
            return true;
        if(thisDiscardsParams > otherDiscardsParams) {
            return false;

        final boolean thisVarArg = thisParamCount == Integer.MAX_VALUE;
        final boolean otherVarArg = otherParamCount == Integer.MAX_VALUE;
        if(!(thisVarArg && otherVarArg && csVarArg)) {
            // At least one of them isn't vararg
            final Type[] thisType = toTypeWithoutCallee(thisMethodType, 0); // Never has callee
            final Type[] otherType = toTypeWithoutCallee(otherMethodType, 0); // Never has callee
            final Type[] callSiteType = toTypeWithoutCallee(callSiteMethodType, 1); // Always has callee

            int narrowWeightDelta = 0;
            int widenWeightDelta = 0;
            final int minParamsCount = Math.min(Math.min(thisParamCount, otherParamCount), callSiteParamCount);
            final boolean convertsNumericArgs = cf.convertsNumericArgs();
            for(int i = 0; i < minParamsCount; ++i) {
                final Type callSiteParamType = getParamType(i, callSiteType, csVarArg);
                final Type thisParamType = getParamType(i, thisType, thisVarArg);
                if (!convertsNumericArgs && callSiteParamType.isBoolean() && thisParamType.isNumeric()) {
                    // When an argument is converted to number by a function it is safe to "widen" booleans to numeric types.
                    // However, we must avoid this conversion for generic functions such as Array.prototype.push.
                    return false;
                final int callSiteParamWeight = callSiteParamType.getWeight();
                // Delta is negative for narrowing, positive for widening
                final int thisParamWeightDelta = thisParamType.getWeight() - callSiteParamWeight;
                final int otherParamWeightDelta = getParamType(i, otherType, otherVarArg).getWeight() - callSiteParamWeight;
                // Only count absolute values of narrowings
                narrowWeightDelta += Math.max(-thisParamWeightDelta, 0) - Math.max(-otherParamWeightDelta, 0);
                // Only count absolute values of widenings
                widenWeightDelta += Math.max(thisParamWeightDelta, 0) - Math.max(otherParamWeightDelta, 0);

            // If both functions accept more arguments than what is passed at the call site, account for ability
            // to receive Undefined un-narrowed in the remaining arguments.
            if(!thisVarArg) {
                for(int i = callSiteParamCount; i < thisParamCount; ++i) {
                    narrowWeightDelta += Math.max(Type.OBJECT.getWeight() - thisType[i].getWeight(), 0);
            if(!otherVarArg) {
                for(int i = callSiteParamCount; i < otherParamCount; ++i) {
                    narrowWeightDelta -= Math.max(Type.OBJECT.getWeight() - otherType[i].getWeight(), 0);

            // Prefer function that narrows less
            if(narrowWeightDelta < 0) {
                return true;
            if(narrowWeightDelta > 0) {
                return false;

            // Prefer function that widens less
            if(widenWeightDelta < 0) {
                return true;
            if(widenWeightDelta > 0) {
                return false;

        // Prefer the function that exactly matches the arity of the call site.
        if(thisParamCount == callSiteParamCount && otherParamCount != callSiteParamCount) {
            return true;
        if(thisParamCount != callSiteParamCount && otherParamCount == callSiteParamCount) {
            return false;

        // Otherwise, neither function matches arity exactly. We also know that at this point, they both can receive
        // more arguments than call site, otherwise we would've already chosen the one that discards less parameters.
        // Note that variable arity methods are preferred, as they actually match the call site arity better, since they
        // really have arbitrary arity.
        if(thisVarArg) {
            if(!otherVarArg) {
                return true; //
        } else if(otherVarArg) {
            return false;

        // Neither is variable arity; chose the one that has less extra parameters.
        final int fnParamDelta = thisParamCount - otherParamCount;
        if(fnParamDelta < 0) {
            return true;
        if(fnParamDelta > 0) {
            return false;

        final int callSiteRetWeight = Type.typeFor(callSiteMethodType.returnType()).getWeight();
        // Delta is negative for narrower return type, positive for wider return type
        final int thisRetWeightDelta = Type.typeFor(thisMethodType.returnType()).getWeight() - callSiteRetWeight;
        final int otherRetWeightDelta = Type.typeFor(otherMethodType.returnType()).getWeight() - callSiteRetWeight;

        // Prefer function that returns a less wide return type
        final int widenRetDelta = Math.max(thisRetWeightDelta, 0) - Math.max(otherRetWeightDelta, 0);
        if(widenRetDelta < 0) {
            return true;
        if(widenRetDelta > 0) {
            return false;

        // Prefer function that returns a less narrow return type
        final int narrowRetDelta = Math.max(-thisRetWeightDelta, 0) - Math.max(-otherRetWeightDelta, 0);
        if(narrowRetDelta < 0) {
            return true;
        if(narrowRetDelta > 0) {
            return false;

        //if they are equal, pick the specialized one first
        if (cf.isSpecialization() != other.isSpecialization()) {
            return cf.isSpecialization(); //always pick the specialized version if we can

        if (cf.isSpecialization() && other.isSpecialization()) {
            return cf.getLinkLogicClass() != null; //pick link logic specialization above generic specializations

        // Signatures are identical
        throw new AssertionError(thisMethodType + " identically applicable to " + otherMethodType + " for " + callSiteMethodType);

    private static Type[] toTypeWithoutCallee(final MethodType type, final int thisIndex) {
        final int paramCount = type.parameterCount();
        final Type[] t = new Type[paramCount - thisIndex];
        for(int i = thisIndex; i < paramCount; ++i) {
            t[i - thisIndex] = Type.typeFor(type.parameterType(i));
        return t;

    private static Type getParamType(final int i, final Type[] paramTypes, final boolean isVarArg) {
        final int fixParamCount = paramTypes.length - (isVarArg ? 1 : 0);
        if(i < fixParamCount) {
            return paramTypes[i];
        assert isVarArg;
        return ((ArrayType)paramTypes[paramTypes.length - 1]).getElementType();

    boolean matchesCallSite(final MethodType other, final boolean pickVarArg) {
        if (other.equals(this.callSiteType)) {
            return true;
        final MethodType type  = type();
        final int fnParamCount = getParamCount(type);
        final boolean isVarArg = fnParamCount == Integer.MAX_VALUE;
        if (isVarArg) {
            return pickVarArg;

        final int csParamCount = getParamCount(other);
        final boolean csIsVarArg = csParamCount == Integer.MAX_VALUE;
        if (csIsVarArg && isApplyToCall()) {
            return false; // apply2call function must be called with exact number of parameters
        final int thisThisIndex = needsCallee() ? 1 : 0; // Index of "this" parameter in this function's type

        final int fnParamCountNoCallee = fnParamCount - thisThisIndex;
        final int minParams = Math.min(csParamCount - 1, fnParamCountNoCallee); // callSiteType always has callee, so subtract 1
        // We must match all incoming parameters, including "this". "this" will usually be Object, but there
        // are exceptions, e.g. when calling functions with primitive "this" in strict mode or through call/apply.
        for(int i = 0; i < minParams; ++i) {
            final Type fnType = Type.typeFor(type.parameterType(i + thisThisIndex));
            final Type csType = csIsVarArg ? Type.OBJECT : Type.typeFor(other.parameterType(i + 1));
            if(!fnType.isEquivalentTo(csType)) {
                return false;

        // Must match any undefined parameters to Object type.
        for(int i = minParams; i < fnParamCountNoCallee; ++i) {
            if(!Type.typeFor(type.parameterType(i + thisThisIndex)).isEquivalentTo(Type.OBJECT)) {
                return false;

        return true;

    private static int getParamCount(final MethodType type) {
        final int paramCount = type.parameterCount();
        return type.parameterType(paramCount - 1).isArray() ? Integer.MAX_VALUE : paramCount;

    private boolean canBeDeoptimized() {
        return optimismInfo != null;

    private MethodHandle createComposableInvoker(final boolean isConstructor) {
        final MethodHandle handle = getInvokerOrConstructor(isConstructor);

        // If compiled function is not optimistic, it can't ever change its invoker/constructor, so just return them
        // directly.
        if(!canBeDeoptimized()) {
            return handle;

        // Otherwise, we need a new level of indirection; need to introduce a mutable call site that can relink itself
        // to the compiled function's changed target whenever the optimistic assumptions are invalidated.
        final CallSite cs = new MutableCallSite(handle.type());
        relinkComposableInvoker(cs, this, isConstructor);
        return cs.dynamicInvoker();

    private static class HandleAndAssumptions {
        final MethodHandle handle;
        final SwitchPoint assumptions;

        HandleAndAssumptions(final MethodHandle handle, final SwitchPoint assumptions) {
            this.handle = handle;
            this.assumptions = assumptions;

        GuardedInvocation createInvocation() {
            return new GuardedInvocation(handle, assumptions);

     * Returns a pair of an invocation created with a passed-in supplier and a non-invalidated switch point for
     * optimistic assumptions (or null for the switch point if the function can not be deoptimized). While the method
     * makes a best effort to return a non-invalidated switch point (compensating for possible deoptimizing
     * recompilation happening on another thread) it is still possible that by the time this method returns the
     * switchpoint has been invalidated by a {@code RewriteException} triggered on another thread for this function.
     * This is not a problem, though, as these switch points are always used to produce call sites that fall back to
     * relinking when they are invalidated, and in this case the execution will end up here again. What this method
     * basically does is minimize such busy-loop relinking while the function is being recompiled on a different thread.
     * @param invocationSupplier the supplier that constructs the actual invocation method handle; should use the
     * {@code CompiledFunction} method itself in some capacity.
     * @return a tuple object containing the method handle as created by the supplier and an optimistic assumptions
     * switch point that is guaranteed to not have been invalidated before the call to this method (or null if the
     * function can't be further deoptimized).
    private synchronized HandleAndAssumptions getValidOptimisticInvocation(final Supplier<MethodHandle> invocationSupplier) {
        for(;;) {
            final MethodHandle handle = invocationSupplier.get();
            final SwitchPoint assumptions = canBeDeoptimized() ? optimismInfo.optimisticAssumptions : null;
            if(assumptions != null && assumptions.hasBeenInvalidated()) {
                // We can be in a situation where one thread is in the middle of a deoptimizing compilation when we hit
                // this and thus, it has invalidated the old switch point, but hasn't created the new one yet. Note that
                // the behavior of invalidating the old switch point before recompilation, and only creating the new one
                // after recompilation is by design. If we didn't wait here for the recompilation to complete, we would
                // be busy looping through the fallback path of the invalidated switch point, relinking the call site
                // again with the same invalidated switch point, invoking the fallback, etc. stealing CPU cycles from
                // the recompilation task we're dependent on. This can still happen if the switch point gets invalidated
                // after we grabbed it here, in which case we'll indeed do one busy relink immediately.
                try {
                } catch (final InterruptedException e) {
                    // Intentionally ignored. There's nothing meaningful we can do if we're interrupted
            } else {
                return new HandleAndAssumptions(handle, assumptions);

    private static void relinkComposableInvoker(final CallSite cs, final CompiledFunction inv, final boolean constructor) {
        final HandleAndAssumptions handleAndAssumptions = inv.getValidOptimisticInvocation(new Supplier<MethodHandle>() {
            public MethodHandle get() {
                return inv.getInvokerOrConstructor(constructor);
        final MethodHandle handle = handleAndAssumptions.handle;
        final SwitchPoint assumptions = handleAndAssumptions.assumptions;
        final MethodHandle target;
        if(assumptions == null) {
            target = handle;
        } else {
            final MethodHandle relink = MethodHandles.insertArguments(RELINK_COMPOSABLE_INVOKER, 0, cs, inv, constructor);
            target = assumptions.guardWithTest(handle, MethodHandles.foldArguments(cs.dynamicInvoker(), relink));

    private MethodHandle getInvokerOrConstructor(final boolean selectCtor) {
        return selectCtor ? getConstructor() : createInvokerForPessimisticCaller();

     * Returns a guarded invocation for this function when not invoked as a constructor. The guarded invocation has no
     * guard but it potentially has an optimistic assumptions switch point. As such, it will probably not be used as a
     * final guarded invocation, but rather as a holder for an invocation handle and switch point to be decomposed and
     * reassembled into a different final invocation by the user of this method. Any recompositions should take care to
     * continue to use the switch point. If that is not possible, use {@link #createComposableInvoker()} instead.
     * @return a guarded invocation for an ordinary (non-constructor) invocation of this function.
    GuardedInvocation createFunctionInvocation(final Class<?> callSiteReturnType, final int callerProgramPoint) {
        return getValidOptimisticInvocation(new Supplier<MethodHandle>() {
            public MethodHandle get() {
                return createInvoker(callSiteReturnType, callerProgramPoint);

     * Returns a guarded invocation for this function when invoked as a constructor. The guarded invocation has no guard
     * but it potentially has an optimistic assumptions switch point. As such, it will probably not be used as a final
     * guarded invocation, but rather as a holder for an invocation handle and switch point to be decomposed and
     * reassembled into a different final invocation by the user of this method. Any recompositions should take care to
     * continue to use the switch point. If that is not possible, use {@link #createComposableConstructor()} instead.
     * @return a guarded invocation for invocation of this function as a constructor.
    GuardedInvocation createConstructorInvocation() {
        return getValidOptimisticInvocation(new Supplier<MethodHandle>() {
            public MethodHandle get() {
                return getConstructor();

    private MethodHandle createInvoker(final Class<?> callSiteReturnType, final int callerProgramPoint) {
        final boolean isOptimistic = canBeDeoptimized();
        MethodHandle handleRewriteException = isOptimistic ? createRewriteExceptionHandler() : null;

        MethodHandle inv = invoker;
        if(isValid(callerProgramPoint)) {
            inv = OptimisticReturnFilters.filterOptimisticReturnValue(inv, callSiteReturnType, callerProgramPoint);
            inv = changeReturnType(inv, callSiteReturnType);
            if(callSiteReturnType.isPrimitive() && handleRewriteException != null) {
                // because handleRewriteException always returns Object
                handleRewriteException = OptimisticReturnFilters.filterOptimisticReturnValue(handleRewriteException,
                        callSiteReturnType, callerProgramPoint);
        } else if(isOptimistic) {
            // Required so that rewrite exception has the same return type. It'd be okay to do it even if we weren't
            // optimistic, but it isn't necessary as the linker upstream will eventually convert the return type.
            inv = changeReturnType(inv, callSiteReturnType);

        if(isOptimistic) {
            assert handleRewriteException != null;
            final MethodHandle typedHandleRewriteException = changeReturnType(handleRewriteException, inv.type().returnType());
            return MH.catchException(inv, RewriteException.class, typedHandleRewriteException);
        return inv;

    private MethodHandle createRewriteExceptionHandler() {
        return MH.foldArguments(RESTOF_INVOKER, MH.insertArguments(HANDLE_REWRITE_EXCEPTION, 0, this, optimismInfo));

    private static MethodHandle changeReturnType(final MethodHandle mh, final Class<?> newReturnType) {
        return Bootstrap.getLinkerServices().asType(mh, mh.type().changeReturnType(newReturnType));

    private static MethodHandle handleRewriteException(final CompiledFunction function, final OptimismInfo oldOptimismInfo, final RewriteException re) {
        return function.handleRewriteException(oldOptimismInfo, re);

     * Debug function for printing out all invalidated program points and their
     * invalidation mapping to next type
     * @param ipp
     * @return string describing the ipp map
    private static List<String> toStringInvalidations(final Map<Integer, Type> ipp) {
        if (ipp == null) {
            return Collections.emptyList();

        final List<String> list = new ArrayList<>();

        for (final Iterator<Map.Entry<Integer, Type>> iter = ipp.entrySet().iterator(); iter.hasNext(); ) {
            final Map.Entry<Integer, Type> entry = iter.next();
            final char bct = entry.getValue().getBytecodeStackType();
            final String type;

            switch (entry.getValue().getBytecodeStackType()) {
            case 'A':
                type = "object";
            case 'I':
                type = "int";
            case 'J':
                type = "long";
            case 'D':
                type = "double";
                type = String.valueOf(bct);

            final StringBuilder sb = new StringBuilder();
                    append("program point: ").
                    append(" -> ").


        return list;

    private void logRecompile(final String reason, final FunctionNode fn, final MethodType type, final Map<Integer, Type> ipp) {
        if (log.isEnabled()) {
            log.info(reason, DebugLogger.quote(fn.getName()), " signature: ", type);
            for (final String str : toStringInvalidations(ipp)) {

     * Handles a {@link RewriteException} raised during the execution of this function by recompiling (if needed) the
     * function with an optimistic assumption invalidated at the program point indicated by the exception, and then
     * executing a rest-of method to complete the execution with the deoptimized version.
     * @param oldOptInfo the optimism info of this function. We must store it explicitly as a bound argument in the
     * method handle, otherwise it could be null for handling a rewrite exception in an outer invocation of a recursive
     * function when recursive invocations of the function have completely deoptimized it.
     * @param re the rewrite exception that was raised
     * @return the method handle for the rest-of method, for folding composition.
    private synchronized MethodHandle handleRewriteException(final OptimismInfo oldOptInfo, final RewriteException re) {
        if (log.isEnabled()) {
                    new RecompilationEvent(
                    "caught RewriteException ",

        final MethodType type = type();

        // Compiler needs a call site type as its input, which always has a callee parameter, so we must add it if
        // this function doesn't have a callee parameter.
        final MethodType ct = type.parameterType(0) == ScriptFunction.class ?
                type :
                type.insertParameterTypes(0, ScriptFunction.class);
        final OptimismInfo currentOptInfo = optimismInfo;
        final boolean shouldRecompile = currentOptInfo != null && currentOptInfo.requestRecompile(re);

        // Effective optimism info, for subsequent use. We'll normally try to use the current (latest) one, but if it
        // isn't available, we'll use the old one bound into the call site.
        final OptimismInfo effectiveOptInfo = currentOptInfo != null ? currentOptInfo : oldOptInfo;
        FunctionNode fn = effectiveOptInfo.reparse();
        final boolean cached = fn.isCached();
        final Compiler compiler = effectiveOptInfo.getCompiler(fn, ct, re); //set to non rest-of

        if (!shouldRecompile) {
            // It didn't necessarily recompile, e.g. for an outer invocation of a recursive function if we already
            // recompiled a deoptimized version for an inner invocation.
            // We still need to do the rest of from the beginning
            logRecompile("Rest-of compilation [STANDALONE] ", fn, ct, effectiveOptInfo.invalidatedProgramPoints);
            return restOfHandle(effectiveOptInfo, compiler.compile(fn, cached ? CompilationPhases.COMPILE_CACHED_RESTOF : CompilationPhases.COMPILE_ALL_RESTOF), currentOptInfo != null);

        logRecompile("Deoptimizing recompilation (up to bytecode) ", fn, ct, effectiveOptInfo.invalidatedProgramPoints);
        fn = compiler.compile(fn, cached ? CompilationPhases.RECOMPILE_CACHED_UPTO_BYTECODE : CompilationPhases.COMPILE_UPTO_BYTECODE);
        log.fine("Reusable IR generated");

        // compile the rest of the function, and install it
        log.info("Generating and installing bytecode from reusable IR...");
        logRecompile("Rest-of compilation [CODE PIPELINE REUSE] ", fn, ct, effectiveOptInfo.invalidatedProgramPoints);
        final FunctionNode normalFn = compiler.compile(fn, CompilationPhases.GENERATE_BYTECODE_AND_INSTALL);

        if (effectiveOptInfo.data.usePersistentCodeCache()) {
            final RecompilableScriptFunctionData data = effectiveOptInfo.data;
            final int functionNodeId = data.getFunctionNodeId();
            final TypeMap typeMap = data.typeMap(ct);
            final Type[] paramTypes = typeMap == null ? null : typeMap.getParameterTypes(functionNodeId);
            final String cacheKey = CodeStore.getCacheKey(functionNodeId, paramTypes);
            compiler.persistClassInfo(cacheKey, normalFn);

        final boolean canBeDeoptimized = normalFn.canBeDeoptimized();

        if (log.isEnabled()) {

            log.info("Recompiled '", fn.getName(), "' (", Debug.id(this), ") ", canBeDeoptimized ? "can still be deoptimized." : " is completely deoptimized.");
            log.finest("Looking up invoker...");

        final MethodHandle newInvoker = effectiveOptInfo.data.lookup(fn);
        invoker     = newInvoker.asType(type.changeReturnType(newInvoker.type().returnType()));
        constructor = null; // Will be regenerated when needed

        log.info("Done: ", invoker);
        final MethodHandle restOf = restOfHandle(effectiveOptInfo, compiler.compile(fn, CompilationPhases.GENERATE_BYTECODE_AND_INSTALL_RESTOF), canBeDeoptimized);

        // Note that we only adjust the switch point after we set the invoker/constructor. This is important.
        if (canBeDeoptimized) {
            effectiveOptInfo.newOptimisticAssumptions(); // Otherwise, set a new switch point.
        } else {
            optimismInfo = null; // If we got to a point where we no longer have optimistic assumptions, let the optimism info go.

        return restOf;

    private MethodHandle restOfHandle(final OptimismInfo info, final FunctionNode restOfFunction, final boolean canBeDeoptimized) {
        assert info != null;
        assert restOfFunction.getCompileUnit().getUnitClassName().contains("restOf");
        final MethodHandle restOf =

        if (!canBeDeoptimized) {
            return restOf;

        // If rest-of is itself optimistic, we must make sure that we can repeat a deoptimization if it, too hits an exception.
        return MH.catchException(restOf, RewriteException.class, createRewriteExceptionHandler());


    private static class OptimismInfo {
        // TODO: this is pointing to its owning ScriptFunctionData. Re-evaluate if that's okay.
        private final RecompilableScriptFunctionData data;
        private final Map<Integer, Type> invalidatedProgramPoints;
        private SwitchPoint optimisticAssumptions;
        private final DebugLogger log;

        OptimismInfo(final RecompilableScriptFunctionData data, final Map<Integer, Type> invalidatedProgramPoints) {
            this.data = data;
            this.log  = data.getLogger();
            this.invalidatedProgramPoints = invalidatedProgramPoints == null ? new TreeMap<>() : invalidatedProgramPoints;

        private void newOptimisticAssumptions() {
            optimisticAssumptions = new SwitchPoint();

        boolean requestRecompile(final RewriteException e) {
            final Type retType            = e.getReturnType();
            final Type previousFailedType = invalidatedProgramPoints.put(e.getProgramPoint(), retType);

            if (previousFailedType != null && !previousFailedType.narrowerThan(retType)) {
                final StackTraceElement[] stack      = e.getStackTrace();
                final String              functionId = stack.length == 0 ?
                        data.getName() :
                        stack[0].getClassName() + "." + stack[0].getMethodName();

                log.info("RewriteException for an already invalidated program point ", e.getProgramPoint(), " in ", functionId, ". This is okay for a recursive function invocation, but a bug otherwise.");

                return false;

            SwitchPoint.invalidateAll(new SwitchPoint[] { optimisticAssumptions });

            return true;

        Compiler getCompiler(final FunctionNode fn, final MethodType actualCallSiteType, final RewriteException e) {
            return data.getCompiler(fn, actualCallSiteType, e.getRuntimeScope(), invalidatedProgramPoints, getEntryPoints(e));

        private static int[] getEntryPoints(final RewriteException e) {
            final int[] prevEntryPoints = e.getPreviousContinuationEntryPoints();
            final int[] entryPoints;
            if (prevEntryPoints == null) {
                entryPoints = new int[1];
            } else {
                final int l = prevEntryPoints.length;
                entryPoints = new int[l + 1];
                System.arraycopy(prevEntryPoints, 0, entryPoints, 1, l);
            entryPoints[0] = e.getProgramPoint();
            return entryPoints;

        FunctionNode reparse() {
            return data.reparse();

    private static Object newFilter(final Object result, final Object allocation) {
        return (result instanceof ScriptObject || !JSType.isPrimitive(result))? result : allocation;

    private static MethodHandle findOwnMH(final String name, final Class<?> rtype, final Class<?>... types) {
        return MH.findStatic(MethodHandles.lookup(), CompiledFunction.class, name, MH.type(rtype, types));



Or download all of them as a single archive file:

File name: jdk.scripting.nashorn-11.0.1-src.zip
File size: 1390965 bytes
Release date: 2018-11-04


JDK 11 jdk.scripting.nashorn.shell.jmod - Scripting Nashorn Shell Module

JDK 11 jdk.rmic.jmod - RMI Compiler Tool

Download and Use JDK 11

⇑⇑ FAQ for JDK (Java Development Kit)

2020-04-25, 83681👍, 0💬