org.aspectj.weaver.bcel.asm.StackMapAdder Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of aspectjweaver Show documentation
The AspectJ weaver applies aspects to Java classes. It can be used as a Java agent in order to apply load-time weaving (LTW) during class-loading and also contains the AspectJ runtime classes.
There is a newer version: 1.9.22.1
Show newest version
/* *******************************************************************
 * Copyright (c) 2008, 2018 Contributors
 * All rights reserved.
 * This program and the accompanying materials are made available
 * under the terms of the Eclipse Public License v 2.0
 * which accompanies this distribution and is available at
 * https://www.eclipse.org/org/documents/epl-2.0/EPL-2.0.txt
 *
 * Contributors:
 *	 Andy Clement
 * ******************************************************************/
package org.aspectj.weaver.bcel.asm;

import org.aspectj.weaver.ResolvedType;
import org.aspectj.weaver.UnresolvedType;
import org.aspectj.weaver.World;

import aj.org.objectweb.asm.ClassReader;
import aj.org.objectweb.asm.ClassVisitor;
import aj.org.objectweb.asm.ClassWriter;
import aj.org.objectweb.asm.MethodVisitor;
import aj.org.objectweb.asm.Opcodes;

/**
 * Uses asm to add the stack map attribute to methods in a class. The class is passed in as pure byte data and then a reader/writer
 * process it. The writer is wired into the world so that types can be resolved and getCommonSuperClass() can be implemented without
 * class loading using the context class loader.
 *
 * It is important that the constant pool is preserved here and asm does not try to remove unused entries.  That is because some
 * entries are refered to from classfile attributes.  Asm cannot see into these attributes so does not realise the constant pool
 * entries are in use.  In order to ensure the copying of cp occurs, we use the variant super constructor call in AspectJConnectClassWriter
 * that passes in the classreader.  However, ordinarily that change causes a further optimization: that if a classreader sees
 * a methodvisitor that has been created by a ClassWriter then it just copies the data across without changing it (and so it
 * fails to attach the stackmapattribute).  In order to avoid this further optimization we use our own minimal MethodVisitor.
 *
 * @author Andy Clement
 */
public class StackMapAdder {

	public static byte[] addStackMaps(World world, String classname, byte[] data) {
		try {
			ClassReader cr = new ClassReader(data);
			ClassWriter cw = new AspectJConnectClassWriter(cr, world);
			ClassVisitor cv = new AspectJClassVisitor(cw);
			cr.accept(cv, 0);
			return cw.toByteArray();
		} catch (Throwable t) {
			// If in here fixing an error about version, change the ASMX in class above!
			System.err.println("AspectJ Internal Error: unable to add stackmap attributes to class '"+classname+"'. " + t.getMessage());
			t.printStackTrace();
			AsmDetector.isAsmAround = false;
			AsmDetector.rootCause = t;
			return data;
		}
	}

	private static class AspectJClassVisitor extends ClassVisitor {

		public AspectJClassVisitor(ClassVisitor classwriter) {
			super(Opcodes.ASM9, classwriter);
		}

		@Override
		public MethodVisitor visitMethod(int access, String name, String desc, String signature, String[] exceptions) {
			MethodVisitor mv = super.visitMethod(access, name, desc, signature, exceptions);
			return new AJMethodVisitor(mv);
		}

		// Minimal pass through MethodVisitor just so that the ClassReader doesn't see one that has been directly
		// created by a ClassWriter (see top level class comment)
		static class AJMethodVisitor extends MethodVisitor {
			public AJMethodVisitor(MethodVisitor mv) {
				super(Opcodes.ASM9,mv);
			}
		}

	}

	private static class AspectJConnectClassWriter extends ClassWriter {
		private final World world;

		public AspectJConnectClassWriter(ClassReader cr, World w) {
			super(cr, ClassWriter.COMPUTE_FRAMES); // passing in cr is necessary so cpool isnt modified (see 2.2.4 of asm doc)
			this.world = w;
		}


		// Implementation of getCommonSuperClass() that avoids Class.forName()
		@Override
		protected String getCommonSuperClass(final String type1, final String type2) {

			ResolvedType resolvedType1 = world.resolve(UnresolvedType.forName(type1.replace('/', '.')));
			ResolvedType resolvedType2 = world.resolve(UnresolvedType.forName(type2.replace('/', '.')));

			if (resolvedType1.isAssignableFrom(resolvedType2)) {
				return type1;
			}

			if (resolvedType2.isAssignableFrom(resolvedType1)) {
				return type2;
			}

			if (resolvedType1.isInterface() || resolvedType2.isInterface()) {
				return "java/lang/Object";
			} else {
				do {
					resolvedType1 = resolvedType1.getSuperclass();
					if (resolvedType1 == null) {
						// This happens if some types are missing, the getSuperclass() call on
						// MissingResolvedTypeWithKnownSignature will return the Missing type which
						// in turn returns a superclass of null. By returning Object here it
						// should surface the cantFindType message raised in the first problematic
						// getSuperclass call
						return "java/lang/Object";
					}
					if (resolvedType1.isParameterizedOrGenericType()) {
						resolvedType1 = resolvedType1.getRawType();
					}
				} while (!resolvedType1.isAssignableFrom(resolvedType2));
				return resolvedType1.getRawName().replace('.', '/');
			}
		}
	}
}