/******************************************************************************* * Copyright (c) 2006 IBM Corporation and others. * All rights reserved. This program and the accompanying materials * are made available under the terms of the Eclipse Public License v1.0 * which accompanies this distribution, and is available at * http://www.eclipse.org/legal/epl-v10.html * * Contributors: * Andy Clement - initial implementation *******************************************************************************/ package org.aspectj.systemtest.ajc152; import java.io.File; import junit.framework.Test; import org.aspectj.testing.XMLBasedAjcTestCase; /** * Work items, phase #1: lock()/unlock() x expose new joinpoints x parse new pcds x fix tjp string x preventing double unlock() * messages/markers in structure model x error messages appropriate for attempting to use around advice on synchronization join * points x making the use of lock/unlock conditional on an -Xjoinpoints:synchronization x activating the -Xjoinpoints options from * LTW configurations rather than through batch/AJDT x ensure the lock/unlock joinpoints only appear when * -Xjoinpoints:synchronization specified TAG: Completion of PHASE1 * * * Work items, phase #2: transformation * * Design: transform all synchronized methods: public synchronized void m() { ... } => public void m() { synchronized (this) { ... } * } * * x transforming synchronized methods x matching execution(synchronized * *(..)) for transformed code x warning message for * execution() hitting a synchronized method x ensure verifier runs over generated code (done by just executing the code as part of * the test spec) - Ant task support for -Xjoinpoints TAG: Completion of PHASE2 * * * TAG: Finished * * Future work items: - optimize matching for transformed methods since we *know* the type we are locking on - supporting type * pattern in lock() unlock() - this is not entirely trivial as kinded pointcuts do not usually have any residue - weaving messages * include 'unusual' strings for the join points, not the same as revealed by thisJoinPoint.getSignature() in code - handler is * probably similar - documentation - lazy translation of synchronized methods, rather than eager - applying execution(* *(..)) * correctly to transformed methods (i.e. inside lock/unlock) - use knowledge of type containing synchronized methods to optimize * matching of (un)lock() - not always needing residue - line number table is incorrect for transformed code (lock joinpoint has no * line number) * * Notes: IllegalMonitorStateException Thrown to indicate that a thread has attempted to wait on an object's monitor or to notify * other threads waiting on an object's monitor without owning the specified monitor. * * around advice won't work on SUN VMs (may be a bug that it does work on other VMs) since the monitor instructions are extracted to * a separate method for proceed() calls and yet the VM seems to want them paired inside a single method. * * Really we only need to restrict the use of around advice on synchronization join points if the advice uses proceed() - but * policing that is a little tough because (DOH) the AdviceAttribute field 'proceedCallSignatures' is never filled in (which records * how many proceeds occur in the advice) - see where it isnt filled in at AdviceDeclaration.resolveStatements() in the loop that * goes over the proceedCalls list. * * * Problems: - Can't run it on a 1.2.1 runtime - just not practical * * * Method transformation, example: * * public synchronized void m(); Code: Stack=2, Locals=1, Args_size=1 0: getstatic #2; //Field * java/lang/System.err:Ljava/io/PrintStream; 3: ldc #3; //String hello 5: invokevirtual #4; //Method * java/io/PrintStream.println:(Ljava/lang/String;)V 8: getstatic #2; //Field java/lang/System.err:Ljava/io/PrintStream; 11: ldc #5; * //String world 13: invokevirtual #4; //Method java/io/PrintStream.println:(Ljava/lang/String;)V 16: return LineNumberTable: line * 4: 0 line 5: 8 line 6: 16 * * public void m2(); Code: Stack=2, Locals=3, Args_size=1 0: aload_0 1: dup 2: astore_1 3: monitorenter 4: getstatic #2; //Field * java/lang/System.err:Ljava/io/PrintStream; 7: ldc #3; //String hello 9: invokevirtual #4; //Method * java/io/PrintStream.println:(Ljava/lang/String;)V 12: getstatic #2; //Field java/lang/System.err:Ljava/io/PrintStream; 15: ldc * #5; //String world 17: invokevirtual #4; //Method java/io/PrintStream.println:(Ljava/lang/String;)V 20: aload_1 21: monitorexit * 22: goto 30 25: astore_2 26: aload_1 27: monitorexit 28: aload_2 29: athrow 30: return Exception table: from to target type 4 22 * 25 any 25 28 25 any * * Factors affecting transformation: - LDC in Java5 supports referring to a class literal, e.g. Foo.class whereas before Java5, it * did not. This means if generating the synchronized() block for a static method from a preJava5 class then we have to generate a * lot of crap to build the class object for locking and unlocking. The object is also stored in a local field of the type (if we * follow the pattern of JDT/JAVAC) */ public class SynchronizationTests extends XMLBasedAjcTestCase { // testing the new join points for monitorenter/monitorexit public void testTheBasics_1() { runTest("basic"); } public void testTheBasics_2() { runTest("basic - within"); } public void testTheBasics_3() { runTest("basic - within plus args"); } public void testTheBasics_4() { runTest("basic - within plus this"); } // this null in static context public void testTheBasics_5() { runTest("basic - within plus target"); } // target null in static context? // testing parsing of the new PCDs lock/unlock public void testParsing_1() { runTest("parsing - lock"); } public void testParsing_2() { runTest("parsing - unlock"); } public void testParsing_errors_1() { runTest("parsing - error - lock"); } public void testParsing_errors_2() { runTest("parsing - error - unlock"); } // testing parsing and matching with the new PCDs public void testParsingAndMatching_1() { runTest("parsing and matching - lock and static context"); } public void testParsingAndMatching_2() { runTest("parsing and matching - unlock and static context"); } public void testParsingAndMatching_3() { runTest("parsing and matching - lock and non-static context"); } public void testParsingAndMatching_4() { runTest("parsing and matching - unlock and non-static context"); } public void testParsingAndMatching_5() { runTest("parsing and matching - lock and non-static context"); } public void testParsingAndMatching_6() { runTest("parsing and matching - unlock and non-static context"); } // using the new PCDs in a LTW environment public void testUsingWithLTW_MissingFlag_1() { runTest("using lock with LTW - missing flag"); } public void testUsingWithLTW_MissingFlag_2() { runTest("using unlock with LTW - missing flag"); } public void testUsingWithLTW_1() { runTest("using lock with LTW"); } public void testUsingWithLTW_2() { runTest("using unlock with LTW"); } // multiple PCDs public void testCombiningPCDs_1() { runTest("combining pcds - lock and this"); } public void testCombiningPCDs_2() { runTest("combining pcds - unlock and this"); } // useful examples public void testUseful_1() { runTest("a useful program"); } // just uses within/args - matching the (un)lock jps public void testUseful_2() { runTest("a useful program - with lock"); } // uses lock/args // all the methods of thisJoinPoint public void testThisJoinPoint_1() { runTest("thisjoinpoint - monitor entry"); } public void testThisJoinPoint_2() { runTest("thisjoinpoint - monitor exit"); } public void testDoubleMessagesOnUnlock() { // AsmManager.setReporting("c:/foo.txt",true,true,true,true); runTest("prevent double unlock weaving messages and model contents"); // checkModel1(); } // targetting 1.2 runtime - signature creation code in LazyClassGen.initializeTjp may not work // different advice kinds public void testBeforeAdvice_1() { runTest("before advice - lock"); } public void testBeforeAdvice_2() { runTest("before advice - unlock"); } public void testAfterAdvice_1() { runTest("after advice - lock"); } public void testAfterAdvice_2() { runTest("after advice - unlock"); } public void testAroundAdvice_1() { runTest("around advice - lock"); } public void testAroundAdvice_2() { runTest("around advice - unlock"); } public void testLockingTJP() { runTest("obtaining locked object through getArgs"); } // binary weaving? // nested locking/unlocking // --- helpers // Half finished - could check there is only one relationship for unlock() rather than two - but // that seems to be the case anyway (peculiar...) // private void checkModel1() { // // Verifies only one unlock relationship, not two // IProgramElement unlockNode = // AsmManager.getDefault().getHierarchy().findElementForLabel(AsmManager.getDefault().getHierarchy().getRoot(), // IProgramElement.Kind.CODE,"unlock(void java.lang.Object.(java.lang.Object))"); // assertTrue("Couldn't find the unlock node",unlockNode!=null); // List l = AsmManager.getDefault().getRelationshipMap().get(unlockNode); // assertTrue("should be one entry :"+l,l!=null && l.size()==1); // IRelationship ir = (IRelationship)l.get(0); // System.err.println(ir); // List targs = ir.getTargets(); // System.err.println(targs.size()); // System.err.println(targs.get(0)); // } // --- public static Test suite() { return XMLBasedAjcTestCase.loadSuite(SynchronizationTests.class); } protected File getSpecFile() { return new File("../tests/src/org/aspectj/systemtest/ajc152/synchronization.xml"); } }