Bugzilla – Full Text Bug Listing
| Summary: | circular-pointcut message when redefining/reusing pertarget pointcut | ||
|---|---|---|---|
| Product: | [Tools] AspectJ | Reporter: | Wes Isberg <wes> |
| Component: | Compiler | Assignee: | aspectj inbox <aspectj-inbox> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | P2 | ||
| Version: | 1.5.0 | ||
| Target Milestone: | --- | ||
| Hardware: | PC | ||
| OS: | Windows XP | ||
| Whiteboard: | |||
In AspectJ 1.5.1a (from AJDT), I now get a ClassCastException rather than a compiler error, so I'm hoisting the priority to P2. Internal compiler error java.lang.ClassCastException: org.aspectj.weaver.MissingResolvedTypeWithKnownSignature at org.aspectj.ajdt.internal.core.builder.AjState.recordClassFile(AjState.java:783) at org.aspectj.ajdt.internal.core.builder.AjState.noteResult(AjState.java:636) at org.aspectj.ajdt.internal.core.builder.AjBuildManager$3.acceptResult(AjBuildManager.java:884) at org.aspectj.ajdt.internal.compiler.AjCompilerAdapter.afterProcessing(AjCompilerAdapter.java:206) at org.aspectj.ajdt.internal.compiler.CompilerAdapter.ajc$afterReturning$org_aspectj_ajdt_internal_compiler_CompilerAdapter$4$6b855184(CompilerAdapter.aj:90) at org.aspectj.org.eclipse.jdt.internal.compiler.Compiler.process(Compiler.java:528) at org.aspectj.org.eclipse.jdt.internal.compiler.Compiler.compile(Compiler.java:329) at org.aspectj.ajdt.internal.core.builder.AjBuildManager.performCompilation(AjBuildManager.java:862) at org.aspectj.ajdt.internal.core.builder.AjBuildManager.doBuild(AjBuildManager.java:269) at org.aspectj.ajdt.internal.core.builder.AjBuildManager.incrementalBuild(AjBuildManager.java:168) at org.aspectj.ajde.internal.CompilerAdapter.compile(CompilerAdapter.java:117) at org.aspectj.ajde.internal.AspectJBuildManager$CompilerThread.run(AspectJBuildManager.java:191) I believe the classcast at that line is fixed (cant remember under which bug we did it though!) Verified fix in head just now. I should have installed test code as test case. |
I get a message about a circular pointcut definition when reusing the superaspect definition in a subaspect redefinition (override), but only if the aspect is declared pertarget on that pointcut. At a minimum we need to decide if the pointcut referred to in the superaspect pertarget clause should resolve to the superaspect pointcut or the subaspect pointcut. I'm inclined to use the overridden pointcut if the pointcut reference is unqualified, in order to support binary library aspects. Conversely, library aspect writers who want to avoid overriding should be able to use a static reference to the library aspect, e.g., abstract aspect Foo pertarget(Foo.pc()) { ... (which would be an error if pc() is itself abstract). More (but not entirely) formally, a reference of the form Foo.pc() is resolved by looking in Foo and supertypes for a poincut pc, while an implicit reference of the form "pc()" is resolved by using the most-concrete type and then up the supertypes. ----------------------------------------------------------------- package bugs; public class ReuseWhileRedefiningPointcut { static void run(){} public static void main(String[] args) { run(); } // use pertarget clause to see error message static abstract aspect Super { //pertarget(voidCalls()){ pointcut voidCalls() : call(void *(..)) && within(ReuseWhileRedefiningPointcut); before() : voidCalls() { System.out.println("here: " + thisJoinPoint); } } static aspect A extends Super { pointcut voidCalls() : Super.voidCalls() && call(void run()); } }