On Wed, 5 Jul 2023 19:14:07 GMT, Mandy Chung <mch...@openjdk.org> wrote:
> Thanks for catching this issue. I agree that `Method::invoke` should be > skipped the caller-sensitive test in this case but the fix isn't quite right. > The caller-sensitive test should apply in any batch. For example, `CSM` calls > `getCallerClass` reflectively, I think the stack would look like this: > > ``` > java.lang.StackWalker::getCallerClass > java.lang.invoke.DirectMethodHandle$Holder::invokeStatic > java.lang.invoke.LambdaForm$MH/0x0000000800002c00::invoke > : > : > jdk.internal.reflect.DirectMethodHandleAccessor::invokeImpl > jdk.internal.reflect.DirectMethodHandleAccessor::invoke > java.lang.reflect.Method::invoke > CSM <--------- caller-sensitive method and UOE should be thrown > ``` > > In this case, UOE should be thrown. Hi @mlchung , Thanks for your comments. I think you're right. More specifically, the caller-sensitive test should be applied to the first non-reflective, non-methodhandle frame, independent on the batch in which it appears. I'll add a new test for the case you've brought up and update the fix accordingly. Best regards, Volker ------------- PR Comment: https://git.openjdk.org/jdk/pull/14773#issuecomment-1623171029
