To resolve JDK-8309671, the short cut taken was to hard code the assumption that Graal does not use precise BCIs for uncommon traps (i.e. deoptimization can restart the interpreter at a bytecode instruction earlier than the one associated with the uncommon trap).

This means the serviceability/jvmti/HeapMonitor/MyPackage/HeapMonitor.java test will break if the assumption is invalidated in future. Instead, the whitebox API should be extended to ask a JIT compiler about this behavior. That is, an uncommonTrapsHavePreciseBCIs() method should be added to jdk.test.whitebox.code.Compiler.