I am wondering why the Exception in the following bytecode (used to throw an Exception) is duplicated.
NEW java/lang/IllegalArgumentException
DUP
INVOKESPECIAL java/lang/IllegalArgumentException <init> ()V
ATHROW
asked Sep 15, 2012 at 15:16
LanguagesNamedAfterCofee
5,9828 gold badges48 silver badges73 bronze badges
2 Answers
I'll analyze this line by line where [] = new stack after that op is used:
- NEW puts a new
IllegalArgumentExceptiononto the stack [SomeIllegalArgumentException] - DUP duplicates it [SomeIllegalArgumentException, SomeIllegalArgumentException]
- INVOKESPECIAL pops off the top one and initializes it by calling it's <init> method [SomeIllegalArgumentException] (The init method will not return the object to put back onto the stack, so the object must first be duplicated so as to keep it on the stack)
- ATHROW Throws the other (a duplicate off the one we initialized) []
1 Comment
obataku
This is easier to understand if you think less of it popping and pushing objects themselves on the stack and instead talk about references.
In byte code, an object is first created by class, and then a constructor is called on that object. The signature of a constructor ends with V for void as it does return anything. This means a copy of the original reference to the object must be kept on the stack (or in a variable) so it can be thrown after the constructor is called.
BTW The internal name for a constructor is <init> and the internal name for a static initialiser code is <clinit>
NEWjust allocates object memeory without calling any constructor,INVOKESPECIALwill call a construction but that call will pop from a stackthispointer and push nothing (constructors returnvoid). By convention compilers targeting JVM will generate aboveNEW/DUP/INVOKESPECIALsequence to create a new object. This is a pattern that is recognized by JVM and may be specially optimized...