Decompilation St Test
Program-Transformation.Org: The Program Transformation Wiki
SourceTec Java Decompiler Simple Tests
SourceTec, also known as Jasmine, is an old decompiler; in fact it's a patch to Mocha, probably the very first Java decompiler. The version I used is 1.10 (from the Readme file), though the decompiler output says Version 1.0. It only works on Java 1.1 files; most Java programs are Java 1.2 (Java 2) these days. I recompiled those tests that I could for 1.1 (using-target 1.1).
Fibo
For source, see DecompilerFiboTestSource. Decompiled source from SourceTec:
/* Decompiled by Jasmine from Fibo.class */
/* Originally compiled from Fibo.java */
import java.io.PrintStream;
synchronized class Fibo
{
Fibo()
{
}
private static int fib(int i)
{
if (i > 1)
return fib(i - 1) + fib(i - 2);
else
return i;
}
public static void main(String astring[])
throws Exception
{
int i = 0;
try
{
i = Integer.parseInt(astring[0]);
}
catch (Exception e)
{
System.out.println("Input error");
System.exit(1);
}
int j = fib(i);
System.out.println("fibonacci(" + i + ") = " + j);
}
}
As you can see, the decompilation looks pretty good, but the synchronized keyword is wrong and stops the output from recompiling. (This would presumably not have happened with true Java 1.1 bytecode files). When this word was removed, the output compiled and ran with no further changes required.
Casting
For source, see DecompilerCastingTestSource. Here is the output from SourceTec:
/* Decompiled by Jasmine from Casting.class */
/* Originally compiled from Casting.java */
import java.io.PrintStream;
public synchronized class Casting
{
public Casting()
{
}
public static void main(String astring[])
{
for (char ch = '\0'; ch < 128; ch = (char)(ch + 1))
System.out.println("ascii " + ch + " character " + ch);
}
}
Again, the synchronized keyword had to be removed for it to recompile. The cast is missing, so the program did not run correctly. With those two changes (removed synchronized, added cast to integer), the program was correct.
Inner classes
For source, see DecompilerInnerClassesTestSource. When decompiled with SourceTec, the result is
Ignoring class attribute InnerClasses
10
/* Decompiled by Jasmine from Usa.class */
/* Originally compiled from Usa.java */
public synchronized class Usa
{
public String name;
public Usa()
{
name = "Detroit";
}
}
It does not understand inner classes. When I tried to decompile Usa$England.class, the result was
Ignoring field attribute Synthetic
0
Ignoring class attribute InnerClasses
18
/* Decompiled by Jasmine from Usa$England.class */
/* Originally compiled from Usa.java */
public synchronized class Usa$England
{
public String name;
private final Usa this$0 = usa;
public Usa$England(Usa usa)
{
name = "London";
}
}
Deuces Wild
This is a 38K applet with two dimensional arrays of integers, some floating point code, and so on. The SourceTec decompiler bombed out with an array out of bounds exception.Sable Test Program
For source, see DecompilerSableTestSource. Here is the result for functionf (as produced by SourceTec):
public static void f(short s)
{
Object object;
boolean flag;
if (s > 10)
{
Rectangle rectangle = new Rectangle(s, s);
flag = rectangle.isFat();
object = rectangle;
}
else
{
Circle circle = new Circle(s);
flag = circle.isFat();
object = circle;
}
if (!flag)
object.draw();
}
It typed the variable, called d in the original source, as type Object, and neglected to cast the variable in the call to drawable. When the cast was added, synchronized was removed, and another cast was added (not shown above), it compiled correctly.
Optimised code.
For source, see DecompilerOptimisedTestSource. This was the result from SourceTec for the methodf:
public static void f(short s)
{
Object object;
if (s > 10)
{
Rectangle rectangle = new Rectangle;
rectangle.<init>(s, s);
s = rectangle.isFat();
object = rectangle;
}
else
{
object = new Circle;
object.<init>(s);
s = object.isFat();
object = object;
}
if (s == 0)
object.draw();
}
As you can see, it is confused with the two constructors, and the cast for the call to draw is still missing. In addition, it is missing another cast to isFat (in the else clause).
When the above problems were corrected (e.g. Rectangle rectangle = new Rectangle(i, i), there are still problems because there is a boolean and a short sharing the same local variable (i and is_fat in the original source, both s in the decompiled output). In the end, quite a few changes were needed to allow the code to compile.
It should be noted that in the "Decompiling Java Bytecode: Problems, Traps and Pitfalls" paper, they optimised the code differently; they report SourceTec (which they label Jasmine) as emitting bytecodes (i.e. failed to decompile the function at all).
Simple control flow
For source, see DecompilerControlFlowTestSource. Output from SourceTec was:
Method foo: Flow analysis could not complete
public int foo(int i, int j)
{
RuntimeException e;
for (i = j++ / i; i < j; i = j++ / i) /* null body */ ;
return j;
pop e
i = 10;
}
The try and catch clauses are not generated, there is bytecode (pop) in the output, and the result is generally quite wrong.
Exceptions
For source, see DecompilerExceptionTestSource. Output from SourceTec was:
public void foo()
{
System.out.println("a");
try
{
System.out.println("b");
try
{
System.out.println("c");
System.out.println("d");
}
catch ()
{
System.out.println("e");
}
}
catch ()
{
System.out.println("g");
}
System.out.println("f");
}
The exception variables are not declared, so the output will not even compile. The exception path from c to g is missing.
Life
This version of Conway's Game of Life was compiled from Ada 95 source code. The top level class file is http://www.appletmagic.com/download/demo/LifeRect.html. WhenLifeRect was decompiled, it emitted these warnings:
Method paint: Flow analysis could not complete Method run: Flow analysis could not completeSimilarly, 6 warnings were emitted for
LR_Colony. These contained bytecode-like code, such as:
compare and if < goto 68
dup 1 over 0
expression 59