ASM 4.0 RC2: Revisiting the NOP ATHROW bug

Hi All,

I've tried to integrate the ASM RC2 with Jython, but I see the NOP
ATHROW behavior still persistent. I saw a previous discussion of this on
this mailing list (at
http://old.nabble.com/NOP-ATHROW-hack-defeated-%21-tt31501819.html) and
thought that the issue was fixed, but maybe not.

Basically if Jython tries to compile this code:

def openfile():
    fullname = "tmp.txt"
    try:
        f = open(fullname, "rU")
    except IOError:
        return
  return f


The following error occurs when trying to run the code:

java.lang.VerifyError: Stack map does not match the one at exception handler
60 in method
org.python.pycode._pyx0._openfile$1(Lorg/python/core/PyFrame;Lorg/python/core/ThreadState;Lorg/python/core/PyObject;)Lorg/python/core/PyObject;
at offset 45
 at java.lang.Class.getDeclaredConstructors0(Native Method)
at java.lang.Class.privateGetDeclaredConstructors(Class.java:2404)
 at java.lang.Class.getConstructor0(Class.java:2714)
at java.lang.Class.getConstructor(Class.java:1674)
 at org.python.core.BytecodeLoader.makeCode(BytecodeLoader.java:68)
at
org.python.compiler.LegacyCompiler$LazyLegacyBundle.loadCode(LegacyCompiler.java:43)
 at org.python.core.CompilerFacade.compile(CompilerFacade.java:34)
at org.python.core.Py.compile_flags(Py.java:1739)
 at org.python.core.Py.compile_flags(Py.java:1744)
at org.python.core.Py.compile_flags(Py.java:1754)
 at org.python.util.PythonInterpreter.execfile(PythonInterpreter.java:235)
at org.python.util.jython.run(jython.java:252)
 at org.python.util.jython.main(jython.java:134)


(notice the nop and athrow at 95-97 of the byte code). If I just turn off
the Java verifier, the code runs fine.

The java byte code generated for this function by Jython (with ASM RC2) is
here: http://pylonshq.com/pasties/d3cc6de01deeed29c459e38e89983ae7
(all the boiler plate has been removed). This is the smallest example
that reproduces the bug.

I saw from the previous threads that the generated ASM code would be of
help. I tried to replace my ClassFile with the ASMifier, but was unable to
get the ASM code of this particular example (the ASMifier always printed
null).
But, I was able to get the output of another python program, similar to
this, the relevant lines of the python program with the ASM code is here:
http://pylonshq.com/pasties/139f695a127a01d5dee1030b9829554b
This also produces the same bug, but this is part of a large program, so
only what I thought was relevant is kept. Of course I could be wrong here
(don't have too much ASM knowledge) so please let me know if you need more
info. I could put up the whole log up there with the generated bytecode.

Also I talked to Remi at the JVM language Summit, and he suggested that
Jython should do some code analysis to not generate this kind of code in the
first place, while I totally agree with that, I just don't want to do that
work if it can be handled by ASM :)

Please let me know if you need any further information and if I'm doing
something wrong.

Thanks,
Shashank Bharadwaj

-- 
You receive this message as a subscriber of the [email protected] mailing list.
To unsubscribe: mailto:[email protected]
For general help: mailto:[email protected]?subject=help
OW2 mailing lists service home page: http://www.ow2.org/wws