Issue1265

I'm working on using the Java library Restlet with Jython (going well,
plan to write an article about it soon) and I'm encountering this problem:

------------------------------------------------------------------
from org.restlet.resource import Resource

class my_resource(Resource):
    def __init__(self):
        Resource.__init__(self)
        # By default, modifiable is False. Let's set it to True.
        self.setModifiable(True)

# Prints True as one would expect
print my_resource().isModifiable()

# Create an instance using Java reflection
obj = my_resource.getClass(my_resource).getConstructor(my_resource,
None).newInstance(None)

# Prints False, which means that when one uses Java reflection, the
Jython subclass' constructor isn't properly called, or something
print obj.isModifiable()
------------------------------------------------------------------

result:
------------------------------------------------------------------
True
False
------------------------------------------------------------------

This is a showstopper for me, because Restlet internally uses reflection
all over the place.

Other than this issue, the project is going well. I look forward to
verification that my observation about this behaviour is correct!

One more thought: the constructor *is* being run, because if I set a
Python attribute on self, such as self.foo = bar, then _that_ *is*
accessible from the method. It's only modifying fields of the Java class
(via a setter) that seems to not function.

I wonder if this might have something to do with the constructor and the
method seeming to have references to different object instances? What I
mean is, if I add 'print str(self)' to the constructor and the method,
they print the Java object ID -- which is different!

Are you able to reduce this to a simple Java class that shows the bug? 
That would help us narrow down the cause.

Sure, I can do that!

attaching zipfile containing a java class and a jython script which
illustrate the problem

Thanks for the test files, that should help.

So I don't think this can really be fixed.  Semantically I think you are
going around Jython by calling getClass() followed by getConstructor().
 I don't see how we could intercept this sort of behavior in a general
way.  Two workarounds:

you could call __init__ manually after calling the constructor, or you
can just call my_reticulator()

OK, I understand. I can work around it in my specific case fairly easily.