is this a python bug?

[Michael Hudson]

paragate at gmx.net (Wolfgang Lipp) writes:

> I'm not so sure about what this is I've found.  I run the code
> snippet below (Py2.1, Win95), and the Python Interpreter blows up.

Yep.  Here too.

> Now, this code does have a few problems 

It's certainly not the clearest code I've ever seen!  Doesn't look
like it should dump core, though.

Oh, I see what's going on.  I think.

The problem is that a dictionary gets mutated while it's being
printed, and thus there are pointers pointing off into never-never
land.  I can reproduce this with the rather shorter:

class Child:
    def __init__(self, parent):
        self.__dict__['parent'] = parent
    def __getattr__(self, attr):
        self.parent.a = 1
        self.parent.b = 1
        self.parent.c = 1
        return getattr(self.parent, attr)

class Parent:
    def __init__(self):
        self.a = Child(self)

print Parent().__dict__

I'm *almost* certain this is the same problem you're seeing.  It
doesn't crash CVS Python, but I think this is more in the details than
anything; this sort of thing depends on eg. the order of items in the
dictionaries internal vector and that's changed recently.

Yep; this one crashes CVS:

class Child:
    def __init__(self, parent):
        self.__dict__['parent'] = parent
    def __getattr__(self, attr):
        self.parent.a = 1
        self.parent.b = 1
        self.parent.c = 1
        self.parent.d = 1
        self.parent.e = 1
        self.parent.f = 1
        self.parent.g = 1
        self.parent.h = 1
        self.parent.i = 1
        return getattr(self.parent, attr)

class Parent:
    def __init__(self):
        self.a = Child(self)

print Parent().__dict__ 

I'll get Tim on the case <wink>.

> The code below is more or less reduced to the max. Don't fight over
> the meaningfulness of it; it's but a tiny senseless portion of the
> original.

That's as may be; but if I was up to this level of __getattr__
hackery, I'd have serious thoughts about the design of my application.

> My issue is here: Granted that the code is dumb, should it
> (or any code fed to python.exe) be able to blow up the interpreter?

No.  This is (as code like yours demonstrates) surprisingly hard to
achieve, though.

Cheers,
M.

-- 
  "declare"?  my bogometer indicates that you're really programming
  in some other language and trying to force Common Lisp into your
  mindset.  this won't work.            -- Erik Naggum, comp.lang.lisp