In Pyside, why does emiting an integer > 0x7FFFFFFF result in "OverflowError" after the signal is processed?

Question

I'm attempting to use signals/slots with large integers ranging from 0 - 2^32-1. I've discovered something a little weird -- once I emit > 7FFFFFFF boundary, I get OverflowError exceptions thrown after the slot is run. I might expect this kind of overflow if I or QT were explicitly using a signed 32 bit integer in another language like C or C++--as we all know 0x80000000 wraps back to -2^31 in 2s complement notation. In python though, its just 2^32 without wrapping. My assumption when writing the code though was that this is python and that the built-in int can grow very large (maybe arbitrarilly so?) and that I don't explicitly need to define something as 32 or 64 bit or signed/unsigned. It would all just work.

The code below demonstrates what I'm seeing (Python 2.7.2 (64 bit), Pyside 1.1.0, Windows 7)

from PySide.QtCore import *

@Slot(int)
def say(i):
    print "Say %i" % i

class Communicate(QObject):
    speak = Signal(int)

someone = Communicate()
someone.speak.connect(say)
someone.speak.emit(0x7FFFFFFF) #works fine
someone.speak.emit(0x80000000) #OverflowError after slot "say" runs
say(0x80000000)                #works fine

The exact output is:

Say 2147483647
Say -2147483648
OverflowError
Say 2147483648

1. Why does Qt seem to treat the signals/slots of type integer as if its dealing with signed 32 bit integers and not python built-in ints?
2. If this is a restriction of Qt, what can I do to mark the int as unsigned or make sure QT can deal with integers > 0x7FFFFFFF?

Answer 1

I'm mostly a PyQt user, but I believe the behavior is similar. int in signal definition is mapped to 4-byte integer (as Qt understands an int).

One possible solution is to force the signal to emit a Python object. This works:

class Communicate(QObject):
    speak = Signal(object)

But be aware that, if you connect this signal to a slot that expects a Qt's version of int (for example QtGui.QSpinBox.setMaximum) you'll see the same behavior. Other than that, using this signal purely on the Python side should be fine.