Daniel wrote: > The binascii crc32 fallback seems to also return signed integers. > Maybe forcing self.value to be unsigned could prevent a future > conflict between binascii crc32 (signed) and zlib crc32 (unsigned).
Reading http://mail.python.org/pipermail/python-3000/2008-March/012615.html http://mail.python.org/pipermail/python-dev/2006-May/065452.html and other sources, I think the most compatible fix is to ride the tide and use signed for py <= 2.6 and unsigned for >= py 3.0 to make it futureproof when 3.0 enters mainstream. For example: def digest(self): # crc32 functions are signed in python <= 2.6.x and unsigned in 3.0+ version = string.split(string.split(sys.version)[0], ".") if map(int, version) < [2, 7]: return struct.pack('>i',self.value) else: return struct.pack('>I',self.value) -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
