> On Jan 19, 2018, at 11:52 AM, Ilya Skriblovsky <ilyaskriblov...@gmail.com> > wrote: > > > Protocols and transports have a fairly defined lifecycle, and as L. Daniel > > Burr already pointed out, it would probably be appropriate to explicitly > > break these reference cycles in connectionLost. > > Explicitly breaking cycle in ProtocolWrapper.connectionLost by any of: > • self.wrappedProtocol = None > • self.wrappedProtocol.transport = None > • self.wrappedProtocol = weakref.proxy(self.wrappedProtocol) > • self.wrappedProtocol.transport = weakref.proxy(self) > > ... breaks some existing tests :( > > Seems like these tests do some post-run checks against protocol instances and > their transports. Not sure whether it is relevant to real-life usage. > Will investigate more... > > - Ilya
Do these tests fail if you only do it in TLSMemoryBIOProtocol instead of WrapperProtocol? If so, this may be worth a compatibility exception. -g _______________________________________________ Twisted-Python mailing list Twisted-Python@twistedmatrix.com https://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python
