Ah, I never tried abortConnection, but I'm doing something similar - I just call loseConnection() and that does the trick. And yes, I believe that the TCP connection does stay alive because I was watching closeSocket() and it was never called.
I wish I could be more helpful with regard to code samples to reproduce this - but right now my stuff is a mess. We're using the wokkel library, calling stopService() when we want everything to be cleaned up. They override that function, but I believe they're doing everything correctly. Basically, they're just calling connection.disconnect() - and eventually the shutdownTLS() gets invoked. Maybe most people just never care about having to clean up their connections during a single session? As for a fix - maybe it's not necessary to do the test for if shutdownSuccess in _shutdownTLS() ? According to the ssl spec, it's not absolutely necessary to wait for the server to return the close handshake. That being said, that seems lousy - and it really seems like something else must be going on. On Thu, Mar 28, 2013 at 4:36 PM, Glyph <gl...@twistedmatrix.com> wrote: > > On Mar 28, 2013, at 9:24 AM, Michael Williamson < > williamson.micha...@gmail.com> wrote: > > > Hi, I'm having an issue where a tls connection never seems to shutdown > properly, and as a result, all the associated cleanup never happens. > > > > Basically we're trying to implement a jabber client and allow the user > to sign out and then re-sign back in without completely quitting the > program. When the user tries to sign back in, Twisted goes nuts because > the original connection was never shutdown properly. > > > > So, I traced through the code with the debugger and I came to the > _shutdownTLS() function in tls.py. I believe that what's supposed to > happen is that the first time through, the self._tlsConnection.shutdown() > returns False because the shutdown is initiated. Then, when the 2nd part > of the shutdown handshake comes back from the server, dataReceived is > called, which calls shutdownTLS again, and then the cleanup can continue. > > > > The problem is that after the initial call to shutdownTLS, I never see > anymore data from the server. I'm not sure if it's a server issue (we're > just connecting to the Google jabber servers) where they don't close down > their SSL connections correctly, or if I'm just doing something silly. So > anyway, I could solve this by overriding a bunch of stuff and ensuring that > self.transport.loseConnection() gets called no matter what, even when the > server doesn't respond with the second part of the close handshake, but I > figured I'd ask here first. > > > > So, any TLS experts out there? Thanks for any help or information - I > really appreciate it, > > Sadly I don't know enough about TLS specifics to tell you for sure. > > It sounds like you may have spotted a bug. More investigation is required > though. Does .abortConnection() do what you expect? In your current > problematic scenario, does the TCP connection to your peer stay alive? > > -glyph > > > > _______________________________________________ > Twisted-Python mailing list > Twisted-Python@twistedmatrix.com > http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python >
_______________________________________________ Twisted-Python mailing list Twisted-Python@twistedmatrix.com http://twistedmatrix.com/cgi-bin/mailman/listinfo/twisted-python
