On Jun 27, 2012, at 4:02 PM, Massimo Di Pierro wrote: > I agree. Please check.
That's better; thanks. > > On Wednesday, 27 June 2012 17:53:25 UTC-5, Jonathan Lundell wrote: > On Jun 27, 2012, at 3:45 PM, Massimo Di Pierro wrote: >> ok.please check the solution in trunk. > > OK, but I think that SocketClosed might be a better exception. And maybe a > comment pointing to this thread or something else explaining what it's doing > there (it's a pretty odd piece of code otherwise). > >> >> On Wednesday, 27 June 2012 08:20:12 UTC-5, Jonathan Lundell wrote: >> On Jun 27, 2012, at 6:12 AM, Massimo Di Pierro wrote: >>> We acn catch it but what do we do? Report "this is a python bug"? >> >> The crash is a python bug. The connection failure isn't (near as I can >> tell). Just treat it as a dropped connection. >> >>> >>> On Tuesday, 26 June 2012 20:01:49 UTC-5, Jonathan Lundell wrote: >>> On Jun 26, 2012, at 5:25 PM, Massimo Di Pierro wrote: >>>> Looks like a buggy socket.py or ssl.py. Is this an occasional error (which >>>> may depend on parameters such as certificates) or it reproducible? >>> >>> This was fixed in Python 2.7. The underlying error is "Transport endpoint >>> is not connected", which of course shouldn't cause a crash. I suppose we >>> could catch the exception in Rocket as a workaround. >>> >> >> > >