On Thu, Aug 30, 2007 at 11:00:37PM +0200, Marco A. Calamari wrote: [...] > Aug 30 10:40:35.952 +0200 [FATAL] Exception while running server > Aug 30 10:40:36.217 +0200 [FATAL] Traceback (most recent call last): > File > "/usr/lib/python2.3/site-packages/mixminion/server/ServerMain.py", line > 1236, in runServer > server.run() > File > "/usr/lib/python2.3/site-packages/mixminion/server/ServerMain.py", line > 922, in run > self.mmtpServer.process(tickInterval) > File > "/usr/lib/python2.3/site-packages/mixminion/server/MMTPServer.py", line > 768, in process > AsyncServer.process(self, timeout) > File > "/usr/lib/python2.3/site-packages/mixminion/server/MMTPServer.py", line > 235, in process > cap) > File > "/usr/lib/python2.3/site-packages/mixminion/server/MMTPServer.py", line > 326, in process > self.connectionFactory(con) > File > "/usr/lib/python2.3/site-packages/mixminion/server/MMTPServer.py", line > 609, in _newMMTPConnection > addr, port = sock.getpeername() > File "<string>", line 1, in getpeername > error: (107, 'Transport endpoint is not connected') > Aug 30 10:40:36.219 +0200 [FATAL] Shutting down because of exception: > socket.error Looks like this (and other similar bugs) were caused by the fact that we didn't handle it correctly if a socket disappeared (say, by closing) in between when we called accept() and when we connected it to our internal socket representation. This probably only happened under heavy load. I've checked a fix into cvs. yrs, -- Nick Mathewson
Attachment:
pgpRXffFwPx3U.pgp
Description: PGP signature