Giampaolo Rodola' <g.rod...@gmail.com> added the comment:
Here's a rewriting attempt (not tested).
Now that I look at it I must say it's quite ugly, so I don't think we should
follow this road.
An alternative I see is to return None in case of errors occurring on accept()
and make this very clear in doc.
Other than accept(), doc should explicitly show how to use handle_accept() in
general, which would end up looking like this:
class SomeServer(asyncore.dispatcher):
...
def handle_accept():
conn = self.accept()
if conn is None:
return
else:
sock, addr = conn
...
...
A completely different approach could be to provide a new "TCPServer" class
which deprecates direct accept() method usage. Something like this:
class TCPServer(asyncore.dispatcher):
def __init__(self, ip, port, handler, interface='', map=None):
asyncore.dispatcher.__init__(self, map=map)
self.create_socket(family=socket.AF_INET, type=socket.SOCK_STREAM)
self.bind((ip, port))
self.listen(5)
self.handler = handler
def handle_accept(self):
try:
sock, addr = self.accept()
except TypeError:
return
except socket.error, err:
if err[0] != errno.ECONNABORTED:
raise
return
else:
if addr == None:
return
handler = self.handler(conn, self._map)
def writable(self):
return 0
...but for some reason I don't like it either. Point is asyncore API design is
fundamentally wrong and there's nothing we can do about it unless we break
backward compatibility or a brand new "asyncore2" module is written.
----------
keywords: +patch
nosy: +pitrou
Added file: http://bugs.python.org/file19001/accept.patch
_______________________________________
Python tracker <rep...@bugs.python.org>
<http://bugs.python.org/issue6706>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
