Title: Different Behaviour in socket.accept() and socket.listen() between jython and python
Type: Severity: normal
Components: Library Versions: 2.2.1rc1
Status: closed Resolution: wont fix
Dependencies: Superseder:
Assigned To: amak Nosy List: amak, asterio
Priority: Keywords:

Created on 2008-04-08.15:43:31 by asterio, last changed 2008-04-08.23:07:24 by amak.

msg3138 (view) Author: Alan Kennedy (amak) Date: 2008-04-08.16:16:41
Need more details, e.g. what behaviour did you expect, and what
behaviour did you actually get.

And ideally, a code snippet which illustrates the problem.
msg3139 (view) Author: asterio (asterio) Date: 2008-04-08.16:49:59
Sorry, was an accident while I was writing the Note.

Consider the following code:

                rnetLoc = ('', 9000)
                fd.bind( rnetLoc ) # python fails here if address is in 
                fd.listen(5)       # but jython fails here

If some code expects to detect an already binded address in bind() 
function then will fails as bind() does send any exception, but 
listen() does.

From my side I guess is a minor detail, just to be as near as possible 
to the python behaviour.

msg3140 (view) Author: Alan Kennedy (amak) Date: 2008-04-08.19:38:36
This problem cannot be fixed.

It happens because jython must present a unified (cpython) interface to
both client and server sockets, which are actually implemented with
different objects on java, namely ServerSocket/Channel and Socket/Channel.

The only way that jython can know what type of underlying socket to
create is to wait for the user to call a method which indicates whether
the socket is a client or a server socket. In this case, that call is
the "listen()" call. The "bind()" is not enough, since both client and
server sockets can bind to an address.

This deferred creation of sockets is why the exception appears only
during the "listen()" call on jython.

This is not an important issue. In practice, running code will most
likely wrap the socket creation, binding and listening in a single
try..except anyway, which will catch the error identically on both
cpython and jython. Consider the following vesion of your code.

import errno, socket

  rnetLoc = ('', 9000)
  fd.bind( rnetLoc )
except socket.error, se:
  print "Exception binding to socket: %s" % se
  if se[0] == errno.EADDRINUSE:
    print "The address is already in use"
  raise SystemExit

while True: pass

So the resolution on this bug is that it is not a bug, or if it is a
bug, then it is not possible to fix it.

However, I will leave the bug open for now, to give the reporter a
chance to complain about a "wontfix" resolution.

Thanks all the same for reporting it; your input is valued.
msg3141 (view) Author: asterio (asterio) Date: 2008-04-08.22:01:41
Thank you Alan, the explanation was very clear and useful.
I agree with you. This is not a bug and can be easily treated.

As far I can I will try to provide valuable feedback for jython.

msg3142 (view) Author: Alan Kennedy (amak) Date: 2008-04-08.23:07:24
Closed by consensus.
Date User Action Args
2008-04-08 23:07:24amaksetstatus: open -> closed
resolution: wont fix
messages: + msg3142
2008-04-08 22:01:41asteriosetmessages: + msg3141
2008-04-08 19:38:37amaksetmessages: + msg3140
2008-04-08 16:49:59asteriosetmessages: + msg3139
2008-04-08 16:16:41amaksetassignee: amak
messages: + msg3138
nosy: + amak
2008-04-08 15:43:31asteriocreate