nntplib hardcodes AF_INET for the socket address family. This prevents
it from using IPv6. Attached is a patch that converts NNTP.__init__() to
use socket.create_connection(), which is IPv6-capable.

This patch doesn't appear to work for python2.5.1 ->

Python 2.5.1 (r251:54863, Jun 15 2008, 18:24:51) 
[GCC 4.3.0 20080428 (Red Hat 4.3.0-8)] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> from nntplib import NNTP
>>> conn = NNTP('newszilla6.xs4all.nl')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python2.5/nntplib.py", line 114, in __init__
    self.sock = socket.create_connection((host, port))
AttributeError: 'module' object has no attribute 'create_connection'

(at least for me it doesn't work...
Linux hostnamehere 2.6.26.6-79.fc9.i686 #1 SMP Fri Oct 17 14:52:14 EDT
2008 i686 i686 i386 GNU/Linux)

I'd be happy to try something else, or debug in other ways it that'd
help... This really ought to get fixed if possible.

Yes. The patch is against 2.6. It uses the socket.create_connection()
helper function, which was added in 2.6. See http://svn.python.org/view?
rev=54546&view=rev for the commit message.

If you really want to apply it to 2.5, it's trivial to adapt the patch.
Just replace the call to create_connection() with something like this:

    msg = "getaddrinfo returns an empty list"
    for res in getaddrinfo(host, port, 0, SOCK_STREAM):
        af, socktype, proto, canonname, sa = res
        sock = None
        try:
            sock = socket(af, socktype, proto)
            if timeout is not _GLOBAL_DEFAULT_TIMEOUT:
                sock.settimeout(timeout)
            sock.connect(sa)
            self.sock = sock

    except error, msg:
        if sock is not None:
            sock.close()

    raise error, msg

oh crap :( I saw the 2.6 AFTER I posted the message :( sorry. grr, have
to find a fix for 2.5 I suppose now.

Thanks.

oy, and I'm not reading emails properly. I'll try the fix you propose
for 2.5.

a possible fix for 2.5 is:

morrowc@tweezer:/tmp$ diff -U3 nntplib.py.orig nntplib.py
--- nntplib.py.orig     2008-12-30 01:06:14.000000000 -0500
+++ nntplib.py  2008-12-30 01:07:33.000000000 -0500
@@ -109,8 +109,19 @@
         """
         self.host = host
         self.port = port
-        self.sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
-        self.sock.connect((self.host, self.port))
+        msg = "getaddrinfo returns an empty list"
+        for res in socket.getaddrinfo(self.host, self.port, 0, 
socket.SOCK_STREAM):
+          af, socktype, proto, canonname, sa = res
+          sock = None
+          try:
+            self.sock = socket.socket(af, socktype, proto)
+            self.sock.connect(sa)
+
+          except error, msg:
+            if self.sock is not None:
+                self.sock.close()
+            raise NNTPError, msg
+
         self.file = self.sock.makefile('rb')
         self.debugging = 0
         self.welcome = self.getresp()

I'll open a bug against 2.5 now with this.

I like nntplib_ipv6.patch: it's a generic solution (may support
address families other than IPv6) and reuse code
(socket.create_connection()) is always a good thing :-)

About Python 2.5: this branch doesn't accept bugfix anymore, only
security issues. So only Python 2.6+ and 3.0+ can be patched.

Are we sure that the 2.6 fix (in the patch) will make it into 2.6? (and
the right upstream patching will happen to the 3.0 code as well?)

I was going to suggest writing a test but I see that nntplib hasn't got
a single unit test :-O

Any chance of this being applied soon? It's been sitting in the
bugtracker for over a year now.

@dmorr: It would be faster if nntplib has some tests :-/

I'm confused by that.

In order to apply a 3 line patch (which replaces one standard library
function with another), you want an entire test suite written for
nntplib?

If we're willing to accept that nttplib works reasonably well now, why
is the testsuite necessary? socket.create_connection() is already used
in several other modules.

This is a little silly and painful... it's utterly broken to hardcode
the AF type in this way, could we please apply a patch (something like
the proposed seems to work fine) and get this rolled into the next release?

It seems really lame to not be able to support normal internet protocols
in a tools language.

Assuming the patch does work, +1 for applying it.

Your patch is committed in r72640. Thanks!

Thanks.

Is there any chance of getting bug 1655 fixed as well? imaplib has the
same issue as nntplib, and the patch is identical.