result = urllib.urlopen("http://docs.python.org/library/urllib.html#OK")
print result.geturl()

result = urllib2.urlopen("http://docs.python.org/library/urllib.html#OK")
print result.geturl()

Python 2.6 returns:
"http://docs.python.org/library/urllib.html#OK"
"http://docs.python.org/library/urllib.html#OK"

Python 2.7 returns:
"http://docs.python.org/library/urllib.html#OK"
"http://docs.python.org/library/urllib.html"

2to3 -w test.py
Python 3 returns:
"http://docs.python.org/library/urllib.html"
"http://docs.python.org/library/urllib.html"

I expect geturl() result with "#OK" in all cases

This is because the Request class' constructor splits the URL into __original and fragment:

    def __init__(self, url, data=None, headers={},
                 origin_req_host=None, unverifiable=False):
        # unwrap('<URL:type://host/path>') --> 'type://host/path'
        self.__original = unwrap(url)
        self.__original, fragment = splittag(self.__original)

And the construction of object that urlopen() returns has its geturl() returns the request object's __original field (by now, minus the fragment).

Attaching patches against 2.7 and 3.1 branches.

Santa4nt, I think you also need to check case with Redirect Response URL:

print urllib2.urlopen("http://16.foobnix-cms.appspot.com/test_base").geturl()

python 2.6 returns OK
http://16.foobnix-cms.appspot.com/test_redirect#json={value:'OK'}

python 2.7 returns KO
http://16.foobnix-cms.appspot.com/test_redirect

It already does. ;-)

Python 2.7.1+ (default, Apr  6 2011, 16:25:38) [MSC v.1500 32 bit (Intel)] on wi
n32
Type "help", "copyright", "credits" or "license" for more information.
>>> import urllib2
[74578 refs]
>>> fp = urllib2.urlopen('http://16.foobnix-cms.appspot.com/test_base')
[75643 refs]
>>> fp.geturl()
"http://16.foobnix-cms.appspot.com/test_redirect#json={value:'OK'}"
[75645 refs]

I'm attaching patches with the appropriate unittest for the redirected case, though.

New changeset 3f240a1cd245 by Senthil Kumaran in branch '3.1':
Fix bpo-11703 - urllib2.geturl() does not return correct url when the original url contains #fragment. Patch Contribution by Santoso Wijaya.
http://hg.python.org/cpython/rev/3f240a1cd245

It should be noted that the bug surfaced in 2.7 and above due to changes made as part of bpo-8280.

New changeset 6e73f75ee034 by Senthil Kumaran in branch '2.7':
Fix bpo-11703 - urllib2.get_url does not handle fragment in url properly.
http://hg.python.org/cpython/rev/6e73f75ee034

This is fixed in all the codelines. Thanks for the patch, Santoso.

New changeset 8ee48ec69844 by Senthil Kumaran in branch '3.1':
Update the News for the fix to bpo-11703.
http://hg.python.org/cpython/rev/8ee48ec69844

New changeset 502bb809b03b by Senthil Kumaran in branch '2.7':
update news in 2.7 for Issue bpo-11703
http://hg.python.org/cpython/rev/502bb809b03b