This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

classification
Title: urlparse.urlunparse forms file urls incorrectly
Type: Stage:
Components: Library (Lib) Versions:
process
Status: closed Resolution: duplicate
Dependencies: Superseder: [CVE-2015-2104] Urlparse insufficient validation leads to open redirect
View: 23505
Assigned To: Nosy List: eigenlambda, epicfaace, martin.panter, orsenthil
Priority: normal Keywords:

Created on 2007-05-20 22:35 by eigenlambda, last changed 2022-04-11 14:56 by admin. This issue is now closed.

Files
File name Uploaded Description Edit
urlparse.diff eigenlambda, 2007-05-20 23:12 diff from urlparse.py shipping with python 2.5 and the version I modified for correctness and efficiency
Messages (5)
msg32097 - (view) Author: Thomas Folz-Donahue (eigenlambda) Date: 2007-05-20 22:35
This is a conversation with the current Python interpreter.

>>> import urlparse
>>> urlparse.urlparse(urlparse.urlunparse(urlparse.urlparse("file:////usr/bin/python")))
('file', 'usr', '/bin/python', '', '', '')

As you can see, the results are incorrect.  The problem is in the urlunsplit function:

def urlunsplit((scheme, netloc, url, query, fragment)):
    if netloc or (scheme and scheme in uses_netloc and url[:2] != '//'):
        if url and url[:1] != '/': url = '/' + url
        url = '//' + (netloc or '') + url
    if scheme:
        url = scheme + ':' + url
    if query:
        url = url + '?' + query
    if fragment:
        url = url + '#' + fragment
    return url

RFC 1808 (see http://www.ietf.org/rfc/rfc1808.txt ) specifies that a URL shall have the following syntax:
<scheme>://<net_loc>/<path>;<params>?<query>#<fragment>

The problem with the current version of urlunsplit is that it tests if there are already two slashes before the 'url' section before outputting a URL.  This is incorrect because (1) RFC 1808 clearly specifies at least three slashes between the end of the scheme portion and the beginning of the path portion and (2) this method will strip the first few slashes from an arbitrary path portion, which may require those slashes.  Removing that url[:2] != '//' causes urlunsplit to behave correctly when dealing with urls like file:////usr/bin/python .
msg32098 - (view) Author: Thomas Folz-Donahue (eigenlambda) Date: 2007-05-20 23:12
Some other issues with the urlparse module.  Several constant lists defined at the beginning of the module should be sets because they are only used for testing if certain strings are in them.  Also, urlunsplit() uses the + operator way too much, creating strings that are immediately thrown away.  IMO, the alternative is actually more readable.  Attaching a patch (diff -u urlparse.py urlparse.py.new > urlparse.diff).
File Added: urlparse.diff
msg32099 - (view) Author: Senthil Kumaran (orsenthil) * (Python committer) Date: 2007-05-23 21:27
Hi Thomas,
Verified the Bug with Python 2.5 and verified the fix as well. Works fine.

>>> urlparse(urlunparse(urlparse('file:////home/ors')))
('file', '', '//home/ors', '', '', '')
>>> urlparse(urlunparse(urlparse('file://///home/ors')))
('file', '', '///home/ors', '', '', '')
>>> urlparse(urlunparse(urlparse('file://////home/ors')))
('file', '', '////home/ors', '', '', '')
>>> urlparse(urlunparse(urlparse(urlunparse(urlparse('file://////home/ors')))))
('file', '', '////home/ors', '', '', '')
>>>

msg67937 - (view) Author: Senthil Kumaran (orsenthil) * (Python committer) Date: 2008-06-11 02:55
This issue no longer exists. I verified the bug report on the trunk and urlparse() and urlunparse methods behave properly on subsquent usages on the same url.
>>> urlparse.urlparse(urlparse.urlunparse(urlparse.urlparse('file:///home/ors/Letter.txt')))
ParseResult(scheme='file', netloc='', path='/home/ors/Letter.txt', params='', query='', fragment='')

Can be closed as Invalid.
msg237833 - (view) Author: Martin Panter (martin.panter) * (Python committer) Date: 2015-03-11 01:46
I believe this was closed incorrectly. The original test case includes four slashes, not three, and still fails in 3.5. However Issue 23505 has been opened in the mean time which should cover this.
History
Date User Action Args
2022-04-11 14:56:24adminsetgithub: 44982
2019-08-15 03:25:46epicfaacesetnosy: + epicfaace
2015-03-11 01:46:49martin.pantersetnosy: + martin.panter
messages: + msg237833
resolution: works for me -> duplicate

superseder: [CVE-2015-2104] Urlparse insufficient validation leads to open redirect
2008-06-11 06:02:39georg.brandlsetstatus: open -> closed
resolution: works for me
2008-06-11 02:55:50orsenthilsetmessages: + msg67937
2007-05-20 22:35:20eigenlambdacreate