Reported by Stefan Behnel in issue22118:

I'm now getting duplicated slashes in URLs, e.g.:

https://new//foo.html
http://my.little.server/url//logo.gif

In both cases, the base URL that gets joined with the postfix had a trailing slash, e.g.

"http://my.little.server/url/" + "logo.gif" -> "http://my.little.server/url//logo.gif"

This patch seems to fix issue #22311 as well (*). However it would be good to add more tests for base URLs with trailing slashes, it seems.

(*) without patch:

>>> base = """https://pypi.python.org/simple/werkzeug/"""
>>> rel = """../../packages/2.3/W/Werkzeug/Werkzeug-0.3.1-py2.3.egg#md5=5f669acf04af135ad8577d99a4387504"""
>>> urllib.parse.urljoin(base, rel)
'https://pypi.python.org/simple/packages/2.3/W/Werkzeug/Werkzeug-0.3.1-py2.3.egg#md5=5f669acf04af135ad8577d99a4387504'

with patch:

>>> urllib.parse.urljoin(base, rel)
'https://pypi.python.org/packages/2.3/W/Werkzeug/Werkzeug-0.3.1-py2.3.egg#md5=5f669acf04af135ad8577d99a4387504'

Were the tests in

http://bugs.python.org/file32591/urischemes.py

merged yet, that Nick Coghlan mentioned in http://bugs.python.org/issue22118#msg225662 ?

Those tests don't seem to bring much. Part of them are straight from the RFC (and therefore already in the current test suite, I assume), part of them are for non-HTTP protocols such as "fred" (!). A couple of them seem to be genuine, although only one fails and it's a corner case.

I'll try to get some time this week to extend the various test cases, thanks for pointing that out Antoine.

I also found that, other than the few RFC-specific blocks in the link that Nick added in the other ticket, not only were they questionable (non-HTTP as Antoine pointed out), but they were also just plain wrong in some cases given the new semantics.

Antoine: On (finally) getting back to this and re-reading your test case, the current behaviour is incorrect and is corrected by the patch. I've added a few more test cases to ensure trailing slashes are handled correctly.

Except for the minor comments made by Antoine in the review, the patch looks good to go.

New changeset 901e4e52b20a by Senthil Kumaran in branch 'default':
Issue #22278: Fix urljoin problem with relative urls, a regression observed
https://hg.python.org/cpython/rev/901e4e52b20a

I addressed Antoine's comments with the patch and committed it. Thank you!

Heh, I'd finally gotten a few minutes to address the comments... And it's already taken care of ;) Thanks Senthil.

I opened Issue 23703 for a regression caused by this commit.