url mangling interacting badly with http sites with trailing /

[Robert Collins]

urlutils.split("http://host/foo/") -> ("http://host/", "foo") rather
than ("http://host/", "foo/")

Also Transport._combine_paths strips trailing '/' in relpath. While
opening a 'foo/' it turns it into 'foo'.

This breaks e.g. 'bzr patch http://pastebin.ubuntu.com/129561/plain/'

The following (wrong because we can just remove the != '' check higher,
but easy for testing) - helps, it at least allows
get_transport(httpurlbase).get('something/') to work.
=== modified file 'bzrlib/transport/__init__.py'
--- bzrlib/transport/__init__.py 2009-02-23 15:42:47 +0000
+++ bzrlib/transport/__init__.py 2009-03-10 23:46:00 +0000
@@ -479,6 +479,8 @@
         path = '/'.join(base_parts)
         if not path.startswith('/'):
             path = '/' + path
+ if relpath.endswith('/'):
+ path += '/'
         return path

     def recommended_page_size(self):

This regularly bites users, so -

 affects bzr
 importance high

[Mary Gardiner (puzzlement)]

I'm assuming this is another example of the same problem:

$ bzr pull http://intertwingly.net/code/venus/
http://intertwingly.net/code/venus is permanently redirected to http://intertwingly.net/code/venus/
No revisions to pull.

[John A Meinel (jameinel)]

So the latter is because of the "get_mergeable_from_url" which is trying to read a Merge Directive from the url that you supply.

Arguably we could try to read it from "venus/" rather than "venus"...

[Robert Collins (lifeless)]

While this does still need to be done, it really isn't as important as most of our highs, and I'm not actively working on it.