'bzr update' fails in an http Launchpad checkout

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin Pool wrote:
>> So ultimately... Do we just need a "urlutils.is_same_location()" which can
>> handle the escaping issues?
> 
> Yes, I think so.  The recent issues about default vs explicit ports
> indicates to me that turning urls to strings and then comparing them
> is not a good approach.
> 

It makes me wonder if:

def is_same_location(location1, location2):
  parts1 = split_url(location1)
  parts2 = split_url(location2)

  return parts1 == parts2

would be a reasonable way to do it.

Maybe ignoring a password, but paying attention to the username, and maybe
urlquoting the path.

Then again, we don't always know whether the path is already quoted... That was
what get_transport() was hiding (because the 'base' could be a local Unicode
path, or a fully escaped URL.)

Certainly the default implementation might still fail, because

urlquote('%7F') != urlquote('~')

So a bigger issue is knowing which urls have been normalized...

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHJks/JdeBCYSNAAMRAufdAJ9wFXJ6YMsg7th94ezDASVKfVzVrQCfQwue
vIJn9FwfWPfe5fIJ8GL1DNc=
=1WTv
-----END PGP SIGNATURE-----