Since we now use utf-8 internally in the database,
we must convert to and fro from user and filesystem
input and our database. Of course to make things worse,
python 3 completely changed the way python handles unicode.

This is an attempt to do things correctly

Conflicts:

	autojump

example:

if you have /tmp/vv/vv and want to specifically target it, you can now type
j vv vv
and it won't match /tmp/vv.

Still needs a migration script

Conflicts:

	autojump.bash
	jumpapplet

This should correctly merge Tanguy's XDG changes (thanks!) into the new
codebase.

Update to Readme

Fix PROMPT_COMMAND for compatibility with OS X Lion

Fix install.sh to check for sourcing of .bashrc

It now uses $AUTOJUMP_HOME if set, else $XDG_DATA_HOME/autojump if set,
else ~/.local/share/autojump.

Instead of updating $PROMPT_COMMAND by appending $AUTOJUMP with a leading
semicolon, prepend $AUTOJUMP followed by a semicolon.

It is unlikely (and frankly wrong) for anything to set $PROMPT_COMMAND to
begin with a semicolon, so this modification should not cause conflicts
where the previous version does.

Use pickle protocol 2.

This ensures that autojump works when the default python interpreter is
changed, e.g. by virtualenv.

Should fix issue #39

autojump is now packaged for Debian too, and will reach Ubuntu.

This should solve
http://github.com/joelthelion/autojump/issues/#issue/34

We are good now except for the shell_utility function which needs to be
refactored.

now that we handle files atomically

(pour les grincheux)

when you "su", the prompt_command can stay set, even without
autojump getting sourced. Then the user's autojump file changes
owner, and a disaster ensues. This is an attempt to fix this
problem.