This file contains more verbose descriptions of the changes between
versions than those in README.

New in 0.7.3:
 + ^T at the start of a line should not crash.

New in 0.7.2:
 + Fix a couple of embarrassing typos in unix_console.py and 
   python_reader.py
 + Ran pychecker over the rest.  Found a few problems, and that pychecker
   really doesn't like some of my idioms...
 + "import no_such_package.[TAB]" should no longer crash.

New in 0.7.1:
 + I learnt how to use distutils properly...
 + Another few days of pygame_console hacking; scrolling support is the 
   biggy.

New in 0.7.0:
 + Moved to a package architecture.
 + Wrote a (very simple!) distutils setup.py script.
 + Changed the keyspec format to be more sensible.  See the docstring in
   pyrepl/keymap.py for more information.
 + Portability fixes (coping without FIONREAD, select.poll, sundry 
   terminal capibilities, probably some other stuff).
 + Various tortuous changes to use 2.2 features where possible but 
   retaining 2.1 support (I hope; haven't got a 2.1 here to test with).
 + Jumping up and down on control-C now shouldn't dump you out of 
   pyrepl (via a large hammer kind of approach).
 + Cancelling a command with ^C left an empty string in the history.
   No longer.  Fixing this proved to require making the history code
   more sensible, which was a good thing in itself.
 + reader.Reader has a new method, bind(), intended to be used by the
   user.
 + The $PYTHONSTARTUP file is executed in a namespace containing Reader.
   This means you can use the presence of the variable 'Reader' to tell
   whether you're executing in a pyrepl shell.
 + The $PYREPLSTARTUP env. var now overrides $PYTHONSTARTUP for pyrepl
   shells if set.
 + A prototype pygame console.  To use this, call python_reader.main(1)
   instead of just main() or main(0), or run 'pythoni pg'
   - input is a bit messy right now
   - printing large gobs of text is really slow
   - you can't ^C running commands (you have to find the controlling 
     terminal and type ^C there).
   - during long running commands we don't pump pygame's event loop, which
     is naughty.  These last two are related and probably require threads 
     (and much thought) to fix.
   - probably lots of other warts!

New in 0.6.0:
 + Rewrote the low level code to be (hopefully) more portable and
   (certainly) clearer.  Most of the code from 0.5.1 is still present,
   but it's been moved around and refactored (and the names have
   gotten more sensible).
 + The above process fixed a fair few bugs.
 + Implemented a few more emacs-mode bindings.
 + Nailed another couple of differences between my top-level and the
   the default one, including playing nice with Tk.
 + Implemented a saner way of handling window resizes (a sane way of 
   handling signals!  Call the Unix police!)

New in 0.5.1:
 + Realizing that it wasn't going to run under 2.0 after all, I ripped
   out some of the 2.0 compatibility code.

Version 0.5.0 is sufficiently historic that I can't remember what it
was like :)