Re: [gnugo-devel] Undo vs. Magic Undo

[bump]

I wrote:

> The problem with magic-undo (maybe just a bug) is that the
> way you'd really like to use undo is to type U repeatedly
> until you get to the board position before you messed up,
> then start playing again. If you use magic undo, after the
> first undo you cannot undo again. (Try it.) To take back
> multiple moves, you have to calculate which move you want
> to go back to and type in the number of moves you want to
> retract or a board location.

I now understand why this happens. After you play a move and
the opponent answers, the mouse location is still where you
played your last move. The way gnugo-magic-undo works is to
undo until the stone at the mouse location is removed from
the board. After it is gone, if you want to undo again with
magic undo you have to first move the mouse and click on the
stone that was to be removed.

This seems obviously nonoptimal to me and I can
think of two solutions.

(1) At the end of gnugo-magic-undo, warp the mouse
pointer to the second stone from the top of the
move history, which will be the next stone of our
color to have been played.

or 

(2) If the mouse pointer is at an empty vertex,
instead of issuing the unhelpful message
(error "%s already clear" pos) instead take the
stone to be undone from the move history.

Either solution would have the effect that if you
type U twice you take back two moves, rather than
taking back one move, then getting an error message.

Dan