Re: move-beginning-of-line

[Richard Stallman]

    In the M-: and M-x cases, doing the right thing is not nearly as obvious as
    you'd think because the "command" that does the work is actually the RET
    which exits from the minibuffer and then runs the actual code, so the
    "start point" and "end point" of the command are not even in the same
    buffer, so inferring the direction of the movement is not done correctly.

Maybe this can be improved if entering the minibuffer saves the "old
position" information and exiting it restores that information.  Want
to try that?