Re: zle: vi mode: wrong undo handling on fresh lines

[Bart Schaefer <schaefer@brasslantern.com>]

On Jan 28,  2:58pm, Peter Stephenson wrote:
}
} > I think this is starting to get somewhere, but I suspect it needs
} > tweaking.  For example, should that synthesised 'i' really be an 'a'?
} 
} Well, I think I'm going to commit it with the "a" because (i) this
} definitely seems better than what was there before in any case and (ii)
} hitting "ESC" and "." after typing the start of the line, which is my
} naive way of getting it to repeat, does repeat what I just inserted
} because of vi's quirk of backing up a character when leaving insert
} mode.  Obviously I'm not claiming this is the end of the matter.

Sorry, I missed this when looking at your previous message.  I agree
that "a" is more correct than "i" here.  In fact "A" might even be a
reasonable choice.

Another thing to check is the interaction of this with push-line.  Is
the right thing done if the buffer isn't empty when entering ZLE?

} > Is it OK to assume we're not in insert mode when vi-repeat is executed?

Wow, good question.  vi-repeat is implemented by replaying (a subset
of) the input keystrokes, not the widget actions they are bound to ...

} > If it's not safe, I don't understand how it ever worked.

If someone were to bind vi-repeat to a keystroke in the viins keymap,
I imagine the wrong thing would happen and always has.  Yep, just now
confirmed this.  If that's a bug, it's not a directly related one.

} I'm not hearing much in the way of knowledgeable direction for the
} completion issue.  It probably needs some hook trying out to see
} if it works.

I'm not familiar enough with the undo mechanics to make and good stabs
at such a hook, I fear.