"J.P." <jp@neverwas.me> writes:

> I've added these changes as
>
>   https://git.savannah.gnu.org/cgit/emacs.git/commit/?id=328a9856
>
> This bug has already been closed. Thanks.

But wait, there's more! (There's always more.)

It seems I went a little overboard in reigning in the frequency of
scroll attempts. This has been walked back slightly in the attached
changes.

There's also a subtle quirk that's come to light involving `point-max'
and (recenter -1) where Emacs calculates point as having drifted off
screen. We could try tackling this with something kludgey, like
decrementing point temporarily during `recenter' attempts. But that
fails if the current input ends in a newline. I've instead decided to
address this by taking a page from other Emacs libraries and setting
`scroll-step' locally to 1 if a user hasn't customized
`scroll-conservatively'. See attached.