Lars Ingebrigtsen <larsi@gnus.org> writes:

> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>
>>> This follows up on changes proposed in bug#45474.
>>
>> Thanks, the first patch looks good to me (assuming it works ;-).
>
> So I've now applied it to Emacs 29.  It didn't lead to any obvious
> regressions (or test suite failures) that I can see, which is a good
> sign.
>
>>> The second patch is a bit more controversial, but is probably required
>>> if we want more reliable usage of completion commands in non-innermost
>>> minibuffers (that is, with minibuffer-follows-selected-frame set
>>> to nil.)
>>
>> The patch is fundamentally right, but as you say it's a bit more
>> controversial because it risks exposing bugs.  Hmm...
>>
>> To be on the safer side, I guess we could replace the
>>
>>     specbind (Qminibuffer_completion_table, Qnil);
>>
>> with a use of `minibuffer-with-setup-hook` that sets the var to nil in
>> the new minibuffer.  But doing it in C is awkward so it would best be
>> done by moving the function to subr.el.
>
> Sounds like a good idea to me.  Miha, could you do that?

Okay, patch attached.