bug#64141: 27.1 newly has clunky type-break y-or-n-p

[Juri Linkov <juri@linkov.net>]

>> >> the clunkiness that i think of that has always been there is this,
>> >> which might or might not be related:
>> >>
>> >>   - if you are using the mouse when the query occurs, the query might
>> >> be in the minibuffer.  it should be in a dialog box.
>> >>   - if you are using the kb when the query occurs, the query might be
>> >> a dialog box.  it should be in the minibuffer.
>> >>
>> >> the reason is that i am almost always either in mouse mode or kb mode.
>> >> switching causes impact on my system as i have to lift my arms,  or
>> >> drop one and possibly lift it again, which can be significantly
>> >> problematic in my case for unrelated reasons.
>> >>
>> >> for isearch, i do not recall previous instances of it in v<=26.1.  i
>> >> was only on 26.1 briefly.  27.1 was a first that i recall.
>> >
>> > Indeed, this is how it was supposed to work.
>> > But there is a regression in 29.0.
>> >
>> > When tried
>> >
>> >   (run-with-timer 5 5 #'yes-or-no-p "OK? ")
>> >
>> > it pops up a dialog box in 26.3, 27.2, 28.2,
>> > but uses the minibuffer in 29.0.
>>
>> Sorry, this is not a regression.  It still pops up the dialog box
>> when you are using a mouse.  So I see no more problems.
>
> Sorry, I'm confused: could you summarize which problems are there, and
> whether any of them are regressions?

There are no problems to summarize.

The "modal" minibuffer is achieved by y-or-n-p-use-read-key in 28.1.