bug#45474: Icomplete exhibiting in recursive minibuffer when it shouldn’t

[Stefan Monnier]

>> >> > Indeed, but you said that 'minibuffer-with-setup-hook' is "fundamentally
>> >> > broken and messy"...
>> >> All schemes using dynamic scoping to pass the information are similarly
>> >> broken&messy
>> > I don't agree.
>> I don't know what that means, here.
> It means I disagree that "all schemes using dynamic scoping to pass
> the information are broken and messy".

;-)

That's the part I understood, indeed.  There are two aspects which make
it rather unclear to me:
- First, from where I stand, what I stated is not really a matter of
  opinion but a mere result of the underlying way the problem works.
  There's a admittedly some amount of "degree" that can depend on some
  of the details, which is why I said "similarly".
- Second I don't know what it implies in terms of your opinion w.r.t the
  various potential problems with:
  - the current way of passing the information (just plain let-binding).
  - the way proposed by Gregory (let-binding of minibuffer-complete-*
    followed by let-binding to other-named vars followed by setq-local
    of minibuffer-complete-*, where the dance is performed in
    `read-from-minibuffer`).
  - the way I suggested (minibuffer-with-setup-hook + setq-local, done
    in `completing-read-default`).
  It seems to suggest that you may not find them all "similarly
  broken&messy", but even that is not sure.


        Stefan