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

[Gregory Heytings]

> I have the impression that I suffer from NIH syndrome with respect to 
> your patch, so I'll refrain from commenting on it by and large

I'm not sure I understand what you mean here.

It uses a nice trick (intermediate variables) to provide a 
backward-compatible behavior for read-from-minibuffer that would last for 
one major Emacs release (to avoid breaking external packages), while 
providing the needed feature (buffer-local completion tables) for code 
those that want to use it.  Doesn't that fit the bill?

> > diff --git a/src/minibuf.c b/src/minibuf.c
> > index 1a637c86ad..fd780bd5c1 100644
> > --- a/src/minibuf.c
> > +++ b/src/minibuf.c
> > @@ -865,6 +865,16 @@ read_minibuf (Lisp_Object map, Lisp_Object initial, 
> > Lisp_Object prompt,
> >    if (STRINGP (input_method) && !NILP (Ffboundp (Qactivate_input_method)))
> >      call1 (Qactivate_input_method, input_method);
> >
> > +  if (! EQ (Vminibuffer_local_completion, Qnil)) {
> > +    Fmake_local_variable (Qminibuffer_completion_table);
> > +    Fset (Qminibuffer_completion_table, Vminibuffer_local_completion_table);
> > +    Fmake_local_variable (Qminibuffer_completion_predicate);
> > +    Fset (Qminibuffer_completion_predicate, 
> > Vminibuffer_local_completion_predicate);
> > +    Fmake_local_variable (Qminibuffer_completion_confirm);
> > +    Fset (Qminibuffer_completion_confirm, 
> > Vminibuffer_local_completion_confirm);
> > +    specbind (Qminibuffer_local_completion, Qnil);
> > +  }
>
> I really don't like adding knowledge of those variables to this 
> function, which so far is completely oblivious to whether the minibuffer 
> is used with completion or not.

Hmmm...  That's not true, see the occurrences of 
minibuffer_completing_file_name.  And that would be temporary anyway, in 
Emacs 29 this code could be removed from read_minibuf() I think.

> >    run_hook (Qminibuffer_setup_hook);
>
> As in my patch, you could use this hook to do the completion-specific 
> part of the setup.

Indeed, but you said that 'minibuffer-with-setup-hook' is "fundamentally 
broken and messy"...