bug#69056: 30.0.50; history-add-new-input and recursive minibuffers

[Stefan Monnier]

> Thanks, that's what I thought too.  Here's an attempt do just that:

Looks pretty good.  I do have some comments/questions:

> @@ -902,6 +903,9 @@ read_minibuf (Lisp_Object map, Lisp_Object initial, 
> Lisp_Object prompt,
>    /* Don't allow the user to undo past this point.  */
>    bset_undo_list (current_buffer, Qnil);
>
> +  /* Cache the buffer-local value. */
> +  nohist = NILP (find_symbol_value (Qhistory_add_new_input));

Why not use `Vhistory_add_new_input`?
[ Also, it's not really "cache" (which implies it impacts only
  performance).  More like "remember".  ]

> @@ -965,7 +969,7 @@ read_minibuf (Lisp_Object map, Lisp_Object initial, 
> Lisp_Object prompt,
>    /* Add the value to the appropriate history list, if any.  This is
>       done after the previous buffer has been made current again, in
>       case the history variable is buffer-local.  */
> -  if (! (NILP (Vhistory_add_new_input) || NILP (histstring)))
> +  if (! (nohist || NILP (histstring)))
>      call2 (Qadd_to_history, histvar, histstring);
>  
>    /* If Lisp form desired instead of string, parse it.  */

IIUC this change is needed because by the time we get here the
buffer-local value of `history-add-new-input` has been flushed by
`minibuffer-inactive-mode` called by `read_minibuf_unwind`,
itself run by the `unbind_to` a few lines above.
So maybe we can simplify this by just moving the above 2 lines before
the `unbind_to`, WDYT?


        Stefan