Re: [Emacs-diffs] master adfb6f1: Continue to fix bug#25607

[Stefan Monnier <monnier@IRO.UMontreal.CA>]

>>> * lisp/ido.el (ido-complete): Let-bind `non-essential' to nil.
>> This looks wrong.  This in an interactive command executing an explicit
>> request from the user, so it's definitely not non-essential.
> Hence "to nil"?

Oh, right, sorry, duh!  Indeed, it's perfectly safe then to bind it to
nil there.

Hmm... then I wonder why this binding would be needed at all: why wouldn't
the variable already be nil?

[...time passes...]

Oh, I see: there is an incorrect binding in ido-read-internal which
means that all commands within IDO are treated as non-essential.

The way I see it, The Right Way would be to do the binding in
ido-exhibit, since that's the code run from post-command-hook, i.e. the
code that's not explicitly requested by the user.


        Stefan


diff --git a/lisp/ido.el b/lisp/ido.el
index e18464d1d6b..bb4c67c7c01 100644
--- a/lisp/ido.el
+++ b/lisp/ido.el
@@ -1882,7 +1882,6 @@ ido-read-internal
        ido-selected
        ido-final-text
        (done nil)
-       (non-essential t) ;; prevent eager Tramp connection
        (icomplete-mode nil) ;; prevent icomplete starting up
        ;; Exported dynamic variables:
        ido-cur-list
@@ -3556,7 +3555,6 @@ ido-file-name-all-completions-1
     ;; Strip method:user@host: part of tramp completions.
     ;; Tramp completions do not include leading slash.
     (let* ((len (1- (length dir)))
-	   (non-essential t)
 	   (compl
 	    (or ;; We do not want to be disturbed by "File does not
                 ;; exist" errors.
@@ -4413,6 +4411,7 @@ ido-exhibit
 
   (when (ido-active)
     (let ((contents (buffer-substring-no-properties (minibuffer-prompt-end) (point-max)))
+          (non-essential t)
 	  (buffer-undo-list t)
 	  try-single-dir-match
 	  refresh)