Re: Error in C++ mode with Emacs 27.0.90

[Dmitry Gutov]

On 29.03.2020 14:50, Alan Mackenzie wrote:

> It seems unlikely that we can get a nested list (in imenu--index-alist)
> with only one element.  I have tried creating a C++ buffer with 25
> functions in it.  This gives a simple list of 25 elements.  With 26
> functions, I get a nested alist with two elements.

What's the worst that can happen in this case? We get a nested submenu 
with just one element?

> However, in an emacs-lisp-mode buffer, we get things like "Variables"
> and "*Rescan*", but I've not yet succeeded in causing a nested list with
> just one element.  I'm not convinced enough that it couldn't happen,
> though.
>
> The latter part of the function looks like this (before my patch):
>
>          (let* ((menu (imenu--split-menu index-alist
>                                          (buffer-name)))
>                 (menu1 (imenu--create-keymap (car menu)
> ========>                                   (cdr (if (< 1 (length (cdr menu)))
>                                                       menu
>                                                     (car (cdr menu))))
>                                              'imenu--menubar-select)))
>            (setcdr imenu--menubar-keymap (cdr menu1)))))))
>
> That `if' form has been there since imenu-update-menubar was first
> written by Karl Heuer in 1997 (commit 0a8e8bc63e3).  Presumably, it
> really was needed back then.

If you're implying we can get rid of the check and always return (cdr 
menu), I say let's try.

> > In any case, imenu--create-keymap is also called from imenu--mouse-menu,
> > which probably needs the same kind of fix.
>
> Possibly.  I've not managed to create the same error in
> imenu--mouse-menu,

  (setq imenu-use-popup-menu t)

followed by

  M-x imenu

did that for me in the example buffer.

> but perhaps it could do with the same correction,
> just in case.  What do you think?

Yup.

> Incidentally, that patch from last night was a bit untidy.  I think it
> needs cleaning up a bit.

So how about this one? In my very brief testing, it seems to work, both 
in the example buffer, and for imenu indices in several bigger files:

diff --git a/lisp/imenu.el b/lisp/imenu.el
index fb8b3de662..809ddd7ed1 100644
--- a/lisp/imenu.el
+++ b/lisp/imenu.el
@@ -814,9 +814,7 @@ imenu--mouse-menu
   (setq index-alist (imenu--split-submenus index-alist))
   (let* ((menu (imenu--split-menu index-alist (or title (buffer-name))))
         (map (imenu--create-keymap (car menu)
-                                   (cdr (if (< 1 (length (cdr menu)))
-                                            menu
-                                          (car (cdr menu)))))))
+                                   (cdr menu))))
     (popup-menu map event)))

 (defun imenu-choose-buffer-index (&optional prompt alist)
@@ -912,9 +910,7 @@ imenu-update-menubar
        (let* ((menu (imenu--split-menu index-alist
                                         (buffer-name)))
                (menu1 (imenu--create-keymap (car menu)
-                                           (cdr (if (< 1 (length (cdr menu)))
-                                                    menu
-                                                  (car (cdr menu))))
+                                           (cdr menu)
                                            'imenu--menubar-select)))
          (setcdr imenu--menubar-keymap (cdr menu1)))))))