Problem binding mouse-1 to Info-mouse-follow-nearest-node

Problem binding mouse-1 to Info-mouse-follow-nearest-node

[martin rudalics]

With yesterday's trunk and emacs -Q yanking the following forms into
*scratch*

(setq debug-on-error t)

(add-hook
  'Info-mode-hook
  '(lambda ()
     (define-key Info-mode-map [mouse-1]
       'Info-mouse-follow-nearest-node)))

and evaluating them, firing up Info via C-h i and doing mouse-1
somewhere in the *info* buffer gets me (with a short delay)

Debugger entered--Lisp error: (args-out-of-range 1 1083)
   buffer-substring-no-properties(1 1083)
   deactivate-mark()

This is part of (probably silly) code that used to work from my .emacs
for many years.  Could a knowing soul please tell me what might have
happened here lately?

Thanks in advance, martin


GNU Emacs 24.0.50.1 (i386-mingw-nt5.1.2600)
  of 2011-04-10 on NESTOR

Re: Problem binding mouse-1 to Info-mouse-follow-nearest-node

[Eli Zaretskii]

> Date: Mon, 11 Apr 2011 11:35:51 +0200
> From: martin rudalics <rudalics@gmx.at>
> 
> With yesterday's trunk and emacs -Q yanking the following forms into
> *scratch*
> 
> (setq debug-on-error t)
> 
> (add-hook
>   'Info-mode-hook
>   '(lambda ()
>      (define-key Info-mode-map [mouse-1]
>        'Info-mouse-follow-nearest-node)))
> 
> and evaluating them, firing up Info via C-h i and doing mouse-1
> somewhere in the *info* buffer gets me (with a short delay)
> 
> Debugger entered--Lisp error: (args-out-of-range 1 1083)
>    buffer-substring-no-properties(1 1083)
>    deactivate-mark()

Can you tell which was the last revision where it worked?

Re: Problem binding mouse-1 to Info-mouse-follow-nearest-node

[martin rudalics]

 > Can you tell which was the last revision where it worked?

I'm quite sure that it worked with my last build before the lexbind
merge, that must have been one or two days before that merge took place.

I'd like to avoid bisecting the revision history since this would
involve repeated bootstrappings which take around 30 minutes here.

Could you reproduce the behavior I described on your machine?

martin

Re: Problem binding mouse-1 to Info-mouse-follow-nearest-node

[Juanma Barranquero]

On Mon, Apr 11, 2011 at 14:36, martin rudalics <rudalics@gmx.at> wrote:

> Could you reproduce the behavior I described on your machine?

I see it too.

    Juanma

Re: Problem binding mouse-1 to Info-mouse-follow-nearest-node

[Eli Zaretskii]

> Date: Mon, 11 Apr 2011 14:36:12 +0200
> From: martin rudalics <rudalics@gmx.at>
> CC: emacs-devel@gnu.org
> 
> Could you reproduce the behavior I described on your machine?

I couldn't try that where I wrote the previous message, but I can
now.  Yes, I see it.  And no, it wasn't introduced after the lex-bind
merge, because I see it in revision 103755 (I happen to have a binary
for that revision in one of my branches).

The immediate problem which triggers the error is that deactivate-mark
is called after Info-mouse-follow-nearest-node already changed the
node displayed in the *info* buffer, which involves narrowing the
buffer to a different range of character positions, and that causes
the region to reference the portion of the buffer outside the current
restriction.

Re: Problem binding mouse-1 to Info-mouse-follow-nearest-node

[martin rudalics]

 >> Could you reproduce the behavior I described on your machine?
 >
 > I couldn't try that where I wrote the previous message, but I can
 > now.  Yes, I see it.  And no, it wasn't introduced after the lex-bind
 > merge, because I see it in revision 103755 (I happen to have a binary
 > for that revision in one of my branches).

For some reason I must have missed it all the time.

 > The immediate problem which triggers the error is that deactivate-mark
 > is called after Info-mouse-follow-nearest-node already changed the
 > node displayed in the *info* buffer, which involves narrowing the
 > buffer to a different range of character positions, and that causes
 > the region to reference the portion of the buffer outside the current
 > restriction.

We could simply try doing

(buffer-substring-no-properties
  (max (region-beginning) (point-min))
  (min (region-end) (point-max)))

in `deactivate-mark'.  But I have no idea how the region should be
defined when it's not entirely within the accessible portion of a
buffer.

martin

Re: Problem binding mouse-1 to Info-mouse-follow-nearest-node

[Eli Zaretskii]

> Date: Mon, 11 Apr 2011 17:53:31 +0200
> From: martin rudalics <rudalics@gmx.at>
> CC: emacs-devel@gnu.org
> 
>  > The immediate problem which triggers the error is that deactivate-mark
>  > is called after Info-mouse-follow-nearest-node already changed the
>  > node displayed in the *info* buffer, which involves narrowing the
>  > buffer to a different range of character positions, and that causes
>  > the region to reference the portion of the buffer outside the current
>  > restriction.
> 
> We could simply try doing
> 
> (buffer-substring-no-properties
>   (max (region-beginning) (point-min))
>   (min (region-end) (point-max)))
> 
> in `deactivate-mark'.  But I have no idea how the region should be
> defined when it's not entirely within the accessible portion of a
> buffer.

If it were up to me, I'd like first to understand (a) how come there's
an active region in this case, (b) why deactivate-mark doesn't try to
validate the region, and (c) why doesn't narrowing clip the region to
the accessible portion.