x_autoselect_window_p

x_autoselect_window_p

[Richard Stallman]

Gerd commented out the code for x_autoselect_window_p because he
saw it was running Lisp code from a signal handler.
ISTR that shortly after the code was installed I pointed
out that it had to be rewritten for this reason, but that
apparently was not done.  It is best for it to stay
commented out until someone rewrites it the right way.

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/emacs-devel

Re: x_autoselect_window_p

[Pavel Janík]

   From: Richard Stallman <rms@gnu.org>
   Date: Wed, 13 Mar 2002 03:58:49 -0700 (MST)

   > Gerd commented out the code for x_autoselect_window_p because he
   > saw it was running Lisp code from a signal handler.
   > ISTR that shortly after the code was installed I pointed
   > out that it had to be rewritten for this reason, but that
   > apparently was not done.  It is best for it to stay
   > commented out until someone rewrites it the right way.

I have it on my TODO list.
-- 
Pavel Janík

Make sure input cannot violate the limits of the program.
                  --  The Elements of Programming Style (Kernighan & Plaugher)

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/emacs-devel

Re: x_autoselect_window_p

[Pavel Janík]

   From: Richard Stallman <rms@gnu.org>
   Date: Wed, 13 Mar 2002 03:58:49 -0700 (MST)

   > Gerd commented out the code for x_autoselect_window_p because he
   > saw it was running Lisp code from a signal handler.
   > ISTR that shortly after the code was installed I pointed
   > out that it had to be rewritten for this reason, but that
   > apparently was not done.  It is best for it to stay
   > commented out until someone rewrites it the right way.

I found some time today and tried to rewrote it. But failed :-( I patched
four files:

  /home/pavel/.Emacs/Work/emacs/src:

  -rw-r--r-- (modified) bře 31 21:23 keyboard.c
  -rw-r--r-- (modified) bře 31 20:59 keyboard.h
  -rw-r--r-- (modified) bře 31 20:55 termhooks.h
  -rw-r--r-- (modified) bře 31 20:58 xterm.c


termhooks.h only adds new event type:

--- termhooks.h.~1.57.~	Sun Mar 10 19:22:28 2002
+++ termhooks.h	Sun Mar 31 20:55:30 2002
@@ -325,6 +325,9 @@
      `switch-frame' events in kbd_buffer_get_event, if necessary.  */
   FOCUS_IN_EVENT,
 
+  /* Generated when mouse moves over window not currently selected.  */
+  WINDOW_IN_EVENT,
+
   /* Queued from XTread_socket when session manager sends
      save yourself before shutdown. */
   save_session_event


xterm.c now generates this event when it should:

--- xterm.c.~1.716.~	Sun Mar 31 18:32:34 2002
+++ xterm.c	Sun Mar 31 20:58:44 2002
@@ -10894,8 +10894,39 @@
 		    clear_mouse_face (dpyinfo);
 		  }
 
-		if (f)
+		if (f) {
+
+		  /* Generate WINDOW_IN_EVENTs when needed.  */
+		  if (x_autoselect_window_p)
+		    {
+		      Lisp_Object window;
+		      int area;
+		      static Lisp_Object last_window;
+
+		      window = window_from_coordinates (f,
+							XINT (event.xmotion.x), XINT (event.xmotion.y),
+							&area, 0);
+
+		      /* Window will be selected only when it is not selected now and
+			 last mouse movement event was not in it.  Minibuffer window
+			 will be selected iff it is active.  */
+		      if (!EQ (window, last_window)
+			  && !EQ (window, selected_window)
+			  && (!MINI_WINDOW_P (XWINDOW (window))
+			      || (EQ (window, minibuf_window) && minibuf_level > 0)))
+			{
+			  fprintf(stderr, "WINDOW_IN_EVENT generated!\n");
+
+			  bufp->kind = WINDOW_IN_EVENT;
+			  XSETFRAME (bufp->frame_or_window, window);
+			  bufp->arg = Qnil;
+			  ++bufp, ++count, --numchars;
+			}
+
+		      last_window=window;
+		    }
 		  note_mouse_movement (f, &event.xmotion);
+		}
 		else
 		  {
 #ifndef USE_TOOLKIT_SCROLL_BARS


So far, it is without problems, WINDOW_IN_EVENT is generated and seen in
keyboard.c without problems. Now, we should process it:

--- keyboard.h.~1.57.~	Wed Mar  6 17:03:10 2002
+++ keyboard.h	Sun Mar 31 20:59:52 2002
@@ -181,6 +181,9 @@
 /* Total number of times read_char has returned.  */
 extern int num_input_events;
 
+/* Defined in xterm.c  */
+extern int x_autoselect_window_p;
+
 /* Total number of times read_char has returned, outside of macros.  */
 extern EMACS_INT num_nonmacro_input_events;
 


--- keyboard.c.~1.665.~	Sun Mar 24 21:28:14 2002
+++ keyboard.c	Sun Mar 31 21:23:08 2002
@@ -544,6 +544,7 @@
 Lisp_Object Qdelete_frame;
 Lisp_Object Qiconify_frame;
 Lisp_Object Qmake_frame_visible;
+Lisp_Object Qselect_window;
 Lisp_Object Qhelp_echo;
 
 /* Symbols to denote kinds of events.  */
@@ -3791,6 +3792,17 @@
 	  internal_last_event_frame = frame;
 	  kbd_fetch_ptr = event + 1;
 	}
+      else if (event->kind == WINDOW_IN_EVENT)
+	{
+
+	  fprintf(stderr, "WINDOW_IN_EVENT received!\n");
+
+	  /* Make an event (select-window (WINDOW)).  */
+	  obj = Fcons (event->frame_or_window, Qnil);
+	  obj = Fcons (Qselect_window, Fcons (obj, Qnil));
+
+	  kbd_fetch_ptr = event + 1;
+	}
       else
 	{
 	  /* If this event is on a different frame, return a switch-frame this
@@ -10302,7 +10314,8 @@
   {&Qswitch_frame,        "switch-frame",        &Qswitch_frame},
   {&Qdelete_frame,        "delete-frame",        &Qdelete_frame},
   {&Qiconify_frame,       "iconify-frame",       &Qiconify_frame},
-  {&Qmake_frame_visible,  "make-frame-visible",  &Qmake_frame_visible}
+  {&Qmake_frame_visible,  "make-frame-visible",  &Qmake_frame_visible},
+  {&Qselect_window,       "select-window",       &Qselect_window}
 };
 
 void
@@ -10968,6 +10981,8 @@
 			    "ignore-event");
   initial_define_lispy_key (Vspecial_event_map, "make-frame-visible",
 			    "ignore-event");
+  initial_define_lispy_key (Vspecial_event_map, "select-window",
+			    "handle-select-window");
   initial_define_lispy_key (Vspecial_event_map, "save-session",
 			    "handle-save-session");
 }



I use this fake function to do the real work:

(defun handle-select-window (event)
  "Handle delete-frame events from the X server."
  (interactive "e")
  (message "Switching to: %s" (car (event-start event)))
  (select-window (car (event-start event))))

To test this, I run emacs -q -eval '(setq x-autoselect-window t)', C-x
2 and move between those two windows. After about 20 calls to
select-window, Emacs will crash in mark_font_cache. There is a problem
somewhere, but I do not see it :-( Can you please help me so we can clear
this issue?
-- 
Pavel Janík

/* These are the most dangerous and useful defines. They do printk() during
 * the interrupt processing routine(s), so if you manage to get "flooded" by
 * irq's, start thinking about the "Power off/on" button... */
                  -- 2.2.16 drivers/sbus/char/aurora.h

Re: x_autoselect_window_p

[Gerd Moellmann]

Pavel@janik.cz (Pavel Janík) writes:

> --- xterm.c.~1.716.~	Sun Mar 31 18:32:34 2002
> +++ xterm.c	Sun Mar 31 20:58:44 2002
> +			  bufp->kind = WINDOW_IN_EVENT;
> +			  XSETFRAME (bufp->frame_or_window, window);
                          ^^^^^^^^^
                          XSETWINDOW

Re: x_autoselect_window_p

[Gerd Moellmann]

Gerd Moellmann <gerd@gnu.org> writes:

> Pavel@janik.cz (Pavel Janík) writes:
> 
> > --- xterm.c.~1.716.~	Sun Mar 31 18:32:34 2002
> > +++ xterm.c	Sun Mar 31 20:58:44 2002
> > +			  bufp->kind = WINDOW_IN_EVENT;
> > +			  XSETFRAME (bufp->frame_or_window, window);
>                           ^^^^^^^^^
>                           XSETWINDOW

Actually, since WINDOW is already a Lisp_Object, TRT would be to just
use `=' here.

Re: x_autoselect_window_p

[Pavel Janík]

   From: gerd.moellmann@t-online.de (Gerd Moellmann)
   Date: 31 Mar 2002 23:48:49 +0200

   > > Pavel@janik.cz (Pavel Janík) writes:
   > > 
   > > > --- xterm.c.~1.716.~	Sun Mar 31 18:32:34 2002
   > > > +++ xterm.c	Sun Mar 31 20:58:44 2002
   > > > +			  bufp->kind = WINDOW_IN_EVENT;
   > > > +			  XSETFRAME (bufp->frame_or_window, window);
   > >                           ^^^^^^^^^
   > >                           XSETWINDOW
   > 
   > Actually, since WINDOW is already a Lisp_Object, TRT would be to just
   > use `=' here.

Thank you, Gerd - it took me about two hours and I did not found
it :-(. BTW - do you think that this patch is correct? I do not have
overview as you have...
-- 
Pavel Janík

How can you say emacs and 'no flame war' in the same sentence :-)
                  -- Jens Axboe in private e-mail

Re: x_autoselect_window_p

[Gerd Moellmann]

Pavel@janik.cz (Pavel Janík) writes:

> BTW - do you think that this patch is correct? I do not have
> overview as you have...

I'll add some notes below.  The principle seems correct to me.


> --- termhooks.h.~1.57.~	Sun Mar 10 19:22:28 2002
> +++ termhooks.h	Sun Mar 31 20:55:30 2002
> @@ -325,6 +325,9 @@
>       `switch-frame' events in kbd_buffer_get_event, if necessary.  */
>    FOCUS_IN_EVENT,
>  
> +  /* Generated when mouse moves over window not currently selected.  */
> +  WINDOW_IN_EVENT,

People might be more happy with a more telling name
(SELECT_WINDOW_EVENT, SWITCH_WINDOW_EVENT, ...?)

> +
>    /* Queued from XTread_socket when session manager sends
>       save yourself before shutdown. */
>    save_session_event
> 
> 
> xterm.c now generates this event when it should:
> 
> --- xterm.c.~1.716.~	Sun Mar 31 18:32:34 2002
> +++ xterm.c	Sun Mar 31 20:58:44 2002
> @@ -10894,8 +10894,39 @@
>  		    clear_mouse_face (dpyinfo);
>  		  }
>  
> -		if (f)
> +		if (f) {

(signal 'coding-convention-error "{") :-)

> +
> +		  /* Generate WINDOW_IN_EVENTs when needed.  */
> +		  if (x_autoselect_window_p)
> +		    {
> +		      Lisp_Object window;
> +		      int area;
> +		      static Lisp_Object last_window;

On some systems, `static' gets defined away, so if you need such a
variable in Emacs, you must move it to global scope.

> +
> +		      window = window_from_coordinates (f,
> +							XINT (event.xmotion.x), XINT (event.xmotion.y),
> +							&area, 0);
> +
> +		      /* Window will be selected only when it is not selected now and
> +			 last mouse movement event was not in it.  Minibuffer window
> +			 will be selected iff it is active.  */
> +		      if (!EQ (window, last_window)
> +			  && !EQ (window, selected_window)
> +			  && (!MINI_WINDOW_P (XWINDOW (window))
> +			      || (EQ (window, minibuf_window) && minibuf_level > 0)))

Maybe it's better to leave this condition testing to the Lisp function
that is called at the end for handling the event?  It would have the
advantage of being more flexible.

> +			{
> +			  fprintf(stderr, "WINDOW_IN_EVENT generated!\n");
> +
> +			  bufp->kind = WINDOW_IN_EVENT;
> +			  XSETFRAME (bufp->frame_or_window, window);
> +			  bufp->arg = Qnil;

Strictly speaking, it might be a good idea to check for numchars > 0
before filling in the event.

> +			  ++bufp, ++count, --numchars;
> +			}
> +
> +		      last_window=window;
> +		    }
>  		  note_mouse_movement (f, &event.xmotion);
> +		}
>  		else
>  		  {
>  #ifndef USE_TOOLKIT_SCROLL_BARS
> 
> 
> So far, it is without problems, WINDOW_IN_EVENT is generated and seen in
> keyboard.c without problems. Now, we should process it:
> 
> --- keyboard.h.~1.57.~	Wed Mar  6 17:03:10 2002
> +++ keyboard.h	Sun Mar 31 20:59:52 2002
> @@ -181,6 +181,9 @@
>  /* Total number of times read_char has returned.  */
>  extern int num_input_events;
>  
> +/* Defined in xterm.c  */
> +extern int x_autoselect_window_p;
> +

If this isn't an X-only feature, the `x_' prefix might be a bit
confusing.

Re: x_autoselect_window_p

[Pavel Janík]

   From: gerd.moellmann@t-online.de (Gerd Moellmann)
   Date: 01 Apr 2002 11:25:16 +0200

   > I'll add some notes below.  The principle seems correct to me.

Thank you.

   > People might be more happy with a more telling name
   > (SELECT_WINDOW_EVENT, SWITCH_WINDOW_EVENT, ...?)

Yes, I have chosen SELECT_WINDOW_EVENT for now because it matches the real
functionality.

   > > -		if (f)
   > > +		if (f) {
   > 
   > (signal 'coding-convention-error "{") :-)

;-) Fixed.

   > On some systems, `static' gets defined away, so if you need such a
   > variable in Emacs, you must move it to global scope.

I moved it up.

   > > +		      /* Window will be selected only when it is not selected now and
   > > +			 last mouse movement event was not in it.  Minibuffer window
   > > +			 will be selected iff it is active.  */
   > > +		      if (!EQ (window, last_window)
   > > +			  && !EQ (window, selected_window)
   > > +			  && (!MINI_WINDOW_P (XWINDOW (window))
   > > +			      || (EQ (window, minibuf_window) && minibuf_level > 0)))
   > 
   > Maybe it's better to leave this condition testing to the Lisp function
   > that is called at the end for handling the event?  It would have the
   > advantage of being more flexible.

Well, we can split this. I do not want to generate an event for every
mouse move, so I think it is OK to have

   +		      if (!EQ (window, last_window)
   +			  && !EQ (window, selected_window)

here, but the rest can in fact be moved to Elisp.

   > 
   > > +			{
   > > +			  fprintf(stderr, "WINDOW_IN_EVENT generated!\n");
   > > +
   > > +			  bufp->kind = WINDOW_IN_EVENT;
   > > +			  XSETFRAME (bufp->frame_or_window, window);
   > > +			  bufp->arg = Qnil;
   > 
   > Strictly speaking, it might be a good idea to check for numchars > 0
   > before filling in the event.

Like this?

+			/* Window will be selected only when it is not selected now and
+			   last mouse movement event was not in it.  Minibuffer window
+			   will be selected iff it is active.  */
+			if (!EQ (window, last_window)
+			    && !EQ (window, selected_window)
+			    && numchars > 0)
+			  {
+			    bufp->kind = SELECT_WINDOW_EVENT;
+			    bufp->frame_or_window = window;
+			    bufp->arg = Qnil;
+			    ++bufp, ++count, --numchars;
+			  }

   > If this isn't an X-only feature, the `x_' prefix might be a bit
   > confusing.

Yes, this was already proposed by both Eli and Richard. I will change that
before really committing to autoselect-window (autoselect_window_p).
-- 
Pavel Janík

kwintv is always crashing. Please fix it fast.. I want to watch Star Trek.
                  -- Thomas Biege <thomas@suse.de>

Re: x_autoselect_window_p

[Gerd Moellmann]

Pavel@janik.cz (Pavel Janík) writes:

>    > > +		      /* Window will be selected only when it is not selected now and
>    > > +			 last mouse movement event was not in it.  Minibuffer window
>    > > +			 will be selected iff it is active.  */
>    > > +		      if (!EQ (window, last_window)
>    > > +			  && !EQ (window, selected_window)
>    > > +			  && (!MINI_WINDOW_P (XWINDOW (window))
>    > > +			      || (EQ (window, minibuf_window) && minibuf_level > 0)))
>    > 
>    > Maybe it's better to leave this condition testing to the Lisp function
>    > that is called at the end for handling the event?  It would have the
>    > advantage of being more flexible.
> 
> Well, we can split this. I do not want to generate an event for every
> mouse move, so I think it is OK to have
> 
>    +		      if (!EQ (window, last_window)
>    +			  && !EQ (window, selected_window)
> 
> here, but the rest can in fact be moved to Elisp.

Yes, probably.

>    > 
>    > > +			{
>    > > +			  fprintf(stderr, "WINDOW_IN_EVENT generated!\n");
>    > > +
>    > > +			  bufp->kind = WINDOW_IN_EVENT;
>    > > +			  XSETFRAME (bufp->frame_or_window, window);
>    > > +			  bufp->arg = Qnil;
>    > 
>    > Strictly speaking, it might be a good idea to check for numchars > 0
>    > before filling in the event.
> 
> Like this?

Yup.

I guess that's all, then.  Except maybe that I forgot to warn that I
didn't see the case handled where the mouse is moved to something
that's not selectable.

Re: x_autoselect_window_p

[Pavel Janík]

   From: gerd.moellmann@t-online.de (Gerd Moellmann)
   Date: 01 Apr 2002 16:55:54 +0200

   > I guess that's all, then.  Except maybe that I forgot to warn that I
   > didn't see the case handled where the mouse is moved to something
   > that's not selectable.

Yup, like tool-bar ;-) I already fixed that.
-- 
Pavel Janík

It is better to keep your mouth shut and be thought a fool, than to open it
and remove all doubt.
                  -- Ion Badulescu in linux-kernel

Re: x_autoselect_window_p

[Eli Zaretskii]

On Mon, 1 Apr 2002, Pavel =?iso-8859-2?q?Jan=EDk?= wrote:

>    > If this isn't an X-only feature, the `x_' prefix might be a bit
>    > confusing.
> 
> Yes, this was already proposed by both Eli and Richard. I will change that
> before really committing to autoselect-window (autoselect_window_p).

Wouldn't it be better to have something in the name of the feature that 
would hint it's connected with the mouse?  Something like 
mouse-autoselect-window-p?

Re: x_autoselect_window_p

[Richard Stallman]

    Wouldn't it be better to have something in the name of the feature that 
    would hint it's connected with the mouse?  Something like 
    mouse-autoselect-window-p?

I agree.  Also, the ending "-p" is not appropriate for a variable.

Re: x_autoselect_window_p

[Pavel Janík]

   From: Richard Stallman <rms@gnu.org>
   Date: Tue, 2 Apr 2002 21:55:40 -0700 (MST)

   >     Wouldn't it be better to have something in the name of the feature that 
   >     would hint it's connected with the mouse?  Something like 
   >     mouse-autoselect-window-p?
   > 
   > I agree.  Also, the ending "-p" is not appropriate for a variable.

So, I will rename it to mouse-autoselect-window. Thank you.
-- 
Pavel Janík

It's known to be very buggy (it comes directly from the hardware
manufacturer).
                  -- Hubert Mantel about some driver for eth. card