When is hourglass shown?

When is hourglass shown?

[Lennart Borgman]

Can someone please tell me when an hourglass is shown on X Emacs? Is it
shown for an operation like ediff-buffers (that takes a long time on my pc)?
Is it shown for indent-region?

Re: When is hourglass shown?

[Jan D.]

>
>
>Can someone please tell me when an hourglass is shown on X Emacs? Is it
>shown for an operation like ediff-buffers (that takes a long time on my pc)?
>Is it shown for indent-region?
>

Not for indent-region (it should be IMHO).  Ediff-buffer is quite fast 
on my machines, so I can't tell.  Constructing the info directory on 
first C-h i is an example where the hourglass is shown.

    Jan D.

Re: When is hourglass shown?

[Stefan Monnier]

>> Can someone please tell me when an hourglass is shown on X Emacs? Is it
>> shown for an operation like ediff-buffers (that takes a long time on my pc)?
>> Is it shown for indent-region?

> Not for indent-region (it should be IMHO).

Are you sure?  I thought it was command-agnostic (i.e. it simply displays
an hourglass if more than X seconds have passed since Emacs was last ready
to process keyboard input).

[ ... trying out (setq indent-line-function (lambda () (while t))) ... ]

Duh! You're right.  Hmm... how does this thing work?


        Stefan

Re: When is hourglass shown?

[Jan D.]

Stefan Monnier wrote:

>>>Can someone please tell me when an hourglass is shown on X Emacs? Is it
>>>shown for an operation like ediff-buffers (that takes a long time on my pc)?
>>>Is it shown for indent-region?
>>>      
>>>
>
>  
>
>>Not for indent-region (it should be IMHO).
>>    
>>
>
>Are you sure?  I thought it was command-agnostic (i.e. it simply displays
>an hourglass if more than X seconds have passed since Emacs was last ready
>to process keyboard input).
>
>[ ... trying out (setq indent-line-function (lambda () (while t))) ... ]
>
>Duh! You're right.  Hmm... how does this thing work?
>

I only tried M-x indent-region.  For that no hourglass is shown.  But if 
invoked with M-C-\ an hourglass is shown.  Turns out that M-x ... calls 
minibuffer-complete-and-exit, and that function throws so the start 
hourglass action is unwind and canceled.  A bit inconsistent here.

    Jan D.

Re: When is hourglass shown?

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Date: Fri, 04 Mar 2005 14:33:19 -0500
> Cc: Emacs-Devel <emacs-devel@gnu.org>
> 
> I thought it was command-agnostic

AFAIK, it is, at least in principle (i.e. barring any bugs).

I see in keyboard.c that one instance of the call to start_hourglass
is ifdef'ed away; perhaps re-enabling that call (and taking care of
the problem described in a comment there, that caused the call to be
disabled) would fix the problem.

Re: When is hourglass shown?

[Eli Zaretskii]

> Date: Fri, 04 Mar 2005 18:34:23 +0100
> From: "Jan D." <jan.h.d@swipnet.se>
> 
> Not for indent-region (it should be IMHO).  Ediff-buffer is quite fast 
> on my machines, so I can't tell.

I think the easiest way to find out whether the hourglass could be
shown in the context of some function is to put a breakpoint on the
place where the hourglass timer is set up.  Then you will know if
Emacs arranges for the hourglass even if the operation itself is fast
enough so that no hourglass is actually shown before it finishes.

(It might be necessary to change the default handling of SIGALRM by
GDB, in order to run this under GDB.)

Re: When is hourglass shown?

[Jan D.]

2005-03-05 kl. 11.45 skrev Eli Zaretskii:

>> Date: Fri, 04 Mar 2005 18:34:23 +0100
>> From: "Jan D." <jan.h.d@swipnet.se>
>>
>> Not for indent-region (it should be IMHO).  Ediff-buffer is quite fast
>> on my machines, so I can't tell.
>
> I think the easiest way to find out whether the hourglass could be
> shown in the context of some function is to put a breakpoint on the
> place where the hourglass timer is set up.  Then you will know if
> Emacs arranges for the hourglass even if the operation itself is fast
> enough so that no hourglass is actually shown before it finishes.
>
> (It might be necessary to change the default handling of SIGALRM by
> GDB, in order to run this under GDB.)

This does not always work, as in the case of M-x indent-region versus 
C-M-\.  Both starts the hourglass timer, but the first stops it 
directly afterwards.

	Jan D.

Re: When is hourglass shown?

[Lennart Borgman]

----- Original Message ----- 
From: "Jan D." <jan.h.d@swipnet.se>

> This does not always work, as in the case of M-x indent-region versus
> C-M-\.  Both starts the hourglass timer, but the first stops it
> directly afterwards.

I am glad you took this up since it is rather confusing when trying to get
things working on w32 (where it currently does not work at all as far as I
can see). What is the reason it is stopped when using M-x? Is this a general
behaviour when using M-x or is it just indent-region?

On w32 I notice that the cursor is changed from an arrow (if outside the
text) to the same cursor that is used when it is over text. I have not been
able to find where this change is done yet. Is this the behaviour under X
Windows too?

Re: When is hourglass shown?

[Jan D.]

> ----- Original Message -----
> From: "Jan D." <jan.h.d@swipnet.se>
>
>> This does not always work, as in the case of M-x indent-region versus
>> C-M-\.  Both starts the hourglass timer, but the first stops it
>> directly afterwards.
>
> I am glad you took this up since it is rather confusing when trying to 
> get
> things working on w32 (where it currently does not work at all as far 
> as I
> can see). What is the reason it is stopped when using M-x? Is this a 
> general
> behaviour when using M-x or is it just indent-region?

AFAIK it is a general M-x behaviour.  Pressing return in the minibuffer 
starts the hourglass, and then invokes minibuffer-complete-and-exit.  
This function throws so the start_hourglass done after return is 
canceled by an unwind-protect (in keyboard.c, command_loop_1(), around 
line 1784).  This may be a bug.  I've checked 21.3 and it is the same 
there so it is not a regression.

	Jan D.

Re: When is hourglass shown?

[Richard Stallman]

    AFAIK it is a general M-x behaviour.  Pressing return in the minibuffer 
    starts the hourglass, and then invokes minibuffer-complete-and-exit.  
    This function throws so the start_hourglass done after return is 
    canceled by an unwind-protect (in keyboard.c, command_loop_1(), around 
    line 1784).  This may be a bug.

I am sure it is a bug.
If you can fix it, please do.

Re: When is hourglass shown?

[Jan D.]

Richard Stallman wrote:

>    AFAIK it is a general M-x behaviour.  Pressing return in the minibuffer 
>    starts the hourglass, and then invokes minibuffer-complete-and-exit.  
>    This function throws so the start_hourglass done after return is 
>    canceled by an unwind-protect (in keyboard.c, command_loop_1(), around 
>    line 1784).  This may be a bug.
>
>I am sure it is a bug.
>If you can fix it, please do.
>

I've checked in a fix.  Note that this fix is only for X, I don't know 
if the other ports may need a similar fix.

     Jan D.

Re: When is hourglass shown?

[Jason Rumney]

"Lennart Borgman" <lennart.borgman.073@student.lu.se> writes:

> On w32 I notice that the cursor is changed from an arrow (if outside the
> text) to the same cursor that is used when it is over text. I have not been
> able to find where this change is done yet. Is this the behaviour under X
> Windows too?

The change to enable mouse cursors on w32 was this one. You should be
able to get the diffs from CVS, or through the HTTP interface on
savannah.gnu.org.


2003-01-21  KOBAYASHI Yasuhiro  <kobayays@otsukakj.co.jp>

	* w32term.h (struct w32_output): New member hand_cursor.
	(WM_EMACS_SETCURSOR): New message definition.

	* w32term.c (note_mode_line_highlight): Delete #if 0 to enable
	function w32_define_cursor.
	(note_mouse_highlight): Initialize, setup cursor accoding to mouse
	position, change member name output_data.x to output_data.w32 and
	add function w32_define_cursor.
	(show_mouse_face): Delete #if 0 to enable function w32_define_cursor
	and change member name output_data.x to output_data.w32.
	(w32_initialize_display_info):
	Setup dpyinfo->vertical_scroll_bar_cursor.

	* w32fns.c (Vx_hand_shape): New variable.
	(w32_wnd_proc): Add message entries for WM_SETCURSOR and
	WM_EMACS_SETCURSOR.
	(x-create-frame): Setup Cursor types.

Re: When is hourglass shown?

[Eli Zaretskii]

> Cc: emacs-devel@gnu.org
> From: "Jan D." <jan.h.d@swipnet.se>
> Date: Sat, 5 Mar 2005 12:17:24 +0100
> 
> > I think the easiest way to find out whether the hourglass could be
> > shown in the context of some function is to put a breakpoint on the
> > place where the hourglass timer is set up.  Then you will know if
> > Emacs arranges for the hourglass even if the operation itself is fast
> > enough so that no hourglass is actually shown before it finishes.
> >
> > (It might be necessary to change the default handling of SIGALRM by
> > GDB, in order to run this under GDB.)
> 
> This does not always work, as in the case of M-x indent-region versus 
> C-M-\.  Both starts the hourglass timer, but the first stops it 
> directly afterwards.

Actually, I think it did work very well: it shows you that the
hourglass was armed, i.e. that Emacs _meant_ for it to appear if the
operation took enough time.  If you want to know why it didn't
actually happen, an additional breakpoint where the timer is disabled
would have shown you the code which did that.

Re: When is hourglass shown?

[Jan D.]

>>
>> This does not always work, as in the case of M-x indent-region versus
>> C-M-\.  Both starts the hourglass timer, but the first stops it
>> directly afterwards.
>
> Actually, I think it did work very well: it shows you that the
> hourglass was armed, i.e. that Emacs _meant_ for it to appear if the
> operation took enough time.  If you want to know why it didn't
> actually happen, an additional breakpoint where the timer is disabled
> would have shown you the code which did that.
>

I know exactly where now :-).  Not sure what to do about though.  Also, 
starting and cancelling hourglass is done frequently, so running with 
those breakpoints enabled makes for a lot of continue in gdb.  Not 
impossible though.

	Jan D.

Re: When is hourglass shown?

[Nick Roberts]

 > I know exactly where now :-).  Not sure what to do about though.  Also, 
 > starting and cancelling hourglass is done frequently, so running with 
 > those breakpoints enabled makes for a lot of continue in gdb.  Not 
 > impossible though.

I'm not sure if this helps, but gdb has 'ignore count', eg.

(gdb) b start_hourglass
Breakpoint 3 at 0x...
(gdb) ignore 3 9
(gdb) b cancel_hourglass
Breakpoint 4 at 0x...
(gdb) ignore 4 9

and gdb will only stop every tenth time.


Nick