Emacs appears to hang

Emacs appears to hang

[Robert J. Chassell]

In today's CVS snapshot, Wed, 2002 Jul 17  15:13 UTC,
GNU Emacs 21.3.50.110 (i686-pc-linux-gnu, X toolkit)


In *scratch* buffer, insert these three lines:

line 1
line 2
line 3

Place cursor over middle line; open line using C-o (open-line).
Kill opened line using C-k (kill-line).

Nothing appears to happen: line appears unkilled.

Type a C-l (recenter) or C-g (keyboard-quit) command 
and then you can see that the line was actually killed.

Here is the backtrace:

(gdb) bt
#0  0x403c27ce in select () from /lib/libc.so.6
#1  0xbffff0cc in ?? ()
#2  0x080571e1 in sit_for (sec=30, usec=0, reading=1, display=1, 
    initial_display=0) at dispnew.c:6246
#3  0x080d65a3 in read_char (commandflag=1, nmaps=2, maps=0xbffff424, 
    prev_event=405373828, used_mouse_menu=0xbffff470) at keyboard.c:2553
#4  0x080dccb0 in read_key_sequence (keybuf=0xbffff574, bufsize=30, 
    prompt=405373828, dont_downcase_last=0, can_return_switch_frame=1, 
    fix_current_buffer=1) at keyboard.c:8368
#5  0x080d4be8 in command_loop_1 () at keyboard.c:1475
#6  0x08128179 in internal_condition_case (bfun=0x80d48e0 <command_loop_1>, 
    handlers=405470388, hfun=0x80d44a4 <cmd_error>) at eval.c:1349
#7  0x080d47b8 in command_loop_2 () at keyboard.c:1276
#8  0x08127d0d in internal_catch (tag=405431716, 
    func=0x80d4794 <command_loop_2>, arg=405373828) at eval.c:1109
#9  0x080d4762 in command_loop () at keyboard.c:1255
#10 0x080d4260 in recursive_edit_1 () at keyboard.c:966
#11 0x080d4390 in Frecursive_edit () at keyboard.c:1022
#12 0x080d3263 in main (argc=5, argv=0xbffffb34, envp=0xbffffb4c)
    at emacs.c:1625



In GNU Emacs 21.3.50.110 (i686-pc-linux-gnu, X toolkit)
 of 2002-07-17 on benthic
configured using `configure --with-type1 --prefix=/usr/local --with-sound=yes'

Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: C
  locale-coding-system: nil
  default-enable-multibyte-characters: t

Recent input:
<help-echo> l i n e SPC 1 C-n <return> l i n e SPC 
2 <return> l i n e SPC 3 C-p C-a C-o C-k C-l M-x r 
e p o r <tab> <return>

Recent messages:
(/usr/local/bin/emacs -q --no-site-file --eval (blink-cursor-mode 0))
Loading disp-table...done
Loading tool-bar...done
Loading image...done
Loading tooltip...done
For information about the GNU Project and its goals, type C-h C-p.
Loading emacsbug...done

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: Emacs appears to hang

[Juanma Barranquero]

On Wed, 17 Jul 2002 16:47:21 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:

> Type a C-l (recenter) or C-g (keyboard-quit) command 
> and then you can see that the line was actually killed.

I've seen the same in ielm. Another thing that's happened to me is
deleting with backspace and not having the display correctly redrawn. I
think both are manifestations of the severe redisplay bug we're
experiencing. 


                                                           /L/e/k/t/u

Re: Emacs appears to hang

[Kim F. Storm]

Juanma Barranquero <lektu@terra.es> writes:

> On Wed, 17 Jul 2002 16:47:21 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:
> 
> > Type a C-l (recenter) or C-g (keyboard-quit) command 
> > and then you can see that the line was actually killed.
> 
> I've seen the same in ielm. Another thing that's happened to me is
> deleting with backspace and not having the display correctly redrawn. I
> think both are manifestations of the severe redisplay bug we're
> experiencing. 

However, this specific bug was not present in the CVS emacs version I built
48 hours ago...  So it must be yet a new redisplay bug...

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

[PATCH] Re: Emacs appears to hang

[Kim F. Storm]

storm@cua.dk (Kim F. Storm) writes:

> Juanma Barranquero <lektu@terra.es> writes:
> 
> > On Wed, 17 Jul 2002 16:47:21 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:
> > 
> > > Type a C-l (recenter) or C-g (keyboard-quit) command 
> > > and then you can see that the line was actually killed.
> > 
> > I've seen the same in ielm. Another thing that's happened to me is
> > deleting with backspace and not having the display correctly redrawn. I
> > think both are manifestations of the severe redisplay bug we're
> > experiencing. 
> 
> However, this specific bug was not present in the CVS emacs version I built
> 48 hours ago...  So it must be yet a new redisplay bug...
> 

The new display bug can be fixed by this small patch:

Index: xdisp.c
===================================================================
RCS file: /cvs/emacs/src/xdisp.c,v
retrieving revision 1.767
diff -c -r1.767 xdisp.c
*** xdisp.c	16 Jul 2002 19:49:12 -0000	1.767
--- xdisp.c	17 Jul 2002 22:35:55 -0000
***************
*** 10170,10175 ****
--- 10170,10176 ----
      = (!NILP (w->window_end_valid)
         && !current_buffer->clip_changed
         && END_UNCHANGED + BEG_UNCHANGED >= Z - BEG
+        && XFASTINT (w->last_modified) >= MODIFF
         && XFASTINT (w->last_overlay_modified) >= OVERLAY_MODIFF);
  
    /* When windows_or_buffers_changed is non-zero, we can't rely on


However, I don't know if that breaks the intention of Richard's 
recent fix (again, I'm not quite sure what that was supposed to fix).

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: Emacs appears to hang

[Juanma Barranquero]

On 18 Jul 2002 01:40:07 +0200, storm@cua.dk (Kim F. Storm) wrote:

> The new display bug can be fixed by this small patch:

Yes, thanks.

> However, I don't know if that breaks the intention of Richard's 
> recent fix (again, I'm not quite sure what that was supposed to fix).

I don't know, but the other problem is still there. The effect in my
emacs is that sometimes scrolling paints each line immediately followed
by the following one, so

line 1
line 2
line 3

apears like

line 1 line 2
line 2 line 3
line 3 ...

etc.


                                                           /L/e/k/t/u

Re: Emacs appears to hang

[Robert J. Chassell]

Today's CVS snapshot, Thu, 2002 Jul 18  11:4 UTC,
GNU Emacs 21.3.50.113 (i686-pc-linux-gnu, X toolkit)
with patch to xdisp.c from storm@cua.dk (Kim F. Storm)
and a 143,044 byte .emacs file that likely also changes Emacs.

Juanma Barranquero <lektu@terra.es> writes

   I don't know, but the other problem is still there. The effect in
   my emacs is that sometimes scrolling paints each line immediately
   followed by the following one, ....

I just spent several hours of heavy editing while using using the new
patch and have not seen the problem that Lektu mentions.  Emacs has
been working fine for me.

   > However, I don't know if [the patch] breaks the intention of
   > Richard's recent fix ...

Neither do I.

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: Emacs appears to hang

[Juanma Barranquero]

On Thu, 18 Jul 2002 15:56:28 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:

> Today's CVS snapshot, Thu, 2002 Jul 18  11:4 UTC,
> GNU Emacs 21.3.50.113 (i686-pc-linux-gnu, X toolkit)
> with patch to xdisp.c from storm@cua.dk (Kim F. Storm)
> and a 143,044 byte .emacs file that likely also changes Emacs.

I'm on Windows 2K, using also the latest CVS snapshot, though my .emacs
is only 59,663 bytes long ;-)

> I just spent several hours of heavy editing while using using the new
> patch and have not seen the problem that Lektu mentions.  Emacs has
> been working fine for me.

With Kim's first patch the redisplay bug that affected redrawing after
backspace deleting, etc. has disappeared.

The second bug (the fist, really, as it started to happen earlier) is
not as severe as the other because it doesn't show very often. In
particular, I can do heavy editing for hours and not see it.

The only way to show it that I know is what I've reported: bring up NEWS,
activate ruler-mode and scroll down. Once the cursor reaches the bottom
line, the redisplay starts to behave incorrectly. In all cases, if you
do <down> and the display is garbled, after three or so seconds of
inactivity a new redisplay happens and all is apparently well.

Definitively the bug is still here.


                                                           /L/e/k/t/u

Re: Emacs appears to hang

[Robert J. Chassell]

Today's CVS snapshot, Thu, 2002 Jul 18  11:4 UTC,
GNU Emacs 21.3.50.113 (i686-pc-linux-gnu, X toolkit)
with patch to xdisp.c from storm@cua.dk (Kim F. Storm)
starting Emacs with:

    emacs -q --no-site-file --eval '(blink-cursor-mode 0)'

Juanma Barranquero <lektu@terra.es> writes:

   The second bug ... bring up NEWS, activate ruler-mode and scroll
   down. Once the cursor reaches the bottom line, the redisplay starts
   to behave incorrectly.

I tried this and may have seen the bug when I was changing margins
and fill columns.  In any case, the bug is not obvious to me, who
has never used ruler-mode.

Please tell us how to create the bug so it's existence is more
obvious.

I think this will help get other people, some of whom may not have
much experience with ruler-mode, to fix the bug. 

(By the way, ruler-mode is rather neat, I think I will use it when I
want to change the value of my fill column.  At the moment, however,
ruler-mode is setting the fill-column to two places less than the
location of my pointer on the ruler.)

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: Emacs appears to hang

[Juanma Barranquero]

On Thu, 18 Jul 2002 16:43:11 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:

> In any case, the bug is not obvious to me, who
> has never used ruler-mode.

I don't use it either. I just found that problem by happenstance,
chasing a problem with having `force-mode-line-update' in
`post-command-hook'.

> Please tell us how to create the bug so it's existence is more
> obvious.

emacs -q --no-site-file
C-h C-n
M-x ruler-mode

<down>...                                   ; all works as expected

M-x set-variable scroll-conservatively 1000 ; any non-zero value

<down>...                                   ; bug happens when cursor
                                            ; is in the last line and
                                            ; you scroll down

                                                           /L/e/k/t/u

Re: Emacs appears to hang

[Robert J. Chassell]

   > Please tell us how to create the bug so it's existence is more
   > obvious.

   emacs -q --no-site-file
   C-h C-n
   M-x ruler-mode

   <down>...                                   ; all works as expected

   M-x set-variable scroll-conservatively 1000 ; any non-zero value

   <down>...                                   ; bug happens when cursor
                                               ; is in the last line and
                                               ; you scroll down

Thank you.  Yes, this recreates the redisplay bug.  

However on my (fairly fast, 850 MHz i686) machine, the repeated lines
show only momentarily, not for several seconds.  After the extra
characters appearance, the extra characters vanish and lines look as
they are supposed to.

Using yesterday's CVS snapshot, Thu, 2002 Jul 18  23:22 UTC,
GNU Emacs 21.3.50.115 (i686-pc-linux-gnu, X toolkit),
with the patch from storm@cua.dk (Kim F. Storm) to xdisp.c 
(that fixes the other problem),
Emacs started with:

    emacs -q --no-site-file --eval '(blink-cursor-mode 0)'

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Redisplay bug (was Re: Emacs appears to hang)

[Juanma Barranquero]

On Fri, 19 Jul 2002 11:45:11 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:

> However on my (fairly fast, 850 MHz i686) machine, the repeated lines
> show only momentarily, not for several seconds.

I have a similar machine. With Kim's last patch I see it several seconds
(though the amount is quite variable). Without the patch, the effect is
permanent, until I do <down> again (is like a flip-flop: one keystroke
you have the bug, the next you don't :)


                                                           /L/e/k/t/u

Re: Emacs appears to hang

[Robert J. Chassell]

Another data point:  no appearance of the bug without ruler-mode:

      > Please tell us how to create the bug so it's existence is more
      > obvious.

      emacs -q --no-site-file
      C-h C-n
      M-x ruler-mode

      <down>...                                   ; all works as expected

      M-x set-variable scroll-conservatively 1000 ; any non-zero value

      <down>...                                   ; bug happens when cursor
                                                  ; is in the last line and
                                                  ; you scroll down

   Thank you.  Yes, this recreates the redisplay bug.  

However, these actions do not recreate the bug if you do not start
ruler-mode, at least, not in a slightly newer CVS snapshot (this
morning's, not yesterday's).  

I did everything else the same as shown above, but did not start
ruler-mode; no bug visible.  But when I did the same as described
above, turning on ruler-mode, the bug appeared.

   Using today's CVS snapshot, Fri, 2002 Jul 19  11:51 UTC
   GNU Emacs 21.3.50.117 (i686-pc-linux-gnu, X toolkit),
   which has 
   the patch from storm@cua.dk (Kim F. Storm) to xdisp.c 
   merged in,

   Emacs started with:

       emacs -q --no-site-file --eval '(blink-cursor-mode 0)'

-- 
    Robert J. Chassell                  bob@rattlesnake.com
    Rattlesnake Enterprises             http://www.rattlesnake.com

Re: Emacs appears to hang

[Juanma Barranquero]

On Fri, 19 Jul 2002 12:01:45 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:

> Another data point:  no appearance of the bug without ruler-mode:

Er... yes. That's why included M-x ruler-mode in the problem description.
What I've reported is the simplest way of triggering it that I know.

                                                           /L/e/k/t/u

Re: [PATCH] Re: Emacs appears to hang

[Richard Stallman]

    The new display bug can be fixed by this small patch:

Thanks.

More clues on the redisplay bug (Re: Emacs appears to hang)

[Károly Lőrentey]

Juanma Barranquero <lektu@terra.es> writes:
> On Fri, 19 Jul 2002 12:01:45 +0000 (UTC), "Robert J. Chassell" <bob@rattlesnake.com> wrote:
>
>> Another data point:  no appearance of the bug without ruler-mode:
>
> Er... yes. That's why included M-x ruler-mode in the problem description.
> What I've reported is the simplest way of triggering it that I know.

The bug is not specific to ruler-mode, but I think it's a general
problem that may be related to header lines.  I noticed something
similar in Info:

	emacs -q --no-site-file
		C-h i
	        C-v C-v
        	M-v M-v
	        <Horribly borken display with concatenated lines.>

This is what I see:

        http://caesar.elte.hu/~lorentey/emacs-redisplay-bug.jpg

If I disable the fancy header line, the bug disappears:

	emacs -q --no-site-file --eval '(setq Info-use-header-line nil)'
		C-h i
	        C-v C-v
	        M-v M-v
	        <Normal behaviour.>

All this with today's CVS Emacs, under X.

A few experiments with my previous builds indicate that the bug may be
related to this change:

2002-07-07  Richard M. Stallman  <rms@gnu.org>

        * xdisp.c (make_cursor_line_fully_visible): Don't try short scrolls.
        Instead just return 0 when there is something to be done.
        (try_scrolling): If make_cursor_line_fully_visible returns 0,
        retry scrolling as if cursor were off the bottom.
        (try_cursor_movement): If make_cursor_line_fully_visible returns 0,
        return CURSOR_MOVEMENT_MUST_SCROLL.
        (redisplay_window): If make_cursor_line_fully_visible returns 0,
        go to try_to_scroll.

-- 
Lőrentey Károly

Re: More clues on the redisplay bug (Re: Emacs appears to hang)

[Juanma Barranquero]

On Wed, 24 Jul 2002 12:32:41 +0200, Károly Lõrentey <lorentey@elte.hu> wrote:

> The bug is not specific to ruler-mode, but I think it's a general
> problem that may be related to header lines.

It is definitely related to trying to scroll while the last line is not
fully drawn. As Emacs resizes the frame at startup (and perhaps other
frame's creation) time to display full lines, the bug doesn't appear
unless you do something to draw partially visible lines. An easy way is
to put a header line whose contents is of a different height that the
lines in the buffer.

> A few experiments with my previous builds indicate that the bug may be
> related to this change:

Yes.

                                                           /L/e/k/t/u