bug#9607: incorrect behavior with "automatic-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

bug#9607: incorrect behavior with "automatic-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[David Lawless]

have "(setq automatic-hscroll-mode nil)" in .emacs file

edit wide output of file copied from /proc/interrupts
(machine has 12 cores)

place cursor at around column 12, then scroll out
to left view with "scroll-left" so ends-of-lines
appear in middle of windows, then from mapped
key issue "next-line" twice

first time it behaves correctly, second time
display scrolls horizontally so cursor is visible,
which it should not do

was upgrading from 20.5a to 23.3a when noticed
this; went back to 22.1 and it works ok


In GNU Emacs 22.1.1 (x86_64-unknown-linux-gnu)
 of 2011-09-26 on siobhan
configured using `configure  '--without-x''

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

Major mode: Lisp Interaction

Minor modes in effect:
  file-name-shadow-mode: t
  unify-8859-on-encoding-mode: t
  utf-translate-cjk-mode: t
  auto-compression-mode: t
  line-number-mode: t

Recent input:
ESC x TAB C-g ESC x r e p TAB o TAB DEL o t TAB DEL 
r TAB C-j

Recent messages:
("emacs")
Loading hi-lock...done
Loading cc-mode...done
Loading make-mode...done
Making completion list...
Loading help-mode...done
Quit [2 times]
Making completion list... [2 times]
Loading emacsbug...done

bug#9607:

[David Lawless]

see that in 22.1 the .emacs file setting is

   (setq automatic-hscrolling nil)

instead of 

   (setq automatic-hscroll-mode nil)

This does not change that the behavior is broken
in version 23 and works ok in 21.1.

Would venture to guess that it stopped
working correctly at the time the code was
modified to effect the above variable name
change.

bug#9607: incorrect behavior with "automatic-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[Glenn Morris]

David Lawless wrote:

> have "(setq automatic-hscroll-mode nil)" in .emacs file

Did you mean `auto-hscroll-mode'?
There is no `automatic-hscroll-mode' in Emacs.

bug#9607: incorrect behavior with "automatic-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[David Lawless]

Yes, miskeyed it.  Emacs didn't spit at it when
it was in .emacs so I did have it right at the
time.  23 is gone so typed that from inaccurate
recollection.


At 08:33 PM 9/26/2011 -0400, Glenn Morris wrote:
>David Lawless wrote:
>
>> have "(setq automatic-hscroll-mode nil)" in .emacs file
>
>Did you mean `auto-hscroll-mode'?
>There is no `automatic-hscroll-mode' in Emacs.

bug#9607: incorrect behavior with "automatic-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[Stefan Monnier]

> Yes, miskeyed it.  Emacs didn't spit at it when
> it was in .emacs so I did have it right at the
> time.

Emacs wouldn't spit at (setq foo-bar-baz nil) either.  It only complains
if the variable does exist and is read-only.


        Stefan

bug#9607: incorrect behavior with "auto-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[David Lawless]

Well in any case I've got it right.

Just bisected the bug and it first appears
in 23.2.  23.1 is ok.  Left the old name
in place during bisection and let emacs
apply the alias.

Definitely broken starting in 23.2.

bug#9607: incorrect behavior with "auto-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23.2 but ok in 23.1

[David Lawless]

Also just reproduced the bug by running
emacs in vanilla mode with 'emacs -q' to
inhibit loading .emacs.  Typed commands
and mode settings manually.

bug#9607: incorrect behavior with "auto-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[Glenn Morris]

Present in 23.2 to 24.0.90 inclusive. Recipe:

emacs -Q --eval '(setq auto-hscroll-mode nil)'

Evaluate:
(dotimes (j 5)
  (insert "\n")
  (dotimes (i 10)
    (insert "0123456890 ")))

Put point at start of first line of digits.
M-x scroll-left RET y
;; Window now scrolled to right-hand end of lines
Down arrow  ; first time is ok
Down arrow
  -> second time window scrolls back to left-hand end of line


(setq line-move-visual nil)

avoids the problem.

I guess it's related to the 2009-07 changes to line-move-visual for hscroll.

bug#9607: incorrect behavior with "auto-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Date: Tue, 27 Sep 2011 00:01:50 -0400
> Cc: David Lawless <david_lawless@flumedata.com>
> 
> emacs -Q --eval '(setq auto-hscroll-mode nil)'
> 
> Evaluate:
> (dotimes (j 5)
>   (insert "\n")
>   (dotimes (i 10)
>     (insert "0123456890 ")))
> 
> Put point at start of first line of digits.
> M-x scroll-left RET y
> ;; Window now scrolled to right-hand end of lines
> Down arrow  ; first time is ok
> Down arrow
>   -> second time window scrolls back to left-hand end of line
> 
> 
> (setq line-move-visual nil)
> 
> avoids the problem.
> 
> I guess it's related to the 2009-07 changes to line-move-visual for hscroll.

The problem is that line-move-visual assumes that the move begins with
cursor positioned at point, which is false when auto-hscroll-mode is
turned off and point is scrolled out of view.

Before I dive into this mess (again), do we even want to support
auto-hscroll-mode nil _and_ visual line movement?  What if we tell
line-move to invoke logical line movement instead when
auto-hscroll-mode is off?  Would someone object?

bug#9607: incorrect behavior with "auto-hscroll-mode nil" in ncurses/vt220 text operation; wrong in 23 but ok in 22.1

[Eli Zaretskii]

> Date: Tue, 27 Sep 2011 02:07:20 -0400
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: 9607@debbugs.gnu.org, david_lawless@flumedata.com
> 
> The problem is that line-move-visual assumes that the move begins with
> cursor positioned at point, which is false when auto-hscroll-mode is
> turned off and point is scrolled out of view.
> 
> Before I dive into this mess (again), do we even want to support
> auto-hscroll-mode nil _and_ visual line movement?  What if we tell
> line-move to invoke logical line movement instead when
> auto-hscroll-mode is off?  Would someone object?

No one objected, so I fixed this in revision 105959 on the trunk by
disabling visual line movement when auto-hscroll-mode is disable and
the window is hscrolled.