Block cursor in a tty

Block cursor in a tty

[Stefan Monnier]

Several people have complained on gnu.emacs.help that Emacs uses an ugly big
block cursor in their terminal emulator even after they've configured their
terminal emulator to use a more subdued cursor.

It turns out that the reason for it is that Emacs uses the termcap "vs"
command to make the cursor visible rather than "ve".  My manpages describe
the two commands as follows:

       cursor_normal                 cnorm      ve        make cursor appear
                                                          normal (undo
                                                          civis/cvvis)
       [...]
       cursor_visible                cvvis      vs        make cursor very
                                                          visible

Does anybody know why Emacs insists on using the big-block-cursor rather
than using whatever the user has selected as default?


        Stefan

Re: Block cursor in a tty

[David Kastrup]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Several people have complained on gnu.emacs.help that Emacs uses an ugly big
> block cursor in their terminal emulator even after they've configured their
> terminal emulator to use a more subdued cursor.
>
> It turns out that the reason for it is that Emacs uses the termcap "vs"
> command to make the cursor visible rather than "ve".  My manpages describe
> the two commands as follows:
>
>        cursor_normal                 cnorm      ve        make cursor appear
>                                                           normal (undo
>                                                           civis/cvvis)
>        [...]
>        cursor_visible                cvvis      vs        make cursor very
>                                                           visible
>
> Does anybody know why Emacs insists on using the big-block-cursor rather
> than using whatever the user has selected as default?

A possible rationale would be that commands like C-x o jump to a
possibly surprising position on the screen, and you would not want to
look too long for locating it.

Personally, I think we should tend to default to the normal cursor,
and maybe provide the visible cursor as an option for people who
really want it.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Block cursor in a tty

[Richard M. Stallman]

    Does anybody know why Emacs insists on using the big-block-cursor rather
    than using whatever the user has selected as default?

vs was defined specifically to be used by full-screen editors to make
the cursor more visible.

The default cursor is for sequential input.  With sequential input,
you know which line to look in, so it is easier to see the cursor;
thus, it does not need extra visibility.

Re: Block cursor in a tty

[Giorgos Keramidas]

        Does anybody know why Emacs insists on using the
        big-block-cursor rather than using whatever the user has
        selected as default?

    vs was defined specifically to be used by full-screen editors to
    make the cursor more visible.

    The default cursor is for sequential input.  With sequential input,
    you know which line to look in, so it is easier to see the cursor;
    thus, it does not need extra visibility.

Can we at least make it a customizable option?  I didn't look very much
into this, but it sure felt annoying when I first set the default cursor
of my consoles to non-block mode and found out that Emacs appeared to
'reset' the cursor to block mode.

- Giorgos

Re: Block cursor in a tty

[Juri Linkov]

>         Does anybody know why Emacs insists on using the
>         big-block-cursor rather than using whatever the user has
>         selected as default?
>
>     vs was defined specifically to be used by full-screen editors to
>     make the cursor more visible.
>
>     The default cursor is for sequential input.  With sequential input,
>     you know which line to look in, so it is easier to see the cursor;
>     thus, it does not need extra visibility.
>
> Can we at least make it a customizable option?  I didn't look very much
> into this, but it sure felt annoying when I first set the default cursor
> of my consoles to non-block mode and found out that Emacs appeared to
> 'reset' the cursor to block mode.

Yes, it is annoying when Emacs overrides the default console cursor.
When the default console cursor is configured to non-blinking,
Emacs changes it to the blinking cursor.

A new customizable option or at least a command line option would be
a good thing.

-- 
Juri Linkov
http://www.jurta.org/emacs/

Re: Block cursor in a tty

[Richard M. Stallman]

    Can we at least make it a customizable option?  I didn't look very much
    into this, but it sure felt annoying when I first set the default cursor
    of my consoles to non-block mode and found out that Emacs appeared to
    'reset' the cursor to block mode.

You can certainly control this by changing the termcap entry.

Could the terminal emulator let you specify how the visual mode
cursor should look?

Re: Block cursor in a tty

[Giorgos Keramidas]

        Can we at least make it a customizable option?  I didn't
        look very much into this, but it sure felt annoying when
        I first set the default cursor of my consoles to
        non-block mode and found out that Emacs appeared to
        'reset' the cursor to block mode.

    You can certainly control this by changing the termcap entry.

    Could the terminal emulator let you specify how the visual mode
    cursor should look?

Possibly.  Not all terminal emulators support modifying the
termcap on the fly, and altering the termcap entry is sometimes
difficult for people who have no write access to the termcap
database or lack the expertise required to hack termcap entries.

Making things a little easier would be fine, IMHO :)

Re: Block cursor in a tty

[Richard M. Stallman]

I would not object to having an option to turn off use of vs.
However, it will be nontrivial to implement one, since term_init is
called before any Lisp code runs.  So it can't be controlled by the
init file.

Re: Block cursor in a tty

[Giorgos Keramidas]

    I would not object to having an option to turn off use of
    vs.  However, it will be nontrivial to implement one, since
    term_init is called before any Lisp code runs.  So it can't
    be controlled by the init file.

Currently, term_init sets TS_cursor_visible to tgetstr("vs", address),
which then gets used in set_terminal_modes().  We could probably add
term_init_cursor() and arrange for it to be called twice, once with the
default "vs" capability before the init file is loaded and once if the
init file specifies something different later on.  By making an option
that the init file can use to make TS_cursor_visible equal to
TS_cursor_normal, we're set :)

I'm not very acquainted with the way the terminal handling works though,
so it may take a time before I have a patch ready and tested.

- Giorgos

Re: Block cursor in a tty

[Richard M. Stallman]

    Currently, term_init sets TS_cursor_visible to tgetstr("vs", address),
    which then gets used in set_terminal_modes().  We could probably add
    term_init_cursor() and arrange for it to be called twice, once with the
    default "vs" capability before the init file is loaded and once if the
    init file specifies something different later on.

Yes, it could be done.  It would not arrange to avoid touching the
cursor type at all, but it could set the cursor type back to ordinary.

Re: Block cursor in a tty

[Stefan Monnier]

>>>>> "Giorgos" == Giorgos Keramidas <keramida@ceid.upatras.gr> writes:

>     I would not object to having an option to turn off use of
>     vs.  However, it will be nontrivial to implement one, since
>     term_init is called before any Lisp code runs.  So it can't
>     be controlled by the init file.

> Currently, term_init sets TS_cursor_visible to tgetstr("vs", address),
> which then gets used in set_terminal_modes().  We could probably add
> term_init_cursor() and arrange for it to be called twice, once with the
> default "vs" capability before the init file is loaded and once if the
> init file specifies something different later on.  By making an option
> that the init file can use to make TS_cursor_visible equal to
> TS_cursor_normal, we're set :)

> I'm not very acquainted with the way the terminal handling works though,
> so it may take a time before I have a patch ready and tested.

> - Giorgos

Guaranteed 100% untested sample patch.
Doesn't include the update to cus-start.el to make it customizable.


        Stefan


--- term.c	20 oct 2005 15:44:27 -0400	1.170
+++ term.c	26 déc 2005 11:17:31 -0500	
@@ -87,6 +87,10 @@
 
 Lisp_Object Vring_bell_function;
 
+/* If non-nil, use "vs", otherwise use "ve" to make the cursor visible.  */
+
+Lisp_Object Vvisible_cursor;
+
 /* Terminal characteristics that higher levels want to look at.
    These are all extern'd in termchar.h */
 
@@ -460,7 +464,7 @@
 	    putchar ('\n');
 	}
 
-      OUTPUT_IF (TS_cursor_visible);
+      OUTPUT_IF (NILP (Vvisible_cursor) ? TS_cursor_normal : TS_cursor_visible);
       OUTPUT_IF (TS_keypad_mode);
       losecursor ();
     }
@@ -615,6 +619,7 @@
     {
       tty_cursor_hidden = 0;
       OUTPUT_IF (TS_cursor_normal);
+      if (!NILP (Vvisible_cursor))
       OUTPUT_IF (TS_cursor_visible);
     }
 }
@@ -2726,6 +2731,10 @@
 The function should accept no arguments.  */);
   Vring_bell_function = Qnil;
 
+  DEFVAR_LISP ("visible-cursor", &Vvisible_cursor,
+	       doc: /* Non-nil means to make the cursor very visible.  */);
+  Vvisible_cursor = Qt;
+
   defsubr (&Stty_display_color_p);
   defsubr (&Stty_display_color_cells);
   defsubr (&Stty_no_underline);

Re: Block cursor in a tty

[Richard M. Stallman]

    Guaranteed 100% untested sample patch.
    Doesn't include the update to cus-start.el to make it customizable.

To make that patch work, something has to call those functions again
after the init file runs.  I doubt that that happens now;
do you see a reason to think it does?

Re: Block cursor in a tty

[Stefan Monnier]

>     Guaranteed 100% untested sample patch.
>     Doesn't include the update to cus-start.el to make it customizable.

> To make that patch work, something has to call those functions again
> after the init file runs.  I doubt that that happens now;
> do you see a reason to think it does?

I do see such reasons: write_glyphs calls tty_hide_cursor and update_end
calls tty_show_cursor.


        Stefan

Re: Block cursor in a tty

[Richard M. Stallman]

    I do see such reasons: write_glyphs calls tty_hide_cursor and update_end
    calls tty_show_cursor.

I'm surprised, but glad, that it is so easy.

Please install your patch.

Re: Block cursor in a tty

[Stefan Monnier]

>     I do see such reasons: write_glyphs calls tty_hide_cursor and update_end
>     calls tty_show_cursor.

> I'm surprised, but glad, that it is so easy.
> Please install your patch.

Actually, as mentioned, the patch is 100% guaranteed untested.
I don't know how to test it.


        Stefan

Re: Block cursor in a tty

[Giorgos Keramidas]

On 2005-12-28 22:02, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
> >     I do see such reasons: write_glyphs calls tty_hide_cursor and update_end
> >     calls tty_show_cursor.
>
> > I'm surprised, but glad, that it is so easy.
> > Please install your patch.
>
> Actually, as mentioned, the patch is 100% guaranteed untested.
> I don't know how to test it.

Running Emacs in a console terminal on FreeBSD, where the default cursor
has been configured in /etc/rc.conf with:

    cursor="destructive"        # Use an underline- instead of block-cursor

turns the cursor to block cursor immediately.  I'll patch my local Emacs
installation and report back later today, Stefan.  Thanks for coming up
with a patch so fast :)

Re: Block cursor in a tty

[Giorgos Keramidas]

On 2005-12-29 12:21, Giorgos Keramidas <keramida@ceid.upatras.gr> wrote:
> On 2005-12-28 22:02, Stefan Monnier <monnier@iro.umontreal.ca> wrote:
> > >     I do see such reasons: write_glyphs calls tty_hide_cursor and
> > >     update_end calls tty_show_cursor.
> >
> > > I'm surprised, but glad, that it is so easy.
> > > Please install your patch.
> >
> > Actually, as mentioned, the patch is 100% guaranteed untested.
> > I don't know how to test it.
>
> Running Emacs in a console terminal on FreeBSD, where the default cursor
> has been configured in /etc/rc.conf with:
>
>     cursor="destructive"        # Use an underline- instead of block-cursor
>
> turns the cursor to block cursor immediately.  I'll patch my local Emacs
> installation and report back later today, Stefan.  Thanks for coming up
> with a patch so fast :)

Excellent.  All works as expected, and changes to cursor-visible from a
Lisp interaction buffer take effect automatically.

Re: Block cursor in a tty

[Stefan Monnier]

>> > >     I do see such reasons: write_glyphs calls tty_hide_cursor and
>> > >     update_end calls tty_show_cursor.
>> >
>> > > I'm surprised, but glad, that it is so easy.
>> > > Please install your patch.
>> >
>> > Actually, as mentioned, the patch is 100% guaranteed untested.
>> > I don't know how to test it.
>> 
>> Running Emacs in a console terminal on FreeBSD, where the default cursor
>> has been configured in /etc/rc.conf with:
>> 
>> cursor="destructive"        # Use an underline- instead of block-cursor
>> 
>> turns the cursor to block cursor immediately.  I'll patch my local Emacs
>> installation and report back later today, Stefan.  Thanks for coming up
>> with a patch so fast :)

> Excellent.  All works as expected, and changes to cursor-visible from a
> Lisp interaction buffer take effect automatically.

Thanks, installed.


        Stefan

Re: Block cursor in a tty

[Richard M. Stallman]

Since Giorgos says this patch works, please install it.