23.0.50; Cannot run calculator on TTY

23.0.50; Cannot run calculator on TTY

[raman]

I can no longer run M-x calculator with emacs running on the
Linux console (outside X).
Error I get is:
ad-Orig-split-window-vertically: Window height 2 too small (after
splitting)

Here is a backtrace:
ad-Orig-split-window-vertically: Window height 2 too small (after splitting)

Debugger entered--Lisp error: (error "Window height 2 too small (after splitting)")
  split-window(nil 10)
  ad-Orig-split-window-vertically(-2)
  split-window-vertically(-2)
  ad-Orig-calculator()
  calculator()
  call-interactively(calculator)
  execute-extended-command(nil)
  call-interactively(execute-extended-command)

In GNU Emacs 23.0.50.26 (i686-pc-linux-gnu, GTK+ Version 2.8.20)
 of 2007-09-27 on labrador
configured using `configure  '--prefix=/usr/local' '--with-gtk' '--with-gpm' '--with-rsvg' '--with-png' '--with-jpeg' '--with-gif''

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: en_US.UTF-8
  locale-coding-system: utf-8
  default-enable-multibyte-characters: nil

Major mode: VM

Minor modes in effect:
  erc-menu-mode: t
  erc-autojoin-mode: t
  erc-ring-mode: t
  erc-pcomplete-mode: t
  erc-button-mode: t
  erc-fill-mode: t
  erc-stamp-mode: t
  erc-netsplit-mode: t
  erc-track-mode: t
  erc-track-minor-mode: t
  erc-match-mode: t
  jabber-activity-mode: t
  erc-irccontrols-mode: t
  erc-noncommands-mode: t
  erc-readonly-mode: t
  dynamic-completion-mode: t
  dired-omit-mode: t
  shell-dirtrack-mode: t
  savehist-mode: t
  display-time-mode: t
  desktop-save-mode: t
  auto-insert-mode: t
  file-name-shadow-mode: t
  global-font-lock-mode: t
  font-lock-mode: t
  unify-8859-on-encoding-mode: t
  utf-translate-cjk-mode: t
  auto-compression-mode: t

Recent input:
RET c d SPC a f g h a n RET w g e t SPC C-y RET m v 
SPC b o TAB m o n . r a m RET C-x C-f m o RET ESC < 
ESC [ [ D ESC > c u r l SPC C-y C-e e ESC b ESC b C-d 
C-d C-d t u e C-e C-e SPC > SPC t u e . r a m RET ESC 
p C-e e ESC b ESC b C-k ESC b ESC b C-k w e d . r p 
m SPC > SPC w e d . r a m RET ESC p C-e C-e ESC b ESC 
b ESC b ESC b C-k t h u . r p m SPC > SPC t h u . r 
a m RET ESC p ESC p C-e C-e ESC b ESC b ESC b ESC b 
C-k f r i . r p m SPC > SPC f r i . r a m RET f o r 
SPC i SPC i n SPC * . r a m RET d o SPC RET r a m p 
3 DEL DEL DEL m 2 p m DEL DEL m p 3 SPC $ i RET d o 
n e RET C-e m C-x @ h m ESC C-s q C-e x m ESC m SPC 
ESC g RET 1 RET N k k k k k k ' V S C-e m N n P j ' 
n j ' n j ' ESC c q n k k k ESC x c a l c u l TAB RET 
C-e a ESC x r e o DEL DEL DEL r e p o r t SPC e m TAB 
b TAB RET

Recent messages:
Decoding MIME message...
Decoding quoted-printable... done
Wrote /tmp/vm-pgg-signature5850T9K
Decoding MIME message... done
Hid 17  citation lines
All messages have been marked as deleted.
Expunging...
Decoding MIME message... done
Deleted messages expunged.
nil
ad-Orig-split-window-vertically: Window height 2 too small (after splitting)

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

> I can no longer run M-x calculator with emacs running on the
> Linux console (outside X).
> Error I get is:
> ad-Orig-split-window-vertically: Window height 2 too small (after
> splitting)
> 
> Here is a backtrace:
> ad-Orig-split-window-vertically: Window height 2 too small (after splitting)
> 
> Debugger entered--Lisp error: (error "Window height 2 too small (after splitting)")
>   split-window(nil 10)
>   ad-Orig-split-window-vertically(-2)
>   split-window-vertically(-2)
>   ad-Orig-calculator()
>   calculator()
>   call-interactively(calculator)
>   execute-extended-command(nil)
>   call-interactively(execute-extended-command)

What do the "ad-Orig-..." do?  What is the value of `window-min-height'
when they attempt to split the window?

Re: 23.0.50; Cannot run calculator on TTY

[T. V. Raman]

Apologies about the Ad-orig -- that's because everything has been
advised by Emacspeak.

window-min-height is 4 -- I've not changed it.

>>>>> "martin" == martin rudalics <rudalics@gmx.at> writes:
    >> I can no longer run M-x calculator with emacs running on
    >> the Linux console (outside X).  Error I get is:
    >> ad-Orig-split-window-vertically: Window height 2 too small
    >> (after splitting)
    >> 
    >> Here is a backtrace: ad-Orig-split-window-vertically:
    >> Window height 2 too small (after splitting)
    >> 
    >> Debugger entered--Lisp error: (error "Window height 2 too
    >> small (after splitting)") split-window(nil 10)
    >> ad-Orig-split-window-vertically(-2)
    >> split-window-vertically(-2) ad-Orig-calculator()
    >> calculator() call-interactively(calculator)
    >> execute-extended-command(nil)
    >> call-interactively(execute-extended-command)
    martin> 
    martin> What do the "ad-Orig-..." do?  What is the value of
    martin> `window-min-height' when they attempt to split the
    martin> window?

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 > Apologies about the Ad-orig -- that's because everything has been
 > advised by Emacspeak.

I understand.  The problem is that apparently

   ad-Orig-calculator()

wants to do

   split-window-vertically(-2)

which eventually has `split-window' complain.  If you bind
`window-min-height' to 2 around this the complaints should go away,
hopefully.

 > window-min-height is 4 -- I've not changed it.

Did you change it earlier?

Re: 23.0.50; Cannot run calculator on TTY

[T. V. Raman]

>>>>> "martin" == martin rudalics <rudalics@gmx.at> writes:
    >> Apologies about the Ad-orig -- that's because everything
    >> has been advised by Emacspeak.
    martin> 
    martin> I understand.  The problem is that apparently
    martin> 
    martin>    ad-Orig-calculator()
    martin> 
    martin> wants to do
    martin> 
    martin>    split-window-vertically(-2)

The problem is in the calculator.el code.
My advice is benign --(will attach it here)
but it's the original code, i.e. ad-Orig-calculator that is
ending up with the bizarre -2.
I tried to understand the code in calculator.el that does this
but failed.

Here is the advice I have on calculator:
(defadvice calculator (after emacspeak pre act comp)
  "Speech enable calculator."
  (when (interactive-p)
    (emacspeak-auditory-icon 'open-object)
    (message "Welcome to the pocket calculator.")))

All the above does is to speak a message and play a sound after
the calculator is up.

The problem lines in calculator.el appear to be:
;;; starting on line 734:

(let ((split-window-keep-point nil)
               (window-min-height 2))
           ;; maybe leave two lines for our window because of the normal
           ;; `raised' modeline in Emacs 21
           (select-window
            (split-window-vertically
             ;; If the modeline might interfere with the calculator buffer,
             ;; use 3 lines instead.
             (if (and (fboundp 'face-attr-construct)
                      (let* ((dh (plist-get (face-attr-construct 'default) :height))
                             (mf (face-attr-construct 'modeline))
                             (mh (plist-get mf :height)))
                        ;; If the modeline is shorter than the default,
                        ;; stick with 2 lines.  (It may be necessary to
                        ;; check how much shorter.)
                        (and
                         (not
                          (or (and (integerp dh)
                                   (integerp mh)
                                   (< mh dh))
                              (and (numberp mh)
                                   (not (integerp mh))
                                   (< mh 1))))
                         (or
                          ;; If the modeline is taller than the default,
                          ;; use 3 lines.
                          (and (integerp dh)
                               (integerp mh)
                               (> mh dh))
                          (and (numberp mh)
                               (not (integerp mh))
                               (> mh 1))
                          ;; If the modeline has a box with non-negative line-width,
                          ;; use 3 lines.
                          (let* ((bx (plist-get mf :box))
                                 (lh (plist-get bx :line-width)))
                            (and bx
                                 (or
                                  (not lh)
                                  (> lh 0))))
                          ;; If the modeline has an overline, use 3 lines.
                          (plist-get (face-attr-construct 'modeline) :overline)))))
               -3 -2)))
           (switch-to-buffer calculator-buffer))

Notice the call to split-window-vertically above, and the
somewhat complex code that determines the numeric argument being
passed to that call.


    martin> 
    martin> which eventually has `split-window' complain.  If you
    martin> bind `window-min-height' to 2 around this the
    martin> complaints should go away, hopefully.
Not sure what you mean.
I got the code to work by deleting the split-window-vertically
    martin> call in my copy of calculator --which is not
    martin> completely right -- because  
obviously, I then end up with the calculator taking the whole
    martin> screen.

    martin> 
    >> window-min-height is 4 -- I've not changed it.
    martin> 
    martin> Did you change it earlier?
No, I dont touch it.

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 > The problem is in the calculator.el code.
 > My advice is benign --(will attach it here)
 > but it's the original code, i.e. ad-Orig-calculator that is
 > ending up with the bizarre -2.

-2 is not bizarre.  It's purpose is to give the calculator window 2
lines.

 > The problem lines in calculator.el appear to be:
 > ;;; starting on line 734:
 >
 > (let ((split-window-keep-point nil)
 >                (window-min-height 2))
 >            ;; maybe leave two lines for our window because of the normal
 >            ;; `raised' modeline in Emacs 21
 >            (select-window
 >             (split-window-vertically

The binding of `window-min-height' is here to avoid that `split-window'
complains about too small windows.

 >              ;; If the modeline might interfere with the calculator buffer,
 >              ;; use 3 lines instead.
 >              (if (and (fboundp 'face-attr-construct)
                    ...

All this shouldn't bother us since we already know that we get the
standard value -2 here.

                    ...
 >                -3 -2)))
 >            (switch-to-buffer calculator-buffer))
 >
 > Notice the call to split-window-vertically above, and the
 > somewhat complex code that determines the numeric argument being
 > passed to that call.

Hence we seem to be left with the fact that your configuration has

(let ((split-window-keep-point nil)
       (window-min-height 2))
   (split-window-vertically -2))

fail for some reason.  Could you please verify by evaluating that form?

Re: 23.0.50; Cannot run calculator on TTY

[T. V. Raman]

    martin> Hence we seem to be left with the fact that your

    martin> configuration has
    martin> 
    martin> (let ((split-window-keep-point nil)
    martin> (window-min-height 2)) (split-window-vertically -2))
    martin> 
    martin> fail for some reason.  Could you please verify by
    martin> evaluating that form?
The form that  computes the argument to split-window-vertically
    martin> (shown below)
evaluates to -2 in my environment.
(if (and (fboundp 'face-attr-construct)
                      (let* ((dh (plist-get (face-attr-construct 'default) :height))
                             (mf (face-attr-construct 'modeline))
                             (mh (plist-get mf :height)))
                        ;; If the modeline is shorter than the default,
                        ;; stick with 2 lines.  (It may be necessary to
                        ;; check how much shorter.)
                        (and
                         (not
                          (or (and (integerp dh)
                                   (integerp mh)
                                   (< mh dh))
                              (and (numberp mh)
                                   (not (integerp mh))
                                   (< mh 1))))
                         (or
                          ;; If the modeline is taller than the default,
                          ;; use 3 lines.
                          (and (integerp dh)
                               (integerp mh)
                               (> mh dh))
                          (and (numberp mh)
                               (not (integerp mh))
                               (> mh 1))
                          ;; If the modeline has a box with non-negative line-width,
                          ;; use 3 lines.
                          (let* ((bx (plist-get mf :box))
                                 (lh (plist-get bx :line-width)))
                            (and bx
                                 (or
                                  (not lh)
                                  (> lh 0))))
                          ;; If the modeline has an overline, use 3 lines.
                          (plist-get (face-attr-construct 'modeline) :overline)))))
               -3 -2)

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 > The form that  computes the argument to split-window-vertically
 >     martin> (shown below)
 > evaluates to -2 in my environment.

Sorry, my question was formulated badly.  I wanted to know what happens
when you evaluate the form below in your environment:

(let ((split-window-keep-point nil)
       (window-min-height 2))
   (split-window-vertically -2))

Re: 23.0.50; Cannot run calculator on TTY

[T. V. Raman]

>>>>> "martin" == martin rudalics <rudalics@gmx.at> writes:
    
    martin> Sorry, my question was formulated badly.  I wanted to
    martin> know what happens when you evaluate the form below in
    martin> your environment:
    martin> 
    martin> (let ((split-window-keep-point nil)
    martin> (window-min-height 2)) (split-window-vertically -2))
Raises the same error as raised by calling M-x calculator shown
    martin> earlier, so we've narrowed it down.
I also get the same error if I just evaluate
     (split-window-vertically -2) in my environment.
Works if I  pass -3

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 >     martin> Sorry, my question was formulated badly.  I wanted to
 >     martin> know what happens when you evaluate the form below in
 >     martin> your environment:
 >     martin>
 >     martin> (let ((split-window-keep-point nil)
 >     martin> (window-min-height 2)) (split-window-vertically -2))
 > Raises the same error as raised by calling M-x calculator shown
 >     martin> earlier, so we've narrowed it down.
 > I also get the same error if I just evaluate
 >      (split-window-vertically -2) in my environment.
 > Works if I  pass -3
 >

This means something's wrong with your `split-window'.  C-x C-e gets me
for

(let ((window-min-height 2))
   (split-window-vertically -3))

=> Make a two lines window at bottom

(split-window-vertically -3)

=> Lisp error: (error "Window height 3 too small (after splitting)")

(let ((window-min-height 2))
   (split-window-vertically -2))

=> Make a one line window at bottom

(split-window-vertically -2)

=> Lisp error: (error "Window height 2 too small (after splitting)")

which is the expected behavior.  Your behavior is wrong because

(let ((window-min-height 2))
   (split-window-vertically -2))

should not_raise an error while

(split-window-vertically -3)

should.  My build is Emacs 22.1 with a few modifications.  Maybe
something changed since then.

Can you run `split_window' with GDB and find out why it fails?

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

> I finally isolated the cause of this problem.

Fine.

> In short, if a buffer has a header-line set  i.e. non-nil value
> for header-line-format,
> then launching calculator there boms with the error shown
> earlier.
> 
> Specifically, compare the following two forms:
> ;;; no error:
> (let ((header-line-format nil)
>       (window-min-height 2))
>   (split-window-vertically -2))
> 
> 
> ;;; Errors:
> (let ((header-line-format "Foo")
>       (window-min-height 2))
>   (split-window-vertically -2))
> 
> Basically when header-line-format is set, the  smallest window
> you can make needs 3 lines.

When you're splitting windows, yes.

> 
> The fix might be to set header-line-format to nil in the
> top-level let form in calculator.

Alternatively we could do something like

(select-window
  (split-window-vertically
   ;; If the modeline might interfere with the calculator buffer,
   ;; use 3 lines instead.
   (- (if (and (fboundp 'face-attr-construct)
	      (let*
		  ...
		(plist-get (face-attr-construct 'modeline) :overline)))
	 -3 -2)
      (if header-line-format 1 0))))

in calculator.  The problem is that `calculator' should do all those
fancy calculations it does for the mode-line for the header-line too.
What do you prefer?  Is a header-line of any use for calculator?

Re: 23.0.50; Cannot run calculator on TTY

[T. V. Raman]

For something like the calculator that wants to claim as small a
window as possible, I would guess that the header line is not too
useful, especially since it has a modeline.

>>>>> "martin" == martin rudalics <rudalics@gmx.at> writes:
    >> I finally isolated the cause of this problem.
    martin> Fine.
    martin> 
    >> In short, if a buffer has a header-line set i.e. non-nil
    >> value for header-line-format, then launching calculator
    >> there boms with the error shown earlier.
    >> 
    >> Specifically, compare the following two forms: ;;; no
    >> error: (let ((header-line-format nil) (window-min-height
    >> 2)) (split-window-vertically -2))
    >> 
    >> 
    >> ;;; Errors: (let ((header-line-format "Foo")
    >> (window-min-height 2)) (split-window-vertically -2))
    >> 
    >> Basically when header-line-format is set, the smallest
    >> window you can make needs 3 lines.
    martin> 
    martin> When you're splitting windows, yes.
    martin> 
    >> 
> The fix might be to set header-line-format to nil in the
    >> top-level let form in calculator.
    martin> 
    martin> Alternatively we could do something like
    martin> 
    martin> (select-window (split-window-vertically ;; If the
    martin> modeline might interfere with the calculator buffer,
    martin> ;; use 3 lines instead.  (- (if (and (fboundp
    martin> 'face-attr-construct) (let* ...  (plist-get
    martin> (face-attr-construct 'modeline) :overline))) -3 -2)
    martin> (if header-line-format 1 0))))
    martin> 
    martin> in calculator.  The problem is that `calculator'
    martin> should do all those fancy calculations it does for
    martin> the mode-line for the header-line too.  What do you
    martin> prefer?  Is a header-line of any use for calculator?
    martin> 

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 > For something like the calculator that wants to claim as small a
 > window as possible, I would guess that the header line is not too
 > useful, especially since it has a modeline.

There should be no problem doing

         ((not (get-buffer-window calculator-buffer))
          (let ((split-window-keep-point nil)
                (window-min-height 2))
	   (setq header-line-format nil)

in `calculator'.

I wonder, however, why `header-line-format' has a non-nil value in the
first place.  Do you have an advice creating a header line for all sorts
of windows?

Re: 23.0.50; Cannot run calculator on TTY

[T. V. Raman]

In the emacspeak environment, I've set header-line-format to
either show just the buffer name or in case of things like my
alsaplayer interface, to show the title of the song that is
playing.
In info mode,  I  use the default behavior of using the header
line to show the node information.

In general, the header-line is one of my more favorite features
in newer emacsuns since it lets me squirrel away summary
information in a 
manner that gets automatically updated.


>>>>> "martin" == martin rudalics <rudalics@gmx.at> writes:
    >> For something like the calculator that wants to claim as
    >> small a window as possible, I would guess that the header
    >> line is not too useful, especially since it has a
    >> modeline.
    martin> 
    martin> There should be no problem doing
    martin> 
    martin>          ((not (get-buffer-window calculator-buffer))
    martin> (let ((split-window-keep-point nil)
    martin> (window-min-height 2)) (setq header-line-format nil)
    martin> 
    martin> in `calculator'.
    martin> 
    martin> I wonder, however, why `header-line-format' has a
    martin> non-nil value in the first place.  Do you have an
    martin> advice creating a header line for all sorts of
    martin> windows?

-- 
Best Regards,
--raman

      
Email:  raman@users.sf.net
WWW:    http://emacspeak.sf.net/raman/
AIM:    emacspeak       GTalk: tv.raman.tv@gmail.com
PGP:    http://emacspeak.sf.net/raman/raman-almaden.asc
Google: tv+raman 
IRC:    irc://irc.freenode.net/#emacs

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 > In the emacspeak environment, I've set header-line-format to
 > either show just the buffer name or in case of things like my
 > alsaplayer interface, to show the title of the song that is
 > playing.
 > In info mode,  I  use the default behavior of using the header
 > line to show the node information.
 >
 > In general, the header-line is one of my more favorite features
 > in newer emacsuns since it lets me squirrel away summary
 > information in a
 > manner that gets automatically updated.

Presently, `split-window' has a problem with mode/header lines and small
heights (presumably also with scroll-bars and fringes).  To check
whether all windows are large enough after splitting it must make an
assumption about the presence of these on the new window.  It currently
bases its judgments on the appearance of the window that shall be split.
This is obviously wrong in your calculator example: You probably have a
header line on the existing window and `split-window' assumes that you
want a header line for the calculator window too.

I'm not sure how `split-window' should handle this:

(1) Use worst-case assumptions (all windows get scroll-bars, header
     lines, ...).

(2) Use assumptions based on default values.

(3) Provide explicit (optional) parameters for `split-window-vertically'
     and `split-window-horizontally'.

(4) Have functions like `calculator' split the window first and
     adjust it afterwards (probably using `fit-window-to-buffer').

Re: 23.0.50; Cannot run calculator on TTY

[Richard Stallman]

    Presently, `split-window' has a problem with mode/header lines and small
    heights (presumably also with scroll-bars and fringes).  To check
    whether all windows are large enough after splitting it must make an
    assumption about the presence of these on the new window.  It currently
    bases its judgments on the appearance of the window that shall be split.

Strictly speaking, that is correct.  Splitting the window shows the
same buffer in both of the resulting windows, so it is valid if that
buffer could be displayed in both.  However, in a larger sense, given
that the aim is to display another window there, this does not really
DTRT.

One idea that occurs to me is that `window-min-height' could count the
header line, if any.  That way, the fact that some buffers have header
lines and others don't will not cause any problems for vertical
splitting of windows.  In the most usual cases, at least -- but not
in all cases, perhaps.

Another solution would be to pass a two buffers to `split-window',
telling it to select those two buffers, and to calculate based
on the assumption that those buffers will be displayed.

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

 >     Presently, `split-window' has a problem with mode/header lines and small
 >     heights (presumably also with scroll-bars and fringes).  To check
 >     whether all windows are large enough after splitting it must make an
 >     assumption about the presence of these on the new window.  It currently
 >     bases its judgments on the appearance of the window that shall be split.
 >
 > Strictly speaking, that is correct.  Splitting the window shows the
 > same buffer in both of the resulting windows, so it is valid if that
 > buffer could be displayed in both.  However, in a larger sense, given
 > that the aim is to display another window there, this does not really
 > DTRT.
 >
 > One idea that occurs to me is that `window-min-height' could count the
 > header line, if any.

Do you mean

"One idea that occurs to me is that the `split-window' could NOT count
the header line, if any."

here?  That is, change

     size = max (window_min_height,
		(MIN_SAFE_WINDOW_HEIGHT
		 + (WINDOW_WANTS_MODELINE_P (w) ? 1 : 0)
		 + (WINDOW_WANTS_HEADER_LINE_P (w) ? 1 : 0 )));

to

     size = max (window_min_height,
		(MIN_SAFE_WINDOW_HEIGHT
		 + (WINDOW_WANTS_MODELINE_P (w) ? 1 : 0)));

in `window_min_size_2'?  If so I think we should do that and maybe add a
short remark in the doc-string of `split-window' and the Elisp manual.
The only drawback I can think of is that splitting may produce a window
showing a header- and a mode-line only.

 > That way, the fact that some buffers have header
 > lines and others don't will not cause any problems for vertical
 > splitting of windows.  In the most usual cases, at least -- but not
 > in all cases, perhaps.
 >
 > Another solution would be to pass a two buffers to `split-window',
 > telling it to select those two buffers, and to calculate based
 > on the assumption that those buffers will be displayed.

That's what I meant with "provide explicit (optional) parameters for
`split-window-vertically' and `split-window-horizontally'".  This would
also mean that the client of these functions (and `split-window') would
have to set the corresponding (buffer-local) variables _before_ calling
these.

Re: 23.0.50; Cannot run calculator on TTY

[Richard Stallman]

     > One idea that occurs to me is that `window-min-height' could count the
     > header line, if any.

    Do you mean

    "One idea that occurs to me is that the `split-window' could NOT count
    the header line, if any."

We may be trying to say the same thing in two ways.  Basically, my
idea is that the minimum allowed number of lines in a window -- in
every sense -- would not depend on whether one of those lines
is used as a header line.

This might require changes in redisplay as well
in order to function correctly.

    The only drawback I can think of is that splitting may produce a window
    showing a header- and a mode-line only.

Yes, that could happen.  It might be a lesser problem than the one we
now have.  At least it would be comprehensible and would follow a
simple rule.

     > That way, the fact that some buffers have header
     > lines and others don't will not cause any problems for vertical
     > splitting of windows.  In the most usual cases, at least -- but not
     > in all cases, perhaps.
     >
     > Another solution would be to pass a two buffers to `split-window',
     > telling it to select those two buffers, and to calculate based
     > on the assumption that those buffers will be displayed.

    That's what I meant with "provide explicit (optional) parameters for
    `split-window-vertically' and `split-window-horizontally'".  This would
    also mean that the client of these functions (and `split-window') would
    have to set the corresponding (buffer-local) variables _before_ calling
    these.

Then we both had the same idea.  If people like this better, it is
fine with me.

Re: 23.0.50; Cannot run calculator on TTY

[martin rudalics]

> We may be trying to say the same thing in two ways.  Basically, my
> idea is that the minimum allowed number of lines in a window -- in
> every sense -- would not depend on whether one of those lines
> is used as a header line.

Since no one opted for the solution below I checked in a fix to do that.

>     That's what I meant with "provide explicit (optional) parameters for
>     `split-window-vertically' and `split-window-horizontally'".  This would
>     also mean that the client of these functions (and `split-window') would
>     have to set the corresponding (buffer-local) variables _before_ calling
>     these.
> 
> Then we both had the same idea.  If people like this better, it is
> fine with me.