>> An observation: When I modify the frame creation like this
 >>
 >> (defun tty-3 ()
 >>    (interactive)
 >>    (setq tty-3
 >> 	(make-frame
 >> 	 `((parent-frame . ,tty-2)
 >> 	   (left . (- 30)) (top . 5)
 >>                     ^^^^^^
 >> 	   (width . 10) (height . 5)
 >>
 >> and then do M-l and C-M-l, tty-3 is displayed at a different location
 >> than after the modify-frame-parameter. This is an indication that there
 >> is an inconsistency somewhere.
 >>
 >> Elisp info says about frame parameter 'left':
 >>
 >>       ‘(- POS)’
 >>            This specifies the position of the right frame edge relative
 >>            to the right edge of the display or parent frame.  The integer
 >>            POS may be positive or negative; a negative value specifies a
 >>            position outside the screen or parent frame or on a monitor
 >>            other than the primary one (for multi-monitor displays).
 >>
 >> Parent frame tty-2 of tty-3 has
 >>
 >> 	 `((parent-frame . ,root-frame)
 >> 	   (left . 40) (top . 5)
 >> 	   (width . 30) (height . 20)
 >>
 >> tty-2's right edges is at 40 + 30 = 70, in root coordinates.

Hopefully.

 >> So, the right edge of tty-3 should land at 70, which makes its left edge
 >> should be 70 - 10 = 60.

The right edge of tty-3 should land at 70 - 30 = 40 in root coordinates
and its left edge at 40 - 10 = 30 in root coordinates which means, given
that tty-2's left is at 40 in root coordinates, tty-3 should be
invisible.

 >> That is absolutely not what I'm seeing, neither with the initial (- 30)
 >> for left nor after modify-frame-parameters. So I guess my reading of
 >> Elisp info is wrong?
 >>
 >> Martin, could you please help me with this? What is the intended effect
 >> of the (- 30), do you know?

The old code in tty_child_pos_param is wrong.  I'm using


tty_child_pos_param (struct frame *f, Lisp_Object key,
		     Lisp_Object params, int pos)
{
   struct frame *p = XFRAME (f->parent_frame);
   Lisp_Object val = Fassq (key, params);

   if (CONSP (val))
     {
       val = XCDR (val);

       if (TYPE_RANGED_FIXNUMP (int, val))
	{
	  pos = XFIXNUM (val);

	  if (pos < 0) ; <------------ see here !!!!!
	    /* Handle negative value. */
	    pos = (EQ (key, Qtop)
		   ? p->pixel_height - f->pixel_height + pos
		   : p->pixel_width - f->pixel_width + pos);
	}
       else if (CONSP (val) && EQ (XCAR (val), Qplus)
	       && CONSP (XCDR (val))
	       && TYPE_RANGED_FIXNUMP (int, XCAR (XCDR (val))))
	pos = XFIXNUM (XCAR (XCDR (val)));
       else if (CONSP (val) && EQ (XCAR (val), Qminus)
	       && CONSP (XCDR (val))
	       && RANGED_FIXNUMP (-INT_MAX, XCAR (XCDR (val)), INT_MAX))
	pos = (EQ (key, Qtop)
	       ? p->pixel_height - f->pixel_height - XFIXNUM (XCAR (XCDR (val)))
	       : p->pixel_width - f->pixel_width - XFIXNUM (XCAR (XCDR (val))));
     }

   return pos;
}


and am still not entirely convinced that it's right.  I attach my
current changes.

martin