Re: modeline face under a tty

[Richard Stallman]

    > Well, my solution is to keep the inverse video attribute and define the
    > reversed colors. The problem of reversed color on X is solved by:
    > 
    > (set-face-foreground 'mode-line
    >            (if window-system "white"    "magenta"))
    > (set-face-background 'mode-line
    >            (if window-system "magenta"  "white"))

    That's what I do as well, but Richard thinks it's wrong.

Doing that isn't wrong; what's wrong is Emacs's behavior.
It is wrong for Emacs to make the user do this.
This is inconsistent behavior, and we need to fix it.

Here is another idea.  Don't set the :inverse-video attribute for
`mode-line' on a tty.  Instead, reverse its colors as usual, so that
the foreground is unspecified-bg and the background is unspecified-fg.

Then, at low level, this combination of colors can be implemented
using stand-out mode, or whatever method would now be used for
inverse-video.

The :inverse-video attribute would still exist, and when used it would
work in a uniform simple way--by swapping the foreground and
background colors in face realization.

I think this makes all the cases work right and consistently.  On X,
the behavior will be unchanged from now.  On a tty or on X, when
colors are specified for mode-line, it will use them as specified (not
the reverse of them).  On a tty, when no colors are specified, it will
still use stand-out for the mode line.

Do you see any flaws in this solution?