Re: mode-line-inactive and face inheritance

[Stefan Monnier]

>     More specifically, it adds an `all' entry which (just like `t')
>     always matches but doesn't prevent subsequent specs from being
>     considered.
> That seems useful.  `common' might be clearer than `all'.

Actually, I think we can just reuse t instead of `all' or `common'.
Currently t can only be meaningfully used for the last entry,
so we can simply extend its meaning a little bit in a backward
compatible way.

>                Also it allows nesting of display-specific specs
>     so you can extract the common part of the `light' and `'dark'
>     settings for `color' displays.
> 
> That seems too complex; I think I would not want to document this
> even if it worked.

I'm ambivalent about it.  I think it could also make things simpler.
Currently defface accepts a SPEC of the form:

        SPEC     ::= (CONDSPEC CONDSPEC ...)
        CONDSPEC ::= (COND PROPS)
                   | (COND PROP PROP ...)

Where (COND PROPS) is an obsolete form kept for backward compatibility only.
The code I posted changes it to

        SPEC     ::= (CONDSPEC CONDSPEC ...)
        CONDSPEC ::= (COND PROPS)
                   | (COND PROP PROP ...)
                   | (COND CONDSPEC CONDSPEC ...)

But an alternative way to get a similar result would be to use

        SPEC     ::= (CONDSPEC CONDSPEC ...)
                   | (PROP PROP ...)
        CONDSPEC ::= (COND PROPS)
                   | (COND . SPEC)

which has the advantage that a simple defface that has no conditional
part could now be written as

        (defface font-lock-doc-face
          '(:inherit font-lock-string-face)
          "Font Lock mode face used to highlight documentation."
          :group 'font-lock-highlighting-faces)

instead of

        (defface font-lock-doc-face
          '((t :inherit font-lock-string-face))
          "Font Lock mode face used to highlight documentation."
          :group 'font-lock-highlighting-faces)


-- Stefan