I notice that my new font-lock port requires the :inherit face property.
The idea seems to be that faces under GNU Emacs are now specified through a
series of attributes, the importance of which is that fonts have been
exploded into a whole series of new properties -- slant, weight, height,
family, etc. In GNU Emacs the code to combine them happens at the C code.
In our C code, the font is just a unified font spec, although there are
nasty functions to `make-font-smaller' etc. There also exist, in our C
code, inheritance which is in some way more powerful than FSF [individual
properties can each inherit from some other face, even from a
differently-named property if there is semantic equivalency]. FSF has only
a single :inherit attribute, which is not as powerful as what is just
mentioned but more convenient to specify and more powerful in that the font
properties have been split out. Currently we have Lisp code to combine
multiple face properties to generate a display face (e.g. multiple faces may
overlay a particular region of text, and in general faces need not specify
values for all possible properties, so one might specify a only foreground
color, another a font, and a third a background pixmap, and the resulting
text has all three properties applied). It would be very easy to add an
`inherit' property (currently inheritance is always from the `default'
face), and we could hook the Lisp code into the C code so that we could
directly set the exploded properties on faces and have the fonts generated
automatically. But it would probably have problems getting proper
on-the-fly inheritance of all the various font properties. Doing this right
would not be too difficult since all the complex algorithmic stuff is
already written, but it would require a bit of work.
So ... There are various paths I could take, from doing nothing at all to
doing the whole thing, complete and correct, to in-between steps that should
be clear from the above discussion.
I'd like to hear people's thoughts on [a] how important is this
functionality, [b] who would use it, [c] how much of it needs to be
implemented, and when [e.g. "While you're at it, do it right" or "Just
in a minimal hack to get font-lock minimally working, then worry about it