XKB Geometry documentation

[Tim Ri]

On Sunday 04 February 2007 13:58, Daniel Stone wrote:

> As with most things XKB, I guess that the most accurate specification
> is, as you've already discovered, the union of all files currently in
> use, and/or the xkbcomp source code

Hi Daniel, thanks for responding.

* X does not use geometry files (xkbcomp knows just enough to 
compile them).
* Without specs the geometry interpretation is entirely up to the 3d-party 
programs using them.
* Hunting those programs down is a chore (writing on-screen keyboards was 
quite the fad some time ago).

> xkbcomp badly needs a rewrite, 
Well, since my attention was mainly on geometry.c it is difficult to judge, 
but what I saw was pretty clean in my opinion.

> if you're that keen ...
Take a wild guess :)
Kidding aside: Since I have no experience in writing pure C code (I prefer 
C++), I doubt I would be the person for the job.
Anyway, I think writing a better spec (read: an up-to-date one) is at the 
moment quite a bit more important, so that's where I will start.

Regards,

Tim

---original message---
> On Sat, Feb 03, 2007 at 11:14:57AM +0100, Tim Ri wrote:
> > This week I started writing a small parser for geometry files. I was
> > wondering if someone could point me to some documentation/specifications
> > for the grammar used in these files.
> >
> > The most useful links I found (http://www.xfree86.org/current/XKBlib.pdf
> > and http://www.charvolant.org/~doug/xkb/index.html) are unfortunately not
> > sufficiently up-to date. Some oddities from the geometry files in my
> > distro (Debian and Ubuntu):
> >
> > * Use of new attributes in text doodads (e.g. setWidth).
> > * Use of cornerRadius attribute for a solid (maybe this is allowed, but I
> > cannot find that explicitly in the spec).
> > * Use of fractional numbers for gaps, coordinates etc. (perhaps this has
> > to do with the "1mm = 0,1 mm" strangeness noted by Doug Palmer).
> > * Use of numerical expressions (e.g. 10 + 12) for attribute values.
> >
> > At the moment I have a working parser, that is, I fiddled with it long
> > enough that it is now capable to recognize all my distro's geometry files
> > as valid, but I would _love_ to see some official/officious specification
> > to back it up.
>
> As with most things XKB, I guess that the most accurate specification
> is, as you've already discovered, the union of all files currently in
> use, and/or the xkbcomp source code[0].
>
> Cheers,
> Daniel
>
> [0]: xkbcomp badly needs a rewrite, if you're that keen ...