[Groff] groff regression tests

[Gaius Mulley]

Hi Werner,

I wonder whether it might be a good idea to introduce another groff
node (say \Q) which allows developers to embed expected horizontal or
vertical positions of glyphs.

The syntax might be:
   \Q[X,id,initialiser]

where X can be x or y
      id is a number
      initialiser is an optional measurement

The idea is that these nodes should be emitted to the device driver
and then the device driver can check that these node actually match
the specified X or Y position.

For example


.ta 1i 2i 3i
\Q[x,1,1i]foo<tab>\Q[x,2,2i]bar<tab>\Q[x,3,3i]mumble
.br
\Q[x,1,]again<tab>\Q[x,2,]this<tab>\Q[x,3,]line up

the benefit of this approach is that we can then place a significant
number of regression tests into the dejagnu format (borrow much tcl
code from gcc regression tests - code reuse is great within the GPL!)
and test, test, test without having to always run a previewer
:-). Certainly I've many .IP tests and manual pages which could easily
be converted to utilise the \Q node modifications..

What do you think?

Gaius