Re: [Groff] Underlining, again

[Peter Schaffter]

On Mon, Sep 20, 2010, Tadziu Hoffmann wrote:
> 
> > As to changing fonts while underlining, it's hard to imagine
> > a situation in which it's likely to happen.  Underlining
> > should at best be an occasional effect, IMHO.  Still,
> > one likes to have solutions that cover every possibility.
> 
> Well, for editorial purposes it's sometimes useful to mark
> up changes to a document, for example by underlining newly
> inserted text and striking through deleted text, and these
> have to be available *in addition* to all other markup.

Yep.  That would be one of those "covers every possibility"
scenarios.

> Here's my second attempt, which works a bit better but still
> has problems once in a while because grops's output is "too
> optimized" (e.g., it sometimes uses "ashow" to produce a space
> and can thus occasionally step (invisibly) beyond the end of
> the line, but the simple underline algorithm doesn't realize
> what grops is doing).

I'll test it out and report back.

> Ultimately I think the only sensible solution is to make grops
> itself underline-aware and make .cu do underlining in troff if not
> in compatibility mode.

As do I, but I recall this issue being discussed a number of years
ago, and someone giving a Very Good Reason why .cu was unlikely ever
to behave this way.  The problem is, I can't remember what said VGR
was.

It seems to me that even were .cu implemented such that it
did constant underlining, it wouldn't offer any flexibility with
respect to rule weight and distance from the baseline unless .cu
were capable of accepting args.  The advantage to your solution is 
that it can easily be modifified with respect to those parameters.

On another topic:

Werner, I can't provide a small example to demonstrate the "blank
page at the end of a diversion" anomaly I posted about earlier, only
a big, unwieldy one.  Aside from having to post several (long) mom
macros that are modified versions of what's in the cvs, I'd have to
include a refer database and an example file deliberately set up
to collect precisely the right number of references to exhibit the
offending behaviour.

I have a no-fail workaround for now, so the matter isn't pressing.
However, bearing in mind that there may be a bug in the C++ source,
I'll try to put together a reasonably sized working (non-working?)
example as soon.

-- 
Peter Schaffter

Author of The Binbrook Caucus
http://www.schaffter.ca