Re: [Groff] Better locale macro file handling?

[Colin Watson]

On Wed, Feb 18, 2009 at 08:10:24AM +0100, Werner LEMBERG wrote:
> Colin Watson wrote:
> >Werner LEMBERG wrote:
> >> [...] Convert the error message of .mso into a new warning category
> >> which is enabled by default.  Then you can easily disable it with
> >> using a -W ... command line option.
> > 
> > Hmm. I thought warning bits were in rather scarcer supply than
> > request names; there are only three bits left at the moment, since
> > it's a signed int, and I can't think of many other plausible places
> > where you'd want the same category of warning as this.
> 
> The last category, WARN_COLOR, has been added in 2001...  This is
> really no problem.  Additionally, we currently have 19 warning
> categories, and groff assumes that integers are 32 bit at least, so we
> have actually 12 bits left.

Whoops, I confused octal with hexadecimal! Of course you're absolutely
right.

> > And, as you say, it is unfortunate from a backward compatibility
> > point of view: man-db would probably end up having to detect the
> > version of groff, since otherwise it would cause a warning on every
> > single manual page unless the macro file is present (which will be
> > the clear majority of pages).
> 
> Then use the `.warn' request directly.  If the `range' warning
> (value 64) isn't set, `.warn' doesn't emit a warning if you supply a
> value larger than WARN_TOTAL.

I hadn't thought of .warn; thanks. In that case simply feeding in this
preamble will work:

  .warn (\n[.warn] - (\n[.warn] / 64 % 2 * 64))
  .warn (\n[.warn] + 1048576 - (\n[.warn] / 1048576 % 2 * 1048576))

> > Nevertheless, here's the necessary patch if you're sure:
> 
> Thanks!  I'll incorporate this in case there aren't strong objections.

Given the above trick, this will work for me.

Thanks,

-- 
Colin Watson                                       address@hidden