Re: \s[-\n(.s] and \s-\n(.s inconsistent

[Gunnar Ritter]

Werner LEMBERG <address@hidden> wrote:

> > > At the same time, I've disallowed `\s-[-...]' and friends.
> >
> > Why? Their previous behavior was just fine mathematically.
>
> First of all, it isn't documented.

Yes it is: "\s+[N] ... N is a numeric expression".

Note that you are now also disallowing code like

.nr a -3
.nr b 5
\s-[\na+\nb]

Then there is

.ps --2

which is compatible to v7 troff.

> Secondly, it creates an ambiguity:
> shall the second minus character start a relative size change (within
> the relative size change), or shall it start a negative number because
> we already are in `relative' mode?

The first alternative makes no sense to me at all.

AT&T troff always only checks the first character of an expression
to determine if it is relative; a sign as the second character is
interpreted as a unary operator.

        Gunnar