Re: [Groff] strange defining of undefined names

[Werner LEMBERG]

> There is a strange anomaly about undefined names in
> commented/ignored environments.  Try the following roff fragments:
>
> 2.
> .rm s
> \c comment on \*[s]
> .ie d s .tm yes
> .el .tm no
> 
> This prints `yes' - strange.

\c only suppresses output of the rest of the line, nevertheless it is
parsed.  `s' isn't found, thus groff emits a warning and defines `s'
to be empty.  For example, this works:

  \c\R'xxx 4'
  .tm xxx = \n[xxx]

    => xxx = 4

This is already documented in groff.texinfo.

Can someone please test Bernd's example with AT&T troff?

> 3.
> .rm s
> .ig
> comment on \*[s]
> ..
> .ie d s .tm yes
> .el .tm no
> 
> This prints `yes' - strange.

.ig is the same as .de except that the parsed data is discarded.  If
you write \\*[s], `s' won't be expanded.

> 4.
> .de c
> ..
> .rm s
> .c comment on \*[s]
> .ie d s .tm yes
> .el .tm no
> 
> This prints `yes' - strange.

Arguments of a macro are always parsed regardless whether the macro is
empty or whether the arguments are used actually.  Since groff tries
to immediately interpolate strings and registers, it checks whether
`s' is defined.


    Werner