Wired behavior of `def()'

Hi list,

After looked into the discuss 2 months ago between Nicolaos and Eric,
I'm interesting to try to make a M4 implement in Rust to better
understand m4 and for fun. A naive version is easy to made (I have
both study the BSD implement and GNU implement together to make a
better architecture). But it seems that both two implement can not
well handle recursive `$@' in O(1) . So I start to study the implement
of 1.6 version of GNU m4 and make a new implementation based on the
idea of that version.

But in the meantime of implementing, I have found a wired behavior of
`defn'. The document said:
> If name is a user-defined macro, the quoted definition is simply the quot=
ed expansion text. If, instead, there is only one name and it is a builtin,=
 the expansion is a special token, which points to the builtin=E2=80=99s in=
ternal definition. This token is only meaningful as the second argument to =
define (and pushdef), and is silently converted to an empty string in most =
other contexts. Combining a builtin with anything else is not supported; a =
warning is issued and the builtin is omitted from the final expansion.

But this text:
define(`foo',    defn(`defn')     bar)
it should be same as:
define(`foo', bar)
as the defn(`defn') expands to empty, and the whitespaces after `,'
will ignored. But I found it defined a wired macro:
foo
=3D> foo
foo()
=3D>

I noticed the statement: "This token is only meaningful as the second
argument to define (and pushdef)", So I tried without `define':
define(`foo', `[$1]-[$2]')
=3D>
foo(   first   , defn(`defn')   second)
=3D>[first   ]-[]

I think it should be expand as:
=3D> [first   ]-[second]
or even:
=3D> [first   ]-[   second]

So which behavior is correct?

--=20
regards,
Xavier Wang.