Re: RFC: m4 2.0 --prepend-include option

[Paul Eggert]

Eric Blake <address@hidden> writes:

> Hmm, that's not quite what we thought earlier:
> http://lists.gnu.org/archive/html/m4-patches/2003-06/msg00035.html

Certainly the spec could be read the other way, as Tim Van Holder
does.  But I don't agree that his reading is the only one.  It's just
as natural to parse the spec as saying that the distinction between
'define' and 'pushdef' is that the former does not preserve the
current definition, whereas the latter does.

When attempting to resolve ambiguities like this, one thing to
consider is what implementations do; and we have genuine disagreement
here among implementations.

One plausible fix is to change the spec to say that choice of behavior
is implementation-specified.  Another possibility is to change POSIX
to clearly require the GNU behavior (since it's better :-), but you'd
have to make a strong case for why users would prefer that.