On Tue, 3 Jan 2006, Joel E. Denny wrote:
> On Tue, 3 Jan 2006, Akim Demaille wrote:
>
> > >>> "Joel" == Joel E Denny <[EMAIL PROTECTED]> writes:
> >
> > > To accommodate YYSTYPE dependencies, YYLTYPE dependencies, a non-%union
> > > YYSTYPE definition, a user YYLTYPE definition, or anything else the user
> > > might wish to place in the header, perhaps there should be some sort of
> > > general `header' qualifier for any literal block:
> >
> > > %header{
> > > /* C code */
> > > %}
> >
> > I dislike the difference that it introduces between a parser with
> > %defines, and another without. That's another thing I enjoy with the
> > current scheme: there's nothing to change.
>
> In the current scheme, adding %defines means you should rearrange your
> bison spec so that no inappropriate code appears in the pre-prologue. If
> you forget, you may unknowingly pollute the header.
>
> In the scheme I'm proposing, adding %defines means you have to rearrange
> so that appropriate code appears in %header literal blocks. If you
> forget, the compiler will complain about missing dependencies from the
> header, so you'll be sure to find out.
>
> In either scheme, removing %defines requires no extra work.
>
> > In fact, I don't like parsers without headers :(
>
> Neither do I.
>
> > Can this wait after 2.2?
>
> Absolutely.
Are we ready to address this now?
The discussion starts here:
http://lists.gnu.org/archive/html/bison-patches/2005-12/msg00000.html
and continues here:
http://lists.gnu.org/archive/html/bison-patches/2006-01/msg00016.html
Joel
