Re: Regression in 2.4 ??

[Nelson Ferreira]

On Nov 9, 2008, at 3:45 PM, Joel E. Denny wrote:

> On Sun, 9 Nov 2008, Nelson Ferreira wrote:
>
> > Attached is the .y file.
> > This is a part of the SXEmacs source, where we had a report of a  
> > build failure
> > with bison 2.4 which we tracked down to this behavior.
> > The relevant files would be under modules/cl.
> > You can get them from http://downloads.sxemacs.org
> >
> > As far as defining manually the YYSTYPE I did it by adding on line  
> > 30:
> >
> > typedef int YYSTYPE;
>
> If YYSTYPE needs to default to an int, then it doesn't make sense to
> declare semantic types, which are fields of a union YYSTYPE.  For  
> example,
> you have:
>
>   %token <forkeys>
>
> Bison 2.3 ignored this.  Bison 2.4 logically assumes you want a  
> union when
> you use a <...> construct.  Moreover, POSIX requires the corrected
> behavior as discussed in the NEWS entry for Bison 2.3a.
>
> I'm going to assume the other issues you mentioned are symptoms of  
> this
> problem.  Let us know if this doesn't fix them.

Ok, I'll try that, because it seems to be a better way to specify the  
grammar,
but I still am confused then about what this actually does:

#if ! defined YYSTYPE && ! defined YYSTYPE_IS_DECLARED

# define yystype YYSTYPE /* obsolescent; will be withdrawn */
# define YYSTYPE_IS_DECLARED 1
#endif

I mean, is this not misleading other code that interfaces with yours?
After all you are stating with #define YYSTYPE_IS_DECLARED 1 that  
YYSTYPE is available
when you cannot actually say that.
AFAICS probably this code should not even be there?