Re: %destructor feedback

[Joel E. Denny]

On Wed, 7 Dec 2005, Frank Heckenbach wrote:

> > > And what about the yyunused variable? If it's declared locally, some
> > > version of a compiler or lint may find that it's assigned a value
> > > that's never used
> >
> > It is used at the next YYUSE() execution:
>
> And for the last (or only) one?

Yeah, sorry, that was stupid. I was still thinking of my global 
declaration.

> > > , and if declared globally, a compiler cannot
> > > readily optimize away the spurious assignment.
> >
> > Unless YYUSE() appears in a short loop (not in bison, I believe), wouldn't
> > the performance penalty be trivial?
>
> Probably, though if the macro is made available to the user, they
> may use it in a short loop ...

My idea of making it available to the user never caught on.  After your 
previous arguments against YYUSE(), I proposed YYBISON_USE() to handle our 
RHS semantic value issue.  It would be available to the user, but it would 
hide its arguments from the compiler.

> I'm wondering if any kind of performance penalty is justified only
> to please lint ...

Personally, I don't ever use lint (although it's looking kind of 
interesting).  But if bison has to deal with this because people want it, 
then I believe the loss of functionality incurred by previous YYUSE 
definitions is worse than a small performance penalty.  Perhaps there 
should be a flag to #define YYUSE(e) as nothing when the user wants to 
swap warnings for small performance improvements.

Joel