Re: avoiding infinite loops

[Hans Aberg]

On 14 Jun 2006, at 00:09, Joel E. Denny wrote:

> On Tue, 13 Jun 2006, Sylvain Schmitz wrote:
>
> > Since we are talking about conflicts and looping parsers, there is a
> > possibility for an infinite loop with bison's conflict resolution  
> > in LALR(1).
> > See attached file loop.y.
>
> Here's one that uses rule order rather than precedence declarations:
>
>   S: A ;
>   B: C ;
>   C: B ;
>   A: B ;
>   C: 'a' ;
>
> Input is just 'a'.

My guess is that it is the strongly connected components (SCC) among  
non-shiftable rules that causes problems. The Bison method of  
resolving reduce/reduce conflicts by only selecting the first one  
only resolves loops of at most length two.

  Hans Aberg