On Mon, 13 Feb 2006, Paul Eggert wrote:

> "Joel E. Denny" <[EMAIL PROTECTED]> writes:
> 
> > You can convert a mid-rule action to a nonterminal, but it seems there is 
> > no other way to prevent its value from being leaked during error recovery. 
> > Is my understanding correct?
> 
> I think so, yes.
> 
> This limitation should be documented, I suppose,

OK, I'll try to get to it later this week.

> until we find a way
> to fix it (probably by having inline destructor declarations or some
> similar monstrosity).

This confusion makes me wonder if %destructor really should be by type and 
if $<type>$ should dynamically record a new type for that state on the 
stack.  $<>$ would remove the type and thus the destructor.  $<type>n and 
$<>n would have no such effect because these semantic values are now in 
the user's hands.

Joel


Reply via email to