Re: %destructor feedback

[Joel E. Denny]

On Tue, 18 Oct 2005, Paul Eggert wrote:

> > If this is implemented, I believe that it should be implemented
> > consistently for all three cases: the user's semantic action completes
> > successfully, it calls YYABORT, or it calls YYERROR.
>
> Yes, that sounds right to me as well.
>
> > In other words, if you have an action that does nothing but invoke
> > YYABORT/YYERROR, clean-up would be automatic.
>
> This sounds a little tricky to explain and/or implement; wouldn't it
> be better to stick with the simpler rules?  Perhaps I'm missing
> something....

I'm not sure what you mean by `the simpler rules'.

As I see it, there would be one rule at first: in all cases (usual 
success, YYACCEPT, YYABORT, or YYERROR), the user must clean up RHS 
semantic values in any semantic action he provides.

Then there's the automatic destructor calls for unmentioned RHS semantic 
values.  This feature is proposed in the bison documentation.  If this 
feature is added, the user could avoid writing the destructor invocation 
for a particular RHS semantic value by simply not mentioning that semantic 
value in the semantic action.  You seem to be agreeing to this feature for 
the case of YYABORT and YYERROR in your first response quoted above.

A logical result of this second rule is my second quote above.  Does this 
make sense?

Please note that, later in this thread, Wolfgang and I pointed out a 
number of potential problems with this proposed unmentioned RHS semantic 
value feature.  I wonder if it should be reconsidered.

Joel