Re: %destructor feedback

[Wolfgang Spraul]

Joel -

> I prefer that the clean-up *not* happen automatically for either YYABORT
> or YYERROR -- as in GLR parsing.  This would allow for a simple consistent
> rule: if the user's semantic action executes, then the user's semantic
> action is responsible for cleaning up RHS semantic values... regardless of
> how the user's semantic action terminates.

the reason I prefer the automatic cleanup is that I generally like 'atomic' 
behavior of functions (actions in this case). Atomic = either the 
function/action _completely_ succeeds, or _completely_ fails. If it fails, 
the resulting state should be as if the function was never called.
If you move the responsibility of cleanup when doing YYERROR/YYABORT into the 
action, you break this atomic design: If the action fails, it must now return 
in a different state than when it was called, otherwise the right-hand 
semantic values will be leaked (never cleaned up by anyone).

On top of this design issue there is a practical advantage: There must be a 
built-in cleanup anyway, so the manual cleanup of RHS values is redundant, 
only increases code size for no good reason.

Regards,
Wolfgang

On Monday 17 October 2005 08:34, Joel E. Denny wrote:
> On Sun, 16 Oct 2005, Wolfgang Spraul wrote:
> > Let's say you have an action that gets an object passed in which has a
> > destructor. When using YYABORT inside the action, the destructor for this
> > object is always called. When using YYERROR, it is never called, i.e. the
> > action needs to destruct passed objects itself before using YYERROR.
>
> It seems like this behavior ought to be consistent between YYERROR and
> YYABORT.
>
> Another inconsistency is that, while your above description is accurate
> for yacc.c, it is not accurate for glr.c.  That is, a GLR parser doesn't
> call destructors in either case.
>
> > One could lean either way on the question whether YYABORT/YYERROR should
> > or should not destruct objects passed to actions. At least with 2.1 the
> > behavior is clean, and I'm fine (I'm mostly using YYABORT now and prefer
> > the built-in cleanup).
>
> I prefer that the clean-up *not* happen automatically for either YYABORT
> or YYERROR -- as in GLR parsing.  This would allow for a simple consistent
> rule: if the user's semantic action executes, then the user's semantic
> action is responsible for cleaning up RHS semantic values... regardless of
> how the user's semantic action terminates.
>
> I've also noticed that the bison documentation says:
>
>    Note that in the future, Bison might also consider that right hand
>    side members that are not mentioned in the action can be destroyed.
>
> 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.
>
> What does everyone else think?
>
> Joel