Re: Bison version 1.875e available for testing

[Laurence Finston]

On Wed, 15 Dec 2004, Akim Demaille wrote:

> >>> "Laurence" == Laurence Finston <address@hidden> writes:
>
>  > On Wed, 15 Dec 2004, Akim Demaille wrote:
>  >> >>> "Laurence" == Laurence Finston <address@hidden> writes:
>  >>
>  >> > Does this make it possible to declare local automatic
>  >> > variables in `yyparse()'?
>  >>
>  >> Nope.  To a similar end you are expected to use %parse-param.
>  >>
>
>  > That's what I do.  The CWEB section
>  > address@hidden declarations for rules@>' contains code that extracts the
>  > values from the parameter.  It just seems a bit wasteful to do this in
>  > every rule, as well as cluttering up my code.
>
> I don't understand what you mean with "every rule".  Automatic
> variables in yyparse are for a whole parse, just as arguments given to
> %parse-param.
>

Perhaps I've misunderstood what you mean by `%parse-param', and I haven't
looked it up.  I generate a reentrant parser and make use of the ability
to pass an object by means of a `void*' to `yyparse()'.  I use a unique
object for each call to `yyparse()' so I can use the data members of that
object in lieu of local automatic variables in `yyparse()'.


>  > For non-threaded applications it would suffice to use static
>  > variables local to the compilation unit for `yyparse()', but this
>  > won't work with threads.
>
> Why is that?  It is up to you to pass to yyparse references to
> automatic variables from its caller.  So I fail to see what actual
> difference it would make.
>

Because if they're static, there will only be one for all instances of
`yyparse()' instead of a separate one for each instance.  I don't
understand what you mean by passing references to automatic variables to
`yyparse()' from its caller.  What I'd like are automatic variables local
to `yyparse()', which cannot be referenced from outside `yyparse()'.

Laurence