Re: Bison/flex compatibility revisited

[Bruce Lilly]

John wrote:
> > One exception that I neglected to mention is that if
> > the flex option for reentrant code is used, YY_G,
> > the related structure, and the definitions of yyout
> > etc. that use YY_G may be required.
>
>
> Any code that explicitly references YY_G is broken. YY_G is an
> internal macro, and subject to change.

But with a reentrant lexical analyzer, yyout is defined as
#define yyout YY_G(yyout_r)
which obviously isn't going to work if there's no definition
of YY_G and/or nothing in which yyout_r is a member.
Any code using yyout therefore needs YY_G; and if
it's subject to change, the definition should come from
the flex-generated header, and not have to be reinvented
by the flex user.