On Wed, 21 Dec 2005, Akim Demaille wrote: > >>> "Paul" == Paul Eggert <[EMAIL PROTECTED]> writes: > > >> - we desire that in all the skeletons no $n is ever freed from an > >> "active" rule, even if we leave it using YYACCEPT, YYERROR, or > >> YYABORT. > > > Yes, that sounds right. > > Well, AFAICS that's already the behavior in glr.c, isn't it? > Otherwise I missed the bit of code that is wrong.
As I recall, it was just YYABORT and YYACCEPT in yacc.c that was a problem. When I looked at the time, glr.c seemed fine. Joel _______________________________________________ Help-bison@gnu.org http://lists.gnu.org/mailman/listinfo/help-bison