Re: push parser

[Bob Rossi]

On Fri, Oct 13, 2006 at 11:17:39AM -0700, Paul Eggert wrote:
> Bob Rossi <address@hidden> writes:
> > If this is the issue, I could force test 90 and 96 to use the
> > skeleton push.c until we merge the two together (probably in a week
> > or so?).
> 
> Yes, that sounds good.  Can you please prepare a revised patch to do that?
> Thanks.

Yes, I'll submit another patch. I do have a question though. This patch
makes it so that yylex takes parameters like:

  static int yylex (YYSTYPE *lvalp, YYLTYPE *llocp)

because now the variables
  YYSTYPE my_lval;
  YYLTYPE my_lloc;

are local. Everything seems to work but it looks like yyerror also
wants to take the   
  YYLTYPE *llocp
parameter, instead of being a void*. Should I make it do that?

That would also require changing b4_yyerror_args macro from

  # b4_yyerror_args
  # ---------------
  # Arguments passed to yyerror: user args plus yylloc.
  m4_define([b4_yyerror_args],
  [b4_yacc_pure_if([b4_locations_if([&yylloc, ])])dnl
  m4_ifset([b4_parse_param], [b4_c_args(b4_parse_param), ])])

to

  # b4_yyerror_args
  # ---------------
  # Arguments passed to yyerror: user args plus yylloc.
  m4_define([b4_yyerror_args],
  [b4_yacc_pure_or_pus_if([b4_locations_if([&yylloc, ])])dnl
  m4_ifset([b4_parse_param], [b4_c_args(b4_parse_param), ])])

Does this make sense?

Bob Rossi