I propose that newer skeletons define the tokens only using enums, not
#defines in addition. That's already the case for C++, but I propose
that we do the same for glr.c. In practice, it only means to apply
the following patch. Any objection?
Index: ChangeLog
from Akim Demaille <[EMAIL PROTECTED]>
* data/glr.c: Use b4_token_enums, not b4_token_enums_defines.
Index: data/glr.c
===================================================================
RCS file: /cvsroot/bison/bison/data/glr.c,v
retrieving revision 1.124
diff -u -u -r1.124 glr.c
--- data/glr.c 16 Sep 2005 22:54:21 -0000 1.124
+++ data/glr.c 19 Sep 2005 07:10:37 -0000
@@ -154,7 +154,7 @@
#define yynerrs b4_prefix[]nerrs
#define yylloc b4_prefix[]lloc])
-b4_token_enums_defines(b4_tokens)
+b4_token_enums(b4_tokens)
/* Copy the first part of user declarations. */
b4_pre_prologue[
@@ -2321,7 +2321,7 @@
b4_copyright([Skeleton parser for GLR parsing with Bison],
[2002, 2003, 2004, 2005])
-b4_token_enums_defines(b4_tokens)
+b4_token_enums(b4_tokens)
#if ! defined (YYSTYPE) && ! defined (YYSTYPE_IS_DECLARED)
m4_ifdef([b4_stype],
