Don't put the pre-prologue in the header file. For the yacc.c code

file and the glr.c header and code files, move the pre-prologue before
the token definitions.  Add new %before-definitions and
%after-definitions to declare code that will go in both the header file
and code file.  Discussed at
<http://lists.gnu.org/archive/html/bison-patches/2005-12/msg00000.html>,
<http://lists.gnu.org/archive/html/bison-patches/2006-01/msg00016.html>,
and
<http://lists.gnu.org/archive/html/bison-patches/2006-06/msg00055.html>.
* NEWS (2.3+): Describe these changes.
* data/glr.c (b4_pre_prologue): Move from within to before...
(b4_shared_declarations): ... this.
Add new b4_before_definitions before b4_token_enums.
Add new b4_after_definitions at the end.
* data/glr.cc (b4_pre_prologue): Replace with...
(b4_before_definitions): ... this in the header file.
(b4_after_definitions): New near the end of the header file.
* data/lalr1.cc (b4_pre_prologue): Move from the header file to the
code file right before including the header file.
(b4_before_definitions): New in the previous position of
b4_pre_prologue in the header file.
(b4_after_definitions): New near the end of the header file.
* data/yacc.c: Clean up some m4 quoting especially in the header file.
(b4_token_enums_defines): In the code file, move to right before
YYSTYPE for consistency with the header file.
(b4_before_definitions): New right before b4_token_enums_defines in
both the header and code file.
(b4_after_definitions): New right after YYLTYPE and yylloc in both the
header and code file.
* doc/bison.texinfo (Prologue): Show use of %before-definitions instead
of prologues for %union dependencies.
(Bison Declaration Summary): In %defines description, mention the
effect of %before-definitions and %after-definitions on the header
file.
(Calc++ Parser): Forward declare driver in a %before-definitions rather
than in the pre-prologue so that make check succeeds.
(Bison Symbols): Add entries for %before-definitions and
%after-definitions.
* src/parse-gram.y (PERCENT_BEFORE_DEFINITIONS): New token for
%before-definitions.
(PERCENT_AFTER_DEFINITIONS): New token for %after-definitions.
(declaration): Parse those declarations and append to
b4_before_definitions and b4_after_definitions, respectively.
* src/reader.c (before_definitions, after_definitions): New bools to
track whether those declarations have been seen.
(prologue_augment): Add to the post-prologue if %union,
%before-definitions, or %after-definitions has been seen.
* src/reader.h (before_definitions, after_definitions): New extern's.
* src/scan-gram.l: Scan the new declarations.
* tests/actions.at (_AT_CHECK_PRINTER_AND_DESTRUCTOR): Place the second
prologue block in a %before-definitions or a %after-definitions based
on whether the %union is declared.
* tests/regression.at (Early token definitions with --yacc, Early token
definitions without --yacc): Move tests for token definitions into the
post-prologue since token names are no longer defined in the
pre-prologue.

NEWS

@@ -9,6 +9,60 @@ Changes in version 2.3+:
   helps to sanitize the global namespace during preprocessing, but POSIX Yacc
   requires them.  Bison still generates an enum for token names in all cases.
 
+* Handling of prologue blocks is now more consistent but potentially backward
+  incompatible.
+
+  As before, you declare prologue blocks in your grammar file with the
+  `%{ ... %}' syntax.  To generate the pre-prologue, Bison concatenates all
+  prologue blocks that you've declared before any %union.  If you've declared a
+  %union, Bison concatenates all prologue blocks that you've declared after it
+  to generate the post-prologue.  (The new %before-definitions and
+  %after-definitions have a similar effect as %union on the prologues.  See
+  below.)
+
+  Previous versions of Bison inserted the pre-prologue into both the header
+  file and the code file in all cases except for LALR(1) parsers in C.  In the
+  latter case, Bison inserted it only into the code file.  For parsers in C++,
+  the point of insertion was before any token definitions (which associate
+  token numbers with names).  For parsers in C, the point of insertion was
+  after the token definitions.
+
+  Now, Bison never inserts the pre-prologue into the header file.  In the code
+  file, it always inserts it before the token definitions.
+
+* Bison now provides the %before-definitions and %after-definitions directives.
+
+  For example, in your grammar file:
+
+    %{
+      /* A pre-prologue block.  For Yacc portability, Bison no longer puts this
+       * in the header file.  In the code file, Bison inserts it before any
+       * %before-definitions blocks.  */
+    %}
+    %before-definitions {
+      /* Bison inserts this into both the header file and code file.  In both
+       * files, the point of insertion is before any Bison-generated token,
+       * semantic type, location type, and class definitions.  This is a good
+       * place to define %union dependencies, for example.  */
+    }
+    %union {
+      /* With previous versions of Bison, the first %union in your grammar file
+       * separated the pre-prologue blocks from the post-prologue blocks.  Now,
+       * the first %union, %before-definitions, or %after-definitions does
+       * that.  */
+    }
+    %after-definitions {
+      /* If you want something in the header file and in the code file and it
+       * depends on any of the Bison-generated definitions in the header file,
+       * put it here.  */
+    }
+    %{
+      /* A post-prologue block.  If you want something in the code file but not
+       * in the header file and it depends on Bison-generated definitions, put
+       * it here.  In the code file, Bison inserts it after any
+       * %after-definitions blocks.  */
+    %}
+
 * The option `--report=look-ahead' has been changed to `--report=lookahead'.
   The old spelling still works, but is not documented and may be removed
   in a future release.