* src/reader.c (parse_union_decl): Instead of handling two obstack

[bison.git] / src / reader.c

diff --git a/src/reader.c b/src/reader.c
index 69619358e1e194bdaf4d12e392c0f7b82583d8af..509aa3cec44fdc43ccb96ae8b986a1c4772496d6 100644 (file)
--- a/src/reader.c
+++ b/src/reader.c
@@ -1,5 +1,5 @@
 /* Input parser for bison
 /* Input parser for bison

-   Copyright 1984, 1986, 1989, 1992, 1998, 2000
+   Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001

    Free Software Foundation, Inc.
 
    This file is part of Bison, the GNU Compiler Compiler.
    Free Software Foundation, Inc.
 
    This file is part of Bison, the GNU Compiler Compiler.


@@ -36,13 +36,14 @@
 #include "conflicts.h"
 #include "muscle_tab.h"
 
 #include "conflicts.h"
 #include "muscle_tab.h"
 

-/* Number of slots allocated (but not necessarily used yet) in `rline'  */
-static int rline_allocated;
-

 typedef struct symbol_list
 {
   struct symbol_list *next;
   bucket *sym;
 typedef struct symbol_list
 {
   struct symbol_list *next;
   bucket *sym;

+  int line;
+  /* The action is attached to the LHS of a rule. */
+  const char *action;
+  int action_line;

   bucket *ruleprec;
 }
 symbol_list;
   bucket *ruleprec;
 }
 symbol_list;


@@ -66,6 +67,19 @@ static int lastprec;
 
 static bucket *errtoken;
 static bucket *undeftoken;
 
 static bucket *errtoken;
 static bucket *undeftoken;

+
+
+static symbol_list *
+symbol_list_new (bucket *sym)
+{
+  symbol_list *res = XMALLOC (symbol_list, 1);
+  res->next = NULL;
+  res->sym = sym;
+  res->line = lineno;
+  res->ruleprec = NULL;
+  return res;
+}
+

 \f
 
 /*===================\
 \f
 
 /*===================\


@@ -123,7 +137,7 @@ read_signed_integer (FILE *stream)
 `--------------------------------------------------------------*/
 
 static char *
 `--------------------------------------------------------------*/
 
 static char *

-get_type_name (int n, symbol_list * rule)
+get_type_name (int n, symbol_list *rule)

 {
   int i;
   symbol_list *rp;
 {
   int i;
   symbol_list *rp;


@@ -219,25 +233,22 @@ copy_identifier (FILE *fin, struct obstack *oout)
   ungetc (c, fin);
 }
 
   ungetc (c, fin);
 }
 

-/*-----------------------------------------------------------------.
-| Dump the wannabee comment from IN to OUT1 and OUT2 (which can be |
-| NULL).  In fact we just saw a `/', which might or might not be a |
-| comment.  In any case, copy what we saw.                         |
-|                                                                  |
-| OUT2 might be NULL.                                              |
-`-----------------------------------------------------------------*/
+
+/*------------------------------------------------------------------.
+| Dump the wannabee comment from IN to OOUT.  In fact we just saw a |
+| `/', which might or might not be a comment.  In any case, copy    |
+| what we saw.                                                      |
+`------------------------------------------------------------------*/

 
 static inline void
 
 static inline void

-copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2)
+copy_comment (FILE *fin, struct obstack *oout)

 {
   int cplus_comment;
   int ended;
   int c;
 
   /* We read a `/', output it. */
 {
   int cplus_comment;
   int ended;
   int c;
 
   /* We read a `/', output it. */

-  obstack_1grow (oout1, '/');
-  if (oout2)
-    obstack_1grow (oout2, '/');
+  obstack_1grow (oout, '/');

 
   switch ((c = getc (fin)))
     {
 
   switch ((c = getc (fin)))
     {


@@ -252,9 +263,7 @@ copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2)
       return;
     }
 
       return;
     }
 

-  obstack_1grow (oout1, c);
-  if (oout2)
-    obstack_1grow (oout2, c);
+  obstack_1grow (oout, c);

   c = getc (fin);
 
   ended = 0;
   c = getc (fin);
 
   ended = 0;


@@ -264,26 +273,20 @@ copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2)
        {
          while (c == '*')
            {
        {
          while (c == '*')
            {

-             obstack_1grow (oout1, c);
-             if (oout2)
-               obstack_1grow (oout2, c);
+             obstack_1grow (oout, c);

              c = getc (fin);
            }
 
          if (c == '/')
            {
              c = getc (fin);
            }
 
          if (c == '/')
            {

-             obstack_1grow (oout1, c);
-             if (oout2)
-               obstack_1grow (oout2, c);
+             obstack_1grow (oout, c);

              ended = 1;
            }
        }
       else if (c == '\n')
        {
          lineno++;
              ended = 1;
            }
        }
       else if (c == '\n')
        {
          lineno++;

-         obstack_1grow (oout1, c);
-         if (oout2)
-           obstack_1grow (oout2, c);
+         obstack_1grow (oout, c);

          if (cplus_comment)
            ended = 1;
          else
          if (cplus_comment)
            ended = 1;
          else


@@ -293,27 +296,13 @@ copy_comment2 (FILE *fin, struct obstack *oout1, struct obstack *oout2)
        fatal (_("unterminated comment"));
       else
        {
        fatal (_("unterminated comment"));
       else
        {

-         obstack_1grow (oout1, c);
-         if (oout2)
-           obstack_1grow (oout2, c);
+         obstack_1grow (oout, c);

          c = getc (fin);
        }
     }
 }
 
 
          c = getc (fin);
        }
     }
 }
 
 

-/*-------------------------------------------------------------------.
-| Dump the comment (actually the current string starting with a `/') |
-| from FIN to OOUT.                                                  |
-`-------------------------------------------------------------------*/
-
-static inline void
-copy_comment (FILE *fin, struct obstack *oout)
-{
-  copy_comment2 (fin, oout, NULL);
-}
-
-

 /*-----------------------------------------------------------------.
 | FIN is pointing to a location (i.e., a `@').  Output to OOUT a   |
 | reference to this location. STACK_OFFSET is the number of values |
 /*-----------------------------------------------------------------.
 | FIN is pointing to a location (i.e., a `@').  Output to OOUT a   |
 | reference to this location. STACK_OFFSET is the number of values |


@@ -426,14 +415,12 @@ copy_definition (void)
   /* -1 while reading a character if prev char was %. */
   int after_percent;
 
   /* -1 while reading a character if prev char was %. */
   int after_percent;
 

-#if 0

   if (!no_lines_flag)
     {
       obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
                      lineno, quotearg_style (c_quoting_style,
                                              muscle_find("filename")));
     }
   if (!no_lines_flag)
     {
       obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
                      lineno, quotearg_style (c_quoting_style,
                                              muscle_find("filename")));
     }

-#endif

 
   after_percent = 0;
 
 
   after_percent = 0;
 


@@ -568,7 +555,8 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not)
       else
        {
          complain (_("`%s' is invalid in %s"),
       else
        {
          complain (_("`%s' is invalid in %s"),

-                   token_buffer, (what_is == token_sym) ? "%token" : "%nterm");
+                   token_buffer,
+                   (what_is == token_sym) ? "%token" : "%nterm");

          skip_to_char ('%');
        }
     }
          skip_to_char ('%');
        }
     }


@@ -724,7 +712,6 @@ token_buffer);
        }
 
       prev = t;
        }
 
       prev = t;

-

     }
 }
 
     }
 }
 


@@ -741,32 +728,23 @@ parse_union_decl (void)
 {
   int c;
   int count = 0;
 {
   int c;
   int count = 0;

+  bool done = FALSE;

   struct obstack union_obstack;
   struct obstack union_obstack;

-

   if (typed)
     complain (_("multiple %s declarations"), "%union");
 
   typed = 1;
 
   if (typed)
     complain (_("multiple %s declarations"), "%union");
 
   typed = 1;
 

-  if (no_lines_flag)
-    obstack_1grow (&attrs_obstack, '\n');
-

   obstack_init (&union_obstack);
   obstack_sgrow (&union_obstack, "union");
   obstack_init (&union_obstack);
   obstack_sgrow (&union_obstack, "union");

-  if (defines_flag)
-    obstack_sgrow (&defines_obstack, "typedef union");
-
-  c = getc (finput);

 
 

-  while (c != EOF)
+  while (!done)

     {
     {

+      c = xgetc (finput);
+

       /* If C contains '/', it is output by copy_comment ().  */
       if (c != '/')
       /* If C contains '/', it is output by copy_comment ().  */
       if (c != '/')

-       {
-         obstack_1grow (&attrs_obstack, c);
-         if (defines_flag)
-           obstack_1grow (&defines_obstack, c);
-       }
+       obstack_1grow (&union_obstack, c);

 
       switch (c)
        {
 
       switch (c)
        {


@@ -775,7 +753,7 @@ parse_union_decl (void)
          break;
 
        case '/':
          break;
 
        case '/':

-         copy_comment2 (finput, &defines_obstack, &union_obstack);
+         copy_comment (finput, &union_obstack);

          break;
 
        case '{':
          break;
 
        case '{':


@@ -783,26 +761,31 @@ parse_union_decl (void)
          break;
 
        case '}':
          break;
 
        case '}':

+         /* FIXME: Errr.  How could this happen???. --akim */

          if (count == 0)
            complain (_("unmatched %s"), "`}'");
          count--;
          if (count == 0)
            complain (_("unmatched %s"), "`}'");
          count--;

-         if (count <= 0)
-           {
-             if (defines_flag)
-               obstack_sgrow (&defines_obstack, " YYSTYPE;\n");
-             /* JF don't choke on trailing semi */
-             c = skip_white_space ();
-             if (c != ';')
-               ungetc (c, finput);
-             obstack_1grow (&union_obstack, 0);
-             muscle_insert ("stype", obstack_finish (&union_obstack));
-             return;
-           }
+         if (!count)
+           done = TRUE;
+         break;

        }
        }

-
-      c = getc (finput);

     }
 
     }
 

+  /* JF don't choke on trailing semi */
+  c = skip_white_space ();
+  if (c != ';')
+    ungetc (c, finput);
+  obstack_1grow (&union_obstack, 0);
+  muscle_insert ("stype", obstack_finish (&union_obstack));
+
+  if (defines_flag)
+    obstack_fgrow1 (&defines_obstack, "\
+#ifndef YYSTYPE\n\
+typedef %s
+yystype;\n\
+# define YYSTYPE yystype\n\
+#endif\n",
+                   muscle_find ("stype"));

 }
 
 
 }
 
 


@@ -850,7 +833,7 @@ parse_thong_decl (void)
   token_t token;
   struct bucket *symbol;
   char *typename = 0;
   token_t token;
   struct bucket *symbol;
   char *typename = 0;

-  int usrtoknum;
+  int usrtoknum = SUNDEF;

 
   token = lex ();              /* fetch typename or first token */
   if (token == tok_typename)
 
   token = lex ();              /* fetch typename or first token */
   if (token == tok_typename)


@@ -881,8 +864,6 @@ parse_thong_decl (void)
       usrtoknum = numval;
       token = lex ();          /* okay, did number, now get literal */
     }
       usrtoknum = numval;
       token = lex ();          /* okay, did number, now get literal */
     }

-  else
-    usrtoknum = 0;

 
   /* process literal string token */
 
 
   /* process literal string token */
 


@@ -944,14 +925,55 @@ parse_muscle_decl (void)
 }
 
 
 }
 
 

+
+/*---------------------------------.
+| Parse a double quoted parameter. |
+`---------------------------------*/
+
+static const char *
+parse_dquoted_param (const char *from)
+{
+  struct obstack param_obstack;
+  const char *param = NULL;
+  int c;
+
+  obstack_init (&param_obstack);
+  c = skip_white_space ();
+
+  if (c != '"')
+    {
+      complain (_("invalid %s declaration"), from);
+      ungetc (c, finput);
+      skip_to_char ('%');
+      return NULL;
+    }
+
+  while ((c = literalchar ()) != '"')
+    obstack_1grow (&param_obstack, c);
+
+  obstack_1grow (&param_obstack, '\0');
+  param = obstack_finish (&param_obstack);
+
+  if (c != '"' || strlen (param) == 0)
+    {
+      complain (_("invalid %s declaration"), from);
+      if (c != '"')
+       ungetc (c, finput);
+      skip_to_char ('%');
+      return NULL;
+    }
+
+  return param;
+}
+

 /*----------------------------------.
 | Parse what comes after %skeleton. |
 `----------------------------------*/
 
 /*----------------------------------.
 | Parse what comes after %skeleton. |
 `----------------------------------*/
 

-void
+static void

 parse_skel_decl (void)
 {
 parse_skel_decl (void)
 {

-  /* Complete with parse_dquoted_param () on the CVS branch 1.29.  */
+  skeleton = parse_dquoted_param ("%skeleton");

 }
 
 /*----------------------------------------------------------------.
 }
 
 /*----------------------------------------------------------------.


@@ -963,16 +985,13 @@ parse_skel_decl (void)
 static void
 read_declarations (void)
 {
 static void
 read_declarations (void)
 {

-  int c;
-  int tok;
-

   for (;;)
     {
   for (;;)
     {

-      c = skip_white_space ();
+      int c = skip_white_space ();

 
       if (c == '%')
        {
 
       if (c == '%')
        {

-         tok = parse_percent_token ();
+         token_t tok = parse_percent_token ();

 
          switch (tok)
            {
 
          switch (tok)
            {


@@ -1034,6 +1053,13 @@ read_declarations (void)
            case tok_noop:
              break;
 
            case tok_noop:
              break;
 

+           case tok_stropt:
+           case tok_intopt:
+           case tok_obsolete:
+             abort ();
+             break;
+
+           case tok_illegal:

            default:
              complain (_("unrecognized: %s"), token_buffer);
              skip_to_char ('%');
            default:
              complain (_("unrecognized: %s"), token_buffer);
              skip_to_char ('%');


@@ -1063,22 +1089,11 @@ copy_action (symbol_list *rule, int stack_offset)
 {
   int c;
   int count;
 {
   int c;
   int count;

-  char buf[4096];

 
   /* offset is always 0 if parser has already popped the stack pointer */
   if (semantic_parser)
     stack_offset = 0;
 
 
   /* offset is always 0 if parser has already popped the stack pointer */
   if (semantic_parser)
     stack_offset = 0;
 

-  obstack_fgrow1 (&action_obstack, "\ncase %d:\n", nrules);
-
-  if (!no_lines_flag)
-    {
-      obstack_fgrow2 (&action_obstack, muscle_find ("linef"),
-                     lineno, quotearg_style (c_quoting_style,
-                                             muscle_find ("filename")));
-    }
-  obstack_1grow (&action_obstack, '{');
-

   count = 1;
   c = getc (finput);
 
   count = 1;
   c = getc (finput);
 


@@ -1136,7 +1151,9 @@ copy_action (symbol_list *rule, int stack_offset)
        }
     }
 
        }
     }
 

-  obstack_sgrow (&action_obstack, ";\n    break;}");
+  obstack_1grow (&action_obstack, '\0');
+  rule->action = obstack_finish (&action_obstack);
+  rule->action_line = lineno;

 }
 \f
 /*-------------------------------------------------------------------.
 }
 \f
 /*-------------------------------------------------------------------.


@@ -1239,20 +1256,6 @@ copy_guard (symbol_list *rule, int stack_offset)
 }
 \f
 
 }
 \f
 

-static void
-record_rule_line (void)
-{
-  /* Record each rule's source line number in rline table.  */
-
-  if (nrules >= rline_allocated)
-    {
-      rline_allocated = nrules * 2;
-      rline = XREALLOC (rline, short, rline_allocated);
-    }
-  rline[nrules] = lineno;
-}
-
-

 /*-------------------------------------------------------------------.
 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
 | with the user's names.                                             |
 /*-------------------------------------------------------------------.
 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
 | with the user's names.                                             |


@@ -1274,337 +1277,246 @@ gensym (void)
   sym->value = nvars++;
   return sym;
 }
   sym->value = nvars++;
   return sym;
 }

-
-#if 0
-/*------------------------------------------------------------------.
-| read in a %type declaration and record its information for        |
-| get_type_name to access.  This is unused.  It is only called from |
-| the #if 0 part of readgram                                        |
-`------------------------------------------------------------------*/
-
-static int
-get_type (void)
-{
-  int k;
-  token_t token;
-  char *name;
-
-  token = lex ();
-
-  if (token != tok_typename)
-    {
-      complain (_("invalid %s declaration"), "%type");
-      return t;
-    }
-
-  name = xstrdup (token_buffer);
-
-  for (;;)
-    {
-      token = lex ();
-
-      switch (token)
-       {
-       case tok_semicolon:
-         return lex ();
-
-       case tok_comma:
-         break;
-
-       case tok_identifier:
-         if (symval->type_name == NULL)
-           symval->type_name = name;
-         else if (strcmp (name, symval->type_name) != 0)
-           complain (_("type redeclaration for %s"), symval->tag);
-
-         break;
-
-       default:
-         return token;
-       }
-    }
-}
-
-#endif

 \f
 \f

-/*------------------------------------------------------------------.
-| Parse the input grammar into a one symbol_list structure.  Each   |
-| rule is represented by a sequence of symbols: the left hand side  |
-| followed by the contents of the right hand side, followed by a    |
-| null pointer instead of a symbol to terminate the rule.  The next |
-| symbol is the lhs of the following rule.                          |
-|                                                                   |
-| All guards and actions are copied out to the appropriate files,   |
-| labelled by the rule number they apply to.                        |
-`------------------------------------------------------------------*/
+/*-------------------------------------------------------------------.
+| Parse the input grammar into a one symbol_list structure.  Each    |
+| rule is represented by a sequence of symbols: the left hand side   |
+| followed by the contents of the right hand side, followed by a     |
+| null pointer instead of a symbol to terminate the rule.  The next  |
+| symbol is the lhs of the following rule.                           |
+|                                                                    |
+| All guards and actions are copied out to the appropriate files,    |
+| labelled by the rule number they apply to.                         |
+|                                                                    |
+| Bison used to allow some %directives in the rules sections, but    |
+| this is no longer consider appropriate: (i) the documented grammar |
+| doesn't claim it, (ii), it would promote bad style, (iii), error   |
+| recovery for %directives consists in skipping the junk until a `%' |
+| is seen and helrp synchronizing.  This scheme is definitely wrong  |
+| in the rules section.                                              |
+`-------------------------------------------------------------------*/

 
 static void
 readgram (void)
 {
   token_t t;
   bucket *lhs = NULL;
 
 static void
 readgram (void)
 {
   token_t t;
   bucket *lhs = NULL;

-  symbol_list *p;
-  symbol_list *p1;
+  symbol_list *p = NULL;
+  symbol_list *p1 = NULL;

   bucket *bp;
 
   /* Points to first symbol_list of current rule. its symbol is the
      lhs of the rule.  */
   bucket *bp;
 
   /* Points to first symbol_list of current rule. its symbol is the
      lhs of the rule.  */

-  symbol_list *crule;
+  symbol_list *crule = NULL;

   /* Points to the symbol_list preceding crule.  */
   /* Points to the symbol_list preceding crule.  */

-  symbol_list *crule1;
-
-  p1 = NULL;
+  symbol_list *crule1 = NULL;

 
   t = lex ();
 
   while (t != tok_two_percents && t != tok_eof)
 
   t = lex ();
 
   while (t != tok_two_percents && t != tok_eof)

-    {
-      if (t == tok_identifier || t == tok_bar)
-       {
-         int action_flag = 0;
-         /* Number of symbols in rhs of this rule so far */
-         int rulelength = 0;
-         int xactions = 0;     /* JF for error checking */
-         bucket *first_rhs = 0;
+    if (t == tok_identifier || t == tok_bar)
+      {
+       int action_flag = 0;
+       /* Number of symbols in rhs of this rule so far */
+       int rulelength = 0;
+       int xactions = 0;       /* JF for error checking */
+       bucket *first_rhs = 0;

 
 

-         if (t == tok_identifier)
-           {
-             lhs = symval;
-
-             if (!start_flag)
-               {
-                 startval = lhs;
-                 start_flag = 1;
-               }
-
-             t = lex ();
-             if (t != tok_colon)
-               {
-                 complain (_("ill-formed rule: initial symbol not followed by colon"));
-                 unlex (t);
-               }
-           }
+       if (t == tok_identifier)
+         {
+           lhs = symval;

 
 

-         if (nrules == 0 && t == tok_bar)
-           {
-             complain (_("grammar starts with vertical bar"));
-             lhs = symval;     /* BOGUS: use a random symval */
-           }
-         /* start a new rule and record its lhs.  */
+           if (!start_flag)
+             {
+               startval = lhs;
+               start_flag = 1;
+             }

 
 

-         nrules++;
-         nitems++;
+           t = lex ();
+           if (t != tok_colon)
+             {
+               complain (_("ill-formed rule: initial symbol not followed by colon"));
+               unlex (t);
+             }
+         }

 
 

-         record_rule_line ();
+       if (nrules == 0 && t == tok_bar)
+         {
+           complain (_("grammar starts with vertical bar"));
+           lhs = symval;       /* BOGUS: use a random symval */
+         }
+       /* start a new rule and record its lhs.  */

 
 

-         p = XCALLOC (symbol_list, 1);
-         p->sym = lhs;
+       nrules++;
+       nitems++;

 
 

-         crule1 = p1;
-         if (p1)
-           p1->next = p;
-         else
-           grammar = p;
+       p = symbol_list_new (lhs);

 
 

-         p1 = p;
-         crule = p;
+       crule1 = p1;
+       if (p1)
+         p1->next = p;
+       else
+         grammar = p;

 
 

-         /* mark the rule's lhs as a nonterminal if not already so.  */
+       p1 = p;
+       crule = p;

 
 

-         if (lhs->class == unknown_sym)
-           {
-             lhs->class = nterm_sym;
-             lhs->value = nvars;
-             nvars++;
-           }
-         else if (lhs->class == token_sym)
-           complain (_("rule given for %s, which is a token"), lhs->tag);
+       /* mark the rule's lhs as a nonterminal if not already so.  */

 
 

-         /* read the rhs of the rule.  */
+       if (lhs->class == unknown_sym)
+         {
+           lhs->class = nterm_sym;
+           lhs->value = nvars;
+           nvars++;
+         }
+       else if (lhs->class == token_sym)
+         complain (_("rule given for %s, which is a token"), lhs->tag);

 
 

-         for (;;)
-           {
-             t = lex ();
-             if (t == tok_prec)
-               {
-                 t = lex ();
-                 crule->ruleprec = symval;
-                 t = lex ();
-               }
-
-             if (!(t == tok_identifier || t == tok_left_curly))
-               break;
-
-             /* If next token is an identifier, see if a colon follows it.
-                If one does, exit this rule now.  */
-             if (t == tok_identifier)
-               {
-                 bucket *ssave;
-                 token_t t1;
-
-                 ssave = symval;
-                 t1 = lex ();
-                 unlex (t1);
-                 symval = ssave;
-                 if (t1 == tok_colon)
-                   break;
-
-                 if (!first_rhs)       /* JF */
-                   first_rhs = symval;
-                 /* Not followed by colon =>
-                    process as part of this rule's rhs.  */
-               }
-
-             /* If we just passed an action, that action was in the middle
-                of a rule, so make a dummy rule to reduce it to a
-                non-terminal.  */
-             if (action_flag)
-               {
-                 bucket *sdummy;
-
-                 /* Since the action was written out with this rule's
-                    number, we must give the new rule this number by
-                    inserting the new rule before it.  */
-
-                 /* Make a dummy nonterminal, a gensym.  */
-                 sdummy = gensym ();
-
-                 /* Make a new rule, whose body is empty,
-                    before the current one, so that the action
-                    just read can belong to it.  */
-                 nrules++;
-                 nitems++;
-                 record_rule_line ();
-                 p = XCALLOC (symbol_list, 1);
-                 if (crule1)
-                   crule1->next = p;
-                 else
-                   grammar = p;
-                 p->sym = sdummy;
-                 crule1 = XCALLOC (symbol_list, 1);
-                 p->next = crule1;
-                 crule1->next = crule;
-
-                 /* Insert the dummy generated by that rule into this
-                    rule.  */
-                 nitems++;
-                 p = XCALLOC (symbol_list, 1);
-                 p->sym = sdummy;
-                 p1->next = p;
-                 p1 = p;
-
-                 action_flag = 0;
-               }
-
-             if (t == tok_identifier)
-               {
-                 nitems++;
-                 p = XCALLOC (symbol_list, 1);
-                 p->sym = symval;
-                 p1->next = p;
-                 p1 = p;
-               }
-             else              /* handle an action.  */
-               {
-                 copy_action (crule, rulelength);
-                 action_flag = 1;
-                 xactions++;   /* JF */
-               }
-             rulelength++;
-           }                   /* end of  read rhs of rule */
-
-         /* Put an empty link in the list to mark the end of this rule  */
-         p = XCALLOC (symbol_list, 1);
-         p1->next = p;
-         p1 = p;
+       /* read the rhs of the rule.  */

 
 

-         if (t == tok_prec)
-           {
-             complain (_("two @prec's in a row"));
-             t = lex ();
-             crule->ruleprec = symval;
-             t = lex ();
-           }
-         if (t == tok_guard)
-           {
-             if (!semantic_parser)
-               complain (_("%%guard present but %%semantic_parser not specified"));
+       for (;;)
+         {
+           t = lex ();
+           if (t == tok_prec)
+             {
+               t = lex ();
+               crule->ruleprec = symval;
+               t = lex ();
+             }
+
+           if (!(t == tok_identifier || t == tok_left_curly))
+             break;

 
 

-             copy_guard (crule, rulelength);
-             t = lex ();
-           }
-         else if (t == tok_left_curly)
-           {
-             /* This case never occurs -wjh */
-             if (action_flag)
-               complain (_("two actions at end of one rule"));
-             copy_action (crule, rulelength);
-             action_flag = 1;
-             xactions++;       /* -wjh */
-             t = lex ();
-           }
-         /* If $$ is being set in default way, report if any type
-            mismatch.  */
-         else if (!xactions
-                  && first_rhs && lhs->type_name != first_rhs->type_name)
-           {
-             if (lhs->type_name == 0
-                 || first_rhs->type_name == 0
-                 || strcmp (lhs->type_name, first_rhs->type_name))
-               complain (_("type clash (`%s' `%s') on default action"),
-                         lhs->type_name ? lhs->type_name : "",
-                         first_rhs->type_name ? first_rhs->type_name : "");
-           }
-         /* Warn if there is no default for $$ but we need one.  */
-         else if (!xactions && !first_rhs && lhs->type_name != 0)
-           complain (_("empty rule for typed nonterminal, and no action"));
-         if (t == tok_semicolon)
+           /* If next token is an identifier, see if a colon follows it.
+              If one does, exit this rule now.  */
+           if (t == tok_identifier)
+             {
+               bucket *ssave;
+               token_t t1;
+
+               ssave = symval;
+               t1 = lex ();
+               unlex (t1);
+               symval = ssave;
+               if (t1 == tok_colon)
+                 break;
+
+               if (!first_rhs) /* JF */
+                 first_rhs = symval;
+               /* Not followed by colon =>
+                  process as part of this rule's rhs.  */
+             }
+
+           /* If we just passed an action, that action was in the middle
+              of a rule, so make a dummy rule to reduce it to a
+              non-terminal.  */
+           if (action_flag)
+             {
+               /* Since the action was written out with this rule's
+                  number, we must give the new rule this number by
+                  inserting the new rule before it.  */
+
+               /* Make a dummy nonterminal, a gensym.  */
+               bucket *sdummy = gensym ();
+
+               /* Make a new rule, whose body is empty, before the
+                  current one, so that the action just read can
+                  belong to it.  */
+               nrules++;
+               nitems++;
+               p = symbol_list_new (sdummy);
+               /* Attach its lineno to that of the host rule. */
+               p->line = crule->line;
+               if (crule1)
+                 crule1->next = p;
+               else
+                 grammar = p;
+               /* End of the rule. */
+               crule1 = symbol_list_new (NULL);
+               crule1->next = crule;
+
+               p->next = crule1;
+
+               /* Insert the dummy generated by that rule into this
+                  rule.  */
+               nitems++;
+               p = symbol_list_new (sdummy);
+               p1->next = p;
+               p1 = p;
+
+               action_flag = 0;
+             }
+
+           if (t == tok_identifier)
+             {
+               nitems++;
+               p = symbol_list_new (symval);
+               p1->next = p;
+               p1 = p;
+             }
+           else                /* handle an action.  */
+             {
+               copy_action (crule, rulelength);
+               action_flag = 1;
+               xactions++;     /* JF */
+             }
+           rulelength++;
+         }                     /* end of  read rhs of rule */
+
+       /* Put an empty link in the list to mark the end of this rule  */
+       p = symbol_list_new (NULL);
+       p1->next = p;
+       p1 = p;
+
+       if (t == tok_prec)
+         {
+           complain (_("two @prec's in a row"));

            t = lex ();
            t = lex ();

-       }
-#if 0
-      /* these things can appear as alternatives to rules.  */
-/* NO, they cannot.
-       a) none of the documentation allows them
-       b) most of them scan forward until finding a next %
-               thus they may swallow lots of intervening rules
-*/
-      else if (t == tok_token)
-       {
-         parse_token_decl (token_sym, nterm_sym);
-         t = lex ();
-       }
-      else if (t == tok_nterm)
-       {
-         parse_token_decl (nterm_sym, token_sym);
-         t = lex ();
-       }
-      else if (t == tok_type)
-       {
-         t = get_type ();
-       }
-      else if (t == tok_union)
-       {
-         parse_union_decl ();
-         t = lex ();
-       }
-      else if (t == tok_expect)
-       {
-         parse_expect_decl ();
-         t = lex ();
-       }
-      else if (t == tok_start)
-       {
-         parse_start_decl ();
-         t = lex ();
-       }
-#endif
+           crule->ruleprec = symval;
+           t = lex ();
+         }
+       if (t == tok_guard)
+         {
+           if (!semantic_parser)
+             complain (_("%%guard present but %%semantic_parser not specified"));

 
 

-      else
-       {
-         complain (_("invalid input: %s"), quote (token_buffer));
+           copy_guard (crule, rulelength);
+           t = lex ();
+         }
+       else if (t == tok_left_curly)
+         {
+           /* This case never occurs -wjh */
+           if (action_flag)
+             complain (_("two actions at end of one rule"));
+           copy_action (crule, rulelength);
+           action_flag = 1;
+           xactions++; /* -wjh */
+           t = lex ();
+         }
+       /* If $$ is being set in default way, report if any type
+          mismatch.  */
+       else if (!xactions
+                && first_rhs && lhs->type_name != first_rhs->type_name)
+         {
+           if (lhs->type_name == 0
+               || first_rhs->type_name == 0
+               || strcmp (lhs->type_name, first_rhs->type_name))
+             complain (_("type clash (`%s' `%s') on default action"),
+                       lhs->type_name ? lhs->type_name : "",
+                       first_rhs->type_name ? first_rhs->type_name : "");
+         }
+       /* Warn if there is no default for $$ but we need one.  */
+       else if (!xactions && !first_rhs && lhs->type_name != 0)
+         complain (_("empty rule for typed nonterminal, and no action"));
+       if (t == tok_semicolon)

          t = lex ();
          t = lex ();

-       }
-    }
+      }
+    else
+      {
+       complain (_("invalid input: %s"), quote (token_buffer));
+       t = lex ();
+      }
+

 
   /* grammar has been read.  Do some checking */
 
 
   /* grammar has been read.  Do some checking */
 


@@ -1640,6 +1552,13 @@ read_additionnal_code (void)
 
   obstack_init (&el_obstack);
 
 
   obstack_init (&el_obstack);
 

+  if (!no_lines_flag)
+    {
+      obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
+                     lineno, quotearg_style (c_quoting_style,
+                                             muscle_find("filename")));
+    }
+

   while ((c = getc (finput)) != EOF)
     obstack_1grow (&el_obstack, c);
 
   while ((c = getc (finput)) != EOF)
     obstack_1grow (&el_obstack, c);
 


@@ -1697,6 +1616,46 @@ output_token_defines (struct obstack *oout)
 }
 
 
 }
 
 

+/*------------------------------------------------------------------.
+| Set TOKEN_TRANSLATIONS.  Check that no two symbols share the same |
+| number.                                                           |
+`------------------------------------------------------------------*/
+
+static void
+token_translations_init (void)
+{
+  bucket *bp = NULL;
+  int i;
+
+  token_translations = XCALLOC (short, max_user_token_number + 1);
+
+  /* Initialize all entries for literal tokens to 2, the internal
+     token number for $undefined., which represents all invalid
+     inputs.  */
+  for (i = 0; i <= max_user_token_number; i++)
+    token_translations[i] = 2;
+
+  for (bp = firstsymbol; bp; bp = bp->next)
+    {
+      /* Non-terminal? */
+      if (bp->value >= ntokens)
+       continue;
+      /* A token string alias? */
+      if (bp->user_token_number == SALIAS)
+       continue;
+
+      assert (bp->user_token_number != SUNDEF);
+
+      /* A token which translation has already been set? */
+      if (token_translations[bp->user_token_number] != 2)
+       complain (_("tokens %s and %s both assigned number %d"),
+                 tags[token_translations[bp->user_token_number]],
+                 bp->tag, bp->user_token_number);
+      token_translations[bp->user_token_number] = bp->value;
+    }
+}
+
+

 /*------------------------------------------------------------------.
 | Assign symbol numbers, and write definition of token names into   |
 | FDEFINES.  Set up vectors TAGS and SPREC of names and precedences |
 /*------------------------------------------------------------------.
 | Assign symbol numbers, and write definition of token names into   |
 | FDEFINES.  Set up vectors TAGS and SPREC of names and precedences |


@@ -1708,18 +1667,19 @@ packsymbols (void)
 {
   bucket *bp = NULL;
   int tokno = 1;
 {
   bucket *bp = NULL;
   int tokno = 1;

-  int i, j;

   int last_user_token_number;
   static char DOLLAR[] = "$";
 
   tags = XCALLOC (char *, nsyms + 1);
   int last_user_token_number;
   static char DOLLAR[] = "$";
 
   tags = XCALLOC (char *, nsyms + 1);

-  tags[0] = DOLLAR;

   user_toknums = XCALLOC (short, nsyms + 1);
   user_toknums = XCALLOC (short, nsyms + 1);

-  user_toknums[0] = 0;

 
   sprec = XCALLOC (short, nsyms);
   sassoc = XCALLOC (short, nsyms);
 
 
   sprec = XCALLOC (short, nsyms);
   sassoc = XCALLOC (short, nsyms);
 

+  /* The EOF token. */
+  tags[0] = DOLLAR;
+  user_toknums[0] = 0;
+

   max_user_token_number = 256;
   last_user_token_number = 256;
 
   max_user_token_number = 256;
   last_user_token_number = 256;
 


@@ -1772,7 +1732,7 @@ packsymbols (void)
 
       if (bp->class == token_sym)
        {
 
       if (bp->class == token_sym)
        {

-         if (!bp->user_token_number)
+         if (bp->user_token_number == SUNDEF)

            bp->user_token_number = ++last_user_token_number;
          if (bp->user_token_number > max_user_token_number)
            max_user_token_number = bp->user_token_number;
            bp->user_token_number = ++last_user_token_number;
          if (bp->user_token_number > max_user_token_number)
            max_user_token_number = bp->user_token_number;


@@ -1784,29 +1744,26 @@ packsymbols (void)
       sassoc[bp->value] = bp->assoc;
     }
 
       sassoc[bp->value] = bp->assoc;
     }
 

-  token_translations = XCALLOC (short, max_user_token_number + 1);
+  token_translations_init ();

 
 

-  /* initialize all entries for literal tokens to 2, the internal
-     token number for $undefined., which represents all invalid
-     inputs.  */
-  for (j = 0; j <= max_user_token_number; j++)
-    token_translations[j] = 2;
+  error_token_number = errtoken->value;

 
 

-  for (bp = firstsymbol; bp; bp = bp->next)
-    {
-      if (bp->value >= ntokens)
-       continue;               /* non-terminal */
-      if (bp->user_token_number == SALIAS)
-       continue;
-      if (token_translations[bp->user_token_number] != 2)
-       complain (_("tokens %s and %s both assigned number %d"),
-                 tags[token_translations[bp->user_token_number]],
-                 bp->tag, bp->user_token_number);
-      token_translations[bp->user_token_number] = bp->value;
-    }
+  if (startval->class == unknown_sym)
+    fatal (_("the start symbol %s is undefined"), startval->tag);
+  else if (startval->class == token_sym)
+    fatal (_("the start symbol %s is a token"), startval->tag);

 
 

-  error_token_number = errtoken->value;
+  start_symbol = startval->value;
+}
+
+
+/*-----------------------------------.
+| Output definition of token names.  |
+`-----------------------------------*/

 
 

+static void
+symbols_output (void)
+{

   {
     struct obstack tokendefs;
     obstack_init (&tokendefs);
   {
     struct obstack tokendefs;
     obstack_init (&tokendefs);


@@ -1816,46 +1773,31 @@ packsymbols (void)
     obstack_free (&tokendefs, NULL);
   }
 
     obstack_free (&tokendefs, NULL);
   }
 

-#if 0
-  if (!no_parser_flag)
-    output_token_defines (&table_obstack);
-#endif
-
-  if (startval->class == unknown_sym)
-    fatal (_("the start symbol %s is undefined"), startval->tag);
-  else if (startval->class == token_sym)
-    fatal (_("the start symbol %s is a token"), startval->tag);
-
-  start_symbol = startval->value;
-

   if (defines_flag)
     {
       output_token_defines (&defines_obstack);
 
       if (!pure_parser)
   if (defines_flag)
     {
       output_token_defines (&defines_obstack);
 
       if (!pure_parser)

+       obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n",
+                       spec_name_prefix);
+      if (semantic_parser)

        {
        {

-         if (spec_name_prefix)
-           obstack_fgrow1 (&defines_obstack, "\nextern YYSTYPE %slval;\n",
-                           spec_name_prefix);
-         else
-           obstack_sgrow (&defines_obstack,
-                                "\nextern YYSTYPE yylval;\n");
-       }
+         int i;

 
 

-      if (semantic_parser)
-       for (i = ntokens; i < nsyms; i++)
-         {
-           /* don't make these for dummy nonterminals made by gensym.  */
-           if (*tags[i] != '@')
-              obstack_fgrow2 (&defines_obstack,
-                              "# define\tNT%s\t%d\n", tags[i], i);
-         }
+         for (i = ntokens; i < nsyms; i++)
+           {
+             /* don't make these for dummy nonterminals made by gensym.  */
+             if (*tags[i] != '@')
+               obstack_fgrow2 (&defines_obstack,
+                               "# define\tNT%s\t%d\n", tags[i], i);
+           }

 #if 0
 #if 0

-      /* `fdefines' is now a temporary file, so we need to copy its
-         contents in `done', so we can't close it here.  */
-      fclose (fdefines);
-      fdefines = NULL;
+         /* `fdefines' is now a temporary file, so we need to copy its
+            contents in `done', so we can't close it here.  */
+         fclose (fdefines);
+         fdefines = NULL;

 #endif
 #endif

+       }

     }
 }
 
     }
 }
 


@@ -1872,14 +1814,8 @@ packgram (void)
   int ruleno;
   symbol_list *p;
 
   int ruleno;
   symbol_list *p;
 

-  bucket *ruleprec;
-

   ritem = XCALLOC (short, nitems + 1);
   ritem = XCALLOC (short, nitems + 1);

-  rlhs = XCALLOC (short, nrules) - 1;
-  rrhs = XCALLOC (short, nrules) - 1;
-  rprec = XCALLOC (short, nrules) - 1;
-  rprecsym = XCALLOC (short, nrules) - 1;
-  rassoc = XCALLOC (short, nrules) - 1;
+  rule_table = XCALLOC (rule_t, nrules) - 1;

 
   itemno = 0;
   ruleno = 1;
 
   itemno = 0;
   ruleno = 1;


@@ -1887,9 +1823,13 @@ packgram (void)
   p = grammar;
   while (p)
     {
   p = grammar;
   while (p)
     {

-      rlhs[ruleno] = p->sym->value;
-      rrhs[ruleno] = itemno;
-      ruleprec = p->ruleprec;
+      bucket *ruleprec = p->ruleprec;
+      rule_table[ruleno].lhs = p->sym->value;
+      rule_table[ruleno].rhs = itemno;
+      rule_table[ruleno].line = p->line;
+      rule_table[ruleno].useful = TRUE;
+      rule_table[ruleno].action = p->action;
+      rule_table[ruleno].action_line = p->action_line;

 
       p = p->next;
       while (p && p->sym)
 
       p = p->next;
       while (p && p->sym)


@@ -1899,8 +1839,8 @@ packgram (void)
             of the last token in it.  */
          if (p->sym->class == token_sym)
            {
             of the last token in it.  */
          if (p->sym->class == token_sym)
            {

-             rprec[ruleno] = p->sym->prec;
-             rassoc[ruleno] = p->sym->assoc;
+             rule_table[ruleno].prec = p->sym->prec;
+             rule_table[ruleno].assoc = p->sym->assoc;

            }
          if (p)
            p = p->next;
            }
          if (p)
            p = p->next;


@@ -1910,9 +1850,9 @@ packgram (void)
          the specified symbol's precedence replaces the default.  */
       if (ruleprec)
        {
          the specified symbol's precedence replaces the default.  */
       if (ruleprec)
        {

-         rprec[ruleno] = ruleprec->prec;
-         rassoc[ruleno] = ruleprec->assoc;
-         rprecsym[ruleno] = ruleprec->value;
+         rule_table[ruleno].prec = ruleprec->prec;
+         rule_table[ruleno].assoc = ruleprec->assoc;
+         rule_table[ruleno].precsym = ruleprec->value;

        }
 
       ritem[itemno++] = -ruleno;
        }
 
       ritem[itemno++] = -ruleno;


@@ -1923,6 +1863,9 @@ packgram (void)
     }
 
   ritem[itemno] = 0;
     }
 
   ritem[itemno] = 0;

+
+  if (trace_flag)
+    ritem_print (stderr);

 }
 \f
 /*-------------------------------------------------------------------.
 }
 \f
 /*-------------------------------------------------------------------.


@@ -1943,8 +1886,6 @@ reader (void)
   nvars = 0;
   nrules = 0;
   nitems = 0;
   nvars = 0;
   nrules = 0;
   nitems = 0;

-  rline_allocated = 10;
-  rline = XCALLOC (short, rline_allocated);

 
   typed = 0;
   lastprec = 0;
 
   typed = 0;
   lastprec = 0;


@@ -1993,4 +1934,6 @@ reader (void)
   packsymbols ();
   /* Convert the grammar into the format described in gram.h.  */
   packgram ();
   packsymbols ();
   /* Convert the grammar into the format described in gram.h.  */
   packgram ();

+  /* Output the headers. */
+  symbols_output ();

 }
 }