X-Git-Url: https://git.saurik.com/bison.git/blobdiff_plain/d945f5cd8e85c0343da52dc720d23624e1c28661..98254360e3cc22aa7a7aecaae477347ee0ecc638:/src/reader.c

diff --git a/src/reader.c b/src/reader.c
index d7d6e86d..cf3e18f3 100644
--- a/src/reader.c
+++ b/src/reader.c
@@ -40,15 +40,18 @@ 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;
+
+  /* The guard is attached to the LHS of a rule. */
+  const char *guard;
+  int guard_line;
   bucket *ruleprec;
 } symbol_list;
 
 int lineno;
-char **tags;
-short *user_toknums;
 static symbol_list *grammar;
 static int start_flag;
 static bucket *startval;
@@ -63,9 +66,10 @@ static int typed;
 /* Incremented for each %left, %right or %nonassoc seen */
 static int lastprec;
 
-static bucket *errtoken;
-static bucket *undeftoken;
-
+static bucket *errtoken = NULL;
+static bucket *undeftoken = NULL;
+static bucket *eoftoken = NULL;
+static bucket *axiom = NULL;
 
 static symbol_list *
 symbol_list_new (bucket *sym)
@@ -76,6 +80,8 @@ symbol_list_new (bucket *sym)
   res->line = lineno;
   res->action = NULL;
   res->action_line = 0;
+  res->guard = NULL;
+  res->guard_line = 0;
   res->ruleprec = NULL;
   return res;
 }
@@ -327,9 +333,16 @@ copy_at (FILE *fin, struct obstack *oout, int stack_offset)
 
       ungetc (c, fin);
       n = read_signed_integer (fin);
-
-      obstack_fgrow1 (oout, "yylsp[%d]", n - stack_offset);
-      locations_flag = 1;
+      if (n > stack_offset)
+	complain (_("invalid value: %s%d"), "@", n);
+      else
+	{
+	  /* Offset is always 0 if parser has already popped the stack
+	     pointer.  */
+	  obstack_fgrow1 (oout, "yylsp[%d]",
+			  n - (semantic_parser ? 0 : stack_offset));
+	  locations_flag = 1;
+	}
     }
   else
     {
@@ -384,16 +397,24 @@ copy_dollar (FILE *fin, struct obstack *oout,
       ungetc (c, fin);
       n = read_signed_integer (fin);
 
-      if (!type_name && n > 0)
-	type_name = get_type_name (n, rule);
-
-      obstack_fgrow1 (oout, "yyvsp[%d]", n - stack_offset);
-
-      if (type_name)
-	obstack_fgrow1 (oout, ".%s", type_name);
-      if (!type_name && typed)
-	complain (_("$%d of `%s' has no declared type"),
-		  n, rule->sym->tag);
+      if (n > stack_offset)
+	complain (_("invalid value: %s%d"), "$", n);
+      else
+	{
+	  if (!type_name && n > 0)
+	    type_name = get_type_name (n, rule);
+
+	  /* Offset is always 0 if parser has already popped the stack
+	     pointer.  */
+	  obstack_fgrow1 (oout, "yyvsp[%d]",
+			  n - (semantic_parser ? 0 : stack_offset));
+
+	  if (type_name)
+	    obstack_fgrow1 (oout, ".%s", type_name);
+	  if (!type_name && typed)
+	    complain (_("$%d of `%s' has no declared type"),
+		      n, rule->sym->tag);
+	}
     }
   else
     {
@@ -419,7 +440,7 @@ copy_definition (void)
     {
       obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
 		      lineno, quotearg_style (c_quoting_style,
-					      muscle_find("filename")));
+					      muscle_find ("filename")));
     }
 
   after_percent = 0;
@@ -551,6 +572,9 @@ parse_token_decl (symbol_class what_is, symbol_class what_is_not)
       else if (symbol && token == tok_number)
 	{
 	  symbol->user_token_number = numval;
+	  /* User defined EOF token? */
+	  if (numval == 0)
+	    eoftoken = symbol;
 	}
       else
 	{
@@ -875,12 +899,13 @@ parse_thong_decl (void)
   nsyms--;
 }
 
+
 static void
 parse_muscle_decl (void)
 {
   int ch = ungetc (skip_white_space (), finput);
-  char* muscle_key;
-  char* muscle_value;
+  char *muscle_key;
+  char *muscle_value;
 
   /* Read key. */
   if (!isalpha (ch) && ch != '_')
@@ -1073,178 +1098,89 @@ read_declarations (void)
 | matching `}' into the actions file.  STACK_OFFSET is the number of |
 | values in the current rule so far, which says where to find `$0'   |
 | with respect to the top of the stack.                              |
+|                                                                    |
+| This routine is used both for actions and guards.  Only            |
+| ACTION_OBSTACK is used, but this is fine, since we use only        |
+| pointers to relevant portions inside this obstack.                 |
 `-------------------------------------------------------------------*/
 
 static void
-copy_action (symbol_list *rule, int stack_offset)
+parse_braces (symbol_list *rule, int stack_offset)
 {
   int c;
   int count;
 
-  /* offset is always 0 if parser has already popped the stack pointer */
-  if (semantic_parser)
-    stack_offset = 0;
-
   count = 1;
-  c = getc (finput);
-
   while (count > 0)
     {
-      while (c != '}')
-	{
-	  switch (c)
-	    {
-	    case '\n':
-	      obstack_1grow (&action_obstack, c);
-	      lineno++;
-	      break;
+      while ((c = getc (finput)) != '}')
+	switch (c)
+	  {
+	  case '\n':
+	    obstack_1grow (&action_obstack, c);
+	    lineno++;
+	    break;
 
-	    case '{':
-	      obstack_1grow (&action_obstack, c);
-	      count++;
-	      break;
+	  case '{':
+	    obstack_1grow (&action_obstack, c);
+	    count++;
+	    break;
 
-	    case '\'':
-	    case '"':
-	      copy_string (finput, &action_obstack, c);
-	      break;
+	  case '\'':
+	  case '"':
+	    copy_string (finput, &action_obstack, c);
+	    break;
 
-	    case '/':
-	      copy_comment (finput, &action_obstack);
-	      break;
+	  case '/':
+	    copy_comment (finput, &action_obstack);
+	    break;
 
-	    case '$':
-	      copy_dollar (finput, &action_obstack,
-			   rule, stack_offset);
-	      break;
+	  case '$':
+	    copy_dollar (finput, &action_obstack,
+			 rule, stack_offset);
+	    break;
 
-	    case '@':
-	      copy_at (finput, &action_obstack,
-		       stack_offset);
-	      break;
+	  case '@':
+	    copy_at (finput, &action_obstack,
+		     stack_offset);
+	    break;
 
-	    case EOF:
-	      fatal (_("unmatched %s"), "`{'");
+	  case EOF:
+	    fatal (_("unmatched %s"), "`{'");
 
-	    default:
-	      obstack_1grow (&action_obstack, c);
-	    }
-
-	  c = getc (finput);
-	}
-
-      /* above loop exits when c is '}' */
+	  default:
+	    obstack_1grow (&action_obstack, c);
+	  }
 
+      /* Above loop exits when C is '}'.  */
       if (--count)
-	{
-	  obstack_1grow (&action_obstack, c);
-	  c = getc (finput);
-	}
+	obstack_1grow (&action_obstack, c);
     }
 
   obstack_1grow (&action_obstack, '\0');
-  rule->action = obstack_finish (&action_obstack);
-  rule->action_line = lineno;
 }
-
-/*-------------------------------------------------------------------.
-| After `%guard' is seen in the input file, copy the actual guard    |
-| into the guards file.  If the guard is followed by an action, copy |
-| that into the actions file.  STACK_OFFSET is the number of values  |
-| in the current rule so far, which says where to find `$0' with     |
-| respect to the top of the stack, for the simple parser in which    |
-| the stack is not popped until after the guard is run.              |
-`-------------------------------------------------------------------*/
+
 
 static void
-copy_guard (symbol_list *rule, int stack_offset)
+parse_action (symbol_list *rule, int stack_offset)
 {
-  int c;
-  int count;
-  int brace_flag = 0;
-
-  /* offset is always 0 if parser has already popped the stack pointer */
-  if (semantic_parser)
-    stack_offset = 0;
-
-  obstack_fgrow1 (&guard_obstack, "\ncase %d:\n", nrules);
-  if (!no_lines_flag)
-    obstack_fgrow2 (&guard_obstack, muscle_find ("linef"),
-		    lineno, quotearg_style (c_quoting_style,
-					    muscle_find ("filename")));
-  obstack_1grow (&guard_obstack, '{');
-
-  count = 0;
-  c = getc (finput);
-
-  while (brace_flag ? (count > 0) : (c != ';'))
-    {
-      switch (c)
-	{
-	case '\n':
-	  obstack_1grow (&guard_obstack, c);
-	  lineno++;
-	  break;
-
-	case '{':
-	  obstack_1grow (&guard_obstack, c);
-	  brace_flag = 1;
-	  count++;
-	  break;
-
-	case '}':
-	  obstack_1grow (&guard_obstack, c);
-	  if (count > 0)
-	    count--;
-	  else
-	    {
-	      complain (_("unmatched %s"), "`}'");
-	      c = getc (finput);	/* skip it */
-	    }
-	  break;
-
-	case '\'':
-	case '"':
-	  copy_string (finput, &guard_obstack, c);
-	  break;
-
-	case '/':
-	  copy_comment (finput, &guard_obstack);
-	  break;
-
-	case '$':
-	  copy_dollar (finput, &guard_obstack, rule, stack_offset);
-	  break;
-
-	case '@':
-	  copy_at (finput, &guard_obstack, stack_offset);
-	  break;
-
-	case EOF:
-	  fatal ("%s", _("unterminated %guard clause"));
-
-	default:
-	  obstack_1grow (&guard_obstack, c);
-	}
-
-      if (c != '}' || count != 0)
-	c = getc (finput);
-    }
+  rule->action_line = lineno;
+  parse_braces (rule, stack_offset);
+  rule->action = obstack_finish (&action_obstack);
+}
 
-  c = skip_white_space ();
 
-  obstack_sgrow (&guard_obstack, ";\n    break;}");
-  if (c == '{')
-    copy_action (rule, stack_offset);
-  else if (c == '=')
-    {
-      c = getc (finput);	/* why not skip_white_space -wjh */
-      if (c == '{')
-	copy_action (rule, stack_offset);
-    }
-  else
-    ungetc (c, finput);
+static void
+parse_guard (symbol_list *rule, int stack_offset)
+{
+  token_t t = lex ();
+  if (t != tok_left_curly)
+    complain (_("invalid %s declaration"), "%guard");
+  rule->guard_line = lineno;
+  parse_braces (rule, stack_offset);
+  rule->guard = obstack_finish (&action_obstack);
 }
+
 
 
 /*-------------------------------------------------------------------.
@@ -1418,6 +1354,11 @@ readgram (void)
 		p = symbol_list_new (sdummy);
 		/* Attach its lineno to that of the host rule. */
 		p->line = crule->line;
+		/* Move the action from the host rule to this one. */
+		p->action = crule->action;
+		p->action_line = crule->action_line;
+		crule->action = NULL;
+
 		if (crule1)
 		  crule1->next = p;
 		else
@@ -1447,7 +1388,7 @@ readgram (void)
 	      }
 	    else		/* handle an action.  */
 	      {
-		copy_action (crule, rulelength);
+		parse_action (crule, rulelength);
 		action_flag = 1;
 		xactions++;	/* JF */
 	      }
@@ -1466,20 +1407,22 @@ readgram (void)
 	    crule->ruleprec = symval;
 	    t = lex ();
 	  }
+
 	if (t == tok_guard)
 	  {
 	    if (!semantic_parser)
 	      complain (_("%%guard present but %%semantic_parser not specified"));
 
-	    copy_guard (crule, rulelength);
+	    parse_guard (crule, rulelength);
 	    t = lex ();
 	  }
-	else if (t == tok_left_curly)
+
+	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);
+	    parse_action (crule, rulelength);
 	    action_flag = 1;
 	    xactions++;	/* -wjh */
 	    t = lex ();
@@ -1508,12 +1451,8 @@ readgram (void)
 	t = lex ();
       }
 
-
   /* grammar has been read.  Do some checking */
 
-  if (nsyms > MAXSHORT)
-    fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
-	   MAXSHORT);
   if (nrules == 0)
     fatal (_("no rules in the input grammar"));
 
@@ -1529,6 +1468,25 @@ readgram (void)
 	bp->value = nvars++;
       }
 
+  /* Insert the initial rule, which line is that of the first rule
+     (not that of the start symbol):
+
+     axiom: %start EOF.  */
+  p = symbol_list_new (axiom);
+  p->line = grammar->line;
+  p->next = symbol_list_new (startval);
+  p->next->next = symbol_list_new (eoftoken);
+  p->next->next->next = symbol_list_new (NULL);
+  p->next->next->next->next = grammar;
+  nrules += 1;
+  nitems += 3;
+  grammar = p;
+  startval = axiom;
+
+  if (nsyms > MAXSHORT)
+    fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
+	   MAXSHORT);
+
   ntokens = nsyms - nvars;
 }
 
@@ -1547,7 +1505,7 @@ read_additionnal_code (void)
     {
       obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
 		      lineno, quotearg_style (c_quoting_style,
-					      muscle_find("filename")));
+					      muscle_find ("filename")));
     }
 
   while ((c = getc (finput)) != EOF)
@@ -1591,18 +1549,17 @@ token_translations_init (void)
       /* 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]],
+		  symbols[token_translations[bp->user_token_number]]->tag,
 		  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 |
-| of symbols.                                                       |
-`------------------------------------------------------------------*/
+/*----------------------------------------------------------------.
+| Assign symbol numbers, and write definition of token names into |
+| FDEFINES.  Set up vectors SYMBOL_TABLE, TAGS of symbols.        |
+`----------------------------------------------------------------*/
 
 static void
 packsymbols (void)
@@ -1610,17 +1567,8 @@ packsymbols (void)
   bucket *bp = NULL;
   int tokno = 1;
   int last_user_token_number;
-  static char DOLLAR[] = "$";
-
-  tags = XCALLOC (char *, nsyms + 1);
-  user_toknums = XCALLOC (short, nsyms + 1);
 
-  sprec = XCALLOC (short, nsyms);
-  sassoc = XCALLOC (short, nsyms);
-
-  /* The EOF token. */
-  tags[0] = DOLLAR;
-  user_toknums[0] = 0;
+  symbols = XCALLOC (bucket *, nsyms);
 
   max_user_token_number = 256;
   last_user_token_number = 256;
@@ -1633,11 +1581,18 @@ packsymbols (void)
 	}
       else if (bp->alias)
 	{
-	  /* this symbol and its alias are a single token defn.
-	     allocate a tokno, and assign to both check agreement of
-	     ->prec and ->assoc fields and make both the same */
-	  if (bp->value == 0)
-	    bp->value = bp->alias->value = tokno++;
+	  /* This symbol and its alias are a single token defn.
+	     Allocate a tokno, and assign to both check agreement of
+	     prec and assoc fields and make both the same */
+	  if (bp->value == -1)
+	    {
+	      if (bp == eoftoken || bp->alias == eoftoken)
+		bp->value = bp->alias->value = 0;
+	      else
+		{
+		  bp->value = bp->alias->value = tokno++;
+		}
+	    }
 
 	  if (bp->prec != bp->alias->prec)
 	    {
@@ -1663,13 +1618,17 @@ packsymbols (void)
 		bp->assoc = bp->alias->assoc;
 	    }
 
+	  /* Do not do processing below for SALIASs.  */
 	  if (bp->user_token_number == SALIAS)
-	    continue;		/* do not do processing below for SALIASs */
+	    continue;
 
 	}
-      else			/* bp->class == token_sym */
+      else /* bp->class == token_sym */
 	{
-	  bp->value = tokno++;
+	  if (bp == eoftoken)
+	    bp->value = 0;
+	  else
+	    bp->value = tokno++;
 	}
 
       if (bp->class == token_sym)
@@ -1680,10 +1639,7 @@ packsymbols (void)
 	    max_user_token_number = bp->user_token_number;
 	}
 
-      tags[bp->value] = bp->tag;
-      user_toknums[bp->value] = bp->user_token_number;
-      sprec[bp->value] = bp->prec;
-      sassoc[bp->value] = bp->assoc;
+      symbols[bp->value] = bp;
     }
 
   token_translations_init ();
@@ -1699,60 +1655,6 @@ packsymbols (void)
 }
 
 
-/*---------------------------------------------------------------.
-| Save the definition of token names in the `TOKENDEFS' muscle.  |
-`---------------------------------------------------------------*/
-
-static void
-symbols_save (void)
-{
-  struct obstack tokendefs;
-  bucket *bp;
-  char *cp, *symbol;
-  char c;
-  obstack_init (&tokendefs);
-
-  for (bp = firstsymbol; bp; bp = bp->next)
-    {
-      symbol = bp->tag;                /* get symbol */
-
-      if (bp->value >= ntokens)
-	continue;
-      if (bp->user_token_number == SALIAS)
-	continue;
-      if ('\'' == *symbol)
-	continue;               /* skip literal character */
-      if (bp == errtoken)
-	continue;               /* skip error token */
-      if ('\"' == *symbol)
-	{
-	  /* use literal string only if given a symbol with an alias */
-	  if (bp->alias)
-	    symbol = bp->alias->tag;
-	  else
-	    continue;
-	}
-
-      /* Don't #define nonliteral tokens whose names contain periods.  */
-      cp = symbol;
-      while ((c = *cp++) && c != '.');
-      if (c != '\0')
-	continue;
-
-      obstack_fgrow2 (&tokendefs, "# define %s\t%d\n",
-		      symbol, bp->user_token_number);
-      if (semantic_parser)
-	/* FIXME: This is probably wrong, and should be just as
-	   above. --akim.  */
-	obstack_fgrow2 (&tokendefs, "# define T%s\t%d\n", symbol, bp->value);
-    }
-
-  obstack_1grow (&tokendefs, 0);
-  muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
-  obstack_free (&tokendefs, NULL);
-}
-
-
 /*---------------------------------------------------------------.
 | Convert the rules into the representation using RRHS, RLHS and |
 | RITEMS.                                                        |
@@ -1765,8 +1667,12 @@ packgram (void)
   int ruleno;
   symbol_list *p;
 
+  /* We use short to index items.  */
+  if (nitems >= MAXSHORT)
+    fatal (_("too many items (max %d)"), MAXSHORT);
+
   ritem = XCALLOC (short, nitems + 1);
-  rule_table = XCALLOC (rule_t, nrules) - 1;
+  rules = XCALLOC (rule_t, nrules) - 1;
 
   itemno = 0;
   ruleno = 1;
@@ -1775,12 +1681,14 @@ packgram (void)
   while (p)
     {
       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;
+      rules[ruleno].lhs = p->sym->value;
+      rules[ruleno].rhs = itemno;
+      rules[ruleno].line = p->line;
+      rules[ruleno].useful = TRUE;
+      rules[ruleno].action = p->action;
+      rules[ruleno].action_line = p->action_line;
+      rules[ruleno].guard = p->guard;
+      rules[ruleno].guard_line = p->guard_line;
 
       p = p->next;
       while (p && p->sym)
@@ -1790,8 +1698,8 @@ packgram (void)
 	     of the last token in it.  */
 	  if (p->sym->class == token_sym)
 	    {
-	      rule_table[ruleno].prec = p->sym->prec;
-	      rule_table[ruleno].assoc = p->sym->assoc;
+	      rules[ruleno].prec = p->sym->prec;
+	      rules[ruleno].assoc = p->sym->assoc;
 	    }
 	  if (p)
 	    p = p->next;
@@ -1801,9 +1709,9 @@ packgram (void)
          the specified symbol's precedence replaces the default.  */
       if (ruleprec)
 	{
-	  rule_table[ruleno].prec = ruleprec->prec;
-	  rule_table[ruleno].assoc = ruleprec->assoc;
-	  rule_table[ruleno].precsym = ruleprec->value;
+	  rules[ruleno].prec = ruleprec->prec;
+	  rules[ruleno].assoc = ruleprec->assoc;
+	  rules[ruleno].precsym = ruleprec->value;
 	}
 
       ritem[itemno++] = -ruleno;
@@ -1814,6 +1722,8 @@ packgram (void)
     }
 
   ritem[itemno] = 0;
+  nritems = itemno;
+  assert (nritems == nitems);
 
   if (trace_flag)
     ritem_print (stderr);
@@ -1833,7 +1743,7 @@ reader (void)
   start_flag = 0;
   startval = NULL;		/* start symbol not specified yet. */
 
-  nsyms = 1;
+  nsyms = 0;
   nvars = 0;
   nrules = 0;
   nitems = 0;
@@ -1855,6 +1765,11 @@ reader (void)
   /* Initialize the symbol table.  */
   tabinit ();
 
+  /* Construct the axiom symbol. */
+  axiom = getsym ("$axiom");
+  axiom->class = nterm_sym;
+  axiom->value = nvars++;
+
   /* Construct the error token */
   errtoken = getsym ("error");
   errtoken->class = token_sym;
@@ -1869,7 +1784,6 @@ reader (void)
   /* Initialize the obstacks. */
   obstack_init (&action_obstack);
   obstack_init (&attrs_obstack);
-  obstack_init (&guard_obstack);
   obstack_init (&output_obstack);
 
   finput = xfopen (infile, "r");
@@ -1878,6 +1792,16 @@ reader (void)
      TABLE_OBSTACK and FDEFINES file.  Also notice any %token, %left,
      etc. found there.  */
   read_declarations ();
+
+  /* If the user did not define her EOFTOKEN, do it now. */
+  if (!eoftoken)
+    {
+      eoftoken = getsym ("$");
+      eoftoken->class = token_sym;
+      /* Value specified by POSIX.  */
+      eoftoken->user_token_number = 0;
+    }
+
   /* Read in the grammar, build grammar in list form.  Write out
      guards and actions.  */
   readgram ();
@@ -1890,9 +1814,10 @@ reader (void)
   /* Assign the symbols their symbol numbers.  Write #defines for the
      token symbols into FDEFINES if requested.  */
   packsymbols ();
-  /* Save them. */
-  symbols_save ();
 
   /* Convert the grammar into the format described in gram.h.  */
   packgram ();
+
+  /* The grammar as a symbol_list is no longer needed. */
+  LIST_FREE (symbol_list, grammar);
 }