* tests/regression.at, tests/torture.at, tests/calc.at: Adjust to

[bison.git] / src / lex.c

diff --git a/src/lex.c b/src/lex.c
index b236ea4393ff254d10f5470933fb78c5c2c431be..1d2424e2b421c73f6e98f4621996262c8a609af2 100644 (file)
--- a/src/lex.c
+++ b/src/lex.c
@@ -1,5 +1,5 @@
 /* Token-reader for Bison's input parser,
 /* Token-reader for Bison's input parser,

-   Copyright (C) 1984, 1986, 1989, 1992, 2000 Free Software Foundation, Inc.
+   Copyright 1984, 1986, 1989, 1992, 2000, 2001 Free Software Foundation, Inc.

 
    This file is part of Bison, the GNU Compiler Compiler.
 
 
    This file is part of Bison, the GNU Compiler Compiler.
 


@@ -18,79 +18,49 @@
    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
    Boston, MA 02111-1307, USA.  */
 
    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
    Boston, MA 02111-1307, USA.  */
 

-
-/*
-   lex is the entry point.  It is called from reader.c.
-   It returns one of the token-type codes defined in lex.h.
-   When an identifier is seen, the code IDENTIFIER is returned
-   and the name is looked up in the symbol table using symtab.c;
-   symval is set to a pointer to the entry found.  */
-

 #include "system.h"
 #include "getargs.h"
 #include "files.h"
 #include "system.h"
 #include "getargs.h"
 #include "files.h"

-#include "getopt.h"            /* for optarg */

 #include "symtab.h"
 #include "lex.h"
 #include "symtab.h"
 #include "lex.h"

-#include "alloc.h"

 #include "complain.h"
 #include "complain.h"

-
-/*spec_outfile is declared in files.h, for -o */
-
-extern int translations;
-
-extern void init_lex PARAMS((void));
-extern char *grow_token_buffer PARAMS((char *));
-extern int skip_white_space PARAMS((void));
-extern void unlex PARAMS((int));
-extern int lex PARAMS((void));
-extern int parse_percent_token PARAMS((void));
-
-static int safegetc PARAMS((FILE *));
-static int literalchar PARAMS((char **, int *, char));
-
-/* functions from main.c */
-extern char *printable_version PARAMS((int));
+#include "gram.h"
+#include "quote.h"

 
 /* Buffer for storing the current token.  */
 
 /* Buffer for storing the current token.  */

-char *token_buffer;
+struct obstack token_obstack;
+const char *token_buffer = NULL;

 
 

-/* Allocated size of token_buffer, not including space for terminator.  */
-int maxtoken;
-
-bucket *symval;
+bucket *symval = NULL;

 int numval;
 
 int numval;
 

-static int unlexed;            /* these two describe a token to be reread */
-static bucket *unlexed_symval; /* by the next call to lex */
-
+/* A token to be reread, see unlex and lex. */
+static token_t unlexed = tok_undef;
+static bucket *unlexed_symval = NULL;
+static const char *unlexed_token_buffer = NULL;

 
 void
 
 void

-init_lex (void)
+lex_init (void)

 {
 {

-  maxtoken = 100;
-  token_buffer = NEW2 (maxtoken + 1, char);
-  unlexed = -1;
+  obstack_init (&token_obstack);
+  unlexed = tok_undef;

 }
 
 
 }
 
 

-char *
-grow_token_buffer (char *p)
+void
+lex_free (void)

 {
 {

-  int offset = p - token_buffer;
-  maxtoken *= 2;
-  token_buffer = (char *) xrealloc(token_buffer, maxtoken + 1);
-  return token_buffer + offset;
+  obstack_free (&token_obstack, NULL);

 }
 
 
 int
 skip_white_space (void)
 {
 }
 
 
 int
 skip_white_space (void)
 {

-  register int c;
-  register int inside;
+  int c;
+  int inside;

 
 

-  c = getc(finput);
+  c = getc (finput);

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


@@ -99,7 +69,8 @@ skip_white_space (void)
       switch (c)
        {
        case '/':
       switch (c)
        {
        case '/':

-         c = getc(finput);
+         /* FIXME: Should probably be merged with copy_comment.  */
+         c = getc (finput);

          if (c != '*' && c != '/')
            {
              complain (_("unexpected `/' found and ignored"));
          if (c != '*' && c != '/')
            {
              complain (_("unexpected `/' found and ignored"));


@@ -107,7 +78,7 @@ skip_white_space (void)
            }
          cplus_comment = (c == '/');
 
            }
          cplus_comment = (c == '/');
 

-         c = getc(finput);
+         c = getc (finput);

 
          inside = 1;
          while (inside)
 
          inside = 1;
          while (inside)


@@ -115,12 +86,12 @@ skip_white_space (void)
              if (!cplus_comment && c == '*')
                {
                  while (c == '*')
              if (!cplus_comment && c == '*')
                {
                  while (c == '*')

-                   c = getc(finput);
+                   c = getc (finput);

 
                  if (c == '/')
                    {
                      inside = 0;
 
                  if (c == '/')
                    {
                      inside = 0;

-                     c = getc(finput);
+                     c = getc (finput);

                    }
                }
              else if (c == '\n')
                    }
                }
              else if (c == '\n')


@@ -128,12 +99,12 @@ skip_white_space (void)
                  lineno++;
                  if (cplus_comment)
                    inside = 0;
                  lineno++;
                  if (cplus_comment)
                    inside = 0;

-                 c = getc(finput);
+                 c = getc (finput);

                }
              else if (c == EOF)
                fatal (_("unterminated comment"));
              else
                }
              else if (c == EOF)
                fatal (_("unterminated comment"));
              else

-               c = getc(finput);
+               c = getc (finput);

            }
 
          break;
            }
 
          break;


@@ -144,7 +115,7 @@ skip_white_space (void)
        case ' ':
        case '\t':
        case '\f':
        case ' ':
        case '\t':
        case '\f':

-         c = getc(finput);
+         c = getc (finput);

          break;
 
        default:
          break;
 
        default:


@@ -153,35 +124,47 @@ skip_white_space (void)
     }
 }
 
     }
 }
 

-/* do a getc, but give error message if EOF encountered */
+
+/*-----------------------------------------------------.
+| Do a getc, but give error message if EOF encountered |
+`-----------------------------------------------------*/
+

 static int
 static int

-safegetc (FILE *f)
+xgetc (FILE *f)

 {
 {

-  register int c = getc(f);
+  int c = getc (f);

   if (c == EOF)
     fatal (_("unexpected end of file"));
   return c;
 }
 
   if (c == EOF)
     fatal (_("unexpected end of file"));
   return c;
 }
 

-/* read one literal character from finput.  process \ escapes.
-   append the normalized string version of the char to *pp.
-   assign the character code to *pcode
-   return 1 unless the character is an unescaped `term' or \n
-       report error for \n
-*/
+
+/*------------------------------------------------------------------.
+| Read one literal character from finput.  Process \ escapes.       |
+| Append the normalized string version of the char to OUT.  Assign  |
+| the character code to *PCODE. Return 1 unless the character is an |
+| unescaped `term' or \n report error for \n.                       |
+`------------------------------------------------------------------*/
+
+/* FIXME: We could directly work in the obstack, but that would make
+   it more difficult to move to quotearg some day.  So for the time
+   being, I prefer have literalchar behave like quotearg, and change
+   my mind later if I was wrong.  */
+

 static int
 static int

-literalchar (char **pp, int *pcode, char term)
+literalchar (struct obstack *out, int *pcode, char term)

 {
 {

-  register int c;
-  register char *p;
-  register int code;
+  int c;
+  char buf[4096];
+  char *cp;
+  int code;

   int wasquote = 0;
 
   int wasquote = 0;
 

-  c = safegetc(finput);
+  c = xgetc (finput);

   if (c == '\n')
     {
       complain (_("unescaped newline in constant"));
   if (c == '\n')
     {
       complain (_("unescaped newline in constant"));

-      ungetc(c, finput);
+      ungetc (c, finput);

       code = '?';
       wasquote = 1;
     }
       code = '?';
       wasquote = 1;
     }


@@ -193,17 +176,27 @@ literalchar (char **pp, int *pcode, char term)
     }
   else
     {
     }
   else
     {

-      c = safegetc(finput);
-      if (c == 't')  code = '\t';
-      else if (c == 'n')  code = '\n';
-      else if (c == 'a')  code = '\007';
-      else if (c == 'r')  code = '\r';
-      else if (c == 'f')  code = '\f';
-      else if (c == 'b')  code = '\b';
-      else if (c == 'v')  code = '\013';
-      else if (c == '\\')  code = '\\';
-      else if (c == '\'')  code = '\'';
-      else if (c == '\"')  code = '\"';
+      c = xgetc (finput);
+      if (c == 't')
+       code = '\t';
+      else if (c == 'n')
+       code = '\n';
+      else if (c == 'a')
+       code = '\007';
+      else if (c == 'r')
+       code = '\r';
+      else if (c == 'f')
+       code = '\f';
+      else if (c == 'b')
+       code = '\b';
+      else if (c == 'v')
+       code = '\013';
+      else if (c == '\\')
+       code = '\\';
+      else if (c == '\'')
+       code = '\'';
+      else if (c == '\"')
+       code = '\"';

       else if (c <= '7' && c >= '0')
        {
          code = 0;
       else if (c <= '7' && c >= '0')
        {
          code = 0;


@@ -217,401 +210,505 @@ literalchar (char **pp, int *pcode, char term)
                  code &= 0xFF;
                  break;
                }
                  code &= 0xFF;
                  break;
                }

-             c = safegetc(finput);
+             c = xgetc (finput);

            }
            }

-         ungetc(c, finput);
+         ungetc (c, finput);

        }
       else if (c == 'x')
        {
        }
       else if (c == 'x')
        {

-         c = safegetc(finput);
+         c = xgetc (finput);

          code = 0;
          while (1)
            {
              if (c >= '0' && c <= '9')
          code = 0;
          while (1)
            {
              if (c >= '0' && c <= '9')

-               code *= 16,  code += c - '0';
+               code *= 16, code += c - '0';

              else if (c >= 'a' && c <= 'f')
              else if (c >= 'a' && c <= 'f')

-               code *= 16,  code += c - 'a' + 10;
+               code *= 16, code += c - 'a' + 10;

              else if (c >= 'A' && c <= 'F')
              else if (c >= 'A' && c <= 'F')

-               code *= 16,  code += c - 'A' + 10;
+               code *= 16, code += c - 'A' + 10;

              else
                break;
              else
                break;

-             if (code >= 256 || code<0)
+             if (code >= 256 || code < 0)

                {
                {

-                 complain (_("hexadecimal value above 255: `\\x%x'"),
-                           code);
+                 complain (_("hexadecimal value above 255: `\\x%x'"), code);

                  code &= 0xFF;
                  break;
                }
                  code &= 0xFF;
                  break;
                }

-             c = safegetc(finput);
+             c = xgetc (finput);

            }
            }

-         ungetc(c, finput);
+         ungetc (c, finput);

        }
       else
        {
        }
       else
        {

+         char badchar [] = "c";
+         badchar[0] = c;

          complain (_("unknown escape sequence: `\\' followed by `%s'"),
          complain (_("unknown escape sequence: `\\' followed by `%s'"),

-                   printable_version(c));
+                   quote (badchar));

          code = '?';
        }
          code = '?';
        }

-    } /* has \ */
+    }                          /* has \ */

 
 

-  /* now fill token_buffer with the canonical name for this character
-     as a literal token.  Do not use what the user typed,
-     so that `\012' and `\n' can be interchangeable.  */
+  /* now fill BUF with the canonical name for this character as a
+     literal token.  Do not use what the user typed, so that `\012'
+     and `\n' can be interchangeable.  */

 
 

-  p = *pp;
+  cp = buf;

   if (code == term && wasquote)
   if (code == term && wasquote)

-    *p++ = code;
-  else if (code == '\\')  {*p++ = '\\'; *p++ = '\\';}
-  else if (code == '\'')  {*p++ = '\\'; *p++ = '\'';}
-  else if (code == '\"')  {*p++ = '\\'; *p++ = '\"';}
+    *cp++ = code;
+  else if (code == '\\')
+    {
+      *cp++ = '\\';
+      *cp++ = '\\';
+    }
+  else if (code == '\'')
+    {
+      *cp++ = '\\';
+      *cp++ = '\'';
+    }
+  else if (code == '\"')
+    {
+      *cp++ = '\\';
+      *cp++ = '\"';
+    }

   else if (code >= 040 && code < 0177)
   else if (code >= 040 && code < 0177)

-    *p++ = code;
-  else if (code == '\t')  {*p++ = '\\'; *p++ = 't';}
-  else if (code == '\n')  {*p++ = '\\'; *p++ = 'n';}
-  else if (code == '\r')  {*p++ = '\\'; *p++ = 'r';}
-  else if (code == '\v')  {*p++ = '\\'; *p++ = 'v';}
-  else if (code == '\b')  {*p++ = '\\'; *p++ = 'b';}
-  else if (code == '\f')  {*p++ = '\\'; *p++ = 'f';}
+    *cp++ = code;
+  else if (code == '\t')
+    {
+      *cp++ = '\\';
+      *cp++ = 't';
+    }
+  else if (code == '\n')
+    {
+      *cp++ = '\\';
+      *cp++ = 'n';
+    }
+  else if (code == '\r')
+    {
+      *cp++ = '\\';
+      *cp++ = 'r';
+    }
+  else if (code == '\v')
+    {
+      *cp++ = '\\';
+      *cp++ = 'v';
+    }
+  else if (code == '\b')
+    {
+      *cp++ = '\\';
+      *cp++ = 'b';
+    }
+  else if (code == '\f')
+    {
+      *cp++ = '\\';
+      *cp++ = 'f';
+    }

   else
     {
   else
     {

-      *p++ = '\\';
-      *p++ = code / 0100 + '0';
-      *p++ = ((code / 010) & 07) + '0';
-      *p++ = (code & 07) + '0';
+      *cp++ = '\\';
+      *cp++ = code / 0100 + '0';
+      *cp++ = ((code / 010) & 07) + '0';
+      *cp++ = (code & 07) + '0';

     }
     }

-  *pp = p;
+  *cp = '\0';
+
+  if (out)
+    obstack_sgrow (out, buf);

   *pcode = code;
   *pcode = code;

-  return  ! wasquote;
+  return !wasquote;

 }
 
 
 void
 }
 
 
 void

-unlex (int token)
+unlex (token_t token)

 {
   unlexed = token;
 {
   unlexed = token;

+  unlexed_token_buffer = token_buffer;

   unlexed_symval = symval;
 }
 
   unlexed_symval = symval;
 }
 

+/*-----------------------------------------------------------------.
+| We just read `<' from FIN.  Store in TOKEN_BUFFER, the type name |
+| specified between the `<...>'.                                   |
+`-----------------------------------------------------------------*/

 
 

-int
+void
+read_type_name (FILE *fin)
+{
+  int c = getc (fin);
+
+  while (c != '>')
+    {
+      if (c == EOF)
+       fatal (_("unterminated type name at end of file"));
+      if (c == '\n')
+       {
+         complain (_("unterminated type name"));
+         ungetc (c, fin);
+         break;
+       }
+
+      obstack_1grow (&token_obstack, c);
+      c = getc (fin);
+    }
+  obstack_1grow (&token_obstack, '\0');
+  token_buffer = obstack_finish (&token_obstack);
+}
+
+
+token_t

 lex (void)
 {
 lex (void)
 {

-  register int c;
-  char *p;
+  int c;

 
 

-  if (unlexed >= 0)
+  /* Just to make sure. */
+  token_buffer = NULL;
+
+  if (unlexed != tok_undef)

     {
     {

+      token_t res = unlexed;

       symval = unlexed_symval;
       symval = unlexed_symval;

-      c = unlexed;
-      unlexed = -1;
-      return c;
+      token_buffer = unlexed_token_buffer;
+      unlexed = tok_undef;
+      return res;

     }
 
     }
 

-  c = skip_white_space();
-  *token_buffer = c;   /* for error messages (token buffer always valid) */
-  token_buffer[1] = 0;
+  c = skip_white_space ();

 
   switch (c)
     {
     case EOF:
 
   switch (c)
     {
     case EOF:

-      strcpy(token_buffer, "EOF");
-      return ENDFILE;
-
-    case 'A':  case 'B':  case 'C':  case 'D':  case 'E':
-    case 'F':  case 'G':  case 'H':  case 'I':  case 'J':
-    case 'K':  case 'L':  case 'M':  case 'N':  case 'O':
-    case 'P':  case 'Q':  case 'R':  case 'S':  case 'T':
-    case 'U':  case 'V':  case 'W':  case 'X':  case 'Y':
+      token_buffer = "EOF";
+      return tok_eof;
+
+    case 'A':    case 'B':    case 'C':    case 'D':    case 'E':
+    case 'F':    case 'G':    case 'H':    case 'I':    case 'J':
+    case 'K':    case 'L':    case 'M':    case 'N':    case 'O':
+    case 'P':    case 'Q':    case 'R':    case 'S':    case 'T':
+    case 'U':    case 'V':    case 'W':    case 'X':    case 'Y':

     case 'Z':
     case 'Z':

-    case 'a':  case 'b':  case 'c':  case 'd':  case 'e':
-    case 'f':  case 'g':  case 'h':  case 'i':  case 'j':
-    case 'k':  case 'l':  case 'm':  case 'n':  case 'o':
-    case 'p':  case 'q':  case 'r':  case 's':  case 't':
-    case 'u':  case 'v':  case 'w':  case 'x':  case 'y':
+    case 'a':    case 'b':    case 'c':    case 'd':    case 'e':
+    case 'f':    case 'g':    case 'h':    case 'i':    case 'j':
+    case 'k':    case 'l':    case 'm':    case 'n':    case 'o':
+    case 'p':    case 'q':    case 'r':    case 's':    case 't':
+    case 'u':    case 'v':    case 'w':    case 'x':    case 'y':

     case 'z':
     case 'z':

-    case '.':  case '_':
-      p = token_buffer;
-      while (isalnum(c) || c == '_' || c == '.')
-       {
-         if (p == token_buffer + maxtoken)
-           p = grow_token_buffer(p);
+    case '.':    case '_':

 
 

-         *p++ = c;
-         c = getc(finput);
+      while (isalnum (c) || c == '_' || c == '.')
+       {
+         obstack_1grow (&token_obstack, c);
+         c = getc (finput);

        }
        }

-
-      *p = 0;
-      ungetc(c, finput);
-      symval = getsym(token_buffer);
-      return IDENTIFIER;
-
-    case '0':  case '1':  case '2':  case '3':  case '4':
-    case '5':  case '6':  case '7':  case '8':  case '9':
+      obstack_1grow (&token_obstack, '\0');
+      token_buffer = obstack_finish (&token_obstack);
+      ungetc (c, finput);
+      symval = getsym (token_buffer);
+      return tok_identifier;
+
+    case '0':    case '1':    case '2':    case '3':    case '4':
+    case '5':    case '6':    case '7':    case '8':    case '9':

       {
        numval = 0;
 
       {
        numval = 0;
 

-       p = token_buffer;
-       while (isdigit(c))
+       while (isdigit (c))

          {
          {

-           if (p == token_buffer + maxtoken)
-             p = grow_token_buffer(p);
-
-           *p++ = c;
-           numval = numval*10 + c - '0';
-           c = getc(finput);
+           obstack_1grow (&token_obstack, c);
+           numval = numval * 10 + c - '0';
+           c = getc (finput);

          }
          }

-       *p = 0;
-       ungetc(c, finput);
-       return NUMBER;
+       obstack_1grow (&token_obstack, '\0');
+       token_buffer = obstack_finish (&token_obstack);
+       ungetc (c, finput);
+       return tok_number;

       }
 
     case '\'':
       }
 
     case '\'':

-

       /* parse the literal token and compute character code in  code  */
 
       /* parse the literal token and compute character code in  code  */
 

-      translations = -1;

       {
       {

-       int code, discode;
-       char discard[10], *dp;
+       int code;

 
 

-       p = token_buffer;
-       *p++ = '\'';
-       literalchar(&p, &code, '\'');
+       obstack_1grow (&token_obstack, '\'');
+       literalchar (&token_obstack, &code, '\'');

 
 

-       c = getc(finput);
+       c = getc (finput);

        if (c != '\'')
          {
        if (c != '\'')
          {

+           int discode;

            complain (_("use \"...\" for multi-character literal tokens"));
            while (1)
            complain (_("use \"...\" for multi-character literal tokens"));
            while (1)

-             {
-               dp = discard;
-               if (! literalchar(&dp, &discode, '\''))
-                 break;
-             }
+             if (!literalchar (0, &discode, '\''))
+               break;

          }
          }

-       *p++ = '\'';
-       *p = 0;
-       symval = getsym(token_buffer);
-       symval->class = STOKEN;
-       if (! symval->user_token_number)
+       obstack_1grow (&token_obstack, '\'');
+       obstack_1grow (&token_obstack, '\0');
+       token_buffer = obstack_finish (&token_obstack);
+       symval = getsym (token_buffer);
+       symval->class = token_sym;
+       if (symval->user_token_number == SUNDEF)

          symval->user_token_number = code;
          symval->user_token_number = code;

-       return IDENTIFIER;
+       return tok_identifier;

       }
 
     case '\"':
       }
 
     case '\"':

-

       /* parse the literal string token and treat as an identifier */
 
       /* parse the literal string token and treat as an identifier */
 

-      translations = -1;

       {
       {

-       int code;       /* ignored here */
-       p = token_buffer;
-       *p++ = '\"';
-       while (literalchar(&p, &code, '\"'))  /* read up to and including " */
-         {
-           if (p >= token_buffer + maxtoken - 4)
-             p = grow_token_buffer(p);
-         }
-       *p = 0;
+       int code;               /* ignored here */
+
+       obstack_1grow (&token_obstack, '\"');
+       /* Read up to and including ".  */
+       while (literalchar (&token_obstack, &code, '\"'))
+         /* nothing */;
+       obstack_1grow (&token_obstack, '\0');
+       token_buffer = obstack_finish (&token_obstack);

 
 

-       symval = getsym(token_buffer);
-       symval->class = STOKEN;
+       symval = getsym (token_buffer);
+       symval->class = token_sym;

 
 

-       return IDENTIFIER;
+       return tok_identifier;

       }
 
     case ',':
       }
 
     case ',':

-      return COMMA;
+      token_buffer = ",";
+      return tok_comma;

 
     case ':':
 
     case ':':

-      return COLON;
+      token_buffer = ":";
+      return tok_colon;

 
     case ';':
 
     case ';':

-      return SEMICOLON;
+      token_buffer = ";";
+      return tok_semicolon;

 
     case '|':
 
     case '|':

-      return BAR;
+      token_buffer = "|";
+      return tok_bar;

 
     case '{':
 
     case '{':

-      return LEFT_CURLY;
+      token_buffer = "{";
+      return tok_left_curly;

 
     case '=':
 
     case '=':

+      obstack_1grow (&token_obstack, c);

       do
        {
       do
        {

-         c = getc(finput);
-         if (c == '\n') lineno++;
+         c = getc (finput);
+         obstack_1grow (&token_obstack, c);
+         if (c == '\n')
+           lineno++;

        }
        }

-      while(c==' ' || c=='\n' || c=='\t');
+      while (c == ' ' || c == '\n' || c == '\t');
+      obstack_1grow (&token_obstack, '\0');
+      token_buffer = obstack_finish (&token_obstack);

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

-         strcpy(token_buffer, "={");
-         return LEFT_CURLY;
+         return tok_left_curly;

        }
       else
        {
        }
       else
        {

-         ungetc(c, finput);
-         return ILLEGAL;
+         ungetc (c, finput);
+         return tok_illegal;

        }
 
     case '<':
        }
 
     case '<':

-      p = token_buffer;
-      c = getc(finput);
-      while (c != '>')
-       {
-         if (c == EOF)
-           fatal (_("unterminated type name at end of file"));
-         if (c == '\n')
-           {
-             complain (_("unterminated type name"));
-             ungetc(c, finput);
-             break;
-           }
-
-         if (p == token_buffer + maxtoken)
-           p = grow_token_buffer(p);
-
-         *p++ = c;
-         c = getc(finput);
-       }
-      *p = 0;
-      return TYPENAME;
-
+      read_type_name (finput);
+      return tok_typename;

 
     case '%':
 
     case '%':

-      return parse_percent_token();
+      return parse_percent_token ();

 
     default:
 
     default:

-      return ILLEGAL;
+      obstack_1grow (&token_obstack, c);
+      obstack_1grow (&token_obstack, '\0');
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_illegal;

     }
 }
 
     }
 }
 

-/* the following table dictates the action taken for the various
-       % directives.  A setflag value causes the named flag to be
-       set.  A retval action returns the code.
-*/
-struct percent_table_struct {
-       const char *name;
-       void *setflag;
-       int retval;
-} percent_table[] =
+/* the following table dictates the action taken for the various %
+   directives.  A set_flag value causes the named flag to be set.  A
+   retval action returns the code.  */
+struct percent_table_struct

 {
 {

-  {"token", NULL, TOKEN},
-  {"term", NULL, TOKEN},
-  {"nterm", NULL, NTERM},
-  {"type", NULL, TYPE},
-  {"guard", NULL, GUARD},
-  {"union", NULL, UNION},
-  {"expect", NULL, EXPECT},
-  {"thong", NULL, THONG},
-  {"start", NULL, START},
-  {"left", NULL, LEFT},
-  {"right", NULL, RIGHT},
-  {"nonassoc", NULL, NONASSOC},
-  {"binary", NULL, NONASSOC},
-  {"semantic_parser", NULL, SEMANTIC_PARSER},
-  {"pure_parser", NULL, PURE_PARSER},
-  {"prec", NULL, PREC},
-
-  {"no_lines", &nolinesflag, NOOP}, /* -l */
-  {"raw", &rawtoknumflag, NOOP}, /* -r */
-  {"token_table", &toknumflag, NOOP}, /* -k */
-
-#if 0
-  /* These can be utilized after main is reoganized so
-     open_files() is deferred 'til after read_declarations().
-     But %{ and %union both put information into files
-     that have to be opened before read_declarations().
-     */
-  {"yacc", &fixed_outfiles, NOOP}, /* -y */
-  {"fixed_output_files", &fixed_outfiles, NOOP}, /* -y */
-  {"defines", &definesflag, NOOP}, /* -d */
-  {"no_parser", &noparserflag, NOOP}, /* -n */
-  {"output_file", &spec_outfile, SETOPT}, /* -o */
-  {"file_prefix", &spec_file_prefix, SETOPT}, /* -b */
-  {"name_prefix", &spec_name_prefix, SETOPT}, /* -p */
-
-  /* These would be acceptable, but they do not affect processing */
-  {"verbose", &verboseflag, NOOP}, /* -v */
-  {"debug", &debugflag, NOOP}, /* -t */
-  /*   {"help", <print usage stmt>, NOOP},*/   /* -h */
-  /*   {"version", <print version number> ,  NOOP},*/  /* -V */
-#endif
-
-  {NULL, NULL, ILLEGAL}
+  const char *name;
+  void *set_flag;
+  token_t retval;
+};
+
+struct percent_table_struct percent_table[] =
+{
+  { "token",           NULL,                   tok_token },
+  { "term",            NULL,                   tok_token },
+  { "nterm",           NULL,                   tok_nterm },
+  { "type",            NULL,                   tok_type },
+  { "guard",           NULL,                   tok_guard },
+  { "union",           NULL,                   tok_union },
+  { "expect",          NULL,                   tok_expect },
+  { "thong",           NULL,                   tok_thong },
+  { "start",           NULL,                   tok_start },
+  { "left",            NULL,                   tok_left },
+  { "right",           NULL,                   tok_right },
+  { "nonassoc",                NULL,                   tok_nonassoc },
+  { "binary",          NULL,                   tok_nonassoc },
+  { "prec",            NULL,                   tok_prec },
+  { "locations",       &locations_flag,        tok_intopt },   /* -l */
+  { "no-lines",                &no_lines_flag,         tok_intopt },   /* -l */
+  { "raw",             NULL,                   tok_obsolete }, /* -r */
+  { "token-table",     &token_table_flag,      tok_intopt },   /* -k */
+  { "yacc",            &yacc_flag,             tok_intopt },   /* -y */
+  { "fixed-output-files",&yacc_flag,           tok_intopt },   /* -y */
+  { "defines",         &defines_flag,          tok_intopt },   /* -d */
+  { "no-parser",       &no_parser_flag,        tok_intopt },   /* -n */
+  { "graph",           &graph_flag,            tok_intopt },   /* -g */
+
+  /* FIXME: semantic parsers which will output an `include' of an
+     output file: be sure that the name included is indeed the name of
+     the output file.  */
+  { "output",          &spec_outfile,          tok_stropt },   /* -o */
+  { "file-prefix",     &spec_file_prefix,      tok_stropt },   /* -b */
+  { "name-prefix",     &spec_name_prefix,      tok_stropt },   /* -p */
+
+  { "verbose",         &verbose_flag,          tok_intopt },   /* -v */
+  { "debug",           &debug_flag,            tok_intopt },   /* -t */
+  { "semantic-parser", &semantic_parser,       tok_intopt },
+  { "pure-parser",     &pure_parser,           tok_intopt },
+
+  { NULL, NULL, tok_illegal}

 };
 
 /* Parse a token which starts with %.
    Assumes the % has already been read and discarded.  */
 
 };
 
 /* Parse a token which starts with %.
    Assumes the % has already been read and discarded.  */
 

-int
+token_t

 parse_percent_token (void)
 {
 parse_percent_token (void)
 {

-  register int c;
-  register char *p;
-  register struct percent_table_struct *tx;
+  struct percent_table_struct *tx = NULL;
+  const char *arg = NULL;
+  /* Where the ARG was found in token_buffer. */
+  size_t arg_offset = 0;

 
 

-  p = token_buffer;
-  c = getc(finput);
-  *p++ = '%';
-  *p++ = c;    /* for error msg */
-  *p = 0;
+  int c = getc (finput);
+  obstack_1grow (&token_obstack, '%');
+  obstack_1grow (&token_obstack, c);

 
   switch (c)
     {
     case '%':
 
   switch (c)
     {
     case '%':

-      return TWO_PERCENTS;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_two_percents;

 
     case '{':
 
     case '{':

-      return PERCENT_LEFT_CURLY;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_percent_left_curly;

 
 

+      /* The following guys are here for backward compatibility with
+        very ancient Yacc versions.  The paper of Johnson mentions
+        them (as ancient :).  */

     case '<':
     case '<':

-      return LEFT;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_left;

 
     case '>':
 
     case '>':

-      return RIGHT;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_right;

 
     case '2':
 
     case '2':

-      return NONASSOC;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_nonassoc;

 
     case '0':
 
     case '0':

-      return TOKEN;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_token;

 
     case '=':
 
     case '=':

-      return PREC;
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_prec;

     }
     }

-  if (!isalpha(c))
-    return ILLEGAL;

 
 

-  p = token_buffer;
-  *p++ = '%';
-  while (isalpha(c) || c == '_' || c == '-')
+  if (!isalpha (c))

     {
     {

-      if (p == token_buffer + maxtoken)
-       p = grow_token_buffer(p);
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_illegal;
+    }

 
 

-      if (c == '-') c = '_';
-      *p++ = c;
-      c = getc(finput);
+  while (c = getc (finput), isalpha (c) || c == '_' || c == '-')
+    {
+      if (c == '_')
+       c = '-';
+      obstack_1grow (&token_obstack, c);

     }
 
     }
 

-  ungetc(c, finput);
+  /* %DIRECTIVE="ARG".  Separate into
+     TOKEN_BUFFER = `%DIRECTIVE\0ARG\0'.
+     This is a bit hackish, but once we move to a Bison parser,
+     things will be cleaned up.  */
+  if (c == '=')
+    {
+      /* End of the directive.  We skip the `='. */
+      obstack_1grow (&token_obstack, '\0');
+      /* Fetch the ARG if present. */
+      c = getc (finput);
+      if (c == '"')
+       {
+         int code;
+         arg_offset = obstack_object_size (&token_obstack);
+         /* Read up to and including `"'.  Do not append the closing
+            `"' in the output: it's not part of the ARG.  */
+         while (literalchar (NULL, &code, '"'))
+           obstack_1grow (&token_obstack, code);
+       }
+      /* else: should be an error. */
+    }
+  else
+    ungetc (c, finput);

 
 

-  *p = 0;
+  obstack_1grow (&token_obstack, '\0');
+  token_buffer = obstack_finish (&token_obstack);
+  if (arg_offset)
+    arg = token_buffer + arg_offset;

 
   /* table lookup % directive */
   for (tx = percent_table; tx->name; tx++)
 
   /* table lookup % directive */
   for (tx = percent_table; tx->name; tx++)

-    if (strcmp(token_buffer+1, tx->name) == 0)
+    if (strcmp (token_buffer + 1, tx->name) == 0)

       break;
       break;

-  if (tx->retval == SETOPT)
-    {
-      *((char **)(tx->setflag)) = optarg;
-      return NOOP;
-    }
-  if (tx->setflag)
+
+  if (arg && tx->retval != tok_stropt)
+    fatal (_("`%s' supports no argument: %s"), token_buffer, quote (arg));
+
+  switch (tx->retval)

     {
     {

-      *((int *)(tx->setflag)) = 1;
-      return NOOP;
+    case tok_stropt:
+      assert (tx->set_flag);
+      if (arg)
+       {
+         /* Keep only the first assignment: command line options have
+            already been processed, and we want them to have
+            precedence.  Side effect: if this %-option is used
+            several times, only the first is honored.  Bah.  */
+         if (!*((char **) (tx->set_flag)))
+           *((char **) (tx->set_flag)) = xstrdup (arg);
+       }
+      else
+       fatal (_("`%s' requires an argument"), token_buffer);
+      return tok_noop;
+      break;
+
+    case tok_intopt:
+      assert (tx->set_flag);
+      *((int *) (tx->set_flag)) = 1;
+      return tok_noop;
+      break;
+
+    case tok_obsolete:
+      fatal (_("`%s' is no longer supported"), token_buffer);
+      return tok_noop;
+      break;
+
+    default:
+      return tx->retval;
+      break;

     }
     }

-  return tx->retval;
+  abort ();

 }
 }