* 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 20d89a1ec6467ec34d0a8d468c3d78b5fa901aff..1d2424e2b421c73f6e98f4621996262c8a609af2 100644 (file)
--- a/src/lex.c
+++ b/src/lex.c
@@ -1,87 +1,66 @@
 /* Token-reader for Bison's input parser,
 /* Token-reader for Bison's input parser,

-   Copyright (C) 1984, 1986, 1989 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.

 
 

-Bison is free software; you can redistribute it and/or modify
-it under the terms of the GNU General Public License as published by
-the Free Software Foundation; either version 2, or (at your option)
-any later version.
+   Bison is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2, or (at your option)
+   any later version.

 
 

-Bison is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU General Public License for more details.
+   Bison is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+   GNU General Public License for more details.

 
 

-You should have received a copy of the GNU General Public License
-along with Bison; see the file COPYING.  If not, write to
-the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.  */
+   You should have received a copy of the GNU General Public License
+   along with Bison; see the file COPYING.  If not, write to
+   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 <stdio.h>
-#include <ctype.h>

 #include "system.h"
 #include "system.h"

+#include "getargs.h"

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

-#include "new.h"
-
-
-extern int lineno;
-extern int translations;
-
-int parse_percent_token();
-
-extern void fatals();
-extern void fatal();
+#include "complain.h"
+#include "gram.h"
+#include "quote.h"

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

-char *token_buffer;
-
-/* Allocated size of token_buffer, not including space for terminator.  */
-static int maxtoken;
+struct obstack token_obstack;
+const char *token_buffer = NULL;

 
 

-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()
+lex_init (void)

 {
 {

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

 }
 
 
 }
 
 

-static char *
-grow_token_buffer (p)
-     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
 }
 
 
 int

-skip_white_space()
+skip_white_space (void)

 {
 {

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

 
 

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

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


@@ -90,12 +69,16 @@ skip_white_space()
       switch (c)
        {
        case '/':
       switch (c)
        {
        case '/':

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

          if (c != '*' && c != '/')
          if (c != '*' && c != '/')

-           fatals("unexpected `/%c' found",c);
+           {
+             complain (_("unexpected `/' found and ignored"));
+             break;
+           }

          cplus_comment = (c == '/');
 
          cplus_comment = (c == '/');
 

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

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


@@ -103,12 +86,12 @@ skip_white_space()
              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')


@@ -116,12 +99,12 @@ skip_white_space()
                  lineno++;
                  if (cplus_comment)
                    inside = 0;
                  lineno++;
                  if (cplus_comment)
                    inside = 0;

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

                }
              else if (c == EOF)
                }
              else if (c == EOF)

-               fatal("unterminated comment");
+               fatal (_("unterminated comment"));

              else
              else

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

            }
 
          break;
            }
 
          break;


@@ -132,373 +115,600 @@ skip_white_space()
        case ' ':
        case '\t':
        case '\f':
        case ' ':
        case '\t':
        case '\f':

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

          break;
 
        default:
          break;
 
        default:

-         return (c);
+         return c;
+       }
+    }
+}
+
+
+/*-----------------------------------------------------.
+| Do a getc, but give error message if EOF encountered |
+`-----------------------------------------------------*/
+
+static int
+xgetc (FILE *f)
+{
+  int c = getc (f);
+  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 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
+literalchar (struct obstack *out, int *pcode, char term)
+{
+  int c;
+  char buf[4096];
+  char *cp;
+  int code;
+  int wasquote = 0;
+
+  c = xgetc (finput);
+  if (c == '\n')
+    {
+      complain (_("unescaped newline in constant"));
+      ungetc (c, finput);
+      code = '?';
+      wasquote = 1;
+    }
+  else if (c != '\\')
+    {
+      code = c;
+      if (c == term)
+       wasquote = 1;
+    }
+  else
+    {
+      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;
+         while (c <= '7' && c >= '0')
+           {
+             code = (code * 8) + (c - '0');
+             if (code >= 256 || code < 0)
+               {
+                 complain (_("octal value outside range 0...255: `\\%o'"),
+                           code);
+                 code &= 0xFF;
+                 break;
+               }
+             c = xgetc (finput);
+           }
+         ungetc (c, finput);
+       }
+      else if (c == 'x')
+       {
+         c = xgetc (finput);
+         code = 0;
+         while (1)
+           {
+             if (c >= '0' && c <= '9')
+               code *= 16, code += c - '0';
+             else if (c >= 'a' && c <= 'f')
+               code *= 16, code += c - 'a' + 10;
+             else if (c >= 'A' && c <= 'F')
+               code *= 16, code += c - 'A' + 10;
+             else
+               break;
+             if (code >= 256 || code < 0)
+               {
+                 complain (_("hexadecimal value above 255: `\\x%x'"), code);
+                 code &= 0xFF;
+                 break;
+               }
+             c = xgetc (finput);
+           }
+         ungetc (c, finput);

        }
        }

+      else
+       {
+         char badchar [] = "c";
+         badchar[0] = c;
+         complain (_("unknown escape sequence: `\\' followed by `%s'"),
+                   quote (badchar));
+         code = '?';
+       }
+    }                          /* has \ */
+
+  /* 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.  */
+
+  cp = buf;
+  if (code == term && wasquote)
+    *cp++ = code;
+  else if (code == '\\')
+    {
+      *cp++ = '\\';
+      *cp++ = '\\';
+    }
+  else if (code == '\'')
+    {
+      *cp++ = '\\';
+      *cp++ = '\'';
+    }
+  else if (code == '\"')
+    {
+      *cp++ = '\\';
+      *cp++ = '\"';
+    }
+  else if (code >= 040 && code < 0177)
+    *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
+    {
+      *cp++ = '\\';
+      *cp++ = code / 0100 + '0';
+      *cp++ = ((code / 010) & 07) + '0';
+      *cp++ = (code & 07) + '0';
+    }
+  *cp = '\0';
+
+  if (out)
+    obstack_sgrow (out, buf);
+  *pcode = code;
+  return !wasquote;

 }
 
 
 void
 }
 
 
 void

-unlex(token)
-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 `<...>'.                                   |
+`-----------------------------------------------------------------*/

 
 

+void
+read_type_name (FILE *fin)
+{
+  int c = getc (fin);

 
 

-int
-lex()
+  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)

 {
 {

-  register int c;
-  register char *p;
+  int c;
+
+  /* Just to make sure. */
+  token_buffer = NULL;

 
 

-  if (unlexed >= 0)
+  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();
+  c = skip_white_space ();

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

-       while (isdigit(c))
+       while (isdigit (c))

          {
          {

-           numval = numval*10 + c - '0';
-           c = getc(finput);
+           obstack_1grow (&token_obstack, c);
+           numval = numval * 10 + c - '0';
+           c = getc (finput);

          }
          }

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

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

-      translations = -1;
-

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

-      c = getc(finput);

       {
       {

-       register int code = 0;
-
-       if (c == '\\')
-         {
-           c = getc(finput);
-
-           if (c <= '7' && c >= '0')
-             {
-               while (c <= '7' && c >= '0')
-                 {
-                   code = (code * 8) + (c - '0');
-                   c = getc(finput);
-                   if (code >= 256 || code < 0)
-                     fatals("malformatted literal token `\\%03o'", code);
-                 }
-             }
-           else
-             {
-               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 == 'x')
-                 {
-                   c = getc(finput);
-                   while ((c <= '9' && c >= '0')
-                          || (c >= 'a' && c <= 'z')
-                          || (c >= 'A' && c <= 'Z'))
-                     {
-                       code *= 16;
-                       if (c <= '9' && c >= '0')
-                         code += c - '0';
-                       else if (c >= 'a' && c <= 'z')
-                         code += c - 'a' + 10;
-                       else if (c >= 'A' && c <= 'Z')
-                         code += c - 'A' + 10;
-                       if (code >= 256 || code<0)/* JF this said if(c>=128) */
-                         fatals("malformatted literal token `\\x%x'",code);
-                       c = getc(finput);
-                     }
-                   ungetc(c, finput);
-                 }
-               else if (c == '\\')
-                 code = '\\';
-               else if (c == '\'')
-                 code = '\'';
-               else if (c == '\"')     /* JF this is a good idea */
-                 code = '\"';
-               else
-                 {
-                   if (c >= 040 && c <= 0177)
-                     fatals ("unknown escape sequence `\\%c'", c);
-                   else
-                     fatals ("unknown escape sequence: `\\' followed by char code 0x%x", c);
-                 }
-
-               c = getc(finput);
-             }
-         }
-       else
-         {
-           code = c;
-           c = getc(finput);
-         }
-       if (c != '\'')
-         fatal("multicharacter literal tokens not supported");
+       int code;

 
 

-       /* 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.  */
+       obstack_1grow (&token_obstack, '\'');
+       literalchar (&token_obstack, &code, '\'');

 
 

-       p = token_buffer;
-       *p++ = '\'';
-       if (code == '\\')
-         {
-           *p++ = '\\';
-           *p++ = '\\';
-         }
-       else if (code == '\'')
-         {
-           *p++ = '\\';
-           *p++ = '\'';
-         }
-       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';
-         }
-        else
+       c = getc (finput);
+       if (c != '\'')

          {
          {

-           *p++ = code / 0100 + '0';
-           *p++ = ((code / 010) & 07) + '0';
-           *p++ = (code & 07) + '0';
+           int discode;
+           complain (_("use \"...\" for multi-character literal tokens"));
+           while (1)
+             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 '\"':
+      /* parse the literal string token and treat as an identifier */
+
+      {
+       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 = token_sym;
+
+       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 == '{')

-       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 == '\n' || c == EOF)
-           fatal("unterminated type name");
-
-         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 set_flag value causes the named flag to be set.  A
+   retval action returns the code.  */
+struct percent_table_struct
+{
+  const char *name;
+  void *set_flag;
+  token_t retval;
+};

 
 

-/* parse a token which starts with %.  Assumes the % has already been read and discarded.  */
-
-int
-parse_percent_token ()
+struct percent_table_struct percent_table[] =

 {
 {

-  register int c;
-  register char *p;
+  { "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.  */
+
+token_t
+parse_percent_token (void)
+{
+  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);
+  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))
+    {
+      token_buffer = obstack_finish (&token_obstack);
+      return tok_illegal;

     }
     }

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

 
 

-  while (isalpha(c) || c == '_')
+  while (c = getc (finput), isalpha (c) || c == '_' || c == '-')

     {
     {

-      if (p == token_buffer + maxtoken)
-       p = grow_token_buffer(p);
+      if (c == '_')
+       c = '-';
+      obstack_1grow (&token_obstack, c);
+    }

 
 

-      *p++ = c;
-      c = getc(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);
+
+  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++)
+    if (strcmp (token_buffer + 1, tx->name) == 0)
+      break;
+
+  if (arg && tx->retval != tok_stropt)
+    fatal (_("`%s' supports no argument: %s"), token_buffer, quote (arg));

 
 

-  ungetc(c, finput);
-
-  *p = 0;
-
-  if (strcmp(token_buffer, "token") == 0
-      ||
-      strcmp(token_buffer, "term") == 0)
-    return (TOKEN);
-  else if (strcmp(token_buffer, "nterm") == 0)
-    return (NTERM);
-  else if (strcmp(token_buffer, "type") == 0)
-    return (TYPE);
-  else if (strcmp(token_buffer, "guard") == 0)
-    return (GUARD);
-  else if (strcmp(token_buffer, "union") == 0)
-    return (UNION);
-  else if (strcmp(token_buffer, "expect") == 0)
-    return (EXPECT);
-  else if (strcmp(token_buffer, "start") == 0)
-    return (START);
-  else if (strcmp(token_buffer, "left") == 0)
-    return (LEFT);
-  else if (strcmp(token_buffer, "right") == 0)
-    return (RIGHT);
-  else if (strcmp(token_buffer, "nonassoc") == 0
-          ||
-          strcmp(token_buffer, "binary") == 0)
-    return (NONASSOC);
-  else if (strcmp(token_buffer, "semantic_parser") == 0)
-    return (SEMANTIC_PARSER);
-  else if (strcmp(token_buffer, "pure_parser") == 0)
-    return (PURE_PARSER);
-  else if (strcmp(token_buffer, "prec") == 0)
-    return (PREC);
-  else return (ILLEGAL);
+  switch (tx->retval)
+    {
+    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;
+    }
+  abort ();

 }
 }