[bison.git] / src / reader.c

/* Input parser for bison
   Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002
   Free Software Foundation, Inc.

   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 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, Inc., 59 Temple Place - Suite 330,
   Boston, MA 02111-1307, USA.  */


#include "system.h"
#include "quotearg.h"
#include "quote.h"
#include "getargs.h"
#include "files.h"
#include "symtab.h"
#include "options.h"
#include "lex.h"
#include "gram.h"
#include "complain.h"
#include "output.h"
#include "reader.h"
#include "conflicts.h"
#include "muscle_tab.h"

typedef struct symbol_list
{
  struct symbol_list *next;
  symbol_t *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;
  symbol_t *ruleprec;
} symbol_list;

int lineno;
static symbol_list *grammar = NULL;
static int start_flag = 0;
static symbol_t *startval = NULL;

/* Nonzero if components of semantic values are used, implying
   they must be unions.  */
static int value_components_used;

/* Nonzero if %union has been seen.  */
static int typed = 0;

/* Incremented for each %left, %right or %nonassoc seen */
static int lastprec = 0;

symbol_t *errtoken = NULL;
symbol_t *undeftoken = NULL;
symbol_t *eoftoken = NULL;
symbol_t *axiom = NULL;

static symbol_list *
symbol_list_new (symbol_t *sym)
{
  symbol_list *res = XMALLOC (symbol_list, 1);
  res->next = NULL;
  res->sym = sym;
  res->line = lineno;
  res->action = NULL;
  res->action_line = 0;
  res->guard = NULL;
  res->guard_line = 0;
  res->ruleprec = NULL;
  return res;
}

/*------------------------.
| Operations on symbols.  |
`------------------------*/


/*-----------------------------------------------------------.
| If THIS is not defined, report an error, and consider it a |
| nonterminal.                                               |
`-----------------------------------------------------------*/

static bool
symbol_check_defined (symbol_t *this)
{
  if (this->class == unknown_sym)
    {
      complain
        (_("symbol %s is used, but is not defined as a token and has no rules"),
         this->tag);
      this->class = nterm_sym;
      this->number = nvars++;
    }

  return TRUE;
}


/*-------------------------------------------------------------------.
| Assign a symbol number, and write the definition of the token name |
| into FDEFINES.  Put in SYMBOLS.                                    |
`-------------------------------------------------------------------*/

static bool
symbol_make_alias (symbol_t *symbol, char *typename)
{
  if (symval->alias)
    warn (_("symbol `%s' used more than once as a literal string"),
          symval->tag);
  else if (symbol->alias)
    warn (_("symbol `%s' given more than one literal string"),
          symbol->tag);
  else
    {
      symval->class = token_sym;
      symval->type_name = typename;
      symval->user_token_number = symbol->user_token_number;
      symbol->user_token_number = SALIAS;
      symval->alias = symbol;
      symbol->alias = symval;
      /* symbol and symval combined are only one symbol */
      nsyms--;
      ntokens--;
      assert (ntokens == symbol->number || ntokens == symval->number);
      symbol->number = symval->number =
        (symval->number < symbol->number) ? symval->number : symbol->number;
    }

  return TRUE;
}

/*---------------------------------------------------------.
| Check that THIS, and its alias, have same precedence and |
| associativity.                                           |
`---------------------------------------------------------*/

static bool
symbol_check_alias_consistence (symbol_t *this)
{
  /* Check only those who _are_ the aliases. */
  if (this->alias && this->user_token_number == SALIAS)
    {
      if (this->prec != this->alias->prec)
        {
          if (this->prec != 0 && this->alias->prec != 0)
            complain (_("conflicting precedences for %s and %s"),
                      this->tag, this->alias->tag);
          if (this->prec != 0)
            this->alias->prec = this->prec;
          else
            this->prec = this->alias->prec;
        }

      if (this->assoc != this->alias->assoc)
        {
          if (this->assoc != 0 && this->alias->assoc != 0)
            complain (_("conflicting assoc values for %s and %s"),
                      this->tag, this->alias->tag);
          if (this->assoc != 0)
            this->alias->assoc = this->assoc;
          else
            this->assoc = this->alias->assoc;
        }
    }
  return TRUE;
}


/*-------------------------------------------------------------------.
| Assign a symbol number, and write the definition of the token name |
| into FDEFINES.  Put in SYMBOLS.                                    |
`-------------------------------------------------------------------*/

static bool
symbol_pack (symbol_t *this)
{
  if (this->class == nterm_sym)
    {
      this->number += ntokens;
    }
  else if (this->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 (this->number == NUMBER_UNDEFINED)
        {
          if (this == eoftoken || this->alias == eoftoken)
            this->number = this->alias->number = 0;
          else
            {
              assert (this->alias->number != NUMBER_UNDEFINED);
              this->number = this->alias->number;
            }
        }
      /* Do not do processing below for SALIASs.  */
      if (this->user_token_number == SALIAS)
        return TRUE;
    }
  else /* this->class == token_sym */
    {
      assert (this->number != NUMBER_UNDEFINED);
    }

  symbols[this->number] = this;
  return TRUE;
}




/*--------------------------------------------------.
| Put THIS in TOKEN_TRANSLATIONS if it is a token.  |
`--------------------------------------------------*/

static bool
symbol_translation (symbol_t *this)
{
  /* Non-terminal? */
  if (this->class == token_sym
      && this->user_token_number != SALIAS)
    {
      /* A token which translation has already been set? */
      if (token_translations[this->user_token_number] != undeftoken->number)
        complain (_("tokens %s and %s both assigned number %d"),
                  symbols[token_translations[this->user_token_number]]->tag,
                  this->tag, this->user_token_number);

      token_translations[this->user_token_number] = this->number;
    }

  return TRUE;
}
\f

/*===================\
| Low level lexing.  |
\===================*/

static void
skip_to_char (int target)
{
  int c;
  if (target == '\n')
    complain (_("   Skipping to next \\n"));
  else
    complain (_("   Skipping to next %c"), target);

  do
    c = skip_white_space ();
  while (c != target && c != EOF);
  if (c != EOF)
    ungetc (c, finput);
}


/*---------------------------------------------------------.
| Read a signed integer from STREAM and return its value.  |
`---------------------------------------------------------*/

static inline int
read_signed_integer (FILE *stream)
{
  int c = getc (stream);
  int sign = 1;
  int n = 0;

  if (c == '-')
    {
      c = getc (stream);
      sign = -1;
    }

  while (isdigit (c))
    {
      n = 10 * n + (c - '0');
      c = getc (stream);
    }

  ungetc (c, stream);

  return sign * n;
}
\f
/*--------------------------------------------------------------.
| Get the data type (alternative in the union) of the value for |
| symbol N in rule RULE.                                        |
`--------------------------------------------------------------*/

static char *
get_type_name (int n, symbol_list *rule)
{
  int i;
  symbol_list *rp;

  if (n < 0)
    {
      complain (_("invalid $ value"));
      return NULL;
    }

  rp = rule;
  i = 0;

  while (i < n)
    {
      rp = rp->next;
      if (rp == NULL || rp->sym == NULL)
        {
          complain (_("invalid $ value"));
          return NULL;
        }
      ++i;
    }

  return rp->sym->type_name;
}
\f
/*------------------------------------------------------------.
| Dump the string from FIN to OOUT if non null.  MATCH is the |
| delimiter of the string (either ' or ").                    |
`------------------------------------------------------------*/

static inline void
copy_string2 (FILE *fin, struct obstack *oout, int match, int store)
{
  int c;

  if (store)
    obstack_1grow (oout, match);

  c = getc (fin);

  while (c != match)
    {
      if (c == EOF)
        fatal (_("unterminated string at end of file"));
      if (c == '\n')
        {
          complain (_("unterminated string"));
          ungetc (c, fin);
          c = match;            /* invent terminator */
          continue;
        }

      obstack_1grow (oout, c);

      if (c == '\\')
        {
          c = getc (fin);
          if (c == EOF)
            fatal (_("unterminated string at end of file"));
          obstack_1grow (oout, c);

          if (c == '\n')
            ++lineno;
        }

      c = getc (fin);
    }

  if (store)
    obstack_1grow (oout, c);
}

/* FIXME. */

static inline void
copy_string (FILE *fin, struct obstack *oout, int match)
{
  copy_string2 (fin, oout, match, 1);
}

/* FIXME. */

static inline void
copy_identifier (FILE *fin, struct obstack *oout)
{
  int c;

  while (isalnum (c = getc (fin)) || c == '_')
    obstack_1grow (oout, c);

  ungetc (c, fin);
}


/*------------------------------------------------------------------.
| 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
copy_comment (FILE *fin, struct obstack *oout)
{
  int cplus_comment;
  int ended;
  int c;

  /* We read a `/', output it. */
  obstack_1grow (oout, '/');

  switch ((c = getc (fin)))
    {
    case '/':
      cplus_comment = 1;
      break;
    case '*':
      cplus_comment = 0;
      break;
    default:
      ungetc (c, fin);
      return;
    }

  obstack_1grow (oout, c);
  c = getc (fin);

  ended = 0;
  while (!ended)
    {
      if (!cplus_comment && c == '*')
        {
          while (c == '*')
            {
              obstack_1grow (oout, c);
              c = getc (fin);
            }

          if (c == '/')
            {
              obstack_1grow (oout, c);
              ended = 1;
            }
        }
      else if (c == '\n')
        {
          ++lineno;
          obstack_1grow (oout, c);
          if (cplus_comment)
            ended = 1;
          else
            c = getc (fin);
        }
      else if (c == EOF)
        fatal (_("unterminated comment"));
      else
        {
          obstack_1grow (oout, c);
          c = getc (fin);
        }
    }
}


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

static inline void
copy_at (FILE *fin, struct obstack *oout, int stack_offset)
{
  int c;

  c = getc (fin);
  if (c == '$')
    {
      obstack_sgrow (oout, "yyloc");
      locations_flag = 1;
    }
  else if (isdigit (c) || c == '-')
    {
      int n;

      ungetc (c, fin);
      n = read_signed_integer (fin);
      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
    {
      char buf[] = "@c";
      buf[1] = c;
      complain (_("%s is invalid"), quote (buf));
    }
}


/*-------------------------------------------------------------------.
| FIN is pointing to a wannabee semantic value (i.e., a `$').        |
|                                                                    |
| Possible inputs: $[<TYPENAME>]($|integer)                          |
|                                                                    |
| Output to OOUT a reference to this semantic value. 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.                 |
`-------------------------------------------------------------------*/

static inline void
copy_dollar (FILE *fin, struct obstack *oout,
             symbol_list *rule, int stack_offset)
{
  int c = getc (fin);
  const char *type_name = NULL;

  /* Get the type name if explicit. */
  if (c == '<')
    {
      read_type_name (fin);
      type_name = token_buffer;
      value_components_used = 1;
      c = getc (fin);
    }

  if (c == '$')
    {
      obstack_sgrow (oout, "yyval");

      if (!type_name)
        type_name = get_type_name (0, rule);
      if (type_name)
        obstack_fgrow1 (oout, ".%s", type_name);
      if (!type_name && typed)
        complain (_("$$ of `%s' has no declared type"),
                  rule->sym->tag);
    }
  else if (isdigit (c) || c == '-')
    {
      int n;
      ungetc (c, fin);
      n = read_signed_integer (fin);

      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 (_("$%d of `%s' has no declared type"),
                      n, rule->sym->tag);
        }
    }
  else
    {
      char buf[] = "$c";
      buf[1] = c;
      complain (_("%s is invalid"), quote (buf));
    }
}
\f
/*-------------------------------------------------------------------.
| Copy the contents of a `%{ ... %}' into the definitions file.  The |
| `%{' has already been read.  Return after reading the `%}'.        |
`-------------------------------------------------------------------*/

static void
copy_definition (void)
{
  int c;
  /* -1 while reading a character if prev char was %. */
  int after_percent;

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

  after_percent = 0;

  c = getc (finput);

  for (;;)
    {
      switch (c)
        {
        case '\n':
          obstack_1grow (&attrs_obstack, c);
          ++lineno;
          break;

        case '%':
          after_percent = -1;
          break;

        case '\'':
        case '"':
          copy_string (finput, &attrs_obstack, c);
          break;

        case '/':
          copy_comment (finput, &attrs_obstack);
          break;

        case EOF:
          fatal ("%s", _("unterminated `%{' definition"));

        default:
          obstack_1grow (&attrs_obstack, c);
        }

      c = getc (finput);

      if (after_percent)
        {
          if (c == '}')
            return;
          obstack_1grow (&attrs_obstack, '%');
        }
      after_percent = 0;
    }
}


/*-------------------------------------------------------------------.
| Parse what comes after %token or %nterm.  For %token, WHAT_IS is   |
| token_sym and WHAT_IS_NOT is nterm_sym.  For %nterm, the arguments |
| are reversed.                                                      |
`-------------------------------------------------------------------*/

static void
parse_token_decl (symbol_class what_is, symbol_class what_is_not)
{
  token_t token = tok_undef;
  char *typename = NULL;

  /* The symbol being defined.  */
  symbol_t *symbol = NULL;

  /* After `%token' and `%nterm', any number of symbols maybe be
     defined.  */
  for (;;)
    {
      int tmp_char = ungetc (skip_white_space (), finput);

      /* `%' (for instance from `%token', or from `%%' etc.) is the
         only valid means to end this declaration.  */
      if (tmp_char == '%')
        return;
      if (tmp_char == EOF)
        fatal (_("Premature EOF after %s"), token_buffer);

      token = lex ();
      if (token == tok_comma)
        {
          symbol = NULL;
          continue;
        }
      if (token == tok_typename)
        {
          typename = xstrdup (token_buffer);
          value_components_used = 1;
          symbol = NULL;
        }
      else if (token == tok_identifier && *symval->tag == '\"' && symbol)
        {
          symbol_make_alias (symbol, typename);
          symbol = NULL;
        }
      else if (token == tok_identifier)
        {
          int oldclass = symval->class;
          symbol = symval;

          if (symbol->class == what_is_not)
            complain (_("symbol %s redefined"), symbol->tag);
          symbol->class = what_is;
          if (what_is == nterm_sym && oldclass != nterm_sym)
            symbol->number = nvars++;
          if (what_is == token_sym && symbol->number == NUMBER_UNDEFINED)
            symbol->number = ntokens++;

          if (typename)
            {
              if (symbol->type_name == NULL)
                symbol->type_name = typename;
              else if (strcmp (typename, symbol->type_name) != 0)
                complain (_("type redeclaration for %s"), symbol->tag);
            }
        }
      else if (symbol && token == tok_number)
        {
          symbol->user_token_number = numval;
          /* User defined EOF token? */
          if (numval == 0)
            {
              eoftoken = symbol;
              eoftoken->number = 0;
              /* It is always mapped to 0, so it was already counted in
                 NTOKENS.  */
              --ntokens;
            }
        }
      else
        {
          complain (_("`%s' is invalid in %s"),
                    token_buffer,
                    (what_is == token_sym) ? "%token" : "%nterm");
          skip_to_char ('%');
        }
    }

}


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

static void
parse_start_decl (void)
{
  if (start_flag)
    complain (_("multiple %s declarations"), "%start");
  if (lex () != tok_identifier)
    complain (_("invalid %s declaration"), "%start");
  else
    {
      start_flag = 1;
      startval = symval;
    }
}

/*-----------------------------------------------------------.
| read in a %type declaration and record its information for |
| get_type_name to access                                    |
`-----------------------------------------------------------*/

static void
parse_type_decl (void)
{
  char *name;

  if (lex () != tok_typename)
    {
      complain ("%s", _("%type declaration has no <typename>"));
      skip_to_char ('%');
      return;
    }

  name = xstrdup (token_buffer);

  for (;;)
    {
      token_t t;
      int tmp_char = ungetc (skip_white_space (), finput);

      if (tmp_char == '%')
        return;
      if (tmp_char == EOF)
        fatal (_("Premature EOF after %s"), token_buffer);

      t = lex ();

      switch (t)
        {

        case tok_comma:
        case tok_semicolon:
          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:
          complain (_("invalid %%type declaration due to item: %s"),
                    token_buffer);
          skip_to_char ('%');
        }
    }
}



/*----------------------------------------------------------------.
| Read in a %left, %right or %nonassoc declaration and record its |
| information.                                                    |
`----------------------------------------------------------------*/

static void
parse_assoc_decl (associativity assoc)
{
  char *name = NULL;
  int prev = 0;

  /* Assign a new precedence level, never 0.  */
  ++lastprec;

  for (;;)
    {
      token_t t;
      int tmp_char = ungetc (skip_white_space (), finput);

      if (tmp_char == '%')
        return;
      if (tmp_char == EOF)
        fatal (_("Premature EOF after %s"), token_buffer);

      t = lex ();

      switch (t)
        {
        case tok_typename:
          name = xstrdup (token_buffer);
          break;

        case tok_comma:
          break;

        case tok_identifier:
          if (symval->prec != 0)
            complain (_("redefining precedence of %s"), symval->tag);
          symval->prec = lastprec;
          symval->assoc = assoc;
          if (symval->class == nterm_sym)
            complain (_("symbol %s redefined"), symval->tag);
          if (symval->number == NUMBER_UNDEFINED)
            {
              symval->number = ntokens++;
              symval->class = token_sym;
            }
          if (name)
            {                   /* record the type, if one is specified */
              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;

        case tok_number:
          if (prev == tok_identifier)
            {
              symval->user_token_number = numval;
            }
          else
            {
              complain
                (_("invalid text (%s) - number should be after identifier"),
                 token_buffer);
              skip_to_char ('%');
            }
          break;

        case tok_semicolon:
          return;

        default:
          complain (_("unexpected item: %s"), token_buffer);
          skip_to_char ('%');
        }

      prev = t;
    }
}



/*--------------------------------------------------------------.
| Copy the union declaration into the stype muscle              |
| (and fdefines),  where it is made into the definition of      |
| YYSTYPE, the type of elements of the parser value stack.      |
`--------------------------------------------------------------*/

static void
parse_union_decl (void)
{
  int c;
  int count = 0;
  bool done = FALSE;
  struct obstack union_obstack;
  if (typed)
    complain (_("multiple %s declarations"), "%union");

  typed = 1;

  MUSCLE_INSERT_INT ("stype_line", lineno);
  obstack_init (&union_obstack);
  obstack_sgrow (&union_obstack, "union");

  while (!done)
    {
      c = xgetc (finput);

      /* If C contains '/', it is output by copy_comment ().  */
      if (c != '/')
        obstack_1grow (&union_obstack, c);

      switch (c)
        {
        case '\n':
          ++lineno;
          break;

        case '/':
          copy_comment (finput, &union_obstack);
          break;

        case '{':
          ++count;
          break;

        case '}':
          /* FIXME: Errr.  How could this happen???. --akim */
          if (count == 0)
            complain (_("unmatched %s"), "`}'");
          count--;
          if (!count)
            done = TRUE;
          break;
        }
    }

  /* 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));
}


/*-------------------------------------------------------.
| Parse the declaration %expect N which says to expect N |
| shift-reduce conflicts.                                |
`-------------------------------------------------------*/

static void
parse_expect_decl (void)
{
  int c = skip_white_space ();
  ungetc (c, finput);

  if (!isdigit (c))
    complain (_("argument of %%expect is not an integer"));
  else
    expected_conflicts = read_signed_integer (finput);
}


/*-------------------------------------------------------------------.
| Parse what comes after %thong.  the full syntax is                 |
|                                                                    |
|                %thong <type> token number literal                  |
|                                                                    |
| the <type> or number may be omitted.  The number specifies the     |
| user_token_number.                                                 |
|                                                                    |
| Two symbols are entered in the table, one for the token symbol and |
| one for the literal.  Both are given the <type>, if any, from the  |
| declaration.  The ->user_token_number of the first is SALIAS and   |
| the ->user_token_number of the second is set to the number, if     |
| any, from the declaration.  The two symbols are linked via         |
| pointers in their ->alias fields.                                  |
|                                                                    |
| During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter,    |
| only the literal string is retained it is the literal string that  |
| is output to yytname                                               |
`-------------------------------------------------------------------*/

static void
parse_thong_decl (void)
{
  token_t token;
  symbol_t *symbol;
  char *typename = 0;
  int usrtoknum = SUNDEF;

  token = lex ();               /* fetch typename or first token */
  if (token == tok_typename)
    {
      typename = xstrdup (token_buffer);
      value_components_used = 1;
      token = lex ();           /* fetch first token */
    }

  /* process first token */

  if (token != tok_identifier)
    {
      complain (_("unrecognized item %s, expected an identifier"),
                token_buffer);
      skip_to_char ('%');
      return;
    }
  symval->class = token_sym;
  symval->type_name = typename;
  symval->user_token_number = SALIAS;
  symbol = symval;

  token = lex ();               /* get number or literal string */

  if (token == tok_number)
    {
      usrtoknum = numval;
      token = lex ();           /* okay, did number, now get literal */
    }

  /* process literal string token */

  if (token != tok_identifier || *symval->tag != '\"')
    {
      complain (_("expected string constant instead of %s"), token_buffer);
      skip_to_char ('%');
      return;
    }
  symval->class = token_sym;
  symval->type_name = typename;
  symval->user_token_number = usrtoknum;

  symval->alias = symbol;
  symbol->alias = symval;

  /* symbol and symval combined are only one symbol.  */
  nsyms--;
}


static void
parse_muscle_decl (void)
{
  int ch = ungetc (skip_white_space (), finput);
  char *muscle_key;
  char *muscle_value;

  /* Read key. */
  if (!isalpha (ch) && ch != '_')
    {
      complain (_("invalid %s declaration"), "%define");
      skip_to_char ('%');
      return;
    }
  copy_identifier (finput, &muscle_obstack);
  obstack_1grow (&muscle_obstack, 0);
  muscle_key = obstack_finish (&muscle_obstack);

  /* Read value. */
  ch = skip_white_space ();
  if (ch != '"')
    {
      ungetc (ch, finput);
      if (ch != EOF)
        {
          complain (_("invalid %s declaration"), "%define");
          skip_to_char ('%');
          return;
        }
      else
        fatal (_("Premature EOF after %s"), "\"");
    }
  copy_string2 (finput, &muscle_obstack, '"', 0);
  obstack_1grow (&muscle_obstack, 0);
  muscle_value = obstack_finish (&muscle_obstack);

  /* Store the (key, value) pair in the environment. */
  muscle_insert (muscle_key, muscle_value);
}



/*---------------------------------.
| 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. |
`----------------------------------*/

static void
parse_skel_decl (void)
{
  skeleton = parse_dquoted_param ("%skeleton");
}

/*----------------------------------------------------------------.
| Read from finput until `%%' is seen.  Discard the `%%'.  Handle |
| any `%' declarations, and copy the contents of any `%{ ... %}'  |
| groups to ATTRS_OBSTACK.                                        |
`----------------------------------------------------------------*/

static void
read_declarations (void)
{
  for (;;)
    {
      int c = skip_white_space ();

      if (c == '%')
        {
          token_t tok = parse_percent_token ();

          switch (tok)
            {
            case tok_two_percents:
              return;

            case tok_percent_left_curly:
              copy_definition ();
              break;

            case tok_token:
              parse_token_decl (token_sym, nterm_sym);
              break;

            case tok_nterm:
              parse_token_decl (nterm_sym, token_sym);
              break;

            case tok_type:
              parse_type_decl ();
              break;

            case tok_start:
              parse_start_decl ();
              break;

            case tok_union:
              parse_union_decl ();
              break;

            case tok_expect:
              parse_expect_decl ();
              break;

            case tok_thong:
              parse_thong_decl ();
              break;

            case tok_left:
              parse_assoc_decl (left_assoc);
              break;

            case tok_right:
              parse_assoc_decl (right_assoc);
              break;

            case tok_nonassoc:
              parse_assoc_decl (non_assoc);
              break;

            case tok_define:
              parse_muscle_decl ();
              break;

            case tok_skel:
              parse_skel_decl ();
              break;

            case tok_noop:
              break;

            case tok_stropt:
            case tok_intopt:
            case tok_obsolete:
              assert (0);
              break;

            case tok_illegal:
            default:
              complain (_("unrecognized: %s"), token_buffer);
              skip_to_char ('%');
            }
        }
      else if (c == EOF)
        fatal (_("no input grammar"));
      else
        {
          char buf[] = "c";
          buf[0] = c;
          complain (_("unknown character: %s"), quote (buf));
          skip_to_char ('%');
        }
    }
}
\f
/*-------------------------------------------------------------------.
| Assuming that a `{' has just been seen, copy everything up to the  |
| 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
parse_braces (symbol_list *rule, int stack_offset)
{
  int c;
  int count;

  count = 1;
  while (count > 0)
    {
      while ((c = getc (finput)) != '}')
        switch (c)
          {
          case '\n':
            obstack_1grow (&action_obstack, c);
            ++lineno;
            break;

          case '{':
            obstack_1grow (&action_obstack, c);
            ++count;
            break;

          case '\'':
          case '"':
            copy_string (finput, &action_obstack, c);
            break;

          case '/':
            copy_comment (finput, &action_obstack);
            break;

          case '$':
            copy_dollar (finput, &action_obstack,
                         rule, stack_offset);
            break;

          case '@':
            copy_at (finput, &action_obstack,
                     stack_offset);
            break;

          case EOF:
            fatal (_("unmatched %s"), "`{'");

          default:
            obstack_1grow (&action_obstack, c);
          }

      /* Above loop exits when C is '}'.  */
      if (--count)
        obstack_1grow (&action_obstack, c);
    }

  obstack_1grow (&action_obstack, '\0');
}


static void
parse_action (symbol_list *rule, int stack_offset)
{
  rule->action_line = lineno;
  parse_braces (rule, stack_offset);
  rule->action = obstack_finish (&action_obstack);
}


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);
}

\f

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

static symbol_t *
gensym (void)
{
  /* Incremented for each generated symbol */
  static int gensym_count = 0;
  static char buf[256];

  symbol_t *sym;

  sprintf (buf, "@%d", ++gensym_count);
  token_buffer = buf;
  sym = getsym (token_buffer);
  sym->class = nterm_sym;
  sym->number = nvars++;
  return sym;
}
\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.                         |
|                                                                    |
| 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;
  symbol_t *lhs = NULL;
  symbol_list *p = NULL;
  symbol_list *p1 = NULL;

  /* Points to first symbol_list of current rule. its symbol is the
     lhs of the rule.  */
  symbol_list *crule = NULL;
  /* Points to the symbol_list preceding crule.  */
  symbol_list *crule1 = NULL;

  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 */
        symbol_t *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 (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.  */

        ++nrules;
        ++nritems;

        p = symbol_list_new (lhs);

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

        p1 = p;
        crule = p;

        /* mark the rule's lhs as a nonterminal if not already so.  */

        if (lhs->class == unknown_sym)
          {
            lhs->class = nterm_sym;
            lhs->number = nvars;
            ++nvars;
          }
        else if (lhs->class == token_sym)
          complain (_("rule given for %s, which is a token"), lhs->tag);

        /* read the rhs of the rule.  */

        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)
              {
                symbol_t *ssave;
                token_t t1;

                ssave = symval;
                t1 = lex ();
                unlex (t1);
                symval = ssave;
                if (t1 == tok_colon)
                  {
                    warn (_("previous rule lacks an ending `;'"));
                    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.  */
                symbol_t *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;
                ++nritems;
                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
                  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.  */
                ++nritems;
                p = symbol_list_new (sdummy);
                p1->next = p;
                p1 = p;

                action_flag = 0;
              }

            if (t == tok_identifier)
              {
                ++nritems;
                p = symbol_list_new (symval);
                p1->next = p;
                p1 = p;
              }
            else                /* handle an action.  */
              {
                parse_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 ();
            crule->ruleprec = symval;
            t = lex ();
          }

        if (t == tok_guard)
          {
            if (!semantic_parser)
              complain (_("%%guard present but %%semantic_parser not specified"));

            parse_guard (crule, rulelength);
            t = lex ();
          }

        if (t == tok_left_curly)
          {
            /* This case never occurs -wjh */
            if (action_flag)
              complain (_("two actions at end of one rule"));
            parse_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_two_percents || t == tok_eof)
          warn (_("previous rule lacks an ending `;'"));
        if (t == tok_semicolon)
          t = lex ();
      }
    else
      {
        complain (_("invalid input: %s"), quote (token_buffer));
        t = lex ();
      }

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

  if (nrules == 0)
    fatal (_("no rules in the input grammar"));

  /* Report any undefined symbols and consider them nonterminals.  */
  symbols_do (symbol_check_defined, NULL);

  /* 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;
  nritems += 3;
  grammar = p;
  startval = axiom;

  if (nsyms > SHRT_MAX)
    fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
           SHRT_MAX);

  assert (nsyms == ntokens + nvars);
}

/* At the end of the grammar file, some C source code must
   be stored. It is going to be associated to the epilogue
   directive.  */
static void
read_additionnal_code (void)
{
  int c;
  struct obstack 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);

  obstack_1grow (&el_obstack, 0);
  muscle_insert ("epilogue", obstack_finish (&el_obstack));
}

\f
/*------------------------------------------------------------------.
| Set TOKEN_TRANSLATIONS.  Check that no two symbols share the same |
| number.                                                           |
`------------------------------------------------------------------*/

static void
token_translations_init (void)
{
  int last_user_token_number = 256;
  int i;

  /* Set the user numbers. */
  for (i = 0; i < ntokens; ++i)
    {
      symbol_t *this = symbols[i];
      if (this->user_token_number == SUNDEF)
        this->user_token_number = ++last_user_token_number;
      if (this->user_token_number > max_user_token_number)
        max_user_token_number = this->user_token_number;
    }

  token_translations = XCALLOC (token_number_t, 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 + 1; i++)
    token_translations[i] = undeftoken->number;

  symbols_do (symbol_translation, NULL);
}


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

static void
packsymbols (void)
{
  symbols = XCALLOC (symbol_t *, nsyms);

  symbols_do (symbol_check_alias_consistence, NULL);
  symbols_do (symbol_pack, NULL);

  token_translations_init ();

  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->number;
}


/*---------------------------------------------------------------.
| Convert the rules into the representation using RRHS, RLHS and |
| RITEM.                                                         |
`---------------------------------------------------------------*/

static void
packgram (void)
{
  int itemno;
  int ruleno;
  symbol_list *p;

  ritem = XCALLOC (item_number_t, nritems + 1);
  rules = XCALLOC (rule_t, nrules) - 1;

  itemno = 0;
  ruleno = 1;

  p = grammar;
  while (p)
    {
      symbol_t *ruleprec = p->ruleprec;
      rules[ruleno].user_number = ruleno;
      rules[ruleno].number = ruleno;
      rules[ruleno].lhs = p->sym;
      rules[ruleno].rhs = ritem + 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)
        {
          /* item_number_t = token_number_t.
             But the former needs to contain more: negative rule numbers. */
          ritem[itemno++] = token_number_as_item_number (p->sym->number);
          /* A rule gets by default the precedence and associativity
             of the last token in it.  */
          if (p->sym->class == token_sym)
            rules[ruleno].prec = p->sym;
          if (p)
            p = p->next;
        }

      /* If this rule has a %prec,
         the specified symbol's precedence replaces the default.  */
      if (ruleprec)
        {
          rules[ruleno].precsym = ruleprec;
          rules[ruleno].prec = ruleprec;
        }
      ritem[itemno++] = -ruleno;
      ++ruleno;

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

  ritem[itemno] = 0;
  assert (itemno == nritems);

  if (trace_flag)
    ritem_print (stderr);
}
\f
/*-------------------------------------------------------------------.
| Read in the grammar specification and record it in the format      |
| described in gram.h.  All guards are copied into the GUARD_OBSTACK |
| and all actions into ACTION_OBSTACK, in each case forming the body |
| of a C function (YYGUARD or YYACTION) which contains a switch      |
| statement to decide which guard or action to execute.              |
`-------------------------------------------------------------------*/

void
reader (void)
{
  lex_init ();
  lineno = 1;

  /* Initialize the muscle obstack.  */
  obstack_init (&muscle_obstack);

  /* Initialize the symbol table.  */
  symbols_new ();

  /* Construct the axiom symbol. */
  axiom = getsym ("$axiom");
  axiom->class = nterm_sym;
  axiom->number = nvars++;

  /* Construct the error token */
  errtoken = getsym ("error");
  errtoken->class = token_sym;
  errtoken->number = ntokens++;
  errtoken->user_token_number = 256;    /* Value specified by POSIX.  */

  /* Construct a token that represents all undefined literal tokens.
     It is always token number 2.  */
  undeftoken = getsym ("$undefined.");
  undeftoken->class = token_sym;
  undeftoken->number = ntokens++;

  /* Initialize the obstacks. */
  obstack_init (&action_obstack);
  obstack_init (&attrs_obstack);
  obstack_init (&output_obstack);

  finput = xfopen (infile, "r");

  /* Read the declaration section.  Copy %{ ... %} groups to
     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;
      eoftoken->number = 0;
      /* Value specified by POSIX.  */
      eoftoken->user_token_number = 0;
    }

  /* Read in the grammar, build grammar in list form.  Write out
     guards and actions.  */
  readgram ();
  /* Some C code is given at the end of the grammar file. */
  read_additionnal_code ();

  lex_free ();
  xfclose (finput);

  /* Assign the symbols their symbol numbers.  Write #defines for the
     token symbols into FDEFINES if requested.  */
  packsymbols ();

  /* 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);
}

void
grammar_free (void)
{
  XFREE (ritem);
  free (rules + 1);
  /* Free the symbol table data structure.  */
  symbols_free ();
}