src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007,
   4    2009-2012 Free Software Foundation, Inc.
   5
   6    This file is part of Bison, the GNU Compiler Compiler.
   7
   8    This program is free software: you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation, either version 3 of the License, or
  11    (at your option) any later version.
  12
  13    This program is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License
  19    along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
  20
  21 #include <config.h>
  22 #include "system.h"
  23
  24 #include <quote.h>
  25
  26 #include "complain.h"
  27 #include "conflicts.h"
  28 #include "files.h"
  29 #include "getargs.h"
  30 #include "gram.h"
  31 #include "muscle-tab.h"
  32 #include "reader.h"
  33 #include "symlist.h"
  34 #include "symtab.h"
  35 #include "scan-gram.h"
  36 #include "scan-code.h"
  37
  38 static void prepare_percent_define_front_end_variables (void);
  39 static void check_and_convert_grammar (void);
  40
  41 static symbol_list *grammar = NULL;
  42 static bool start_flag = false;
  43 merger_list *merge_functions;
  44
  45 /* Was %union seen?  */
  46 bool union_seen = false;
  47
  48 /* Was a tag seen?  */
  49 bool tag_seen = false;
  50
  51 /* Should rules have a default precedence?  */
  52 bool default_prec = true;
  53 \f
  54 /*-----------------------.
  55 | Set the start symbol.  |
  56 `-----------------------*/
  57
  58 void
  59 grammar_start_symbol_set (symbol *sym, location loc)
  60 {
  61   if (start_flag)
  62     complain_at (loc, complaint, _("multiple %s declarations"), "%start");
  63   else
  64     {
  65       start_flag = true;
  66       startsymbol = sym;
  67       startsymbol_location = loc;
  68     }
  69 }
  70
  71 \f
  72
  73 /*------------------------------------------------------------------------.
  74 | Return the merger index for a merging function named NAME.  Records the |
  75 | function, if new, in MERGER_LIST.                                       |
  76 `------------------------------------------------------------------------*/
  77
  78 static int
  79 get_merge_function (uniqstr name)
  80 {
  81   merger_list *syms;
  82   merger_list head;
  83   int n;
  84
  85   if (! glr_parser)
  86     return 0;
  87
  88   head.next = merge_functions;
  89   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
  90     if (UNIQSTR_EQ (name, syms->next->name))
  91       break;
  92   if (syms->next == NULL)
  93     {
  94       syms->next = xmalloc (sizeof syms->next[0]);
  95       syms->next->name = uniqstr_new (name);
  96       /* After all symbol type declarations have been parsed, packgram invokes
  97          record_merge_function_type to set the type.  */
  98       syms->next->type = NULL;
  99       syms->next->next = NULL;
 100       merge_functions = head.next;
 101     }
 102   return n;
 103 }
 104
 105 /*-------------------------------------------------------------------------.
 106 | For the existing merging function with index MERGER, record the result   |
 107 | type as TYPE as required by the lhs of the rule whose %merge declaration |
 108 | is at DECLARATION_LOC.                                                   |
 109 `-------------------------------------------------------------------------*/
 110
 111 static void
 112 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
 113 {
 114   int merger_find;
 115   merger_list *merge_function;
 116
 117   if (merger <= 0)
 118     return;
 119
 120   if (type == NULL)
 121     type = uniqstr_new ("");
 122
 123   merger_find = 1;
 124   for (merge_function = merge_functions;
 125        merge_function != NULL && merger_find != merger;
 126        merge_function = merge_function->next)
 127     merger_find += 1;
 128   aver (merge_function != NULL && merger_find == merger);
 129   if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
 130     {
 131       complain_at (declaration_loc, complaint,
 132                    _("result type clash on merge function %s: <%s> != <%s>"),
 133                    quote (merge_function->name), type, merge_function->type);
 134       complain_at (merge_function->type_declaration_location, complaint,
 135                    _("previous declaration"));
 136     }
 137   merge_function->type = uniqstr_new (type);
 138   merge_function->type_declaration_location = declaration_loc;
 139 }
 140
 141 /*--------------------------------------.
 142 | Free all merge-function definitions.  |
 143 `--------------------------------------*/
 144
 145 void
 146 free_merger_functions (void)
 147 {
 148   merger_list *L0 = merge_functions;
 149   while (L0)
 150     {
 151       merger_list *L1 = L0->next;
 152       free (L0);
 153       L0 = L1;
 154     }
 155 }
 156
 157 \f
 158 /*-------------------------------------------------------------------.
 159 | Parse the input grammar into a one symbol_list structure.  Each    |
 160 | rule is represented by a sequence of symbols: the left hand side   |
 161 | followed by the contents of the right hand side, followed by a     |
 162 | null pointer instead of a symbol to terminate the rule.  The next  |
 163 | symbol is the lhs of the following rule.                           |
 164 |                                                                    |
 165 | All actions are copied out, labelled by the rule number they apply |
 166 | to.                                                                |
 167 `-------------------------------------------------------------------*/
 168
 169 /* The (currently) last symbol of GRAMMAR. */
 170 static symbol_list *grammar_end = NULL;
 171
 172 /* Append SYM to the grammar.  */
 173 static symbol_list *
 174 grammar_symbol_append (symbol *sym, location loc)
 175 {
 176   symbol_list *p = symbol_list_sym_new (sym, loc);
 177
 178   if (grammar_end)
 179     grammar_end->next = p;
 180   else
 181     grammar = p;
 182
 183   grammar_end = p;
 184
 185   /* A null SYM stands for an end of rule; it is not an actual
 186      part of it.  */
 187   if (sym)
 188     ++nritems;
 189
 190   return p;
 191 }
 192
 193 static void
 194 assign_named_ref (symbol_list *p, named_ref *name)
 195 {
 196   symbol *sym = p->content.sym;
 197
 198   if (name->id == sym->tag)
 199     {
 200       complain_at (name->loc, Wother,
 201                    _("duplicated symbol name for %s ignored"),
 202                    quote (sym->tag));
 203       named_ref_free (name);
 204     }
 205   else
 206     p->named_ref = name;
 207 }
 208
 209
 210 /* The rule currently being defined, and the previous rule.
 211    CURRENT_RULE points to the first LHS of the current rule, while
 212    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 213 static symbol_list *current_rule = NULL;
 214 static symbol_list *previous_rule_end = NULL;
 215
 216
 217 /*----------------------------------------------.
 218 | Create a new rule for LHS in to the GRAMMAR.  |
 219 `----------------------------------------------*/
 220
 221 void
 222 grammar_current_rule_begin (symbol *lhs, location loc,
 223                             named_ref *lhs_name)
 224 {
 225   symbol_list* p;
 226
 227   /* Start a new rule and record its lhs.  */
 228   ++nrules;
 229   previous_rule_end = grammar_end;
 230
 231   p = grammar_symbol_append (lhs, loc);
 232   if (lhs_name)
 233     assign_named_ref (p, named_ref_copy (lhs_name));
 234
 235   current_rule = grammar_end;
 236
 237   /* Mark the rule's lhs as a nonterminal if not already so.  */
 238   if (lhs->class == unknown_sym)
 239     {
 240       lhs->class = nterm_sym;
 241       lhs->number = nvars;
 242       ++nvars;
 243     }
 244   else if (lhs->class == token_sym)
 245     complain_at (loc, complaint, _("rule given for %s, which is a token"),
 246                  lhs->tag);
 247 }
 248
 249
 250 /*----------------------------------------------------------------------.
 251 | A symbol should be used if either:                                    |
 252 |   1. It has a destructor.                                             |
 253 |   2. The symbol is a mid-rule symbol (i.e., the generated LHS         |
 254 |      replacing a mid-rule action) that was assigned to or used, as in |
 255 |      "exp: { $$ = 1; } { $$ = $1; }".                                 |
 256 `----------------------------------------------------------------------*/
 257
 258 static bool
 259 symbol_should_be_used (symbol_list const *s, bool *midrule_warning)
 260 {
 261   if (symbol_code_props_get (s->content.sym, destructor)->code)
 262     return true;
 263   if ((s->midrule && s->midrule->action_props.is_value_used)
 264       || (s->midrule_parent_rule
 265           && symbol_list_n_get (s->midrule_parent_rule,
 266                                 s->midrule_parent_rhs_index)
 267                ->action_props.is_value_used))
 268     {
 269       *midrule_warning = true;
 270       return true;
 271     }
 272   return false;
 273 }
 274
 275 /*----------------------------------------------------------------.
 276 | Check that the rule R is properly defined.  For instance, there |
 277 | should be no type clash on the default action.                  |
 278 `----------------------------------------------------------------*/
 279
 280 static void
 281 grammar_rule_check (const symbol_list *r)
 282 {
 283   /* Type check.
 284
 285      If there is an action, then there is nothing we can do: the user
 286      is allowed to shoot herself in the foot.
 287
 288      Don't worry about the default action if $$ is untyped, since $$'s
 289      value can't be used.  */
 290   if (!r->action_props.code && r->content.sym->type_name)
 291     {
 292       symbol *first_rhs = r->next->content.sym;
 293       /* If $$ is being set in default way, report if any type mismatch.  */
 294       if (first_rhs)
 295         {
 296           char const *lhs_type = r->content.sym->type_name;
 297           const char *rhs_type =
 298             first_rhs->type_name ? first_rhs->type_name : "";
 299           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 300             complain_at (r->location, Wother,
 301                          _("type clash on default action: <%s> != <%s>"),
 302                          lhs_type, rhs_type);
 303         }
 304       /* Warn if there is no default for $$ but we need one.  */
 305       else
 306         complain_at (r->location, Wother,
 307                      _("empty rule for typed nonterminal, and no action"));
 308     }
 309
 310   /* Check that symbol values that should be used are in fact used.  */
 311   {
 312     symbol_list const *l = r;
 313     int n = 0;
 314     for (; l && l->content.sym; l = l->next, ++n)
 315       {
 316         bool midrule_warning = false;
 317         if (!l->action_props.is_value_used
 318             && symbol_should_be_used (l, &midrule_warning)
 319             /* The default action, $$ = $1, `uses' both.  */
 320             && (r->action_props.code || (n != 0 && n != 1)))
 321           {
 322             warnings warn_flag = midrule_warning ? Wmidrule_values : Wother;
 323             if (n)
 324               complain_at (r->location, warn_flag, _("unused value: $%d"), n);
 325             else
 326               complain_at (r->location, warn_flag, _("unset value: $$"));
 327           }
 328       }
 329   }
 330
 331   /* See comments in grammar_current_rule_prec_set for how POSIX
 332      mandates this complaint.  It's only for identifiers, so skip
 333      it for char literals and strings, which are always tokens.  */
 334   if (r->ruleprec
 335       && r->ruleprec->tag[0] != '\'' && r->ruleprec->tag[0] != '"'
 336       && r->ruleprec->status != declared && !r->ruleprec->prec)
 337     complain_at (r->location, Wother,
 338                  _("token for %%prec is not defined: %s"), r->ruleprec->tag);
 339 }
 340
 341
 342 /*-------------------------------------.
 343 | End the currently being grown rule.  |
 344 `-------------------------------------*/
 345
 346 void
 347 grammar_current_rule_end (location loc)
 348 {
 349   /* Put an empty link in the list to mark the end of this rule  */
 350   grammar_symbol_append (NULL, grammar_end->location);
 351   current_rule->location = loc;
 352 }
 353
 354
 355 /*-------------------------------------------------------------------.
 356 | The previous action turns out the be a mid-rule action.  Attach it |
 357 | to the current rule, i.e., create a dummy symbol, attach it this   |
 358 | mid-rule action, and append this dummy nonterminal to the current  |
 359 | rule.                                                              |
 360 `-------------------------------------------------------------------*/
 361
 362 void
 363 grammar_midrule_action (void)
 364 {
 365   /* Since the action was written out with this rule's number, we must
 366      give the new rule this number by inserting the new rule before
 367      it.  */
 368
 369   /* Make a DUMMY nonterminal, whose location is that of the midrule
 370      action.  Create the MIDRULE.  */
 371   location dummy_location = current_rule->action_props.location;
 372   symbol *dummy = dummy_symbol_get (dummy_location);
 373   symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
 374
 375   /* Remember named_ref of previous action. */
 376   named_ref *action_name = current_rule->action_props.named_ref;
 377
 378   /* Make a new rule, whose body is empty, before the current one, so
 379      that the action just read can belong to it.  */
 380   ++nrules;
 381   ++nritems;
 382   /* Attach its location and actions to that of the DUMMY.  */
 383   midrule->location = dummy_location;
 384   code_props_rule_action_init (&midrule->action_props,
 385                                current_rule->action_props.code,
 386                                current_rule->action_props.location,
 387                                midrule, 0,
 388                                current_rule->action_props.is_predicate);
 389   code_props_none_init (&current_rule->action_props);
 390
 391   if (previous_rule_end)
 392     previous_rule_end->next = midrule;
 393   else
 394     grammar = midrule;
 395
 396   /* End the dummy's rule.  */
 397   midrule->next = symbol_list_sym_new (NULL, dummy_location);
 398   midrule->next->next = current_rule;
 399
 400   previous_rule_end = midrule->next;
 401
 402   /* Insert the dummy nonterminal replacing the midrule action into
 403      the current rule.  Bind it to its dedicated rule.  */
 404   grammar_current_rule_symbol_append (dummy, dummy_location,
 405                                       action_name);
 406   grammar_end->midrule = midrule;
 407   midrule->midrule_parent_rule = current_rule;
 408   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 409 }
 410
 411 /* Set the precedence symbol of the current rule to PRECSYM. */
 412
 413 void
 414 grammar_current_rule_prec_set (symbol *precsym, location loc)
 415 {
 416   /* POSIX says that any identifier is a nonterminal if it does not
 417      appear on the LHS of a grammar rule and is not defined by %token
 418      or by one of the directives that assigns precedence to a token.  We
 419      ignore this here because the only kind of identifier that POSIX
 420      allows to follow a %prec is a token and because assuming it's a
 421      token now can produce more logical error messages.  Nevertheless,
 422      grammar_rule_check does obey what we believe is the real intent of
 423      POSIX here: that an error be reported for any identifier that
 424      appears after %prec but that is not defined separately as a
 425      token.  */
 426   symbol_class_set (precsym, token_sym, loc, false);
 427   if (current_rule->ruleprec)
 428     complain_at (loc, complaint, _("only one %s allowed per rule"), "%prec");
 429   current_rule->ruleprec = precsym;
 430 }
 431
 432 /* Attach dynamic precedence DPREC to the current rule. */
 433
 434 void
 435 grammar_current_rule_dprec_set (int dprec, location loc)
 436 {
 437   if (! glr_parser)
 438     complain_at (loc, Wother, _("%s affects only GLR parsers"),
 439                  "%dprec");
 440   if (dprec <= 0)
 441     complain_at (loc, complaint, _("%s must be followed by positive number"),
 442                  "%dprec");
 443   else if (current_rule->dprec != 0)
 444     complain_at (loc, complaint, _("only one %s allowed per rule"), "%dprec");
 445   current_rule->dprec = dprec;
 446 }
 447
 448 /* Attach a merge function NAME with argument type TYPE to current
 449    rule. */
 450
 451 void
 452 grammar_current_rule_merge_set (uniqstr name, location loc)
 453 {
 454   if (! glr_parser)
 455     complain_at (loc, Wother, _("%s affects only GLR parsers"),
 456                  "%merge");
 457   if (current_rule->merger != 0)
 458     complain_at (loc, complaint, _("only one %s allowed per rule"), "%merge");
 459   current_rule->merger = get_merge_function (name);
 460   current_rule->merger_declaration_location = loc;
 461 }
 462
 463 /* Attach SYM to the current rule.  If needed, move the previous
 464    action as a mid-rule action.  */
 465
 466 void
 467 grammar_current_rule_symbol_append (symbol *sym, location loc,
 468                                     named_ref *name)
 469 {
 470   symbol_list *p;
 471   if (current_rule->action_props.code)
 472     grammar_midrule_action ();
 473   p = grammar_symbol_append (sym, loc);
 474   if (name)
 475     assign_named_ref(p, name);
 476   if (sym->status == undeclared || sym->status == used)
 477     sym->status = needed;
 478 }
 479
 480 /* Attach an ACTION to the current rule.  */
 481
 482 void
 483 grammar_current_rule_action_append (const char *action, location loc,
 484                                     named_ref *name, bool is_predicate)
 485 {
 486   if (current_rule->action_props.code)
 487     grammar_midrule_action ();
 488   /* After all symbol declarations have been parsed, packgram invokes
 489      code_props_translate_code.  */
 490   code_props_rule_action_init (&current_rule->action_props, action, loc,
 491                                current_rule, name, is_predicate);
 492 }
 493
 494 \f
 495 /*---------------------------------------------------------------.
 496 | Convert the rules into the representation using RRHS, RLHS and |
 497 | RITEM.                                                         |
 498 `---------------------------------------------------------------*/
 499
 500 static void
 501 packgram (void)
 502 {
 503   unsigned int itemno = 0;
 504   rule_number ruleno = 0;
 505   symbol_list *p = grammar;
 506
 507   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 508
 509   /* This sentinel is used by build_relations in gram.c.  */
 510   *ritem++ = 0;
 511
 512   rules = xnmalloc (nrules, sizeof *rules);
 513
 514   while (p)
 515     {
 516       int rule_length = 0;
 517       symbol *ruleprec = p->ruleprec;
 518       record_merge_function_type (p->merger, p->content.sym->type_name,
 519                                   p->merger_declaration_location);
 520       rules[ruleno].user_number = ruleno;
 521       rules[ruleno].number = ruleno;
 522       rules[ruleno].lhs = p->content.sym;
 523       rules[ruleno].rhs = ritem + itemno;
 524       rules[ruleno].prec = NULL;
 525       rules[ruleno].dprec = p->dprec;
 526       rules[ruleno].merger = p->merger;
 527       rules[ruleno].precsym = NULL;
 528       rules[ruleno].location = p->location;
 529       rules[ruleno].useful = true;
 530       rules[ruleno].action = p->action_props.code;
 531       rules[ruleno].action_location = p->action_props.location;
 532       rules[ruleno].is_predicate = p->action_props.is_predicate;
 533
 534       /* If the midrule's $$ is set or its $n is used, remove the `$' from the
 535          symbol name so that it's a user-defined symbol so that the default
 536          %destructor and %printer apply.  */
 537       if (p->midrule_parent_rule
 538           && (p->action_props.is_value_used
 539               || symbol_list_n_get (p->midrule_parent_rule,
 540                                     p->midrule_parent_rhs_index)
 541                    ->action_props.is_value_used))
 542         p->content.sym->tag += 1;
 543
 544       /* Don't check the generated rule 0.  It has no action, so some rhs
 545          symbols may appear unused, but the parsing algorithm ensures that
 546          %destructor's are invoked appropriately.  */
 547       if (p != grammar)
 548         grammar_rule_check (p);
 549
 550       for (p = p->next; p && p->content.sym; p = p->next)
 551         {
 552           ++rule_length;
 553
 554           /* Don't allow rule_length == INT_MAX, since that might
 555              cause confusion with strtol if INT_MAX == LONG_MAX.  */
 556           if (rule_length == INT_MAX)
 557             complain_at (rules[ruleno].location, fatal, _("rule is too long"));
 558
 559           /* item_number = symbol_number.
 560              But the former needs to contain more: negative rule numbers. */
 561           ritem[itemno++] =
 562             symbol_number_as_item_number (p->content.sym->number);
 563           /* A rule gets by default the precedence and associativity
 564              of its last token.  */
 565           if (p->content.sym->class == token_sym && default_prec)
 566             rules[ruleno].prec = p->content.sym;
 567         }
 568
 569       /* If this rule has a %prec,
 570          the specified symbol's precedence replaces the default.  */
 571       if (ruleprec)
 572         {
 573           rules[ruleno].precsym = ruleprec;
 574           rules[ruleno].prec = ruleprec;
 575         }
 576       /* An item ends by the rule number (negated).  */
 577       ritem[itemno++] = rule_number_as_item_number (ruleno);
 578       aver (itemno < ITEM_NUMBER_MAX);
 579       ++ruleno;
 580       aver (ruleno < RULE_NUMBER_MAX);
 581
 582       if (p)
 583         p = p->next;
 584     }
 585
 586   aver (itemno == nritems);
 587
 588   if (trace_flag & trace_sets)
 589     ritem_print (stderr);
 590 }
 591 \f
 592 /*------------------------------------------------------------------.
 593 | Read in the grammar specification and record it in the format     |
 594 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 595 | in each case forming the body of a C function (YYACTION) which    |
 596 | contains a switch statement to decide which action to execute.    |
 597 `------------------------------------------------------------------*/
 598
 599 void
 600 reader (void)
 601 {
 602   /* Initialize the symbol table.  */
 603   symbols_new ();
 604
 605   /* Construct the accept symbol. */
 606   accept = symbol_get ("$accept", empty_location);
 607   accept->class = nterm_sym;
 608   accept->number = nvars++;
 609
 610   /* Construct the error token */
 611   errtoken = symbol_get ("error", empty_location);
 612   errtoken->class = token_sym;
 613   errtoken->number = ntokens++;
 614
 615   /* Construct a token that represents all undefined literal tokens.
 616      It is always token number 2.  */
 617   undeftoken = symbol_get ("$undefined", empty_location);
 618   undeftoken->class = token_sym;
 619   undeftoken->number = ntokens++;
 620
 621   gram_in = xfopen (grammar_file, "r");
 622
 623   gram__flex_debug = trace_flag & trace_scan;
 624   gram_debug = trace_flag & trace_parse;
 625   gram_scanner_initialize ();
 626   gram_parse ();
 627   prepare_percent_define_front_end_variables ();
 628
 629   if (! complaint_issued)
 630     check_and_convert_grammar ();
 631
 632   xfclose (gram_in);
 633 }
 634
 635 static void
 636 prepare_percent_define_front_end_variables (void)
 637 {
 638   /* Set %define front-end variable defaults.  */
 639   muscle_percent_define_default ("lr.keep-unreachable-states", "false");
 640   {
 641     char *lr_type;
 642     /* IELR would be a better default, but LALR is historically the
 643        default.  */
 644     muscle_percent_define_default ("lr.type", "lalr");
 645     lr_type = muscle_percent_define_get ("lr.type");
 646     if (STRNEQ (lr_type, "canonical-lr"))
 647       muscle_percent_define_default ("lr.default-reductions", "most");
 648     else
 649       muscle_percent_define_default ("lr.default-reductions", "accepting");
 650     free (lr_type);
 651   }
 652
 653   /* Check %define front-end variables.  */
 654   {
 655     static char const * const values[] = {
 656       "lr.type", "lalr", "ielr", "canonical-lr", NULL,
 657       "lr.default-reductions", "most", "consistent", "accepting", NULL,
 658       NULL
 659     };
 660     muscle_percent_define_check_values (values);
 661   }
 662 }
 663
 664
 665 /*-------------------------------------------------------------.
 666 | Check the grammar that has just been read, and convert it to |
 667 | internal form.                                               |
 668 `-------------------------------------------------------------*/
 669
 670 static void
 671 check_and_convert_grammar (void)
 672 {
 673   /* Grammar has been read.  Do some checking.  */
 674   if (nrules == 0)
 675     complain (fatal, _("no rules in the input grammar"));
 676
 677   /* If the user did not define her ENDTOKEN, do it now. */
 678   if (!endtoken)
 679     {
 680       endtoken = symbol_get ("$end", empty_location);
 681       endtoken->class = token_sym;
 682       endtoken->number = 0;
 683       /* Value specified by POSIX.  */
 684       endtoken->user_token_number = 0;
 685     }
 686
 687   /* Report any undefined symbols and consider them nonterminals.  */
 688   symbols_check_defined ();
 689
 690   /* Find the start symbol if no %start.  */
 691   if (!start_flag)
 692     {
 693       symbol_list *node;
 694       for (node = grammar;
 695            node != NULL && symbol_is_dummy (node->content.sym);
 696            node = node->next)
 697         {
 698           for (node = node->next;
 699                node != NULL && node->content.sym != NULL;
 700                node = node->next)
 701             ;
 702         }
 703       aver (node != NULL);
 704       grammar_start_symbol_set (node->content.sym,
 705                                 node->content.sym->location);
 706     }
 707
 708   /* Insert the initial rule, whose line is that of the first rule
 709      (not that of the start symbol):
 710
 711      accept: %start EOF.  */
 712   {
 713     symbol_list *p = symbol_list_sym_new (accept, empty_location);
 714     p->location = grammar->location;
 715     p->next = symbol_list_sym_new (startsymbol, empty_location);
 716     p->next->next = symbol_list_sym_new (endtoken, empty_location);
 717     p->next->next->next = symbol_list_sym_new (NULL, empty_location);
 718     p->next->next->next->next = grammar;
 719     nrules += 1;
 720     nritems += 3;
 721     grammar = p;
 722   }
 723
 724   aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 725
 726   /* Assign the symbols their symbol numbers.  Write #defines for the
 727      token symbols into FDEFINES if requested.  */
 728   symbols_pack ();
 729
 730   /* Scan rule actions after invoking symbol_check_alias_consistency (in
 731      symbols_pack above) so that token types are set correctly before the rule
 732      action type checking.
 733
 734      Before invoking grammar_rule_check (in packgram below) on any rule, make
 735      sure all actions have already been scanned in order to set `used' flags.
 736      Otherwise, checking that a midrule's $$ should be set will not always work
 737      properly because the check must forward-reference the midrule's parent
 738      rule.  For the same reason, all the `used' flags must be set before
 739      checking whether to remove `$' from any midrule symbol name (also in
 740      packgram).  */
 741   {
 742     symbol_list *sym;
 743     for (sym = grammar; sym; sym = sym->next)
 744       code_props_translate_code (&sym->action_props);
 745   }
 746
 747   /* Convert the grammar into the format described in gram.h.  */
 748   packgram ();
 749
 750   /* The grammar as a symbol_list is no longer needed. */
 751   symbol_list_free (grammar);
 752 }