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