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, _("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,
 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,
 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       warn_at (name->loc,
 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, _("rule given for %s, which is a token"), lhs->tag);
 246 }
 247
 248
 249 /*----------------------------------------------------------------------.
 250 | A symbol should be used if either:                                    |
 251 |   1. It has a destructor.                                             |
 252 |   2. The symbol is a mid-rule symbol (i.e., the generated LHS         |
 253 |      replacing a mid-rule action) that was assigned to or used, as in |
 254 |      "exp: { $$ = 1; } { $$ = $1; }".                                 |
 255 `----------------------------------------------------------------------*/
 256
 257 static bool
 258 symbol_should_be_used (symbol_list const *s, bool *midrule_warning)
 259 {
 260   if (symbol_destructor_get (s->content.sym)->code)
 261     return true;
 262   if ((s->midrule && s->midrule->action_props.is_value_used)
 263       || (s->midrule_parent_rule
 264           && symbol_list_n_get (s->midrule_parent_rule,
 265                                 s->midrule_parent_rhs_index)
 266                ->action_props.is_value_used))
 267     {
 268       *midrule_warning = true;
 269       return true;
 270     }
 271   return false;
 272 }
 273
 274 /*----------------------------------------------------------------.
 275 | Check that the rule R is properly defined.  For instance, there |
 276 | should be no type clash on the default action.                  |
 277 `----------------------------------------------------------------*/
 278
 279 static void
 280 grammar_rule_check (const symbol_list *r)
 281 {
 282   /* Type check.
 283
 284      If there is an action, then there is nothing we can do: the user
 285      is allowed to shoot herself in the foot.
 286
 287      Don't worry about the default action if $$ is untyped, since $$'s
 288      value can't be used.  */
 289   if (!r->action_props.code && r->content.sym->type_name)
 290     {
 291       symbol *first_rhs = r->next->content.sym;
 292       /* If $$ is being set in default way, report if any type mismatch.  */
 293       if (first_rhs)
 294         {
 295           char const *lhs_type = r->content.sym->type_name;
 296           const char *rhs_type =
 297             first_rhs->type_name ? first_rhs->type_name : "";
 298           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 299             warn_at (r->location,
 300                      _("type clash on default action: <%s> != <%s>"),
 301                      lhs_type, rhs_type);
 302         }
 303       /* Warn if there is no default for $$ but we need one.  */
 304       else
 305         warn_at (r->location,
 306                  _("empty rule for typed nonterminal, and no action"));
 307     }
 308
 309   /* Check that symbol values that should be used are in fact used.  */
 310   {
 311     symbol_list const *l = r;
 312     int n = 0;
 313     for (; l && l->content.sym; l = l->next, ++n)
 314       {
 315         bool midrule_warning = false;
 316         if (!l->action_props.is_value_used
 317             && symbol_should_be_used (l, &midrule_warning)
 318             /* The default action, $$ = $1, `uses' both.  */
 319             && (r->action_props.code || (n != 0 && n != 1)))
 320           {
 321             void (*warn_at_ptr)(location, char const*, ...) =
 322               midrule_warning ? midrule_value_at : warn_at;
 323             if (n)
 324               warn_at_ptr (r->location, _("unused value: $%d"), n);
 325             else
 326               warn_at_ptr (r->location, _("unset value: $$"));
 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->declared && !r->ruleprec->prec)
 337     warn_at (r->location, _("token for %%prec is not defined: %s"),
 338              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   code_props_none_init (&current_rule->action_props);
 389
 390   if (previous_rule_end)
 391     previous_rule_end->next = midrule;
 392   else
 393     grammar = midrule;
 394
 395   /* End the dummy's rule.  */
 396   midrule->next = symbol_list_sym_new (NULL, dummy_location);
 397   midrule->next->next = current_rule;
 398
 399   previous_rule_end = midrule->next;
 400
 401   /* Insert the dummy nonterminal replacing the midrule action into
 402      the current rule.  Bind it to its dedicated rule.  */
 403   grammar_current_rule_symbol_append (dummy, dummy_location,
 404                                       action_name);
 405   grammar_end->midrule = midrule;
 406   midrule->midrule_parent_rule = current_rule;
 407   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 408 }
 409
 410 /* Set the precedence symbol of the current rule to PRECSYM. */
 411
 412 void
 413 grammar_current_rule_prec_set (symbol *precsym, location loc)
 414 {
 415   /* POSIX says that any identifier is a nonterminal if it does not
 416      appear on the LHS of a grammar rule and is not defined by %token
 417      or by one of the directives that assigns precedence to a token.  We
 418      ignore this here because the only kind of identifier that POSIX
 419      allows to follow a %prec is a token and because assuming it's a
 420      token now can produce more logical error messages.  Nevertheless,
 421      grammar_rule_check does obey what we believe is the real intent of
 422      POSIX here: that an error be reported for any identifier that
 423      appears after %prec but that is not defined separately as a
 424      token.  */
 425   symbol_class_set (precsym, token_sym, loc, false);
 426   if (current_rule->ruleprec)
 427     complain_at (loc, _("only one %s allowed per rule"), "%prec");
 428   current_rule->ruleprec = precsym;
 429 }
 430
 431 /* Attach dynamic precedence DPREC to the current rule. */
 432
 433 void
 434 grammar_current_rule_dprec_set (int dprec, location loc)
 435 {
 436   if (! glr_parser)
 437     warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
 438   if (dprec <= 0)
 439     complain_at (loc, _("%s must be followed by positive number"), "%dprec");
 440   else if (current_rule->dprec != 0)
 441     complain_at (loc, _("only one %s allowed per rule"), "%dprec");
 442   current_rule->dprec = dprec;
 443 }
 444
 445 /* Attach a merge function NAME with argument type TYPE to current
 446    rule. */
 447
 448 void
 449 grammar_current_rule_merge_set (uniqstr name, location loc)
 450 {
 451   if (! glr_parser)
 452     warn_at (loc, _("%s affects only GLR parsers"), "%merge");
 453   if (current_rule->merger != 0)
 454     complain_at (loc, _("only one %s allowed per rule"), "%merge");
 455   current_rule->merger = get_merge_function (name);
 456   current_rule->merger_declaration_location = loc;
 457 }
 458
 459 /* Attach SYM to the current rule.  If needed, move the previous
 460    action as a mid-rule action.  */
 461
 462 void
 463 grammar_current_rule_symbol_append (symbol *sym, location loc,
 464                                     named_ref *name)
 465 {
 466   symbol_list *p;
 467   if (current_rule->action_props.code)
 468     grammar_midrule_action ();
 469   p = grammar_symbol_append (sym, loc);
 470   if (name)
 471     assign_named_ref(p, name);
 472 }
 473
 474 /* Attach an ACTION to the current rule.  */
 475
 476 void
 477 grammar_current_rule_action_append (const char *action, location loc,
 478                                     named_ref *name)
 479 {
 480   if (current_rule->action_props.code)
 481     grammar_midrule_action ();
 482   /* After all symbol declarations have been parsed, packgram invokes
 483      code_props_translate_code.  */
 484   code_props_rule_action_init (&current_rule->action_props, action, loc,
 485                                current_rule, name);
 486 }
 487
 488 \f
 489 /*---------------------------------------------------------------.
 490 | Convert the rules into the representation using RRHS, RLHS and |
 491 | RITEM.                                                         |
 492 `---------------------------------------------------------------*/
 493
 494 static void
 495 packgram (void)
 496 {
 497   unsigned int itemno = 0;
 498   rule_number ruleno = 0;
 499   symbol_list *p = grammar;
 500
 501   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 502
 503   /* This sentinel is used by build_relations in gram.c.  */
 504   *ritem++ = 0;
 505
 506   rules = xnmalloc (nrules, sizeof *rules);
 507
 508   while (p)
 509     {
 510       int rule_length = 0;
 511       symbol *ruleprec = p->ruleprec;
 512       record_merge_function_type (p->merger, p->content.sym->type_name,
 513                                   p->merger_declaration_location);
 514       rules[ruleno].user_number = ruleno;
 515       rules[ruleno].number = ruleno;
 516       rules[ruleno].lhs = p->content.sym;
 517       rules[ruleno].rhs = ritem + itemno;
 518       rules[ruleno].prec = NULL;
 519       rules[ruleno].dprec = p->dprec;
 520       rules[ruleno].merger = p->merger;
 521       rules[ruleno].precsym = NULL;
 522       rules[ruleno].location = p->location;
 523       rules[ruleno].useful = true;
 524       rules[ruleno].action = p->action_props.code;
 525       rules[ruleno].action_location = p->action_props.location;
 526
 527       /* If the midrule's $$ is set or its $n is used, remove the `$' from the
 528          symbol name so that it's a user-defined symbol so that the default
 529          %destructor and %printer apply.  */
 530       if (p->midrule_parent_rule
 531           && (p->action_props.is_value_used
 532               || symbol_list_n_get (p->midrule_parent_rule,
 533                                     p->midrule_parent_rhs_index)
 534                    ->action_props.is_value_used))
 535         p->content.sym->tag += 1;
 536
 537       /* Don't check the generated rule 0.  It has no action, so some rhs
 538          symbols may appear unused, but the parsing algorithm ensures that
 539          %destructor's are invoked appropriately.  */
 540       if (p != grammar)
 541         grammar_rule_check (p);
 542
 543       for (p = p->next; p && p->content.sym; p = p->next)
 544         {
 545           ++rule_length;
 546
 547           /* Don't allow rule_length == INT_MAX, since that might
 548              cause confusion with strtol if INT_MAX == LONG_MAX.  */
 549           if (rule_length == INT_MAX)
 550               fatal_at (rules[ruleno].location, _("rule is too long"));
 551
 552           /* item_number = symbol_number.
 553              But the former needs to contain more: negative rule numbers. */
 554           ritem[itemno++] =
 555             symbol_number_as_item_number (p->content.sym->number);
 556           /* A rule gets by default the precedence and associativity
 557              of its last token.  */
 558           if (p->content.sym->class == token_sym && default_prec)
 559             rules[ruleno].prec = p->content.sym;
 560         }
 561
 562       /* If this rule has a %prec,
 563          the specified symbol's precedence replaces the default.  */
 564       if (ruleprec)
 565         {
 566           rules[ruleno].precsym = ruleprec;
 567           rules[ruleno].prec = ruleprec;
 568         }
 569       /* An item ends by the rule number (negated).  */
 570       ritem[itemno++] = rule_number_as_item_number (ruleno);
 571       aver (itemno < ITEM_NUMBER_MAX);
 572       ++ruleno;
 573       aver (ruleno < RULE_NUMBER_MAX);
 574
 575       if (p)
 576         p = p->next;
 577     }
 578
 579   aver (itemno == nritems);
 580
 581   if (trace_flag & trace_sets)
 582     ritem_print (stderr);
 583 }
 584 \f
 585 /*------------------------------------------------------------------.
 586 | Read in the grammar specification and record it in the format     |
 587 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 588 | in each case forming the body of a C function (YYACTION) which    |
 589 | contains a switch statement to decide which action to execute.    |
 590 `------------------------------------------------------------------*/
 591
 592 void
 593 reader (void)
 594 {
 595   /* Initialize the symbol table.  */
 596   symbols_new ();
 597
 598   /* Construct the accept symbol. */
 599   accept = symbol_get ("$accept", empty_location);
 600   accept->class = nterm_sym;
 601   accept->number = nvars++;
 602
 603   /* Construct the error token */
 604   errtoken = symbol_get ("error", empty_location);
 605   errtoken->class = token_sym;
 606   errtoken->number = ntokens++;
 607
 608   /* Construct a token that represents all undefined literal tokens.
 609      It is always token number 2.  */
 610   undeftoken = symbol_get ("$undefined", empty_location);
 611   undeftoken->class = token_sym;
 612   undeftoken->number = ntokens++;
 613
 614   gram_in = xfopen (grammar_file, "r");
 615
 616   gram__flex_debug = trace_flag & trace_scan;
 617   gram_debug = trace_flag & trace_parse;
 618   gram_scanner_initialize ();
 619   gram_parse ();
 620   prepare_percent_define_front_end_variables ();
 621
 622   if (! complaint_issued)
 623     check_and_convert_grammar ();
 624
 625   xfclose (gram_in);
 626 }
 627
 628 static void
 629 prepare_percent_define_front_end_variables (void)
 630 {
 631   /* Set %define front-end variable defaults.  */
 632   muscle_percent_define_default ("lr.keep-unreachable-states", "false");
 633   {
 634     char *lr_type;
 635     /* IELR would be a better default, but LALR is historically the
 636        default.  */
 637     muscle_percent_define_default ("lr.type", "lalr");
 638     lr_type = muscle_percent_define_get ("lr.type");
 639     if (0 != strcmp (lr_type, "canonical-lr"))
 640       muscle_percent_define_default ("lr.default-reductions", "most");
 641     else
 642       muscle_percent_define_default ("lr.default-reductions", "accepting");
 643     free (lr_type);
 644   }
 645
 646   /* Check %define front-end variables.  */
 647   {
 648     static char const * const values[] = {
 649       "lr.type", "lalr", "ielr", "canonical-lr", NULL,
 650       "lr.default-reductions", "most", "consistent", "accepting", NULL,
 651       NULL
 652     };
 653     muscle_percent_define_check_values (values);
 654   }
 655 }
 656
 657
 658 /*-------------------------------------------------------------.
 659 | Check the grammar that has just been read, and convert it to |
 660 | internal form.                                               |
 661 `-------------------------------------------------------------*/
 662
 663 static void
 664 check_and_convert_grammar (void)
 665 {
 666   /* Grammar has been read.  Do some checking.  */
 667   if (nrules == 0)
 668     fatal (_("no rules in the input grammar"));
 669
 670   /* If the user did not define her ENDTOKEN, do it now. */
 671   if (!endtoken)
 672     {
 673       endtoken = symbol_get ("$end", empty_location);
 674       endtoken->class = token_sym;
 675       endtoken->number = 0;
 676       /* Value specified by POSIX.  */
 677       endtoken->user_token_number = 0;
 678     }
 679
 680   /* Report any undefined symbols and consider them nonterminals.  */
 681   symbols_check_defined ();
 682
 683   /* Find the start symbol if no %start.  */
 684   if (!start_flag)
 685     {
 686       symbol_list *node;
 687       for (node = grammar;
 688            node != NULL && symbol_is_dummy (node->content.sym);
 689            node = node->next)
 690         {
 691           for (node = node->next;
 692                node != NULL && node->content.sym != NULL;
 693                node = node->next)
 694             ;
 695         }
 696       aver (node != NULL);
 697       grammar_start_symbol_set (node->content.sym,
 698                                 node->content.sym->location);
 699     }
 700
 701   /* Insert the initial rule, whose line is that of the first rule
 702      (not that of the start symbol):
 703
 704      accept: %start EOF.  */
 705   {
 706     symbol_list *p = symbol_list_sym_new (accept, empty_location);
 707     p->location = grammar->location;
 708     p->next = symbol_list_sym_new (startsymbol, empty_location);
 709     p->next->next = symbol_list_sym_new (endtoken, empty_location);
 710     p->next->next->next = symbol_list_sym_new (NULL, empty_location);
 711     p->next->next->next->next = grammar;
 712     nrules += 1;
 713     nritems += 3;
 714     grammar = p;
 715   }
 716
 717   aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
 718
 719   /* Assign the symbols their symbol numbers.  Write #defines for the
 720      token symbols into FDEFINES if requested.  */
 721   symbols_pack ();
 722
 723   /* Scan rule actions after invoking symbol_check_alias_consistency (in
 724      symbols_pack above) so that token types are set correctly before the rule
 725      action type checking.
 726
 727      Before invoking grammar_rule_check (in packgram below) on any rule, make
 728      sure all actions have already been scanned in order to set `used' flags.
 729      Otherwise, checking that a midrule's $$ should be set will not always work
 730      properly because the check must forward-reference the midrule's parent
 731      rule.  For the same reason, all the `used' flags must be set before
 732      checking whether to remove `$' from any midrule symbol name (also in
 733      packgram).  */
 734   {
 735     symbol_list *sym;
 736     for (sym = grammar; sym; sym = sym->next)
 737       code_props_translate_code (&sym->action_props);
 738   }
 739
 740   /* Convert the grammar into the format described in gram.h.  */
 741   packgram ();
 742
 743   /* The grammar as a symbol_list is no longer needed. */
 744   symbol_list_free (grammar);
 745 }