src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007,
   4    2009-2015, 2018-2021 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 <https://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 "fixits.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 = NULL;
  45
  46 /* Was %union seen?  */
  47 bool union_seen = false;
  48
  49 /* Should rules have a default precedence?  */
  50 bool default_prec = true;
  51 \f
  52 /*-----------------------.
  53 | Set the start symbol.  |
  54 `-----------------------*/
  55
  56 void
  57 grammar_start_symbol_set (symbol *sym, location loc)
  58 {
  59   if (start_flag)
  60     complain (&loc, complaint, _("multiple %s declarations"), "%start");
  61   else
  62     {
  63       start_flag = true;
  64       startsymbol = sym;
  65       startsymbol_loc = loc;
  66     }
  67 }
  68
  69 \f
  70
  71 /*------------------------------------------------------------------------.
  72 | Return the merger index for a merging function named NAME.  Records the |
  73 | function, if new, in MERGER_LIST.                                       |
  74 `------------------------------------------------------------------------*/
  75
  76 static int
  77 get_merge_function (uniqstr name)
  78 {
  79   if (! glr_parser)
  80     return 0;
  81
  82   merger_list *syms;
  83   merger_list head;
  84   int n;
  85
  86   head.next = merge_functions;
  87   for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
  88     if (UNIQSTR_EQ (name, syms->next->name))
  89       break;
  90   if (syms->next == NULL)
  91     {
  92       syms->next = xmalloc (sizeof syms->next[0]);
  93       syms->next->name = uniqstr_new (name);
  94       /* After all symbol type declarations have been parsed, packgram invokes
  95          record_merge_function_type to set the type.  */
  96       syms->next->sym = NULL;
  97       syms->next->next = NULL;
  98       merge_functions = head.next;
  99     }
 100   return n;
 101 }
 102
 103 /*-------------------------------------------------------------------.
 104 | For the existing merging function with index MERGER, record that   |
 105 | the result type is that of SYM, as required by the lhs (i.e., SYM) |
 106 | of the rule whose %merge declaration is at DECLARATION_LOC.        |
 107 `-------------------------------------------------------------------*/
 108
 109 static void
 110 record_merge_function_type (int merger, symbol *sym, location declaration_loc)
 111 {
 112   if (merger <= 0)
 113     return;
 114
 115   uniqstr type
 116     = sym->content->type_name ? sym->content->type_name : uniqstr_new ("");
 117
 118   merger_list *merge_function;
 119   int merger_find = 1;
 120   for (merge_function = merge_functions;
 121        merge_function != NULL && merger_find != merger;
 122        merge_function = merge_function->next)
 123     merger_find += 1;
 124   aver (merge_function != NULL && merger_find == merger);
 125   if (merge_function->sym && merge_function->sym->content->type_name)
 126     {
 127       if (!UNIQSTR_EQ (merge_function->sym->content->type_name, type))
 128         {
 129           complain (&declaration_loc, complaint,
 130                     _("result type clash on merge function %s: "
 131                       "<%s> != <%s>"),
 132                     quote (merge_function->name), type,
 133                     merge_function->sym->content->type_name);
 134           subcomplain (&merge_function->type_declaration_loc, complaint,
 135                        _("previous declaration"));
 136         }
 137     }
 138   else
 139     {
 140       merge_function->sym = sym;
 141       merge_function->type_declaration_loc = declaration_loc;
 142     }
 143 }
 144
 145 /*--------------------------------------.
 146 | Free all merge-function definitions.  |
 147 `--------------------------------------*/
 148
 149 void
 150 free_merger_functions (void)
 151 {
 152   merger_list *L0 = merge_functions;
 153   while (L0)
 154     {
 155       merger_list *L1 = L0->next;
 156       free (L0);
 157       L0 = L1;
 158     }
 159 }
 160
 161 \f
 162 /*-------------------------------------------------------------------.
 163 | Parse the input grammar into a one symbol_list structure.  Each    |
 164 | rule is represented by a sequence of symbols: the left hand side   |
 165 | followed by the contents of the right hand side, followed by a     |
 166 | null pointer instead of a symbol to terminate the rule.  The next  |
 167 | symbol is the lhs of the following rule.                           |
 168 |                                                                    |
 169 | All actions are copied out, labelled by the rule number they apply |
 170 | to.                                                                |
 171 `-------------------------------------------------------------------*/
 172
 173 /* The (currently) last symbol of GRAMMAR. */
 174 static symbol_list *grammar_end = NULL;
 175
 176 /* Append SYM to the grammar.  */
 177 static symbol_list *
 178 grammar_symbol_append (symbol *sym, location loc)
 179 {
 180   symbol_list *p = symbol_list_sym_new (sym, loc);
 181
 182   if (grammar_end)
 183     grammar_end->next = p;
 184   else
 185     grammar = p;
 186
 187   grammar_end = p;
 188
 189   /* A null SYM stands for an end of rule; it is not an actual
 190      part of it.  */
 191   if (sym)
 192     ++nritems;
 193
 194   return p;
 195 }
 196
 197 static void
 198 assign_named_ref (symbol_list *p, named_ref *name)
 199 {
 200   symbol *sym = p->content.sym;
 201
 202   if (name->id == sym->tag)
 203     {
 204       complain (&name->loc, Wother,
 205                 _("duplicated symbol name for %s ignored"),
 206                 quote (sym->tag));
 207       named_ref_free (name);
 208     }
 209   else
 210     p->named_ref = name;
 211 }
 212
 213
 214 /* The rule currently being defined, and the previous rule.
 215    CURRENT_RULE points to the first LHS of the current rule, while
 216    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 217 static symbol_list *current_rule = NULL;
 218 static symbol_list *previous_rule_end = NULL;
 219
 220
 221 /*----------------------------------------------.
 222 | Create a new rule for LHS in to the GRAMMAR.  |
 223 `----------------------------------------------*/
 224
 225 void
 226 grammar_current_rule_begin (symbol *lhs, location loc,
 227                             named_ref *lhs_name)
 228 {
 229   /* Start a new rule and record its lhs.  */
 230   ++nrules;
 231   previous_rule_end = grammar_end;
 232
 233   current_rule = grammar_symbol_append (lhs, loc);
 234   if (lhs_name)
 235     assign_named_ref (current_rule, named_ref_copy (lhs_name));
 236
 237   /* Mark the rule's lhs as a nonterminal if not already so.  */
 238   if (lhs->content->class == unknown_sym || lhs->content->class == pct_type_sym)
 239     symbol_class_set (lhs, nterm_sym, empty_loc, false);
 240   else if (lhs->content->class == token_sym)
 241     complain (&loc, complaint, _("rule given for %s, which is a token"),
 242               lhs->tag);
 243 }
 244
 245
 246 /*----------------------------------------------------------------------.
 247 | A symbol should be used if either:                                    |
 248 |   1. It has a destructor.                                             |
 249 |   2. The symbol is a midrule symbol (i.e., the generated LHS          |
 250 |      replacing a midrule action) that was assigned to or used, as in  |
 251 |      "exp: { $$ = 1; } { $$ = $1; }".                                 |
 252 `----------------------------------------------------------------------*/
 253
 254 static bool
 255 symbol_should_be_used (symbol_list const *s, bool *midrule_warning)
 256 {
 257   if (symbol_code_props_get (s->content.sym, destructor)->code)
 258     return true;
 259   if ((s->midrule && s->midrule->action_props.is_value_used)
 260       || (s->midrule_parent_rule
 261           && (symbol_list_n_get (s->midrule_parent_rule,
 262                                  s->midrule_parent_rhs_index)
 263               ->action_props.is_value_used)))
 264     {
 265       *midrule_warning = true;
 266       return true;
 267     }
 268   return false;
 269 }
 270
 271 /*-----------------------------------------------------------------.
 272 | Check that the rule R is properly defined.  For instance, there  |
 273 | should be no type clash on the default action.  Possibly install |
 274 | the default action.                                              |
 275 `-----------------------------------------------------------------*/
 276
 277 static void
 278 grammar_rule_check_and_complete (symbol_list *r)
 279 {
 280   /* Type check.
 281
 282      If there is an action, then there is nothing we can do: the user
 283      is allowed to shoot herself in the foot.
 284
 285      Don't worry about the default action if $$ is untyped, since $$'s
 286      value can't be used.  */
 287   if (!r->action_props.code && r->content.sym->content->type_name)
 288     {
 289       symbol *first_rhs = r->next->content.sym;
 290       /* If $$ is being set in default way, report if any type mismatch.  */
 291       if (first_rhs)
 292         {
 293           char const *lhs_type = r->content.sym->content->type_name;
 294           char const *rhs_type =
 295             first_rhs->content->type_name ? first_rhs->content->type_name : "";
 296           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 297             complain (&r->rhs_loc, Wother,
 298                       _("type clash on default action: <%s> != <%s>"),
 299                       lhs_type, rhs_type);
 300           else
 301             {
 302               /* Install the default action only for C++.  */
 303               const bool is_cxx =
 304                 STREQ (language->language, "c++")
 305                 || (skeleton && (STREQ (skeleton, "glr.cc")
 306                                  || STREQ (skeleton, "lalr1.cc")));
 307               if (is_cxx)
 308                 {
 309                   code_props_rule_action_init (&r->action_props, "{ $$ = $1; }",
 310                                                r->rhs_loc, r,
 311                                                /* name */ NULL,
 312                                                /* type */ NULL,
 313                                                /* is_predicate */ false);
 314                   code_props_translate_code (&r->action_props);
 315                 }
 316             }
 317         }
 318       /* Warn if there is no default for $$ but we need one.  */
 319       else
 320         complain (&r->rhs_loc, Wother,
 321                   _("empty rule for typed nonterminal, and no action"));
 322     }
 323
 324   /* Check that symbol values that should be used are in fact used.  */
 325   {
 326     int n = 0;
 327     for (symbol_list const *l = r; l && l->content.sym; l = l->next, ++n)
 328       {
 329         bool midrule_warning = false;
 330         if (!l->action_props.is_value_used
 331             && symbol_should_be_used (l, &midrule_warning)
 332             /* The default action, $$ = $1, 'uses' both.  */
 333             && (r->action_props.code || (n != 0 && n != 1)))
 334           {
 335             warnings warn_flag = midrule_warning ? Wmidrule_values : Wother;
 336             if (n)
 337               complain (&l->sym_loc, warn_flag, _("unused value: $%d"), n);
 338             else
 339               complain (&l->rhs_loc, warn_flag, _("unset value: $$"));
 340           }
 341       }
 342   }
 343
 344   /* Check that %empty => empty rule.  */
 345   if (r->percent_empty_loc.start.file
 346       && r->next && r->next->content.sym)
 347     {
 348       complain (&r->percent_empty_loc, complaint,
 349                 _("%%empty on non-empty rule"));
 350       fixits_register (&r->percent_empty_loc, "");
 351     }
 352
 353   /* Check that empty rule => %empty.  */
 354   if (!(r->next && r->next->content.sym)
 355       && !r->midrule_parent_rule
 356       && !r->percent_empty_loc.start.file
 357       && warning_is_enabled (Wempty_rule))
 358     {
 359       complain (&r->rhs_loc, Wempty_rule, _("empty rule without %%empty"));
 360       if (feature_flag & feature_caret)
 361         location_caret_suggestion (r->rhs_loc, "%empty", stderr);
 362       location loc = r->rhs_loc;
 363       loc.end = loc.start;
 364       fixits_register (&loc, " %empty ");
 365     }
 366
 367   /* See comments in grammar_current_rule_prec_set for how POSIX
 368      mandates this complaint.  It's only for identifiers, so skip
 369      it for char literals and strings, which are always tokens.  */
 370   if (r->ruleprec
 371       && r->ruleprec->tag[0] != '\'' && r->ruleprec->tag[0] != '"'
 372       && r->ruleprec->content->status != declared
 373       && !r->ruleprec->content->prec)
 374     complain (&r->rhs_loc, Wother,
 375               _("token for %%prec is not defined: %s"), r->ruleprec->tag);
 376
 377   /* Check that the (main) action was not typed.  */
 378   if (r->action_props.type)
 379     complain (&r->rhs_loc, Wother,
 380               _("only midrule actions can be typed: %s"), r->action_props.type);
 381 }
 382
 383
 384 /*-------------------------------------.
 385 | End the currently being grown rule.  |
 386 `-------------------------------------*/
 387
 388 void
 389 grammar_current_rule_end (location loc)
 390 {
 391   /* Put an empty link in the list to mark the end of this rule  */
 392   grammar_symbol_append (NULL, grammar_end->rhs_loc);
 393   current_rule->rhs_loc = loc;
 394 }
 395
 396
 397 /*-------------------------------------------------------------------.
 398 | The previous action turns out to be a midrule action.  Attach it   |
 399 | to the current rule, i.e., create a dummy symbol, attach it this   |
 400 | midrule action, and append this dummy nonterminal to the current   |
 401 | rule.                                                              |
 402 `-------------------------------------------------------------------*/
 403
 404 void
 405 grammar_midrule_action (void)
 406 {
 407   /* Since the action was written out with this rule's number, we must
 408      give the new rule this number by inserting the new rule before
 409      it.  */
 410
 411   /* Make a DUMMY nonterminal, whose location is that of the midrule
 412      action.  Create the MIDRULE.  */
 413   location dummy_loc = current_rule->action_props.location;
 414   symbol *dummy = dummy_symbol_get (dummy_loc);
 415   symbol_type_set (dummy,
 416                    current_rule->action_props.type, current_rule->action_props.location);
 417   symbol_list *midrule = symbol_list_sym_new (dummy, dummy_loc);
 418
 419   /* Remember named_ref of previous action. */
 420   named_ref *action_name = current_rule->action_props.named_ref;
 421
 422   /* Make a new rule, whose body is empty, before the current one, so
 423      that the action just read can belong to it.  */
 424   ++nrules;
 425   ++nritems;
 426   /* Attach its location and actions to that of the DUMMY.  */
 427   midrule->rhs_loc = dummy_loc;
 428   code_props_rule_action_init (&midrule->action_props,
 429                                current_rule->action_props.code,
 430                                current_rule->action_props.location,
 431                                midrule,
 432                                /* name_ref */ NULL,
 433                                /* type */ NULL,
 434                                current_rule->action_props.is_predicate);
 435   code_props_none_init (&current_rule->action_props);
 436
 437   midrule->expected_sr_conflicts = current_rule->expected_sr_conflicts;
 438   midrule->expected_rr_conflicts = current_rule->expected_rr_conflicts;
 439   current_rule->expected_sr_conflicts = -1;
 440   current_rule->expected_rr_conflicts = -1;
 441
 442   if (previous_rule_end)
 443     previous_rule_end->next = midrule;
 444   else
 445     grammar = midrule;
 446
 447   /* End the dummy's rule.  */
 448   midrule->next = symbol_list_sym_new (NULL, dummy_loc);
 449   midrule->next->next = current_rule;
 450
 451   previous_rule_end = midrule->next;
 452
 453   /* Insert the dummy nonterminal replacing the midrule action into
 454      the current rule.  Bind it to its dedicated rule.  */
 455   grammar_current_rule_symbol_append (dummy, dummy_loc,
 456                                       action_name);
 457   grammar_end->midrule = midrule;
 458   midrule->midrule_parent_rule = current_rule;
 459   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 460 }
 461
 462 /* Set the precedence symbol of the current rule to PRECSYM. */
 463
 464 void
 465 grammar_current_rule_prec_set (symbol *precsym, location loc)
 466 {
 467   /* POSIX says that any identifier is a nonterminal if it does not
 468      appear on the LHS of a grammar rule and is not defined by %token
 469      or by one of the directives that assigns precedence to a token.
 470      We ignore this here because the only kind of identifier that
 471      POSIX allows to follow a %prec is a token and because assuming
 472      it's a token now can produce more logical error messages.
 473      Nevertheless, grammar_rule_check_and_complete does obey what we
 474      believe is the real intent of POSIX here: that an error be
 475      reported for any identifier that appears after %prec but that is
 476      not defined separately as a token.  */
 477   symbol_class_set (precsym, token_sym, loc, false);
 478   if (current_rule->ruleprec)
 479     duplicate_rule_directive ("%prec",
 480                               current_rule->ruleprec->location, loc);
 481   else
 482     current_rule->ruleprec = precsym;
 483 }
 484
 485 /* Set %empty for the current rule. */
 486
 487 void
 488 grammar_current_rule_empty_set (location loc)
 489 {
 490   /* If %empty is used and -Wno-empty-rule is not, then enable
 491      -Wempty-rule.  */
 492   if (warning_is_unset (Wempty_rule))
 493     warning_argmatch ("empty-rule", 0, 0);
 494   if (current_rule->percent_empty_loc.start.file)
 495     duplicate_rule_directive ("%empty",
 496                               current_rule->percent_empty_loc, loc);
 497   else
 498     current_rule->percent_empty_loc = loc;
 499 }
 500
 501 /* Attach dynamic precedence DPREC to the current rule. */
 502
 503 void
 504 grammar_current_rule_dprec_set (int dprec, location loc)
 505 {
 506   if (! glr_parser)
 507     complain (&loc, Wother, _("%s affects only GLR parsers"),
 508               "%dprec");
 509   if (dprec <= 0)
 510     complain (&loc, complaint, _("%s must be followed by positive number"),
 511               "%dprec");
 512   else if (current_rule->dprec != 0)
 513     duplicate_rule_directive ("%dprec",
 514                               current_rule->dprec_loc, loc);
 515   else
 516     {
 517       current_rule->dprec = dprec;
 518       current_rule->dprec_loc = loc;
 519     }
 520 }
 521
 522 /* Attach a merge function NAME with argument type TYPE to current
 523    rule. */
 524
 525 void
 526 grammar_current_rule_merge_set (uniqstr name, location loc)
 527 {
 528   if (! glr_parser)
 529     complain (&loc, Wother, _("%s affects only GLR parsers"),
 530               "%merge");
 531   if (current_rule->merger != 0)
 532     duplicate_rule_directive ("%merge",
 533                               current_rule->merger_declaration_loc, loc);
 534   else
 535     {
 536       current_rule->merger = get_merge_function (name);
 537       current_rule->merger_declaration_loc = loc;
 538     }
 539 }
 540
 541 /* Attach SYM to the current rule.  If needed, move the previous
 542    action as a midrule action.  */
 543
 544 void
 545 grammar_current_rule_symbol_append (symbol *sym, location loc,
 546                                     named_ref *name)
 547 {
 548   if (current_rule->action_props.code)
 549     grammar_midrule_action ();
 550   symbol_list *p = grammar_symbol_append (sym, loc);
 551   if (name)
 552     assign_named_ref (p, name);
 553   if (sym->content->status == undeclared || sym->content->status == used)
 554     sym->content->status = needed;
 555 }
 556
 557 void
 558 grammar_current_rule_action_append (const char *action, location loc,
 559                                     named_ref *name, uniqstr type)
 560 {
 561   if (current_rule->action_props.code)
 562     grammar_midrule_action ();
 563   if (type)
 564     complain (&loc, Wyacc,
 565               _("POSIX Yacc does not support typed midrule actions"));
 566   /* After all symbol declarations have been parsed, packgram invokes
 567      code_props_translate_code.  */
 568   code_props_rule_action_init (&current_rule->action_props, action, loc,
 569                                current_rule,
 570                                name, type,
 571                                /* is_predicate */ false);
 572 }
 573
 574 void
 575 grammar_current_rule_predicate_append (const char *pred, location loc)
 576 {
 577   if (current_rule->action_props.code)
 578     grammar_midrule_action ();
 579   code_props_rule_action_init (&current_rule->action_props, pred, loc,
 580                                current_rule,
 581                                NULL, NULL,
 582                                /* is_predicate */ true);
 583 }
 584
 585 /* Set the expected number of shift/reduce (reduce/reduce) conflicts
 586  * for the current rule.  If a midrule is encountered later, the count
 587  * is transferred to it and reset in the current rule to -1. */
 588
 589 void
 590 grammar_current_rule_expect_sr (int count, location loc)
 591 {
 592   (void) loc;
 593   current_rule->expected_sr_conflicts = count;
 594 }
 595
 596 void
 597 grammar_current_rule_expect_rr (int count, location loc)
 598 {
 599   if (! glr_parser)
 600     complain (&loc, Wother, _("%s affects only GLR parsers"),
 601               "%expect-rr");
 602   else
 603     current_rule->expected_rr_conflicts = count;
 604 }
 605
 606 \f
 607 /*---------------------------------------------.
 608 | Build RULES and RITEM from what was parsed.  |
 609 `---------------------------------------------*/
 610
 611 static void
 612 packgram (void)
 613 {
 614   int itemno = 0;
 615   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 616   /* This sentinel is used by build_relations() in lalr.c.  */
 617   *ritem++ = 0;
 618
 619   rule_number ruleno = 0;
 620   rules = xnmalloc (nrules, sizeof *rules);
 621
 622   for (symbol_list *p = grammar; p; p = p->next)
 623     {
 624       symbol_list *lhs = p;
 625       record_merge_function_type (lhs->merger, lhs->content.sym,
 626                                   lhs->merger_declaration_loc);
 627       /* If the midrule's $$ is set or its $n is used, remove the '$' from the
 628          symbol name so that it's a user-defined symbol so that the default
 629          %destructor and %printer apply.  */
 630       if (lhs->midrule_parent_rule /* i.e., symbol_is_dummy (lhs->content.sym).  */
 631           && (lhs->action_props.is_value_used
 632               || (symbol_list_n_get (lhs->midrule_parent_rule,
 633                                      lhs->midrule_parent_rhs_index)
 634                   ->action_props.is_value_used)))
 635         lhs->content.sym->tag += 1;
 636
 637       /* Don't check the generated rule 0.  It has no action, so some rhs
 638          symbols may appear unused, but the parsing algorithm ensures that
 639          %destructor's are invoked appropriately.  */
 640       if (lhs != grammar)
 641         grammar_rule_check_and_complete (lhs);
 642
 643       rules[ruleno].code = ruleno;
 644       rules[ruleno].number = ruleno;
 645       rules[ruleno].lhs = lhs->content.sym->content;
 646       rules[ruleno].rhs = ritem + itemno;
 647       rules[ruleno].prec = NULL;
 648       rules[ruleno].dprec = lhs->dprec;
 649       rules[ruleno].merger = lhs->merger;
 650       rules[ruleno].precsym = NULL;
 651       rules[ruleno].location = lhs->rhs_loc;
 652       rules[ruleno].useful = true;
 653       rules[ruleno].action = lhs->action_props.code;
 654       rules[ruleno].action_loc = lhs->action_props.location;
 655       rules[ruleno].is_predicate = lhs->action_props.is_predicate;
 656       rules[ruleno].expected_sr_conflicts = lhs->expected_sr_conflicts;
 657       rules[ruleno].expected_rr_conflicts = lhs->expected_rr_conflicts;
 658
 659       /* Traverse the rhs.  */
 660       {
 661         size_t rule_length = 0;
 662         for (p = lhs->next; p->content.sym; p = p->next)
 663           {
 664             ++rule_length;
 665
 666             /* Don't allow rule_length == INT_MAX, since that might
 667                cause confusion with strtol if INT_MAX == LONG_MAX.  */
 668             if (rule_length == INT_MAX)
 669               complain (&rules[ruleno].location, fatal, _("rule is too long"));
 670
 671             /* item_number = symbol_number.
 672                But the former needs to contain more: negative rule numbers. */
 673             ritem[itemno++] =
 674               symbol_number_as_item_number (p->content.sym->content->number);
 675             /* A rule gets by default the precedence and associativity
 676                of its last token.  */
 677             if (p->content.sym->content->class == token_sym && default_prec)
 678               rules[ruleno].prec = p->content.sym->content;
 679           }
 680       }
 681
 682       /* If this rule has a %prec,
 683          the specified symbol's precedence replaces the default.  */
 684       if (lhs->ruleprec)
 685         {
 686           rules[ruleno].precsym = lhs->ruleprec->content;
 687           rules[ruleno].prec = lhs->ruleprec->content;
 688         }
 689
 690       /* An item ends by the rule number (negated).  */
 691       ritem[itemno++] = rule_number_as_item_number (ruleno);
 692       aver (itemno < ITEM_NUMBER_MAX);
 693       ++ruleno;
 694       aver (ruleno < RULE_NUMBER_MAX);
 695     }
 696
 697   aver (itemno == nritems);
 698
 699   if (trace_flag & trace_sets)
 700     ritem_print (stderr);
 701 }
 702
 703
 704 /*--------------------------------------------------------------.
 705 | Read in the grammar specification and record it in the format |
 706 | described in gram.h.                                          |
 707 `--------------------------------------------------------------*/
 708
 709 void
 710 reader (const char *gram)
 711 {
 712   /* Set up symbol_table, semantic_type_table, and the built-in
 713      symbols.  */
 714   symbols_new ();
 715
 716   gram_scanner_open (gram);
 717   parser_init ();
 718   gram_parse ();
 719   gram_scanner_close ();
 720
 721   prepare_percent_define_front_end_variables ();
 722
 723   if (complaint_status  < status_complaint)
 724     check_and_convert_grammar ();
 725 }
 726
 727 static void
 728 prepare_percent_define_front_end_variables (void)
 729 {
 730   /* Set %define front-end variable defaults.  */
 731   muscle_percent_define_default ("lr.keep-unreachable-state", "false");
 732   {
 733     /* IELR would be a better default, but LALR is historically the
 734        default.  */
 735     muscle_percent_define_default ("lr.type", "lalr");
 736     char *lr_type = muscle_percent_define_get ("lr.type");
 737     if (STRNEQ (lr_type, "canonical-lr"))
 738       muscle_percent_define_default ("lr.default-reduction", "most");
 739     else
 740       muscle_percent_define_default ("lr.default-reduction", "accepting");
 741     free (lr_type);
 742   }
 743
 744   /* Check %define front-end variables.  */
 745   {
 746     static char const * const values[] =
 747       {
 748        "lr.type", "lr""(0)", "lalr", "ielr", "canonical-lr", NULL,
 749        "lr.default-reduction", "most", "consistent", "accepting", NULL,
 750        NULL
 751       };
 752     muscle_percent_define_check_values (values);
 753   }
 754 }
 755
 756 /* Find the first LHS which is not a dummy.  */
 757
 758 static symbol *
 759 find_start_symbol (void)
 760 {
 761   symbol_list *res = grammar;
 762   /* Skip all the possible dummy rules of the first rule.  */
 763   for (; symbol_is_dummy (res->content.sym); res = res->next)
 764     /* Skip the LHS, and then all the RHS of the dummy rule.  */
 765     for (res = res->next; res->content.sym; res = res->next)
 766       continue;
 767   return res->content.sym;
 768 }
 769
 770
 771 /*-------------------------------------------------------------.
 772 | Check the grammar that has just been read, and convert it to |
 773 | internal form.                                               |
 774 `-------------------------------------------------------------*/
 775
 776 static void
 777 check_and_convert_grammar (void)
 778 {
 779   /* Grammar has been read.  Do some checking.  */
 780   if (nrules == 0)
 781     complain (NULL, fatal, _("no rules in the input grammar"));
 782
 783   /* If the user did not define her EOFTOKEN, do it now. */
 784   if (!eoftoken)
 785     {
 786       eoftoken = symbol_get ("YYEOF", empty_loc);
 787       eoftoken->content->class = token_sym;
 788       eoftoken->content->number = 0;
 789       /* Value specified by POSIX.  */
 790       eoftoken->content->code = 0;
 791       {
 792         symbol *alias = symbol_get ("$end", empty_loc);
 793         symbol_class_set (alias, token_sym, empty_loc, false);
 794         symbol_make_alias (eoftoken, alias, empty_loc);
 795       }
 796     }
 797
 798   /* Report any undefined symbols and consider them nonterminals.  */
 799   symbols_check_defined ();
 800
 801   /* Find the start symbol if no %start.  */
 802   if (!start_flag)
 803     {
 804       symbol *start = find_start_symbol ();
 805       grammar_start_symbol_set (start, start->location);
 806     }
 807
 808   /* Insert the initial rule, whose line is that of the first rule
 809      (not that of the start symbol):
 810
 811      $accept: %start $end.  */
 812   {
 813     symbol_list *p = symbol_list_sym_new (acceptsymbol, empty_loc);
 814     p->rhs_loc = grammar->rhs_loc;
 815     p->next = symbol_list_sym_new (startsymbol, empty_loc);
 816     p->next->next = symbol_list_sym_new (eoftoken, empty_loc);
 817     p->next->next->next = symbol_list_sym_new (NULL, empty_loc);
 818     p->next->next->next->next = grammar;
 819     nrules += 1;
 820     nritems += 3;
 821     grammar = p;
 822   }
 823
 824   if (SYMBOL_NUMBER_MAXIMUM - nnterms < ntokens)
 825     complain (NULL, fatal, "too many symbols in input grammar (limit is %d)",
 826               SYMBOL_NUMBER_MAXIMUM);
 827
 828   nsyms = ntokens + nnterms;
 829
 830   /* Assign the symbols their symbol numbers.  */
 831   symbols_pack ();
 832
 833   /* Scan rule actions after invoking symbol_check_alias_consistency
 834      (in symbols_pack above) so that token types are set correctly
 835      before the rule action type checking.
 836
 837      Before invoking grammar_rule_check_and_complete (in packgram
 838      below) on any rule, make sure all actions have already been
 839      scanned in order to set 'used' flags.  Otherwise, checking that a
 840      midrule's $$ should be set will not always work properly because
 841      the check must forward-reference the midrule's parent rule.  For
 842      the same reason, all the 'used' flags must be set before checking
 843      whether to remove '$' from any midrule symbol name (also in
 844      packgram).  */
 845   for (symbol_list *sym = grammar; sym; sym = sym->next)
 846     code_props_translate_code (&sym->action_props);
 847
 848   /* Convert the grammar into the format described in gram.h.  */
 849   packgram ();
 850
 851   /* The grammar as a symbol_list is no longer needed. */
 852   symbol_list_free (grammar);
 853 }