src/reader.c

   1 /* Input parser for Bison
   2
   3    Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000-2003, 2005-2007,
   4    2009-2013 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 (&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       unsigned indent = 0;
 132       complain_indent (&declaration_loc, complaint, &indent,
 133                        _("result type clash on merge function %s: "
 134                          "<%s> != <%s>"),
 135                        quote (merge_function->name), type,
 136                        merge_function->type);
 137       indent += SUB_INDENT;
 138       complain_indent (&merge_function->type_declaration_location, complaint,
 139                        &indent,
 140                        _("previous declaration"));
 141     }
 142   merge_function->type = uniqstr_new (type);
 143   merge_function->type_declaration_location = declaration_loc;
 144 }
 145
 146 /*--------------------------------------.
 147 | Free all merge-function definitions.  |
 148 `--------------------------------------*/
 149
 150 void
 151 free_merger_functions (void)
 152 {
 153   merger_list *L0 = merge_functions;
 154   while (L0)
 155     {
 156       merger_list *L1 = L0->next;
 157       free (L0);
 158       L0 = L1;
 159     }
 160 }
 161
 162 \f
 163 /*-------------------------------------------------------------------.
 164 | Parse the input grammar into a one symbol_list structure.  Each    |
 165 | rule is represented by a sequence of symbols: the left hand side   |
 166 | followed by the contents of the right hand side, followed by a     |
 167 | null pointer instead of a symbol to terminate the rule.  The next  |
 168 | symbol is the lhs of the following rule.                           |
 169 |                                                                    |
 170 | All actions are copied out, labelled by the rule number they apply |
 171 | to.                                                                |
 172 `-------------------------------------------------------------------*/
 173
 174 /* The (currently) last symbol of GRAMMAR. */
 175 static symbol_list *grammar_end = NULL;
 176
 177 /* Append SYM to the grammar.  */
 178 static symbol_list *
 179 grammar_symbol_append (symbol *sym, location loc)
 180 {
 181   symbol_list *p = symbol_list_sym_new (sym, loc);
 182
 183   if (grammar_end)
 184     grammar_end->next = p;
 185   else
 186     grammar = p;
 187
 188   grammar_end = p;
 189
 190   /* A null SYM stands for an end of rule; it is not an actual
 191      part of it.  */
 192   if (sym)
 193     ++nritems;
 194
 195   return p;
 196 }
 197
 198 static void
 199 assign_named_ref (symbol_list *p, named_ref *name)
 200 {
 201   symbol *sym = p->content.sym;
 202
 203   if (name->id == sym->tag)
 204     {
 205       complain (&name->loc, Wother,
 206                 _("duplicated symbol name for %s ignored"),
 207                 quote (sym->tag));
 208       named_ref_free (name);
 209     }
 210   else
 211     p->named_ref = name;
 212 }
 213
 214
 215 /* The rule currently being defined, and the previous rule.
 216    CURRENT_RULE points to the first LHS of the current rule, while
 217    PREVIOUS_RULE_END points to the *end* of the previous rule (NULL).  */
 218 static symbol_list *current_rule = NULL;
 219 static symbol_list *previous_rule_end = NULL;
 220
 221
 222 /*----------------------------------------------.
 223 | Create a new rule for LHS in to the GRAMMAR.  |
 224 `----------------------------------------------*/
 225
 226 void
 227 grammar_current_rule_begin (symbol *lhs, location loc,
 228                             named_ref *lhs_name)
 229 {
 230   symbol_list* p;
 231
 232   /* Start a new rule and record its lhs.  */
 233   ++nrules;
 234   previous_rule_end = grammar_end;
 235
 236   p = grammar_symbol_append (lhs, loc);
 237   if (lhs_name)
 238     assign_named_ref (p, named_ref_copy (lhs_name));
 239
 240   current_rule = grammar_end;
 241
 242   /* Mark the rule's lhs as a nonterminal if not already so.  */
 243   if (lhs->class == unknown_sym)
 244     {
 245       lhs->class = nterm_sym;
 246       lhs->number = nvars;
 247       ++nvars;
 248     }
 249   else if (lhs->class == token_sym)
 250     complain (&loc, complaint, _("rule given for %s, which is a token"),
 251               lhs->tag);
 252 }
 253
 254
 255 /*----------------------------------------------------------------------.
 256 | A symbol should be used if either:                                    |
 257 |   1. It has a destructor.                                             |
 258 |   2. The symbol is a mid-rule symbol (i.e., the generated LHS         |
 259 |      replacing a mid-rule action) that was assigned to or used, as in |
 260 |      "exp: { $$ = 1; } { $$ = $1; }".                                 |
 261 `----------------------------------------------------------------------*/
 262
 263 static bool
 264 symbol_should_be_used (symbol_list const *s, bool *midrule_warning)
 265 {
 266   if (symbol_code_props_get (s->content.sym, destructor)->code)
 267     return true;
 268   if ((s->midrule && s->midrule->action_props.is_value_used)
 269       || (s->midrule_parent_rule
 270           && (symbol_list_n_get (s->midrule_parent_rule,
 271                                  s->midrule_parent_rhs_index)
 272               ->action_props.is_value_used)))
 273     {
 274       *midrule_warning = true;
 275       return true;
 276     }
 277   return false;
 278 }
 279
 280 /*----------------------------------------------------------------.
 281 | Check that the rule R is properly defined.  For instance, there |
 282 | should be no type clash on the default action.                  |
 283 `----------------------------------------------------------------*/
 284
 285 static void
 286 grammar_rule_check (const symbol_list *r)
 287 {
 288   /* Type check.
 289
 290      If there is an action, then there is nothing we can do: the user
 291      is allowed to shoot herself in the foot.
 292
 293      Don't worry about the default action if $$ is untyped, since $$'s
 294      value can't be used.  */
 295   if (!r->action_props.code && r->content.sym->type_name)
 296     {
 297       symbol *first_rhs = r->next->content.sym;
 298       /* If $$ is being set in default way, report if any type mismatch.  */
 299       if (first_rhs)
 300         {
 301           char const *lhs_type = r->content.sym->type_name;
 302           const char *rhs_type =
 303             first_rhs->type_name ? first_rhs->type_name : "";
 304           if (!UNIQSTR_EQ (lhs_type, rhs_type))
 305             complain (&r->location, Wother,
 306                       _("type clash on default action: <%s> != <%s>"),
 307                       lhs_type, rhs_type);
 308         }
 309       /* Warn if there is no default for $$ but we need one.  */
 310       else
 311         complain (&r->location, Wother,
 312                   _("empty rule for typed nonterminal, and no action"));
 313     }
 314
 315   /* Check that symbol values that should be used are in fact used.  */
 316   {
 317     symbol_list const *l = r;
 318     int n = 0;
 319     for (; l && l->content.sym; l = l->next, ++n)
 320       {
 321         bool midrule_warning = false;
 322         if (!l->action_props.is_value_used
 323             && symbol_should_be_used (l, &midrule_warning)
 324             /* The default action, $$ = $1, `uses' both.  */
 325             && (r->action_props.code || (n != 0 && n != 1)))
 326           {
 327             warnings warn_flag = midrule_warning ? Wmidrule_values : Wother;
 328             if (n)
 329               complain (&l->location, warn_flag, _("unused value: $%d"), n);
 330             else
 331               complain (&l->location, warn_flag, _("unset value: $$"));
 332           }
 333       }
 334   }
 335
 336   /* Check that %empty => empty rule.  */
 337   if (r->percent_empty_loc.start.file
 338       && r->next && r->next->content.sym)
 339     complain (&r->percent_empty_loc, complaint,
 340               _("%%empty on non-empty rule"));
 341
 342   /* See comments in grammar_current_rule_prec_set for how POSIX
 343      mandates this complaint.  It's only for identifiers, so skip
 344      it for char literals and strings, which are always tokens.  */
 345   if (r->ruleprec
 346       && r->ruleprec->tag[0] != '\'' && r->ruleprec->tag[0] != '"'
 347       && r->ruleprec->status != declared && !r->ruleprec->prec)
 348     complain (&r->location, Wother,
 349               _("token for %%prec is not defined: %s"), r->ruleprec->tag);
 350 }
 351
 352
 353 /*-------------------------------------.
 354 | End the currently being grown rule.  |
 355 `-------------------------------------*/
 356
 357 void
 358 grammar_current_rule_end (location loc)
 359 {
 360   /* Put an empty link in the list to mark the end of this rule  */
 361   grammar_symbol_append (NULL, grammar_end->location);
 362   current_rule->location = loc;
 363 }
 364
 365
 366 /*-------------------------------------------------------------------.
 367 | The previous action turns out the be a mid-rule action.  Attach it |
 368 | to the current rule, i.e., create a dummy symbol, attach it this   |
 369 | mid-rule action, and append this dummy nonterminal to the current  |
 370 | rule.                                                              |
 371 `-------------------------------------------------------------------*/
 372
 373 void
 374 grammar_midrule_action (void)
 375 {
 376   /* Since the action was written out with this rule's number, we must
 377      give the new rule this number by inserting the new rule before
 378      it.  */
 379
 380   /* Make a DUMMY nonterminal, whose location is that of the midrule
 381      action.  Create the MIDRULE.  */
 382   location dummy_location = current_rule->action_props.location;
 383   symbol *dummy = dummy_symbol_get (dummy_location);
 384   symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
 385
 386   /* Remember named_ref of previous action. */
 387   named_ref *action_name = current_rule->action_props.named_ref;
 388
 389   /* Make a new rule, whose body is empty, before the current one, so
 390      that the action just read can belong to it.  */
 391   ++nrules;
 392   ++nritems;
 393   /* Attach its location and actions to that of the DUMMY.  */
 394   midrule->location = dummy_location;
 395   code_props_rule_action_init (&midrule->action_props,
 396                                current_rule->action_props.code,
 397                                current_rule->action_props.location,
 398                                midrule, 0,
 399                                current_rule->action_props.is_predicate);
 400   code_props_none_init (&current_rule->action_props);
 401
 402   if (previous_rule_end)
 403     previous_rule_end->next = midrule;
 404   else
 405     grammar = midrule;
 406
 407   /* End the dummy's rule.  */
 408   midrule->next = symbol_list_sym_new (NULL, dummy_location);
 409   midrule->next->next = current_rule;
 410
 411   previous_rule_end = midrule->next;
 412
 413   /* Insert the dummy nonterminal replacing the midrule action into
 414      the current rule.  Bind it to its dedicated rule.  */
 415   grammar_current_rule_symbol_append (dummy, dummy_location,
 416                                       action_name);
 417   grammar_end->midrule = midrule;
 418   midrule->midrule_parent_rule = current_rule;
 419   midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
 420 }
 421
 422 /* Set the precedence symbol of the current rule to PRECSYM. */
 423
 424 void
 425 grammar_current_rule_prec_set (symbol *precsym, location loc)
 426 {
 427   /* POSIX says that any identifier is a nonterminal if it does not
 428      appear on the LHS of a grammar rule and is not defined by %token
 429      or by one of the directives that assigns precedence to a token.  We
 430      ignore this here because the only kind of identifier that POSIX
 431      allows to follow a %prec is a token and because assuming it's a
 432      token now can produce more logical error messages.  Nevertheless,
 433      grammar_rule_check does obey what we believe is the real intent of
 434      POSIX here: that an error be reported for any identifier that
 435      appears after %prec but that is not defined separately as a
 436      token.  */
 437   symbol_class_set (precsym, token_sym, loc, false);
 438   if (current_rule->ruleprec)
 439     complain (&loc, complaint, _("only one %s allowed per rule"), "%prec");
 440   current_rule->ruleprec = precsym;
 441 }
 442
 443 /* Set %empty for the current rule. */
 444
 445 void
 446 grammar_current_rule_empty_set (location loc)
 447 {
 448   if (current_rule->percent_empty_loc.start.file)
 449     complain (&loc, complaint, _("only one %s allowed per rule"), "%empty");
 450   else
 451     current_rule->percent_empty_loc = loc;
 452 }
 453
 454 /* Attach dynamic precedence DPREC to the current rule. */
 455
 456 void
 457 grammar_current_rule_dprec_set (int dprec, location loc)
 458 {
 459   if (! glr_parser)
 460     complain (&loc, Wother, _("%s affects only GLR parsers"),
 461               "%dprec");
 462   if (dprec <= 0)
 463     complain (&loc, complaint, _("%s must be followed by positive number"),
 464               "%dprec");
 465   else if (current_rule->dprec != 0)
 466     complain (&loc, complaint, _("only one %s allowed per rule"), "%dprec");
 467   current_rule->dprec = dprec;
 468 }
 469
 470 /* Attach a merge function NAME with argument type TYPE to current
 471    rule. */
 472
 473 void
 474 grammar_current_rule_merge_set (uniqstr name, location loc)
 475 {
 476   if (! glr_parser)
 477     complain (&loc, Wother, _("%s affects only GLR parsers"),
 478               "%merge");
 479   if (current_rule->merger != 0)
 480     complain (&loc, complaint, _("only one %s allowed per rule"), "%merge");
 481   current_rule->merger = get_merge_function (name);
 482   current_rule->merger_declaration_location = loc;
 483 }
 484
 485 /* Attach SYM to the current rule.  If needed, move the previous
 486    action as a mid-rule action.  */
 487
 488 void
 489 grammar_current_rule_symbol_append (symbol *sym, location loc,
 490                                     named_ref *name)
 491 {
 492   symbol_list *p;
 493   if (current_rule->action_props.code)
 494     grammar_midrule_action ();
 495   p = grammar_symbol_append (sym, loc);
 496   if (name)
 497     assign_named_ref(p, name);
 498   if (sym->status == undeclared || sym->status == used)
 499     sym->status = needed;
 500 }
 501
 502 /* Attach an ACTION to the current rule.  */
 503
 504 void
 505 grammar_current_rule_action_append (const char *action, location loc,
 506                                     named_ref *name, bool is_predicate)
 507 {
 508   if (current_rule->action_props.code)
 509     grammar_midrule_action ();
 510   /* After all symbol declarations have been parsed, packgram invokes
 511      code_props_translate_code.  */
 512   code_props_rule_action_init (&current_rule->action_props, action, loc,
 513                                current_rule, name, is_predicate);
 514 }
 515
 516 \f
 517 /*---------------------------------------------------------------.
 518 | Convert the rules into the representation using RRHS, RLHS and |
 519 | RITEM.                                                         |
 520 `---------------------------------------------------------------*/
 521
 522 static void
 523 packgram (void)
 524 {
 525   unsigned int itemno = 0;
 526   rule_number ruleno = 0;
 527   symbol_list *p;
 528
 529   ritem = xnmalloc (nritems + 1, sizeof *ritem);
 530
 531   /* This sentinel is used by build_relations in gram.c.  */
 532   *ritem++ = 0;
 533
 534   rules = xnmalloc (nrules, sizeof *rules);
 535
 536   for (p = grammar; p; p = p->next)
 537     {
 538       symbol *ruleprec = p->ruleprec;
 539       record_merge_function_type (p->merger, p->content.sym->type_name,
 540                                   p->merger_declaration_location);
 541       rules[ruleno].user_number = ruleno;
 542       rules[ruleno].number = ruleno;
 543       rules[ruleno].lhs = p->content.sym;
 544       rules[ruleno].rhs = ritem + itemno;
 545       rules[ruleno].prec = NULL;
 546       rules[ruleno].dprec = p->dprec;
 547       rules[ruleno].merger = p->merger;
 548       rules[ruleno].precsym = NULL;
 549       rules[ruleno].location = p->location;
 550       rules[ruleno].useful = true;
 551       rules[ruleno].action = p->action_props.code;
 552       rules[ruleno].action_location = p->action_props.location;
 553       rules[ruleno].is_predicate = p->action_props.is_predicate;
 554
 555       /* If the midrule's $$ is set or its $n is used, remove the `$' from the
 556          symbol name so that it's a user-defined symbol so that the default
 557          %destructor and %printer apply.  */
 558       if (p->midrule_parent_rule
 559           && (p->action_props.is_value_used
 560               || (symbol_list_n_get (p->midrule_parent_rule,
 561                                      p->midrule_parent_rhs_index)
 562                   ->action_props.is_value_used)))
 563         p->content.sym->tag += 1;
 564
 565       /* Don't check the generated rule 0.  It has no action, so some rhs
 566          symbols may appear unused, but the parsing algorithm ensures that
 567          %destructor's are invoked appropriately.  */
 568       if (p != grammar)
 569         grammar_rule_check (p);
 570
 571       {
 572         size_t rule_length = 0;
 573         for (p = p->next; p->content.sym; p = p->next)
 574           {
 575             ++rule_length;
 576
 577             /* Don't allow rule_length == INT_MAX, since that might
 578                cause confusion with strtol if INT_MAX == LONG_MAX.  */
 579             if (rule_length == INT_MAX)
 580               complain (&rules[ruleno].location, fatal, _("rule is too long"));
 581
 582             /* item_number = symbol_number.
 583                But the former needs to contain more: negative rule numbers. */
 584             ritem[itemno++] =
 585               symbol_number_as_item_number (p->content.sym->number);
 586             /* A rule gets by default the precedence and associativity
 587                of its last token.  */
 588             if (p->content.sym->class == token_sym && default_prec)
 589               rules[ruleno].prec = p->content.sym;
 590           }
 591       }
 592
 593       /* If this rule has a %prec,
 594          the specified symbol's precedence replaces the default.  */
 595       if (ruleprec)
 596         {
 597           rules[ruleno].precsym = ruleprec;
 598           rules[ruleno].prec = ruleprec;
 599         }
 600       /* An item ends by the rule number (negated).  */
 601       ritem[itemno++] = rule_number_as_item_number (ruleno);
 602       aver (itemno < ITEM_NUMBER_MAX);
 603       ++ruleno;
 604       aver (ruleno < RULE_NUMBER_MAX);
 605     }
 606
 607   aver (itemno == nritems);
 608
 609   if (trace_flag & trace_sets)
 610     ritem_print (stderr);
 611 }
 612 \f
 613 /*------------------------------------------------------------------.
 614 | Read in the grammar specification and record it in the format     |
 615 | described in gram.h.  All actions are copied into ACTION_OBSTACK, |
 616 | in each case forming the body of a C function (YYACTION) which    |
 617 | contains a switch statement to decide which action to execute.    |
 618 `------------------------------------------------------------------*/
 619
 620 void
 621 reader (void)
 622 {
 623   /* Initialize the symbol table.  */
 624   symbols_new ();
 625
 626   /* Construct the accept symbol. */
 627   accept = symbol_get ("$accept", empty_location);
 628   accept->class = nterm_sym;
 629   accept->number = nvars++;
 630
 631   /* Construct the error token */
 632   errtoken = symbol_get ("error", empty_location);
 633   errtoken->class = token_sym;
 634   errtoken->number = ntokens++;
 635
 636   /* Construct a token that represents all undefined literal tokens.
 637      It is always token number 2.  */
 638   undeftoken = symbol_get ("$undefined", empty_location);
 639   undeftoken->class = token_sym;
 640   undeftoken->number = ntokens++;
 641
 642   gram_in = xfopen (grammar_file, "r");
 643
 644   gram__flex_debug = trace_flag & trace_scan;
 645   gram_debug = trace_flag & trace_parse;
 646   gram_scanner_initialize ();
 647   gram_parse ();
 648   prepare_percent_define_front_end_variables ();
 649
 650   if (complaint_status  < status_complaint)
 651     check_and_convert_grammar ();
 652
 653   xfclose (gram_in);
 654 }
 655
 656 static void
 657 prepare_percent_define_front_end_variables (void)
 658 {
 659   /* Set %define front-end variable defaults.  */
 660   muscle_percent_define_default ("lr.keep-unreachable-state", "false");
 661   {
 662     char *lr_type;
 663     /* IELR would be a better default, but LALR is historically the
 664        default.  */
 665     muscle_percent_define_default ("lr.type", "lalr");
 666     lr_type = muscle_percent_define_get ("lr.type");
 667     if (STRNEQ (lr_type, "canonical-lr"))
 668       muscle_percent_define_default ("lr.default-reduction", "most");
 669     else
 670       muscle_percent_define_default ("lr.default-reduction", "accepting");
 671     free (lr_type);
 672   }
 673
 674   /* Check %define front-end variables.  */
 675   {
 676     static char const * const values[] = {
 677       "lr.type", "lalr", "ielr", "canonical-lr", NULL,
 678       "lr.default-reduction", "most", "consistent", "accepting", NULL,
 679       NULL
 680     };
 681     muscle_percent_define_check_values (values);
 682   }
 683 }
 684
 685
 686 /*-------------------------------------------------------------.
 687 | Check the grammar that has just been read, and convert it to |
 688 | internal form.                                               |
 689 `-------------------------------------------------------------*/
 690
 691 static void
 692 check_and_convert_grammar (void)
 693 {
 694   /* Grammar has been read.  Do some checking.  */
 695   if (nrules == 0)
 696     complain (NULL, fatal, _("no rules in the input grammar"));
 697
 698   /* If the user did not define her ENDTOKEN, do it now. */
 699   if (!endtoken)
 700     {
 701       endtoken = symbol_get ("$end", empty_location);
 702       endtoken->class = token_sym;
 703       endtoken->number = 0;
 704       /* Value specified by POSIX.  */
 705       endtoken->user_token_number = 0;
 706     }
 707
 708   /* Report any undefined symbols and consider them nonterminals.  */
 709   symbols_check_defined ();
 710
 711   /* Find the start symbol if no %start.  */
 712   if (!start_flag)
 713     {
 714       symbol_list *node;
 715       for (node = grammar;
 716            node != NULL && symbol_is_dummy (node->content.sym);
 717            node = node->next)
 718         {
 719           for (node = node->next;
 720                node != NULL && node->content.sym != NULL;
 721                node = node->next)
 722             ;
 723         }
 724       aver (node != NULL);
 725       grammar_start_symbol_set (node->content.sym,
 726                                 node->content.sym->location);
 727     }
 728
 729   /* Insert the initial rule, whose line is that of the first rule
 730      (not that of the start symbol):
 731
 732      $accept: %start $end.  */
 733   {
 734     symbol_list *p = symbol_list_sym_new (accept, empty_location);
 735     p->location = grammar->location;
 736     p->next = symbol_list_sym_new (startsymbol, empty_location);
 737     p->next->next = symbol_list_sym_new (endtoken, empty_location);
 738     p->next->next->next = symbol_list_sym_new (NULL, empty_location);
 739     p->next->next->next->next = grammar;
 740     nrules += 1;
 741     nritems += 3;
 742     grammar = p;
 743   }
 744
 745   aver (nsyms <= SYMBOL_NUMBER_MAXIMUM);
 746   aver (nsyms == ntokens + nvars);
 747
 748   /* Assign the symbols their symbol numbers.  Write #defines for the
 749      token symbols into FDEFINES if requested.  */
 750   symbols_pack ();
 751
 752   /* Scan rule actions after invoking symbol_check_alias_consistency (in
 753      symbols_pack above) so that token types are set correctly before the rule
 754      action type checking.
 755
 756      Before invoking grammar_rule_check (in packgram below) on any rule, make
 757      sure all actions have already been scanned in order to set `used' flags.
 758      Otherwise, checking that a midrule's $$ should be set will not always work
 759      properly because the check must forward-reference the midrule's parent
 760      rule.  For the same reason, all the `used' flags must be set before
 761      checking whether to remove `$' from any midrule symbol name (also in
 762      packgram).  */
 763   {
 764     symbol_list *sym;
 765     for (sym = grammar; sym; sym = sym->next)
 766       code_props_translate_code (&sym->action_props);
 767   }
 768
 769   /* Convert the grammar into the format described in gram.h.  */
 770   packgram ();
 771
 772   /* The grammar as a symbol_list is no longer needed. */
 773   symbol_list_free (grammar);
 774 }