src/reader.c

   1 /* Input parser for bison
   2    Copyright 1984, 1986, 1989, 1992, 1998, 2000, 2001
   3    Free Software Foundation, Inc.
   4
   5    This file is part of Bison, the GNU Compiler Compiler.
   6
   7    Bison is free software; you can redistribute it and/or modify
   8    it under the terms of the GNU General Public License as published by
   9    the Free Software Foundation; either version 2, or (at your option)
  10    any later version.
  11
  12    Bison is distributed in the hope that it will be useful,
  13    but WITHOUT ANY WARRANTY; without even the implied warranty of
  14    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15    GNU General Public License for more details.
  16
  17    You should have received a copy of the GNU General Public License
  18    along with Bison; see the file COPYING.  If not, write to
  19    the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  20    Boston, MA 02111-1307, USA.  */
  21
  22
  23 #include "system.h"
  24 #include "quotearg.h"
  25 #include "quote.h"
  26 #include "getargs.h"
  27 #include "files.h"
  28 #include "symtab.h"
  29 #include "options.h"
  30 #include "lex.h"
  31 #include "gram.h"
  32 #include "complain.h"
  33 #include "output.h"
  34 #include "reader.h"
  35 #include "conflicts.h"
  36 #include "muscle_tab.h"
  37
  38 typedef struct symbol_list
  39 {
  40   struct symbol_list *next;
  41   bucket *sym;
  42   int line;
  43
  44   /* The action is attached to the LHS of a rule. */
  45   const char *action;
  46   int action_line;
  47
  48   /* The guard is attached to the LHS of a rule. */
  49   const char *guard;
  50   int guard_line;
  51   bucket *ruleprec;
  52 } symbol_list;
  53
  54 int lineno;
  55 char **tags;
  56 short *user_toknums;
  57 static symbol_list *grammar;
  58 static int start_flag;
  59 static bucket *startval;
  60
  61 /* Nonzero if components of semantic values are used, implying
  62    they must be unions.  */
  63 static int value_components_used;
  64
  65 /* Nonzero if %union has been seen.  */
  66 static int typed;
  67
  68 /* Incremented for each %left, %right or %nonassoc seen */
  69 static int lastprec;
  70
  71 static bucket *errtoken = NULL;
  72 static bucket *undeftoken = NULL;
  73 static bucket *eoftoken = NULL;
  74 static bucket *axiom = NULL;
  75
  76 static symbol_list *
  77 symbol_list_new (bucket *sym)
  78 {
  79   symbol_list *res = XMALLOC (symbol_list, 1);
  80   res->next = NULL;
  81   res->sym = sym;
  82   res->line = lineno;
  83   res->action = NULL;
  84   res->action_line = 0;
  85   res->guard = NULL;
  86   res->guard_line = 0;
  87   res->ruleprec = NULL;
  88   return res;
  89 }
  90
  91 \f
  92
  93 /*===================\
  94 | Low level lexing.  |
  95 \===================*/
  96
  97 static void
  98 skip_to_char (int target)
  99 {
 100   int c;
 101   if (target == '\n')
 102     complain (_("   Skipping to next \\n"));
 103   else
 104     complain (_("   Skipping to next %c"), target);
 105
 106   do
 107     c = skip_white_space ();
 108   while (c != target && c != EOF);
 109   if (c != EOF)
 110     ungetc (c, finput);
 111 }
 112
 113
 114 /*---------------------------------------------------------.
 115 | Read a signed integer from STREAM and return its value.  |
 116 `---------------------------------------------------------*/
 117
 118 static inline int
 119 read_signed_integer (FILE *stream)
 120 {
 121   int c = getc (stream);
 122   int sign = 1;
 123   int n = 0;
 124
 125   if (c == '-')
 126     {
 127       c = getc (stream);
 128       sign = -1;
 129     }
 130
 131   while (isdigit (c))
 132     {
 133       n = 10 * n + (c - '0');
 134       c = getc (stream);
 135     }
 136
 137   ungetc (c, stream);
 138
 139   return sign * n;
 140 }
 141 \f
 142 /*--------------------------------------------------------------.
 143 | Get the data type (alternative in the union) of the value for |
 144 | symbol N in rule RULE.                                        |
 145 `--------------------------------------------------------------*/
 146
 147 static char *
 148 get_type_name (int n, symbol_list *rule)
 149 {
 150   int i;
 151   symbol_list *rp;
 152
 153   if (n < 0)
 154     {
 155       complain (_("invalid $ value"));
 156       return NULL;
 157     }
 158
 159   rp = rule;
 160   i = 0;
 161
 162   while (i < n)
 163     {
 164       rp = rp->next;
 165       if (rp == NULL || rp->sym == NULL)
 166         {
 167           complain (_("invalid $ value"));
 168           return NULL;
 169         }
 170       i++;
 171     }
 172
 173   return rp->sym->type_name;
 174 }
 175 \f
 176 /*------------------------------------------------------------.
 177 | Dump the string from FIN to OOUT if non null.  MATCH is the |
 178 | delimiter of the string (either ' or ").                    |
 179 `------------------------------------------------------------*/
 180
 181 static inline void
 182 copy_string2 (FILE *fin, struct obstack *oout, int match, int store)
 183 {
 184   int c;
 185
 186   if (store)
 187     obstack_1grow (oout, match);
 188
 189   c = getc (fin);
 190
 191   while (c != match)
 192     {
 193       if (c == EOF)
 194         fatal (_("unterminated string at end of file"));
 195       if (c == '\n')
 196         {
 197           complain (_("unterminated string"));
 198           ungetc (c, fin);
 199           c = match;            /* invent terminator */
 200           continue;
 201         }
 202
 203       obstack_1grow (oout, c);
 204
 205       if (c == '\\')
 206         {
 207           c = getc (fin);
 208           if (c == EOF)
 209             fatal (_("unterminated string at end of file"));
 210           obstack_1grow (oout, c);
 211
 212           if (c == '\n')
 213             lineno++;
 214         }
 215
 216       c = getc (fin);
 217     }
 218
 219   if (store)
 220     obstack_1grow (oout, c);
 221 }
 222
 223 /* FIXME. */
 224
 225 static inline void
 226 copy_string (FILE *fin, struct obstack *oout, int match)
 227 {
 228   copy_string2 (fin, oout, match, 1);
 229 }
 230
 231 /* FIXME. */
 232
 233 static inline void
 234 copy_identifier (FILE *fin, struct obstack *oout)
 235 {
 236   int c;
 237
 238   while (isalnum (c = getc (fin)) || c == '_')
 239     obstack_1grow (oout, c);
 240
 241   ungetc (c, fin);
 242 }
 243
 244
 245 /*------------------------------------------------------------------.
 246 | Dump the wannabee comment from IN to OOUT.  In fact we just saw a |
 247 | `/', which might or might not be a comment.  In any case, copy    |
 248 | what we saw.                                                      |
 249 `------------------------------------------------------------------*/
 250
 251 static inline void
 252 copy_comment (FILE *fin, struct obstack *oout)
 253 {
 254   int cplus_comment;
 255   int ended;
 256   int c;
 257
 258   /* We read a `/', output it. */
 259   obstack_1grow (oout, '/');
 260
 261   switch ((c = getc (fin)))
 262     {
 263     case '/':
 264       cplus_comment = 1;
 265       break;
 266     case '*':
 267       cplus_comment = 0;
 268       break;
 269     default:
 270       ungetc (c, fin);
 271       return;
 272     }
 273
 274   obstack_1grow (oout, c);
 275   c = getc (fin);
 276
 277   ended = 0;
 278   while (!ended)
 279     {
 280       if (!cplus_comment && c == '*')
 281         {
 282           while (c == '*')
 283             {
 284               obstack_1grow (oout, c);
 285               c = getc (fin);
 286             }
 287
 288           if (c == '/')
 289             {
 290               obstack_1grow (oout, c);
 291               ended = 1;
 292             }
 293         }
 294       else if (c == '\n')
 295         {
 296           lineno++;
 297           obstack_1grow (oout, c);
 298           if (cplus_comment)
 299             ended = 1;
 300           else
 301             c = getc (fin);
 302         }
 303       else if (c == EOF)
 304         fatal (_("unterminated comment"));
 305       else
 306         {
 307           obstack_1grow (oout, c);
 308           c = getc (fin);
 309         }
 310     }
 311 }
 312
 313
 314 /*-----------------------------------------------------------------.
 315 | FIN is pointing to a location (i.e., a `@').  Output to OOUT a   |
 316 | reference to this location. STACK_OFFSET is the number of values |
 317 | in the current rule so far, which says where to find `$0' with   |
 318 | respect to the top of the stack.                                 |
 319 `-----------------------------------------------------------------*/
 320
 321 static inline void
 322 copy_at (FILE *fin, struct obstack *oout, int stack_offset)
 323 {
 324   int c;
 325
 326   c = getc (fin);
 327   if (c == '$')
 328     {
 329       obstack_sgrow (oout, "yyloc");
 330       locations_flag = 1;
 331     }
 332   else if (isdigit (c) || c == '-')
 333     {
 334       int n;
 335
 336       ungetc (c, fin);
 337       n = read_signed_integer (fin);
 338       if (n > stack_offset)
 339         complain (_("invalid value: %s%d"), "@", n);
 340       else
 341         {
 342           /* Offset is always 0 if parser has already popped the stack
 343              pointer.  */
 344           obstack_fgrow1 (oout, "yylsp[%d]",
 345                           n - (semantic_parser ? 0 : stack_offset));
 346           locations_flag = 1;
 347         }
 348     }
 349   else
 350     {
 351       char buf[] = "@c";
 352       buf[1] = c;
 353       complain (_("%s is invalid"), quote (buf));
 354     }
 355 }
 356
 357
 358 /*-------------------------------------------------------------------.
 359 | FIN is pointing to a wannabee semantic value (i.e., a `$').        |
 360 |                                                                    |
 361 | Possible inputs: $[<TYPENAME>]($|integer)                          |
 362 |                                                                    |
 363 | Output to OOUT a reference to this semantic value. STACK_OFFSET is |
 364 | the number of values in the current rule so far, which says where  |
 365 | to find `$0' with respect to the top of the stack.                 |
 366 `-------------------------------------------------------------------*/
 367
 368 static inline void
 369 copy_dollar (FILE *fin, struct obstack *oout,
 370              symbol_list *rule, int stack_offset)
 371 {
 372   int c = getc (fin);
 373   const char *type_name = NULL;
 374
 375   /* Get the type name if explicit. */
 376   if (c == '<')
 377     {
 378       read_type_name (fin);
 379       type_name = token_buffer;
 380       value_components_used = 1;
 381       c = getc (fin);
 382     }
 383
 384   if (c == '$')
 385     {
 386       obstack_sgrow (oout, "yyval");
 387
 388       if (!type_name)
 389         type_name = get_type_name (0, rule);
 390       if (type_name)
 391         obstack_fgrow1 (oout, ".%s", type_name);
 392       if (!type_name && typed)
 393         complain (_("$$ of `%s' has no declared type"),
 394                   rule->sym->tag);
 395     }
 396   else if (isdigit (c) || c == '-')
 397     {
 398       int n;
 399       ungetc (c, fin);
 400       n = read_signed_integer (fin);
 401
 402       if (n > stack_offset)
 403         complain (_("invalid value: %s%d"), "$", n);
 404       else
 405         {
 406           if (!type_name && n > 0)
 407             type_name = get_type_name (n, rule);
 408
 409           /* Offset is always 0 if parser has already popped the stack
 410              pointer.  */
 411           obstack_fgrow1 (oout, "yyvsp[%d]",
 412                           n - (semantic_parser ? 0 : stack_offset));
 413
 414           if (type_name)
 415             obstack_fgrow1 (oout, ".%s", type_name);
 416           if (!type_name && typed)
 417             complain (_("$%d of `%s' has no declared type"),
 418                       n, rule->sym->tag);
 419         }
 420     }
 421   else
 422     {
 423       char buf[] = "$c";
 424       buf[1] = c;
 425       complain (_("%s is invalid"), quote (buf));
 426     }
 427 }
 428 \f
 429 /*-------------------------------------------------------------------.
 430 | Copy the contents of a `%{ ... %}' into the definitions file.  The |
 431 | `%{' has already been read.  Return after reading the `%}'.        |
 432 `-------------------------------------------------------------------*/
 433
 434 static void
 435 copy_definition (void)
 436 {
 437   int c;
 438   /* -1 while reading a character if prev char was %. */
 439   int after_percent;
 440
 441   if (!no_lines_flag)
 442     {
 443       obstack_fgrow2 (&attrs_obstack, muscle_find ("linef"),
 444                       lineno, quotearg_style (c_quoting_style,
 445                                               muscle_find ("filename")));
 446     }
 447
 448   after_percent = 0;
 449
 450   c = getc (finput);
 451
 452   for (;;)
 453     {
 454       switch (c)
 455         {
 456         case '\n':
 457           obstack_1grow (&attrs_obstack, c);
 458           lineno++;
 459           break;
 460
 461         case '%':
 462           after_percent = -1;
 463           break;
 464
 465         case '\'':
 466         case '"':
 467           copy_string (finput, &attrs_obstack, c);
 468           break;
 469
 470         case '/':
 471           copy_comment (finput, &attrs_obstack);
 472           break;
 473
 474         case EOF:
 475           fatal ("%s", _("unterminated `%{' definition"));
 476
 477         default:
 478           obstack_1grow (&attrs_obstack, c);
 479         }
 480
 481       c = getc (finput);
 482
 483       if (after_percent)
 484         {
 485           if (c == '}')
 486             return;
 487           obstack_1grow (&attrs_obstack, '%');
 488         }
 489       after_percent = 0;
 490     }
 491 }
 492
 493
 494 /*-------------------------------------------------------------------.
 495 | Parse what comes after %token or %nterm.  For %token, WHAT_IS is   |
 496 | token_sym and WHAT_IS_NOT is nterm_sym.  For %nterm, the arguments |
 497 | are reversed.                                                      |
 498 `-------------------------------------------------------------------*/
 499
 500 static void
 501 parse_token_decl (symbol_class what_is, symbol_class what_is_not)
 502 {
 503   token_t token = tok_undef;
 504   char *typename = NULL;
 505
 506   /* The symbol being defined.  */
 507   struct bucket *symbol = NULL;
 508
 509   /* After `%token' and `%nterm', any number of symbols maybe be
 510      defined.  */
 511   for (;;)
 512     {
 513       int tmp_char = ungetc (skip_white_space (), finput);
 514
 515       /* `%' (for instance from `%token', or from `%%' etc.) is the
 516          only valid means to end this declaration.  */
 517       if (tmp_char == '%')
 518         return;
 519       if (tmp_char == EOF)
 520         fatal (_("Premature EOF after %s"), token_buffer);
 521
 522       token = lex ();
 523       if (token == tok_comma)
 524         {
 525           symbol = NULL;
 526           continue;
 527         }
 528       if (token == tok_typename)
 529         {
 530           typename = xstrdup (token_buffer);
 531           value_components_used = 1;
 532           symbol = NULL;
 533         }
 534       else if (token == tok_identifier && *symval->tag == '\"' && symbol)
 535         {
 536           if (symval->alias)
 537             warn (_("symbol `%s' used more than once as a literal string"),
 538                   symval->tag);
 539           else if (symbol->alias)
 540             warn (_("symbol `%s' given more than one literal string"),
 541                   symbol->tag);
 542           else
 543             {
 544               symval->class = token_sym;
 545               symval->type_name = typename;
 546               symval->user_token_number = symbol->user_token_number;
 547               symbol->user_token_number = SALIAS;
 548               symval->alias = symbol;
 549               symbol->alias = symval;
 550               /* symbol and symval combined are only one symbol */
 551               nsyms--;
 552             }
 553           symbol = NULL;
 554         }
 555       else if (token == tok_identifier)
 556         {
 557           int oldclass = symval->class;
 558           symbol = symval;
 559
 560           if (symbol->class == what_is_not)
 561             complain (_("symbol %s redefined"), symbol->tag);
 562           symbol->class = what_is;
 563           if (what_is == nterm_sym && oldclass != nterm_sym)
 564             symbol->value = nvars++;
 565
 566           if (typename)
 567             {
 568               if (symbol->type_name == NULL)
 569                 symbol->type_name = typename;
 570               else if (strcmp (typename, symbol->type_name) != 0)
 571                 complain (_("type redeclaration for %s"), symbol->tag);
 572             }
 573         }
 574       else if (symbol && token == tok_number)
 575         {
 576           symbol->user_token_number = numval;
 577           /* User defined EOF token? */
 578           if (numval == 0)
 579             eoftoken = symbol;
 580         }
 581       else
 582         {
 583           complain (_("`%s' is invalid in %s"),
 584                     token_buffer,
 585                     (what_is == token_sym) ? "%token" : "%nterm");
 586           skip_to_char ('%');
 587         }
 588     }
 589
 590 }
 591
 592
 593 /*------------------------------.
 594 | Parse what comes after %start |
 595 `------------------------------*/
 596
 597 static void
 598 parse_start_decl (void)
 599 {
 600   if (start_flag)
 601     complain (_("multiple %s declarations"), "%start");
 602   if (lex () != tok_identifier)
 603     complain (_("invalid %s declaration"), "%start");
 604   else
 605     {
 606       start_flag = 1;
 607       startval = symval;
 608     }
 609 }
 610
 611 /*-----------------------------------------------------------.
 612 | read in a %type declaration and record its information for |
 613 | get_type_name to access                                    |
 614 `-----------------------------------------------------------*/
 615
 616 static void
 617 parse_type_decl (void)
 618 {
 619   char *name;
 620
 621   if (lex () != tok_typename)
 622     {
 623       complain ("%s", _("%type declaration has no <typename>"));
 624       skip_to_char ('%');
 625       return;
 626     }
 627
 628   name = xstrdup (token_buffer);
 629
 630   for (;;)
 631     {
 632       token_t t;
 633       int tmp_char = ungetc (skip_white_space (), finput);
 634
 635       if (tmp_char == '%')
 636         return;
 637       if (tmp_char == EOF)
 638         fatal (_("Premature EOF after %s"), token_buffer);
 639
 640       t = lex ();
 641
 642       switch (t)
 643         {
 644
 645         case tok_comma:
 646         case tok_semicolon:
 647           break;
 648
 649         case tok_identifier:
 650           if (symval->type_name == NULL)
 651             symval->type_name = name;
 652           else if (strcmp (name, symval->type_name) != 0)
 653             complain (_("type redeclaration for %s"), symval->tag);
 654
 655           break;
 656
 657         default:
 658           complain (_("invalid %%type declaration due to item: %s"),
 659                     token_buffer);
 660           skip_to_char ('%');
 661         }
 662     }
 663 }
 664
 665
 666
 667 /*----------------------------------------------------------------.
 668 | Read in a %left, %right or %nonassoc declaration and record its |
 669 | information.                                                    |
 670 `----------------------------------------------------------------*/
 671
 672 static void
 673 parse_assoc_decl (associativity assoc)
 674 {
 675   char *name = NULL;
 676   int prev = 0;
 677
 678   lastprec++;                   /* Assign a new precedence level, never 0.  */
 679
 680   for (;;)
 681     {
 682       token_t t;
 683       int tmp_char = ungetc (skip_white_space (), finput);
 684
 685       if (tmp_char == '%')
 686         return;
 687       if (tmp_char == EOF)
 688         fatal (_("Premature EOF after %s"), token_buffer);
 689
 690       t = lex ();
 691
 692       switch (t)
 693         {
 694         case tok_typename:
 695           name = xstrdup (token_buffer);
 696           break;
 697
 698         case tok_comma:
 699           break;
 700
 701         case tok_identifier:
 702           if (symval->prec != 0)
 703             complain (_("redefining precedence of %s"), symval->tag);
 704           symval->prec = lastprec;
 705           symval->assoc = assoc;
 706           if (symval->class == nterm_sym)
 707             complain (_("symbol %s redefined"), symval->tag);
 708           symval->class = token_sym;
 709           if (name)
 710             {                   /* record the type, if one is specified */
 711               if (symval->type_name == NULL)
 712                 symval->type_name = name;
 713               else if (strcmp (name, symval->type_name) != 0)
 714                 complain (_("type redeclaration for %s"), symval->tag);
 715             }
 716           break;
 717
 718         case tok_number:
 719           if (prev == tok_identifier)
 720             {
 721               symval->user_token_number = numval;
 722             }
 723           else
 724             {
 725               complain (_
 726                         ("invalid text (%s) - number should be after identifier"),
 727 token_buffer);
 728               skip_to_char ('%');
 729             }
 730           break;
 731
 732         case tok_semicolon:
 733           return;
 734
 735         default:
 736           complain (_("unexpected item: %s"), token_buffer);
 737           skip_to_char ('%');
 738         }
 739
 740       prev = t;
 741     }
 742 }
 743
 744
 745
 746 /*--------------------------------------------------------------.
 747 | Copy the union declaration into the stype muscle              |
 748 | (and fdefines),  where it is made into the definition of      |
 749 | YYSTYPE, the type of elements of the parser value stack.      |
 750 `--------------------------------------------------------------*/
 751
 752 static void
 753 parse_union_decl (void)
 754 {
 755   int c;
 756   int count = 0;
 757   bool done = FALSE;
 758   struct obstack union_obstack;
 759   if (typed)
 760     complain (_("multiple %s declarations"), "%union");
 761
 762   typed = 1;
 763
 764   obstack_init (&union_obstack);
 765   obstack_sgrow (&union_obstack, "union");
 766
 767   while (!done)
 768     {
 769       c = xgetc (finput);
 770
 771       /* If C contains '/', it is output by copy_comment ().  */
 772       if (c != '/')
 773         obstack_1grow (&union_obstack, c);
 774
 775       switch (c)
 776         {
 777         case '\n':
 778           lineno++;
 779           break;
 780
 781         case '/':
 782           copy_comment (finput, &union_obstack);
 783           break;
 784
 785         case '{':
 786           count++;
 787           break;
 788
 789         case '}':
 790           /* FIXME: Errr.  How could this happen???. --akim */
 791           if (count == 0)
 792             complain (_("unmatched %s"), "`}'");
 793           count--;
 794           if (!count)
 795             done = TRUE;
 796           break;
 797         }
 798     }
 799
 800   /* JF don't choke on trailing semi */
 801   c = skip_white_space ();
 802   if (c != ';')
 803     ungetc (c, finput);
 804   obstack_1grow (&union_obstack, 0);
 805   muscle_insert ("stype", obstack_finish (&union_obstack));
 806 }
 807
 808
 809 /*-------------------------------------------------------.
 810 | Parse the declaration %expect N which says to expect N |
 811 | shift-reduce conflicts.                                |
 812 `-------------------------------------------------------*/
 813
 814 static void
 815 parse_expect_decl (void)
 816 {
 817   int c = skip_white_space ();
 818   ungetc (c, finput);
 819
 820   if (!isdigit (c))
 821     complain (_("argument of %%expect is not an integer"));
 822   else
 823     expected_conflicts = read_signed_integer (finput);
 824 }
 825
 826
 827 /*-------------------------------------------------------------------.
 828 | Parse what comes after %thong.  the full syntax is                 |
 829 |                                                                    |
 830 |                %thong <type> token number literal                  |
 831 |                                                                    |
 832 | the <type> or number may be omitted.  The number specifies the     |
 833 | user_token_number.                                                 |
 834 |                                                                    |
 835 | Two symbols are entered in the table, one for the token symbol and |
 836 | one for the literal.  Both are given the <type>, if any, from the  |
 837 | declaration.  The ->user_token_number of the first is SALIAS and   |
 838 | the ->user_token_number of the second is set to the number, if     |
 839 | any, from the declaration.  The two symbols are linked via         |
 840 | pointers in their ->alias fields.                                  |
 841 |                                                                    |
 842 | During OUTPUT_DEFINES_TABLE, the symbol is reported thereafter,    |
 843 | only the literal string is retained it is the literal string that  |
 844 | is output to yytname                                               |
 845 `-------------------------------------------------------------------*/
 846
 847 static void
 848 parse_thong_decl (void)
 849 {
 850   token_t token;
 851   struct bucket *symbol;
 852   char *typename = 0;
 853   int usrtoknum = SUNDEF;
 854
 855   token = lex ();               /* fetch typename or first token */
 856   if (token == tok_typename)
 857     {
 858       typename = xstrdup (token_buffer);
 859       value_components_used = 1;
 860       token = lex ();           /* fetch first token */
 861     }
 862
 863   /* process first token */
 864
 865   if (token != tok_identifier)
 866     {
 867       complain (_("unrecognized item %s, expected an identifier"),
 868                 token_buffer);
 869       skip_to_char ('%');
 870       return;
 871     }
 872   symval->class = token_sym;
 873   symval->type_name = typename;
 874   symval->user_token_number = SALIAS;
 875   symbol = symval;
 876
 877   token = lex ();               /* get number or literal string */
 878
 879   if (token == tok_number)
 880     {
 881       usrtoknum = numval;
 882       token = lex ();           /* okay, did number, now get literal */
 883     }
 884
 885   /* process literal string token */
 886
 887   if (token != tok_identifier || *symval->tag != '\"')
 888     {
 889       complain (_("expected string constant instead of %s"), token_buffer);
 890       skip_to_char ('%');
 891       return;
 892     }
 893   symval->class = token_sym;
 894   symval->type_name = typename;
 895   symval->user_token_number = usrtoknum;
 896
 897   symval->alias = symbol;
 898   symbol->alias = symval;
 899
 900   /* symbol and symval combined are only one symbol.  */
 901   nsyms--;
 902 }
 903
 904
 905 static void
 906 parse_muscle_decl (void)
 907 {
 908   int ch = ungetc (skip_white_space (), finput);
 909   char *muscle_key;
 910   char *muscle_value;
 911
 912   /* Read key. */
 913   if (!isalpha (ch) && ch != '_')
 914     {
 915       complain (_("invalid %s declaration"), "%define");
 916       skip_to_char ('%');
 917       return;
 918     }
 919   copy_identifier (finput, &muscle_obstack);
 920   obstack_1grow (&muscle_obstack, 0);
 921   muscle_key = obstack_finish (&muscle_obstack);
 922
 923   /* Read value. */
 924   ch = skip_white_space ();
 925   if (ch != '"')
 926     {
 927       ungetc (ch, finput);
 928       if (ch != EOF)
 929         {
 930           complain (_("invalid %s declaration"), "%define");
 931           skip_to_char ('%');
 932           return;
 933         }
 934       else
 935         fatal (_("Premature EOF after %s"), "\"");
 936     }
 937   copy_string2 (finput, &muscle_obstack, '"', 0);
 938   obstack_1grow (&muscle_obstack, 0);
 939   muscle_value = obstack_finish (&muscle_obstack);
 940
 941   /* Store the (key, value) pair in the environment. */
 942   muscle_insert (muscle_key, muscle_value);
 943 }
 944
 945
 946
 947 /*---------------------------------.
 948 | Parse a double quoted parameter. |
 949 `---------------------------------*/
 950
 951 static const char *
 952 parse_dquoted_param (const char *from)
 953 {
 954   struct obstack param_obstack;
 955   const char *param = NULL;
 956   int c;
 957
 958   obstack_init (&param_obstack);
 959   c = skip_white_space ();
 960
 961   if (c != '"')
 962     {
 963       complain (_("invalid %s declaration"), from);
 964       ungetc (c, finput);
 965       skip_to_char ('%');
 966       return NULL;
 967     }
 968
 969   while ((c = literalchar ()) != '"')
 970     obstack_1grow (&param_obstack, c);
 971
 972   obstack_1grow (&param_obstack, '\0');
 973   param = obstack_finish (&param_obstack);
 974
 975   if (c != '"' || strlen (param) == 0)
 976     {
 977       complain (_("invalid %s declaration"), from);
 978       if (c != '"')
 979         ungetc (c, finput);
 980       skip_to_char ('%');
 981       return NULL;
 982     }
 983
 984   return param;
 985 }
 986
 987 /*----------------------------------.
 988 | Parse what comes after %skeleton. |
 989 `----------------------------------*/
 990
 991 static void
 992 parse_skel_decl (void)
 993 {
 994   skeleton = parse_dquoted_param ("%skeleton");
 995 }
 996
 997 /*----------------------------------------------------------------.
 998 | Read from finput until `%%' is seen.  Discard the `%%'.  Handle |
 999 | any `%' declarations, and copy the contents of any `%{ ... %}'  |
1000 | groups to ATTRS_OBSTACK.                                        |
1001 `----------------------------------------------------------------*/
1002
1003 static void
1004 read_declarations (void)
1005 {
1006   for (;;)
1007     {
1008       int c = skip_white_space ();
1009
1010       if (c == '%')
1011         {
1012           token_t tok = parse_percent_token ();
1013
1014           switch (tok)
1015             {
1016             case tok_two_percents:
1017               return;
1018
1019             case tok_percent_left_curly:
1020               copy_definition ();
1021               break;
1022
1023             case tok_token:
1024               parse_token_decl (token_sym, nterm_sym);
1025               break;
1026
1027             case tok_nterm:
1028               parse_token_decl (nterm_sym, token_sym);
1029               break;
1030
1031             case tok_type:
1032               parse_type_decl ();
1033               break;
1034
1035             case tok_start:
1036               parse_start_decl ();
1037               break;
1038
1039             case tok_union:
1040               parse_union_decl ();
1041               break;
1042
1043             case tok_expect:
1044               parse_expect_decl ();
1045               break;
1046
1047             case tok_thong:
1048               parse_thong_decl ();
1049               break;
1050
1051             case tok_left:
1052               parse_assoc_decl (left_assoc);
1053               break;
1054
1055             case tok_right:
1056               parse_assoc_decl (right_assoc);
1057               break;
1058
1059             case tok_nonassoc:
1060               parse_assoc_decl (non_assoc);
1061               break;
1062
1063             case tok_define:
1064               parse_muscle_decl ();
1065               break;
1066
1067             case tok_skel:
1068               parse_skel_decl ();
1069               break;
1070
1071             case tok_noop:
1072               break;
1073
1074             case tok_stropt:
1075             case tok_intopt:
1076             case tok_obsolete:
1077               abort ();
1078               break;
1079
1080             case tok_illegal:
1081             default:
1082               complain (_("unrecognized: %s"), token_buffer);
1083               skip_to_char ('%');
1084             }
1085         }
1086       else if (c == EOF)
1087         fatal (_("no input grammar"));
1088       else
1089         {
1090           char buf[] = "c";
1091           buf[0] = c;
1092           complain (_("unknown character: %s"), quote (buf));
1093           skip_to_char ('%');
1094         }
1095     }
1096 }
1097 \f
1098 /*-------------------------------------------------------------------.
1099 | Assuming that a `{' has just been seen, copy everything up to the  |
1100 | matching `}' into the actions file.  STACK_OFFSET is the number of |
1101 | values in the current rule so far, which says where to find `$0'   |
1102 | with respect to the top of the stack.                              |
1103 |                                                                    |
1104 | This routine is used both for actions and guards.  Only            |
1105 | ACTION_OBSTACK is used, but this is fine, since we use only        |
1106 | pointers to relevant portions inside this obstack.                 |
1107 `-------------------------------------------------------------------*/
1108
1109 static void
1110 parse_braces (symbol_list *rule, int stack_offset)
1111 {
1112   int c;
1113   int count;
1114
1115   count = 1;
1116   while (count > 0)
1117     {
1118       while ((c = getc (finput)) != '}')
1119         switch (c)
1120           {
1121           case '\n':
1122             obstack_1grow (&action_obstack, c);
1123             lineno++;
1124             break;
1125
1126           case '{':
1127             obstack_1grow (&action_obstack, c);
1128             count++;
1129             break;
1130
1131           case '\'':
1132           case '"':
1133             copy_string (finput, &action_obstack, c);
1134             break;
1135
1136           case '/':
1137             copy_comment (finput, &action_obstack);
1138             break;
1139
1140           case '$':
1141             copy_dollar (finput, &action_obstack,
1142                          rule, stack_offset);
1143             break;
1144
1145           case '@':
1146             copy_at (finput, &action_obstack,
1147                      stack_offset);
1148             break;
1149
1150           case EOF:
1151             fatal (_("unmatched %s"), "`{'");
1152
1153           default:
1154             obstack_1grow (&action_obstack, c);
1155           }
1156
1157       /* Above loop exits when C is '}'.  */
1158       if (--count)
1159         {
1160           obstack_1grow (&action_obstack, c);
1161           c = getc (finput);
1162         }
1163     }
1164
1165   obstack_1grow (&action_obstack, '\0');
1166 }
1167
1168
1169 static void
1170 parse_action (symbol_list *rule, int stack_offset)
1171 {
1172   rule->action_line = lineno;
1173   parse_braces (rule, stack_offset);
1174   rule->action = obstack_finish (&action_obstack);
1175 }
1176
1177
1178 static void
1179 parse_guard (symbol_list *rule, int stack_offset)
1180 {
1181   token_t t = lex ();
1182   if (t != tok_left_curly)
1183     complain (_("invalid %s declaration"), "%guard");
1184   rule->guard_line = lineno;
1185   parse_braces (rule, stack_offset);
1186   rule->guard = obstack_finish (&action_obstack);
1187 }
1188
1189 \f
1190
1191 /*-------------------------------------------------------------------.
1192 | Generate a dummy symbol, a nonterminal, whose name cannot conflict |
1193 | with the user's names.                                             |
1194 `-------------------------------------------------------------------*/
1195
1196 static bucket *
1197 gensym (void)
1198 {
1199   /* Incremented for each generated symbol */
1200   static int gensym_count = 0;
1201   static char buf[256];
1202
1203   bucket *sym;
1204
1205   sprintf (buf, "@%d", ++gensym_count);
1206   token_buffer = buf;
1207   sym = getsym (token_buffer);
1208   sym->class = nterm_sym;
1209   sym->value = nvars++;
1210   return sym;
1211 }
1212 \f
1213 /*-------------------------------------------------------------------.
1214 | Parse the input grammar into a one symbol_list structure.  Each    |
1215 | rule is represented by a sequence of symbols: the left hand side   |
1216 | followed by the contents of the right hand side, followed by a     |
1217 | null pointer instead of a symbol to terminate the rule.  The next  |
1218 | symbol is the lhs of the following rule.                           |
1219 |                                                                    |
1220 | All guards and actions are copied out to the appropriate files,    |
1221 | labelled by the rule number they apply to.                         |
1222 |                                                                    |
1223 | Bison used to allow some %directives in the rules sections, but    |
1224 | this is no longer consider appropriate: (i) the documented grammar |
1225 | doesn't claim it, (ii), it would promote bad style, (iii), error   |
1226 | recovery for %directives consists in skipping the junk until a `%' |
1227 | is seen and helrp synchronizing.  This scheme is definitely wrong  |
1228 | in the rules section.                                              |
1229 `-------------------------------------------------------------------*/
1230
1231 static void
1232 readgram (void)
1233 {
1234   token_t t;
1235   bucket *lhs = NULL;
1236   symbol_list *p = NULL;
1237   symbol_list *p1 = NULL;
1238   bucket *bp;
1239
1240   /* Points to first symbol_list of current rule. its symbol is the
1241      lhs of the rule.  */
1242   symbol_list *crule = NULL;
1243   /* Points to the symbol_list preceding crule.  */
1244   symbol_list *crule1 = NULL;
1245
1246   t = lex ();
1247
1248   while (t != tok_two_percents && t != tok_eof)
1249     if (t == tok_identifier || t == tok_bar)
1250       {
1251         int action_flag = 0;
1252         /* Number of symbols in rhs of this rule so far */
1253         int rulelength = 0;
1254         int xactions = 0;       /* JF for error checking */
1255         bucket *first_rhs = 0;
1256
1257         if (t == tok_identifier)
1258           {
1259             lhs = symval;
1260
1261             if (!start_flag)
1262               {
1263                 startval = lhs;
1264                 start_flag = 1;
1265               }
1266
1267             t = lex ();
1268             if (t != tok_colon)
1269               {
1270                 complain (_("ill-formed rule: initial symbol not followed by colon"));
1271                 unlex (t);
1272               }
1273           }
1274
1275         if (nrules == 0 && t == tok_bar)
1276           {
1277             complain (_("grammar starts with vertical bar"));
1278             lhs = symval;       /* BOGUS: use a random symval */
1279           }
1280         /* start a new rule and record its lhs.  */
1281
1282         nrules++;
1283         nitems++;
1284
1285         p = symbol_list_new (lhs);
1286
1287         crule1 = p1;
1288         if (p1)
1289           p1->next = p;
1290         else
1291           grammar = p;
1292
1293         p1 = p;
1294         crule = p;
1295
1296         /* mark the rule's lhs as a nonterminal if not already so.  */
1297
1298         if (lhs->class == unknown_sym)
1299           {
1300             lhs->class = nterm_sym;
1301             lhs->value = nvars;
1302             nvars++;
1303           }
1304         else if (lhs->class == token_sym)
1305           complain (_("rule given for %s, which is a token"), lhs->tag);
1306
1307         /* read the rhs of the rule.  */
1308
1309         for (;;)
1310           {
1311             t = lex ();
1312             if (t == tok_prec)
1313               {
1314                 t = lex ();
1315                 crule->ruleprec = symval;
1316                 t = lex ();
1317               }
1318
1319             if (!(t == tok_identifier || t == tok_left_curly))
1320               break;
1321
1322             /* If next token is an identifier, see if a colon follows it.
1323                If one does, exit this rule now.  */
1324             if (t == tok_identifier)
1325               {
1326                 bucket *ssave;
1327                 token_t t1;
1328
1329                 ssave = symval;
1330                 t1 = lex ();
1331                 unlex (t1);
1332                 symval = ssave;
1333                 if (t1 == tok_colon)
1334                   break;
1335
1336                 if (!first_rhs) /* JF */
1337                   first_rhs = symval;
1338                 /* Not followed by colon =>
1339                    process as part of this rule's rhs.  */
1340               }
1341
1342             /* If we just passed an action, that action was in the middle
1343                of a rule, so make a dummy rule to reduce it to a
1344                non-terminal.  */
1345             if (action_flag)
1346               {
1347                 /* Since the action was written out with this rule's
1348                    number, we must give the new rule this number by
1349                    inserting the new rule before it.  */
1350
1351                 /* Make a dummy nonterminal, a gensym.  */
1352                 bucket *sdummy = gensym ();
1353
1354                 /* Make a new rule, whose body is empty, before the
1355                    current one, so that the action just read can
1356                    belong to it.  */
1357                 nrules++;
1358                 nitems++;
1359                 p = symbol_list_new (sdummy);
1360                 /* Attach its lineno to that of the host rule. */
1361                 p->line = crule->line;
1362                 if (crule1)
1363                   crule1->next = p;
1364                 else
1365                   grammar = p;
1366                 /* End of the rule. */
1367                 crule1 = symbol_list_new (NULL);
1368                 crule1->next = crule;
1369
1370                 p->next = crule1;
1371
1372                 /* Insert the dummy generated by that rule into this
1373                    rule.  */
1374                 nitems++;
1375                 p = symbol_list_new (sdummy);
1376                 p1->next = p;
1377                 p1 = p;
1378
1379                 action_flag = 0;
1380               }
1381
1382             if (t == tok_identifier)
1383               {
1384                 nitems++;
1385                 p = symbol_list_new (symval);
1386                 p1->next = p;
1387                 p1 = p;
1388               }
1389             else                /* handle an action.  */
1390               {
1391                 parse_action (crule, rulelength);
1392                 action_flag = 1;
1393                 xactions++;     /* JF */
1394               }
1395             rulelength++;
1396           }                     /* end of  read rhs of rule */
1397
1398         /* Put an empty link in the list to mark the end of this rule  */
1399         p = symbol_list_new (NULL);
1400         p1->next = p;
1401         p1 = p;
1402
1403         if (t == tok_prec)
1404           {
1405             complain (_("two @prec's in a row"));
1406             t = lex ();
1407             crule->ruleprec = symval;
1408             t = lex ();
1409           }
1410
1411         if (t == tok_guard)
1412           {
1413             if (!semantic_parser)
1414               complain (_("%%guard present but %%semantic_parser not specified"));
1415
1416             parse_guard (crule, rulelength);
1417             t = lex ();
1418           }
1419
1420         if (t == tok_left_curly)
1421           {
1422             /* This case never occurs -wjh */
1423             if (action_flag)
1424               complain (_("two actions at end of one rule"));
1425             parse_action (crule, rulelength);
1426             action_flag = 1;
1427             xactions++; /* -wjh */
1428             t = lex ();
1429           }
1430         /* If $$ is being set in default way, report if any type
1431            mismatch.  */
1432         else if (!xactions
1433                  && first_rhs && lhs->type_name != first_rhs->type_name)
1434           {
1435             if (lhs->type_name == 0
1436                 || first_rhs->type_name == 0
1437                 || strcmp (lhs->type_name, first_rhs->type_name))
1438               complain (_("type clash (`%s' `%s') on default action"),
1439                         lhs->type_name ? lhs->type_name : "",
1440                         first_rhs->type_name ? first_rhs->type_name : "");
1441           }
1442         /* Warn if there is no default for $$ but we need one.  */
1443         else if (!xactions && !first_rhs && lhs->type_name != 0)
1444           complain (_("empty rule for typed nonterminal, and no action"));
1445         if (t == tok_semicolon)
1446           t = lex ();
1447       }
1448     else
1449       {
1450         complain (_("invalid input: %s"), quote (token_buffer));
1451         t = lex ();
1452       }
1453
1454   /* grammar has been read.  Do some checking */
1455
1456   if (nrules == 0)
1457     fatal (_("no rules in the input grammar"));
1458
1459   /* Report any undefined symbols and consider them nonterminals.  */
1460
1461   for (bp = firstsymbol; bp; bp = bp->next)
1462     if (bp->class == unknown_sym)
1463       {
1464         complain (_
1465                   ("symbol %s is used, but is not defined as a token and has no rules"),
1466                   bp->tag);
1467         bp->class = nterm_sym;
1468         bp->value = nvars++;
1469       }
1470
1471   /* Insert the initial rule, which line is that of the first rule
1472      (not that of the start symbol):
1473
1474      axiom: %start EOF.  */
1475   p = symbol_list_new (axiom);
1476   p->line = grammar->line;
1477   p->next = symbol_list_new (startval);
1478   p->next->next = symbol_list_new (eoftoken);
1479   p->next->next->next = symbol_list_new (NULL);
1480   p->next->next->next->next = grammar;
1481   nrules += 1;
1482   nitems += 3;
1483   grammar = p;
1484   startval = axiom;
1485
1486   if (nsyms > MAXSHORT)
1487     fatal (_("too many symbols (tokens plus nonterminals); maximum %d"),
1488            MAXSHORT);
1489
1490   ntokens = nsyms - nvars;
1491 }
1492
1493 /* At the end of the grammar file, some C source code must
1494    be stored. It is going to be associated to the epilogue
1495    directive.  */
1496 static void
1497 read_additionnal_code (void)
1498 {
1499   char c;
1500   struct obstack el_obstack;
1501
1502   obstack_init (&el_obstack);
1503
1504   if (!no_lines_flag)
1505     {
1506       obstack_fgrow2 (&el_obstack, muscle_find ("linef"),
1507                       lineno, quotearg_style (c_quoting_style,
1508                                               muscle_find ("filename")));
1509     }
1510
1511   while ((c = getc (finput)) != EOF)
1512     obstack_1grow (&el_obstack, c);
1513
1514   obstack_1grow (&el_obstack, 0);
1515   muscle_insert ("epilogue", obstack_finish (&el_obstack));
1516 }
1517
1518 \f
1519 /*------------------------------------------------------------------.
1520 | Set TOKEN_TRANSLATIONS.  Check that no two symbols share the same |
1521 | number.                                                           |
1522 `------------------------------------------------------------------*/
1523
1524 static void
1525 token_translations_init (void)
1526 {
1527   bucket *bp = NULL;
1528   int i;
1529
1530   token_translations = XCALLOC (short, max_user_token_number + 1);
1531
1532   /* Initialize all entries for literal tokens to 2, the internal
1533      token number for $undefined., which represents all invalid
1534      inputs.  */
1535   for (i = 0; i <= max_user_token_number; i++)
1536     token_translations[i] = 2;
1537
1538   for (bp = firstsymbol; bp; bp = bp->next)
1539     {
1540       /* Non-terminal? */
1541       if (bp->value >= ntokens)
1542         continue;
1543       /* A token string alias? */
1544       if (bp->user_token_number == SALIAS)
1545         continue;
1546
1547       assert (bp->user_token_number != SUNDEF);
1548
1549       /* A token which translation has already been set? */
1550       if (token_translations[bp->user_token_number] != 2)
1551         complain (_("tokens %s and %s both assigned number %d"),
1552                   tags[token_translations[bp->user_token_number]],
1553                   bp->tag, bp->user_token_number);
1554       token_translations[bp->user_token_number] = bp->value;
1555     }
1556 }
1557
1558
1559 /*------------------------------------------------------------------.
1560 | Assign symbol numbers, and write definition of token names into   |
1561 | FDEFINES.  Set up vectors TAGS and SPREC of names and precedences |
1562 | of symbols.                                                       |
1563 `------------------------------------------------------------------*/
1564
1565 static void
1566 packsymbols (void)
1567 {
1568   bucket *bp = NULL;
1569   int tokno = 1;
1570   int last_user_token_number;
1571
1572   tags = XCALLOC (char *, nsyms + 1);
1573   user_toknums = XCALLOC (short, nsyms + 1);
1574
1575   sprec = XCALLOC (short, nsyms);
1576   sassoc = XCALLOC (short, nsyms);
1577
1578   max_user_token_number = 256;
1579   last_user_token_number = 256;
1580
1581   for (bp = firstsymbol; bp; bp = bp->next)
1582     {
1583       if (bp->class == nterm_sym)
1584         {
1585           bp->value += ntokens;
1586         }
1587       else if (bp->alias)
1588         {
1589           /* This symbol and its alias are a single token defn.
1590              Allocate a tokno, and assign to both check agreement of
1591              prec and assoc fields and make both the same */
1592           if (bp->value == -1)
1593             {
1594               if (bp == eoftoken || bp->alias == eoftoken)
1595                 bp->value = bp->alias->value = 0;
1596               else
1597                 {
1598                   bp->value = bp->alias->value = tokno++;
1599                 }
1600             }
1601
1602           if (bp->prec != bp->alias->prec)
1603             {
1604               if (bp->prec != 0 && bp->alias->prec != 0
1605                   && bp->user_token_number == SALIAS)
1606                 complain (_("conflicting precedences for %s and %s"),
1607                           bp->tag, bp->alias->tag);
1608               if (bp->prec != 0)
1609                 bp->alias->prec = bp->prec;
1610               else
1611                 bp->prec = bp->alias->prec;
1612             }
1613
1614           if (bp->assoc != bp->alias->assoc)
1615             {
1616               if (bp->assoc != 0 && bp->alias->assoc != 0
1617                   && bp->user_token_number == SALIAS)
1618                 complain (_("conflicting assoc values for %s and %s"),
1619                           bp->tag, bp->alias->tag);
1620               if (bp->assoc != 0)
1621                 bp->alias->assoc = bp->assoc;
1622               else
1623                 bp->assoc = bp->alias->assoc;
1624             }
1625
1626           /* Do not do processing below for SALIASs.  */
1627           if (bp->user_token_number == SALIAS)
1628             continue;
1629
1630         }
1631       else /* bp->class == token_sym */
1632         {
1633           if (bp == eoftoken)
1634             bp->value = 0;
1635           else
1636             bp->value = tokno++;
1637         }
1638
1639       if (bp->class == token_sym)
1640         {
1641           if (bp->user_token_number == SUNDEF)
1642             bp->user_token_number = ++last_user_token_number;
1643           if (bp->user_token_number > max_user_token_number)
1644             max_user_token_number = bp->user_token_number;
1645         }
1646
1647       tags[bp->value] = bp->tag;
1648       user_toknums[bp->value] = bp->user_token_number;
1649       sprec[bp->value] = bp->prec;
1650       sassoc[bp->value] = bp->assoc;
1651     }
1652
1653   token_translations_init ();
1654
1655   error_token_number = errtoken->value;
1656
1657   if (startval->class == unknown_sym)
1658     fatal (_("the start symbol %s is undefined"), startval->tag);
1659   else if (startval->class == token_sym)
1660     fatal (_("the start symbol %s is a token"), startval->tag);
1661
1662   start_symbol = startval->value;
1663 }
1664
1665
1666 /*---------------------------------------------------------------.
1667 | Save the definition of token names in the `TOKENDEFS' muscle.  |
1668 `---------------------------------------------------------------*/
1669
1670 static void
1671 symbols_save (void)
1672 {
1673   struct obstack tokendefs;
1674   bucket *bp;
1675   obstack_init (&tokendefs);
1676
1677   for (bp = firstsymbol; bp; bp = bp->next)
1678     {
1679       char *symbol = bp->tag;                /* get symbol */
1680
1681       if (bp->value >= ntokens)
1682         continue;
1683       if (bp->user_token_number == SALIAS)
1684         continue;
1685       if ('\'' == *symbol)
1686         continue;               /* skip literal character */
1687       if (bp == errtoken)
1688         continue;               /* skip error token */
1689       if ('\"' == *symbol)
1690         {
1691           /* use literal string only if given a symbol with an alias */
1692           if (bp->alias)
1693             symbol = bp->alias->tag;
1694           else
1695             continue;
1696         }
1697
1698       /* Don't #define nonliteral tokens whose names contain periods.  */
1699       if (strchr (symbol, '.'))
1700         continue;
1701
1702       obstack_fgrow2 (&tokendefs, "# define %s\t%d\n",
1703                       symbol, bp->user_token_number);
1704       if (semantic_parser)
1705         /* FIXME: This is probably wrong, and should be just as
1706            above. --akim.  */
1707         obstack_fgrow2 (&tokendefs, "# define T%s\t%d\n", symbol, bp->value);
1708     }
1709
1710   obstack_1grow (&tokendefs, 0);
1711   muscle_insert ("tokendef", xstrdup (obstack_finish (&tokendefs)));
1712   obstack_free (&tokendefs, NULL);
1713 }
1714
1715
1716 /*---------------------------------------------------------------.
1717 | Convert the rules into the representation using RRHS, RLHS and |
1718 | RITEMS.                                                        |
1719 `---------------------------------------------------------------*/
1720
1721 static void
1722 packgram (void)
1723 {
1724   int itemno;
1725   int ruleno;
1726   symbol_list *p;
1727
1728   /* We use short to index items.  */
1729   if (nitems >= MAXSHORT)
1730     fatal (_("too many items (max %d)"), MAXSHORT);
1731
1732   ritem = XCALLOC (short, nitems + 1);
1733   rule_table = XCALLOC (rule_t, nrules) - 1;
1734
1735   itemno = 0;
1736   ruleno = 1;
1737
1738   p = grammar;
1739   while (p)
1740     {
1741       bucket *ruleprec = p->ruleprec;
1742       rule_table[ruleno].lhs = p->sym->value;
1743       rule_table[ruleno].rhs = itemno;
1744       rule_table[ruleno].line = p->line;
1745       rule_table[ruleno].useful = TRUE;
1746       rule_table[ruleno].action = p->action;
1747       rule_table[ruleno].action_line = p->action_line;
1748       rule_table[ruleno].guard = p->guard;
1749       rule_table[ruleno].guard_line = p->guard_line;
1750
1751       p = p->next;
1752       while (p && p->sym)
1753         {
1754           ritem[itemno++] = p->sym->value;
1755           /* A rule gets by default the precedence and associativity
1756              of the last token in it.  */
1757           if (p->sym->class == token_sym)
1758             {
1759               rule_table[ruleno].prec = p->sym->prec;
1760               rule_table[ruleno].assoc = p->sym->assoc;
1761             }
1762           if (p)
1763             p = p->next;
1764         }
1765
1766       /* If this rule has a %prec,
1767          the specified symbol's precedence replaces the default.  */
1768       if (ruleprec)
1769         {
1770           rule_table[ruleno].prec = ruleprec->prec;
1771           rule_table[ruleno].assoc = ruleprec->assoc;
1772           rule_table[ruleno].precsym = ruleprec->value;
1773         }
1774
1775       ritem[itemno++] = -ruleno;
1776       ruleno++;
1777
1778       if (p)
1779         p = p->next;
1780     }
1781
1782   ritem[itemno] = 0;
1783   nritems = itemno;
1784   assert (nritems == nitems);
1785
1786   if (trace_flag)
1787     ritem_print (stderr);
1788 }
1789 \f
1790 /*-------------------------------------------------------------------.
1791 | Read in the grammar specification and record it in the format      |
1792 | described in gram.h.  All guards are copied into the GUARD_OBSTACK |
1793 | and all actions into ACTION_OBSTACK, in each case forming the body |
1794 | of a C function (YYGUARD or YYACTION) which contains a switch      |
1795 | statement to decide which guard or action to execute.              |
1796 `-------------------------------------------------------------------*/
1797
1798 void
1799 reader (void)
1800 {
1801   start_flag = 0;
1802   startval = NULL;              /* start symbol not specified yet. */
1803
1804   nsyms = 0;
1805   nvars = 0;
1806   nrules = 0;
1807   nitems = 0;
1808
1809   typed = 0;
1810   lastprec = 0;
1811
1812   semantic_parser = 0;
1813   pure_parser = 0;
1814
1815   grammar = NULL;
1816
1817   lex_init ();
1818   lineno = 1;
1819
1820   /* Initialize the muscle obstack.  */
1821   obstack_init (&muscle_obstack);
1822
1823   /* Initialize the symbol table.  */
1824   tabinit ();
1825
1826   /* Construct the axiom symbol. */
1827   axiom = getsym ("$axiom");
1828   axiom->class = nterm_sym;
1829   axiom->value = nvars++;
1830
1831   /* Construct the error token */
1832   errtoken = getsym ("error");
1833   errtoken->class = token_sym;
1834   errtoken->user_token_number = 256;    /* Value specified by POSIX.  */
1835
1836   /* Construct a token that represents all undefined literal tokens.
1837      It is always token number 2.  */
1838   undeftoken = getsym ("$undefined.");
1839   undeftoken->class = token_sym;
1840   undeftoken->user_token_number = 2;
1841
1842   /* Initialize the obstacks. */
1843   obstack_init (&action_obstack);
1844   obstack_init (&attrs_obstack);
1845   obstack_init (&output_obstack);
1846
1847   finput = xfopen (infile, "r");
1848
1849   /* Read the declaration section.  Copy %{ ... %} groups to
1850      TABLE_OBSTACK and FDEFINES file.  Also notice any %token, %left,
1851      etc. found there.  */
1852   read_declarations ();
1853
1854   /* If the user did not define her EOFTOKEN, do it now. */
1855   if (!eoftoken)
1856     {
1857       eoftoken = getsym ("$");
1858       eoftoken->class = token_sym;
1859       /* Value specified by POSIX.  */
1860       eoftoken->user_token_number = 0;
1861     }
1862
1863   /* Read in the grammar, build grammar in list form.  Write out
1864      guards and actions.  */
1865   readgram ();
1866   /* Some C code is given at the end of the grammar file. */
1867   read_additionnal_code ();
1868
1869   lex_free ();
1870   xfclose (finput);
1871
1872   /* Assign the symbols their symbol numbers.  Write #defines for the
1873      token symbols into FDEFINES if requested.  */
1874   packsymbols ();
1875   /* Save them. */
1876   symbols_save ();
1877
1878   /* Convert the grammar into the format described in gram.h.  */
1879   packgram ();
1880 }