]> git.saurik.com Git - bison.git/blob - src/reader.c
Fix --enable-gcc-warnings problems.
[bison.git] / src / reader.c
1 /* Input parser for Bison
2
3 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003,
4 2005, 2006, 2007, 2009 Free Software Foundation, Inc.
5
6 This file is part of Bison, the GNU Compiler Compiler.
7
8 This program is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation, either version 3 of the License, or
11 (at your option) any later version.
12
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
17
18 You should have received a copy of the GNU General Public License
19 along with this program. If not, see <http://www.gnu.org/licenses/>. */
20
21 #include <config.h>
22 #include "system.h"
23
24 #include <quote.h>
25 #include <quotearg.h>
26
27 #include "complain.h"
28 #include "conflicts.h"
29 #include "files.h"
30 #include "getargs.h"
31 #include "gram.h"
32 #include "muscle-tab.h"
33 #include "reader.h"
34 #include "symlist.h"
35 #include "symtab.h"
36 #include "scan-gram.h"
37 #include "scan-code.h"
38
39 static void check_and_convert_grammar (void);
40
41 static symbol_list *grammar = NULL;
42 static bool start_flag = false;
43 merger_list *merge_functions;
44
45 /* Was %union seen? */
46 bool union_seen = false;
47
48 /* Was a tag seen? */
49 bool tag_seen = false;
50
51 /* Should rules have a default precedence? */
52 bool default_prec = true;
53 \f
54 /*-----------------------.
55 | Set the start symbol. |
56 `-----------------------*/
57
58 void
59 grammar_start_symbol_set (symbol *sym, location loc)
60 {
61 if (start_flag)
62 complain_at (loc, _("multiple %s declarations"), "%start");
63 else
64 {
65 start_flag = true;
66 startsymbol = sym;
67 startsymbol_location = loc;
68 }
69 }
70
71 \f
72
73 /*------------------------------------------------------------------------.
74 | Return the merger index for a merging function named NAME. Records the |
75 | function, if new, in MERGER_LIST. |
76 `------------------------------------------------------------------------*/
77
78 static int
79 get_merge_function (uniqstr name)
80 {
81 merger_list *syms;
82 merger_list head;
83 int n;
84
85 if (! glr_parser)
86 return 0;
87
88 head.next = merge_functions;
89 for (syms = &head, n = 1; syms->next; syms = syms->next, n += 1)
90 if (UNIQSTR_EQ (name, syms->next->name))
91 break;
92 if (syms->next == NULL)
93 {
94 syms->next = xmalloc (sizeof syms->next[0]);
95 syms->next->name = uniqstr_new (name);
96 /* After all symbol type declarations have been parsed, packgram invokes
97 record_merge_function_type to set the type. */
98 syms->next->type = NULL;
99 syms->next->next = NULL;
100 merge_functions = head.next;
101 }
102 return n;
103 }
104
105 /*-------------------------------------------------------------------------.
106 | For the existing merging function with index MERGER, record the result |
107 | type as TYPE as required by the lhs of the rule whose %merge declaration |
108 | is at DECLARATION_LOC. |
109 `-------------------------------------------------------------------------*/
110
111 static void
112 record_merge_function_type (int merger, uniqstr type, location declaration_loc)
113 {
114 int merger_find;
115 merger_list *merge_function;
116
117 if (merger <= 0)
118 return;
119
120 if (type == NULL)
121 type = uniqstr_new ("");
122
123 merger_find = 1;
124 for (merge_function = merge_functions;
125 merge_function != NULL && merger_find != merger;
126 merge_function = merge_function->next)
127 merger_find += 1;
128 aver (merge_function != NULL && merger_find == merger);
129 if (merge_function->type != NULL && !UNIQSTR_EQ (merge_function->type, type))
130 {
131 complain_at (declaration_loc,
132 _("result type clash on merge function `%s': <%s> != <%s>"),
133 merge_function->name, type, merge_function->type);
134 complain_at (merge_function->type_declaration_location,
135 _("previous declaration"));
136 }
137 merge_function->type = uniqstr_new (type);
138 merge_function->type_declaration_location = declaration_loc;
139 }
140
141 /*--------------------------------------.
142 | Free all merge-function definitions. |
143 `--------------------------------------*/
144
145 void
146 free_merger_functions (void)
147 {
148 merger_list *L0 = merge_functions;
149 while (L0)
150 {
151 merger_list *L1 = L0->next;
152 free (L0);
153 L0 = L1;
154 }
155 }
156
157 \f
158 /*-------------------------------------------------------------------.
159 | Parse the input grammar into a one symbol_list structure. Each |
160 | rule is represented by a sequence of symbols: the left hand side |
161 | followed by the contents of the right hand side, followed by a |
162 | null pointer instead of a symbol to terminate the rule. The next |
163 | symbol is the lhs of the following rule. |
164 | |
165 | All actions are copied out, labelled by the rule number they apply |
166 | to. |
167 `-------------------------------------------------------------------*/
168
169 /* The (currently) last symbol of GRAMMAR. */
170 static symbol_list *grammar_end = NULL;
171
172 /* Append SYM to the grammar. */
173 static symbol_list *
174 grammar_symbol_append (symbol *sym, location loc)
175 {
176 symbol_list *p = symbol_list_sym_new (sym, loc);
177
178 if (grammar_end)
179 grammar_end->next = p;
180 else
181 grammar = p;
182
183 grammar_end = p;
184
185 /* A null SYM stands for an end of rule; it is not an actual
186 part of it. */
187 if (sym)
188 ++nritems;
189
190 return p;
191 }
192
193 static void
194 assign_named_ref (symbol_list *p, named_ref *name)
195 {
196 symbol *sym = p->content.sym;
197
198 if (name->id == sym->tag)
199 {
200 warn_at (name->loc,
201 _("duplicated symbol name for %s ignored"),
202 quote (sym->tag));
203 named_ref_free (name);
204 }
205 else
206 p->named_ref = name;
207 }
208
209
210 /* The rule currently being defined, and the previous rule.
211 CURRENT_RULE points to the first LHS of the current rule, while
212 PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
213 static symbol_list *current_rule = NULL;
214 static symbol_list *previous_rule_end = NULL;
215
216
217 /*----------------------------------------------.
218 | Create a new rule for LHS in to the GRAMMAR. |
219 `----------------------------------------------*/
220
221 void
222 grammar_current_rule_begin (symbol *lhs, location loc,
223 named_ref *lhs_name)
224 {
225 symbol_list* p;
226
227 /* Start a new rule and record its lhs. */
228 ++nrules;
229 previous_rule_end = grammar_end;
230
231 p = grammar_symbol_append (lhs, loc);
232 if (lhs_name)
233 assign_named_ref(p, lhs_name);
234
235 current_rule = grammar_end;
236
237 /* Mark the rule's lhs as a nonterminal if not already so. */
238 if (lhs->class == unknown_sym)
239 {
240 lhs->class = nterm_sym;
241 lhs->number = nvars;
242 ++nvars;
243 }
244 else if (lhs->class == token_sym)
245 complain_at (loc, _("rule given for %s, which is a token"), lhs->tag);
246 }
247
248
249 /*----------------------------------------------------------------------.
250 | A symbol should be used if either: |
251 | 1. It has a destructor. |
252 | 2. --warnings=midrule-values and the symbol is a mid-rule symbol |
253 | (i.e., the generated LHS replacing a mid-rule action) that was |
254 | assigned to or used, as in "exp: { $$ = 1; } { $$ = $1; }". |
255 `----------------------------------------------------------------------*/
256
257 static bool
258 symbol_should_be_used (symbol_list const *s)
259 {
260 if (symbol_destructor_get (s->content.sym)->code)
261 return true;
262 if (warnings_flag & warnings_midrule_values)
263 return ((s->midrule && s->midrule->action_props.is_value_used)
264 || (s->midrule_parent_rule
265 && symbol_list_n_get (s->midrule_parent_rule,
266 s->midrule_parent_rhs_index)
267 ->action_props.is_value_used));
268 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. |
274 `----------------------------------------------------------------*/
275
276 static void
277 grammar_rule_check (const symbol_list *r)
278 {
279 /* Type check.
280
281 If there is an action, then there is nothing we can do: the user
282 is allowed to shoot herself in the foot.
283
284 Don't worry about the default action if $$ is untyped, since $$'s
285 value can't be used. */
286 if (!r->action_props.code && r->content.sym->type_name)
287 {
288 symbol *first_rhs = r->next->content.sym;
289 /* If $$ is being set in default way, report if any type mismatch. */
290 if (first_rhs)
291 {
292 char const *lhs_type = r->content.sym->type_name;
293 const char *rhs_type =
294 first_rhs->type_name ? first_rhs->type_name : "";
295 if (!UNIQSTR_EQ (lhs_type, rhs_type))
296 warn_at (r->location,
297 _("type clash on default action: <%s> != <%s>"),
298 lhs_type, rhs_type);
299 }
300 /* Warn if there is no default for $$ but we need one. */
301 else
302 warn_at (r->location,
303 _("empty rule for typed nonterminal, and no action"));
304 }
305
306 /* Check that symbol values that should be used are in fact used. */
307 {
308 symbol_list const *l = r;
309 int n = 0;
310 for (; l && l->content.sym; l = l->next, ++n)
311 if (! (l->action_props.is_value_used
312 || !symbol_should_be_used (l)
313 /* The default action, $$ = $1, `uses' both. */
314 || (!r->action_props.code && (n == 0 || n == 1))))
315 {
316 if (n)
317 warn_at (r->location, _("unused value: $%d"), n);
318 else
319 warn_at (r->location, _("unset value: $$"));
320 }
321 }
322 }
323
324
325 /*-------------------------------------.
326 | End the currently being grown rule. |
327 `-------------------------------------*/
328
329 void
330 grammar_current_rule_end (location loc)
331 {
332 /* Put an empty link in the list to mark the end of this rule */
333 grammar_symbol_append (NULL, grammar_end->location);
334 current_rule->location = loc;
335 }
336
337
338 /*-------------------------------------------------------------------.
339 | The previous action turns out the be a mid-rule action. Attach it |
340 | to the current rule, i.e., create a dummy symbol, attach it this |
341 | mid-rule action, and append this dummy nonterminal to the current |
342 | rule. |
343 `-------------------------------------------------------------------*/
344
345 void
346 grammar_midrule_action (void)
347 {
348 /* Since the action was written out with this rule's number, we must
349 give the new rule this number by inserting the new rule before
350 it. */
351
352 /* Make a DUMMY nonterminal, whose location is that of the midrule
353 action. Create the MIDRULE. */
354 location dummy_location = current_rule->action_props.location;
355 symbol *dummy = dummy_symbol_get (dummy_location);
356 symbol_list *midrule = symbol_list_sym_new (dummy, dummy_location);
357
358 /* Remember named_ref of previous action. */
359 named_ref *action_name = current_rule->action_props.named_ref;
360
361 /* Make a new rule, whose body is empty, before the current one, so
362 that the action just read can belong to it. */
363 ++nrules;
364 ++nritems;
365 /* Attach its location and actions to that of the DUMMY. */
366 midrule->location = dummy_location;
367 code_props_rule_action_init (&midrule->action_props,
368 current_rule->action_props.code,
369 current_rule->action_props.location,
370 midrule, 0);
371 code_props_none_init (&current_rule->action_props);
372
373 if (previous_rule_end)
374 previous_rule_end->next = midrule;
375 else
376 grammar = midrule;
377
378 /* End the dummy's rule. */
379 midrule->next = symbol_list_sym_new (NULL, dummy_location);
380 midrule->next->next = current_rule;
381
382 previous_rule_end = midrule->next;
383
384 /* Insert the dummy nonterminal replacing the midrule action into
385 the current rule. Bind it to its dedicated rule. */
386 grammar_current_rule_symbol_append (dummy, dummy_location,
387 action_name);
388 grammar_end->midrule = midrule;
389 midrule->midrule_parent_rule = current_rule;
390 midrule->midrule_parent_rhs_index = symbol_list_length (current_rule->next);
391 }
392
393 /* Set the precedence symbol of the current rule to PRECSYM. */
394
395 void
396 grammar_current_rule_prec_set (symbol *precsym, location loc)
397 {
398 symbol_class_set (precsym, token_sym, loc, false);
399 if (current_rule->ruleprec)
400 complain_at (loc, _("only one %s allowed per rule"), "%prec");
401 current_rule->ruleprec = precsym;
402 }
403
404 /* Attach dynamic precedence DPREC to the current rule. */
405
406 void
407 grammar_current_rule_dprec_set (int dprec, location loc)
408 {
409 if (! glr_parser)
410 warn_at (loc, _("%s affects only GLR parsers"), "%dprec");
411 if (dprec <= 0)
412 complain_at (loc, _("%s must be followed by positive number"), "%dprec");
413 else if (current_rule->dprec != 0)
414 complain_at (loc, _("only one %s allowed per rule"), "%dprec");
415 current_rule->dprec = dprec;
416 }
417
418 /* Attach a merge function NAME with argument type TYPE to current
419 rule. */
420
421 void
422 grammar_current_rule_merge_set (uniqstr name, location loc)
423 {
424 if (! glr_parser)
425 warn_at (loc, _("%s affects only GLR parsers"), "%merge");
426 if (current_rule->merger != 0)
427 complain_at (loc, _("only one %s allowed per rule"), "%merge");
428 current_rule->merger = get_merge_function (name);
429 current_rule->merger_declaration_location = loc;
430 }
431
432 /* Attach SYM to the current rule. If needed, move the previous
433 action as a mid-rule action. */
434
435 void
436 grammar_current_rule_symbol_append (symbol *sym, location loc,
437 named_ref *name)
438 {
439 symbol_list *p;
440 if (current_rule->action_props.code)
441 grammar_midrule_action ();
442 p = grammar_symbol_append (sym, loc);
443 if (name)
444 assign_named_ref(p, name);
445 }
446
447 /* Attach an ACTION to the current rule. */
448
449 void
450 grammar_current_rule_action_append (const char *action, location loc,
451 named_ref *name)
452 {
453 if (current_rule->action_props.code)
454 grammar_midrule_action ();
455 /* After all symbol declarations have been parsed, packgram invokes
456 code_props_translate_code. */
457 code_props_rule_action_init (&current_rule->action_props, action, loc,
458 current_rule, name);
459 }
460
461 \f
462 /*---------------------------------------------------------------.
463 | Convert the rules into the representation using RRHS, RLHS and |
464 | RITEM. |
465 `---------------------------------------------------------------*/
466
467 static void
468 packgram (void)
469 {
470 unsigned int itemno = 0;
471 rule_number ruleno = 0;
472 symbol_list *p = grammar;
473
474 ritem = xnmalloc (nritems + 1, sizeof *ritem);
475
476 /* This sentinel is used by build_relations in gram.c. */
477 *ritem++ = 0;
478
479 rules = xnmalloc (nrules, sizeof *rules);
480
481 while (p)
482 {
483 int rule_length = 0;
484 symbol *ruleprec = p->ruleprec;
485 record_merge_function_type (p->merger, p->content.sym->type_name,
486 p->merger_declaration_location);
487 rules[ruleno].user_number = ruleno;
488 rules[ruleno].number = ruleno;
489 rules[ruleno].lhs = p->content.sym;
490 rules[ruleno].rhs = ritem + itemno;
491 rules[ruleno].prec = NULL;
492 rules[ruleno].dprec = p->dprec;
493 rules[ruleno].merger = p->merger;
494 rules[ruleno].precsym = NULL;
495 rules[ruleno].location = p->location;
496 rules[ruleno].useful = true;
497 rules[ruleno].action = p->action_props.code;
498 rules[ruleno].action_location = p->action_props.location;
499
500 /* If the midrule's $$ is set or its $n is used, remove the `$' from the
501 symbol name so that it's a user-defined symbol so that the default
502 %destructor and %printer apply. */
503 if (p->midrule_parent_rule
504 && (p->action_props.is_value_used
505 || symbol_list_n_get (p->midrule_parent_rule,
506 p->midrule_parent_rhs_index)
507 ->action_props.is_value_used))
508 p->content.sym->tag += 1;
509
510 /* Don't check the generated rule 0. It has no action, so some rhs
511 symbols may appear unused, but the parsing algorithm ensures that
512 %destructor's are invoked appropriately. */
513 if (p != grammar)
514 grammar_rule_check (p);
515
516 for (p = p->next; p && p->content.sym; p = p->next)
517 {
518 ++rule_length;
519
520 /* Don't allow rule_length == INT_MAX, since that might
521 cause confusion with strtol if INT_MAX == LONG_MAX. */
522 if (rule_length == INT_MAX)
523 fatal_at (rules[ruleno].location, _("rule is too long"));
524
525 /* item_number = symbol_number.
526 But the former needs to contain more: negative rule numbers. */
527 ritem[itemno++] =
528 symbol_number_as_item_number (p->content.sym->number);
529 /* A rule gets by default the precedence and associativity
530 of its last token. */
531 if (p->content.sym->class == token_sym && default_prec)
532 rules[ruleno].prec = p->content.sym;
533 }
534
535 /* If this rule has a %prec,
536 the specified symbol's precedence replaces the default. */
537 if (ruleprec)
538 {
539 rules[ruleno].precsym = ruleprec;
540 rules[ruleno].prec = ruleprec;
541 }
542 /* An item ends by the rule number (negated). */
543 ritem[itemno++] = rule_number_as_item_number (ruleno);
544 aver (itemno < ITEM_NUMBER_MAX);
545 ++ruleno;
546 aver (ruleno < RULE_NUMBER_MAX);
547
548 if (p)
549 p = p->next;
550 }
551
552 aver (itemno == nritems);
553
554 if (trace_flag & trace_sets)
555 ritem_print (stderr);
556 }
557 \f
558 /*------------------------------------------------------------------.
559 | Read in the grammar specification and record it in the format |
560 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
561 | in each case forming the body of a C function (YYACTION) which |
562 | contains a switch statement to decide which action to execute. |
563 `------------------------------------------------------------------*/
564
565 void
566 reader (void)
567 {
568 /* Initialize the symbol table. */
569 symbols_new ();
570
571 /* Construct the accept symbol. */
572 accept = symbol_get ("$accept", empty_location);
573 accept->class = nterm_sym;
574 accept->number = nvars++;
575
576 /* Construct the error token */
577 errtoken = symbol_get ("error", empty_location);
578 errtoken->class = token_sym;
579 errtoken->number = ntokens++;
580
581 /* Construct a token that represents all undefined literal tokens.
582 It is always token number 2. */
583 undeftoken = symbol_get ("$undefined", empty_location);
584 undeftoken->class = token_sym;
585 undeftoken->number = ntokens++;
586
587 gram_in = xfopen (grammar_file, "r");
588
589 gram__flex_debug = trace_flag & trace_scan;
590 gram_debug = trace_flag & trace_parse;
591 gram_scanner_initialize ();
592 gram_parse ();
593
594 /* Set front-end %define variable defaults. */
595 muscle_percent_define_default ("lr.keep-unreachable-states", "false");
596 {
597 char *lr_type;
598 /* IELR would be a better default, but LALR is historically the
599 default. */
600 muscle_percent_define_default ("lr.type", "LALR");
601 lr_type = muscle_percent_define_get ("lr.type");
602 if (0 != strcmp (lr_type, "canonical LR"))
603 muscle_percent_define_default ("lr.default-reductions", "all");
604 else
605 muscle_percent_define_default ("lr.default-reductions", "accepting");
606 free (lr_type);
607 }
608
609 /* Check front-end %define variables. */
610 {
611 static char const * const values[] = {
612 "lr.type", "LALR", "IELR", "canonical LR", NULL,
613 "lr.default-reductions", "all", "consistent", "accepting", NULL,
614 NULL
615 };
616 muscle_percent_define_check_values (values);
617 }
618
619 if (! complaint_issued)
620 check_and_convert_grammar ();
621
622 xfclose (gram_in);
623 }
624
625
626 /*-------------------------------------------------------------.
627 | Check the grammar that has just been read, and convert it to |
628 | internal form. |
629 `-------------------------------------------------------------*/
630
631 static void
632 check_and_convert_grammar (void)
633 {
634 /* Grammar has been read. Do some checking. */
635 if (nrules == 0)
636 fatal (_("no rules in the input grammar"));
637
638 /* Report any undefined symbols and consider them nonterminals. */
639 symbols_check_defined ();
640
641 /* If the user did not define her ENDTOKEN, do it now. */
642 if (!endtoken)
643 {
644 endtoken = symbol_get ("$end", empty_location);
645 endtoken->class = token_sym;
646 endtoken->number = 0;
647 /* Value specified by POSIX. */
648 endtoken->user_token_number = 0;
649 }
650
651 /* Find the start symbol if no %start. */
652 if (!start_flag)
653 {
654 symbol_list *node;
655 for (node = grammar;
656 node != NULL && symbol_is_dummy (node->content.sym);
657 node = node->next)
658 {
659 for (node = node->next;
660 node != NULL && node->content.sym != NULL;
661 node = node->next)
662 ;
663 }
664 aver (node != NULL);
665 grammar_start_symbol_set (node->content.sym,
666 node->content.sym->location);
667 }
668
669 /* Insert the initial rule, whose line is that of the first rule
670 (not that of the start symbol):
671
672 accept: %start EOF. */
673 {
674 symbol_list *p = symbol_list_sym_new (accept, empty_location);
675 p->location = grammar->location;
676 p->next = symbol_list_sym_new (startsymbol, empty_location);
677 p->next->next = symbol_list_sym_new (endtoken, empty_location);
678 p->next->next->next = symbol_list_sym_new (NULL, empty_location);
679 p->next->next->next->next = grammar;
680 nrules += 1;
681 nritems += 3;
682 grammar = p;
683 }
684
685 aver (nsyms <= SYMBOL_NUMBER_MAXIMUM && nsyms == ntokens + nvars);
686
687 /* Assign the symbols their symbol numbers. Write #defines for the
688 token symbols into FDEFINES if requested. */
689 symbols_pack ();
690
691 /* Scan rule actions after invoking symbol_check_alias_consistency (in
692 symbols_pack above) so that token types are set correctly before the rule
693 action type checking.
694
695 Before invoking grammar_rule_check (in packgram below) on any rule, make
696 sure all actions have already been scanned in order to set `used' flags.
697 Otherwise, checking that a midrule's $$ should be set will not always work
698 properly because the check must forward-reference the midrule's parent
699 rule. For the same reason, all the `used' flags must be set before
700 checking whether to remove `$' from any midrule symbol name (also in
701 packgram). */
702 {
703 symbol_list *sym;
704 for (sym = grammar; sym; sym = sym->next)
705 code_props_translate_code (&sym->action_props);
706 }
707
708 /* Convert the grammar into the format described in gram.h. */
709 packgram ();
710
711 /* The grammar as a symbol_list is no longer needed. */
712 symbol_list_free (grammar);
713 }