?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
validation: delete obsolete dev_hold() check
[smatch.git] / pre-process.c
blob9c925e7af809cfbad9b17bb412e0f9f68b967b44
1 /*
2 * Do C preprocessing, based on a token list gathered by
3 * the tokenizer.
4 *
5 * This may not be the smartest preprocessor on the planet.
6 *
7 * Copyright (C) 2003 Transmeta Corp.
8 * 2003-2004 Linus Torvalds
9 *
10 * Permission is hereby granted, free of charge, to any person obtaining a copy
11 * of this software and associated documentation files (the "Software"), to deal
12 * in the Software without restriction, including without limitation the rights
13 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
14 * copies of the Software, and to permit persons to whom the Software is
15 * furnished to do so, subject to the following conditions:
16 *
17 * The above copyright notice and this permission notice shall be included in
18 * all copies or substantial portions of the Software.
19 *
20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
23 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
24 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
25 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
26 * THE SOFTWARE.
27 */
28 #include <stdio.h>
29 #include <stdlib.h>
30 #include <stdarg.h>
31 #include <stddef.h>
32 #include <string.h>
33 #include <ctype.h>
34 #include <unistd.h>
35 #include <fcntl.h>
36 #include <limits.h>
37 #include <time.h>
38 #include <dirent.h>
39 #include <sys/stat.h>
40
41 #include "lib.h"
42 #include "allocate.h"
43 #include "parse.h"
44 #include "token.h"
45 #include "symbol.h"
46 #include "expression.h"
47 #include "scope.h"
48
49 static struct ident_list *macros; // only needed for -dD
50 static int false_nesting = 0;
51 static int counter_macro = 0; // __COUNTER__ expansion
52 static int include_level = 0;
53 static int expanding = 0;
54
55 #define INCLUDEPATHS 300
56 const char *includepath[INCLUDEPATHS+1] = {
57 "",
58 "/usr/include",
59 "/usr/local/include",
60 NULL
61 };
62
63 static const char **quote_includepath = includepath;
64 static const char **angle_includepath = includepath + 1;
65 static const char **isys_includepath = includepath + 1;
66 static const char **sys_includepath = includepath + 1;
67 static const char **dirafter_includepath = includepath + 3;
68
69 #define dirty_stream(stream) \
70 do { \
71 if (!stream->dirty) { \
72 stream->dirty = 1; \
73 if (!stream->ifndef) \
74 stream->protect = NULL; \
75 } \
76 } while(0)
77
78 #define end_group(stream) \
79 do { \
80 if (stream->ifndef == stream->top_if) { \
81 stream->ifndef = NULL; \
82 if (!stream->dirty) \
83 stream->protect = NULL; \
84 else if (stream->protect) \
85 stream->dirty = 0; \
86 } \
87 } while(0)
88
89 #define nesting_error(stream) \
90 do { \
91 stream->dirty = 1; \
92 stream->ifndef = NULL; \
93 stream->protect = NULL; \
94 } while(0)
95
96 static struct token *alloc_token(struct position *pos)
97 {
98 struct token *token = __alloc_token(0);
99
100 token->pos.stream = pos->stream;
101 token->pos.line = pos->line;
102 token->pos.pos = pos->pos;
103 token->pos.whitespace = 1;
104 return token;
105 }
106
107 /* Expand symbol 'sym' at '*list' */
108 static int expand(struct token **, struct symbol *);
109
110 static void replace_with_string(struct token *token, const char *str)
111 {
112 int size = strlen(str) + 1;
113 struct string *s = __alloc_string(size);
114
115 s->length = size;
116 memcpy(s->data, str, size);
117 token_type(token) = TOKEN_STRING;
118 token->string = s;
119 }
120
121 static void replace_with_integer(struct token *token, unsigned int val)
122 {
123 char *buf = __alloc_bytes(11);
124 sprintf(buf, "%u", val);
125 token_type(token) = TOKEN_NUMBER;
126 token->number = buf;
127 }
128
129 static struct symbol *lookup_macro(struct ident *ident)
130 {
131 struct symbol *sym = lookup_symbol(ident, NS_MACRO | NS_UNDEF);
132 if (sym && sym->namespace != NS_MACRO)
133 sym = NULL;
134 return sym;
135 }
136
137 static int token_defined(struct token *token)
138 {
139 if (token_type(token) == TOKEN_IDENT) {
140 struct symbol *sym = lookup_macro(token->ident);
141 if (sym) {
142 sym->used_in = file_scope;
143 return 1;
144 }
145 return 0;
146 }
147
148 sparse_error(token->pos, "expected preprocessor identifier");
149 return 0;
150 }
151
152 static void replace_with_bool(struct token *token, bool val)
153 {
154 static const char *string[] = { "0", "1" };
155
156 token_type(token) = TOKEN_NUMBER;
157 token->number = string[val];
158 }
159
160 static void replace_with_defined(struct token *token)
161 {
162 replace_with_bool(token, token_defined(token));
163 }
164
165 static void expand_line(struct token *token)
166 {
167 replace_with_integer(token, token->pos.line);
168 }
169
170 static void expand_file(struct token *token)
171 {
172 replace_with_string(token, stream_name(token->pos.stream));
173 }
174
175 static void expand_basefile(struct token *token)
176 {
177 replace_with_string(token, base_filename);
178 }
179
180 static time_t t = 0;
181 static void expand_date(struct token *token)
182 {
183 static char buffer[12]; /* __DATE__: 3 + ' ' + 2 + ' ' + 4 + '\0' */
184
185 if (!t)
186 time(&t);
187 strftime(buffer, 12, "%b %e %Y", localtime(&t));
188 replace_with_string(token, buffer);
189 }
190
191 static void expand_time(struct token *token)
192 {
193 static char buffer[9]; /* __TIME__: 2 + ':' + 2 + ':' + 2 + '\0' */
194
195 if (!t)
196 time(&t);
197 strftime(buffer, 9, "%T", localtime(&t));
198 replace_with_string(token, buffer);
199 }
200
201 static void expand_counter(struct token *token)
202 {
203 replace_with_integer(token, counter_macro++);
204 }
205
206 static void expand_include_level(struct token *token)
207 {
208 replace_with_integer(token, include_level - 1);
209 }
210
211 static int expand_one_symbol(struct token **list)
212 {
213 struct token *token = *list;
214 struct symbol *sym;
215
216 if (token->pos.noexpand)
217 return 1;
218
219 sym = lookup_macro(token->ident);
220 if (!sym)
221 return 1;
222 store_macro_pos(token);
223 if (sym->expand_simple) {
224 sym->expand_simple(token);
225 return 1;
226 } else {
227 int rc;
228
229 sym->used_in = file_scope;
230 expanding = 1;
231 rc = expand(list, sym);
232 expanding = 0;
233 return rc;
234 }
235 }
236
237 static inline struct token *scan_next(struct token **where)
238 {
239 struct token *token = *where;
240 if (token_type(token) != TOKEN_UNTAINT)
241 return token;
242 do {
243 token->ident->tainted = 0;
244 token = token->next;
245 } while (token_type(token) == TOKEN_UNTAINT);
246 *where = token;
247 return token;
248 }
249
250 static void expand_list(struct token **list)
251 {
252 struct token *next;
253 while (!eof_token(next = scan_next(list))) {
254 if (token_type(next) != TOKEN_IDENT || expand_one_symbol(list))
255 list = &next->next;
256 }
257 }
258
259 static void preprocessor_line(struct stream *stream, struct token **line);
260
261 static struct token *collect_arg(struct token *prev, int vararg, struct position *pos, int count)
262 {
263 struct stream *stream = input_streams + prev->pos.stream;
264 struct token **p = &prev->next;
265 struct token *next;
266 int nesting = 0;
267
268 while (!eof_token(next = scan_next(p))) {
269 if (next->pos.newline && match_op(next, '#')) {
270 if (!next->pos.noexpand) {
271 preprocessor_line(stream, p);
272 __free_token(next); /* Free the '#' token */
273 continue;
274 }
275 }
276 switch (token_type(next)) {
277 case TOKEN_STREAMEND:
278 case TOKEN_STREAMBEGIN:
279 *p = &eof_token_entry;
280 return next;
281 case TOKEN_STRING:
282 case TOKEN_WIDE_STRING:
283 if (count > 1)
284 next->string->immutable = 1;
285 break;
286 }
287 if (false_nesting) {
288 *p = next->next;
289 __free_token(next);
290 continue;
291 }
292 if (match_op(next, '(')) {
293 nesting++;
294 } else if (match_op(next, ')')) {
295 if (!nesting--)
296 break;
297 } else if (match_op(next, ',') && !nesting && !vararg) {
298 break;
299 }
300 next->pos.stream = pos->stream;
301 next->pos.line = pos->line;
302 next->pos.pos = pos->pos;
303 next->pos.newline = 0;
304 p = &next->next;
305 }
306 *p = &eof_token_entry;
307 return next;
308 }
309
310 /*
311 * We store arglist as <counter> [arg1] <number of uses for arg1> ... eof
312 */
313
314 struct arg {
315 struct token *arg;
316 struct token *expanded;
317 struct token *str;
318 int n_normal;
319 int n_quoted;
320 int n_str;
321 };
322
323 static int collect_arguments(struct token *start, struct token *arglist, struct arg *args, struct token *what)
324 {
325 int wanted = arglist->count.normal;
326 struct token *next = NULL;
327 int count = 0;
328
329 arglist = arglist->next; /* skip counter */
330
331 if (!wanted) {
332 next = collect_arg(start, 0, &what->pos, 0);
333 if (eof_token(next))
334 goto Eclosing;
335 if (!eof_token(start->next) || !match_op(next, ')')) {
336 count++;
337 goto Emany;
338 }
339 } else {
340 for (count = 0; count < wanted; count++) {
341 struct argcount *p = &arglist->next->count;
342 next = collect_arg(start, p->vararg, &what->pos, p->normal);
343 if (eof_token(next))
344 goto Eclosing;
345 if (p->vararg && wanted == 1 && eof_token(start->next))
346 break;
347 arglist = arglist->next->next;
348 args[count].arg = start->next;
349 args[count].n_normal = p->normal;
350 args[count].n_quoted = p->quoted;
351 args[count].n_str = p->str;
352 if (match_op(next, ')')) {
353 count++;
354 break;
355 }
356 start = next;
357 }
358 if (count == wanted && !match_op(next, ')'))
359 goto Emany;
360 if (count == wanted - 1) {
361 struct argcount *p = &arglist->next->count;
362 if (!p->vararg)
363 goto Efew;
364 args[count].arg = NULL;
365 args[count].n_normal = p->normal;
366 args[count].n_quoted = p->quoted;
367 args[count].n_str = p->str;
368 }
369 if (count < wanted - 1)
370 goto Efew;
371 }
372 what->next = next->next;
373 return 1;
374
375 Efew:
376 sparse_error(what->pos, "macro \"%s\" requires %d arguments, but only %d given",
377 show_token(what), wanted, count);
378 goto out;
379 Emany:
380 while (match_op(next, ',')) {
381 next = collect_arg(next, 0, &what->pos, 0);
382 count++;
383 }
384 if (eof_token(next))
385 goto Eclosing;
386 sparse_error(what->pos, "macro \"%s\" passed %d arguments, but takes just %d",
387 show_token(what), count, wanted);
388 goto out;
389 Eclosing:
390 sparse_error(what->pos, "unterminated argument list invoking macro \"%s\"",
391 show_token(what));
392 out:
393 what->next = next->next;
394 return 0;
395 }
396
397 static struct token *dup_list(struct token *list)
398 {
399 struct token *res = NULL;
400 struct token **p = &res;
401
402 while (!eof_token(list)) {
403 struct token *newtok = __alloc_token(0);
404 *newtok = *list;
405 *p = newtok;
406 p = &newtok->next;
407 list = list->next;
408 }
409 return res;
410 }
411
412 static const char *show_token_sequence(struct token *token, int quote)
413 {
414 static char buffer[MAX_STRING];
415 char *ptr = buffer;
416 int whitespace = 0;
417
418 if (!token && !quote)
419 return "<none>";
420 while (!eof_token(token)) {
421 const char *val = quote ? quote_token(token) : show_token(token);
422 int len = strlen(val);
423
424 if (ptr + whitespace + len >= buffer + sizeof(buffer)) {
425 sparse_error(token->pos, "too long token expansion");
426 break;
427 }
428
429 if (whitespace)
430 *ptr++ = ' ';
431 memcpy(ptr, val, len);
432 ptr += len;
433 token = token->next;
434 whitespace = token->pos.whitespace;
435 }
436 *ptr = 0;
437 return buffer;
438 }
439
440 static struct token *stringify(struct token *arg)
441 {
442 const char *s = show_token_sequence(arg, 1);
443 int size = strlen(s)+1;
444 struct token *token = __alloc_token(0);
445 struct string *string = __alloc_string(size);
446
447 memcpy(string->data, s, size);
448 string->length = size;
449 token->pos = arg->pos;
450 token_type(token) = TOKEN_STRING;
451 token->string = string;
452 token->next = &eof_token_entry;
453 return token;
454 }
455
456 static void expand_arguments(int count, struct arg *args)
457 {
458 int i;
459 for (i = 0; i < count; i++) {
460 struct token *arg = args[i].arg;
461 if (!arg)
462 arg = &eof_token_entry;
463 if (args[i].n_str)
464 args[i].str = stringify(arg);
465 if (args[i].n_normal) {
466 if (!args[i].n_quoted) {
467 args[i].expanded = arg;
468 args[i].arg = NULL;
469 } else if (eof_token(arg)) {
470 args[i].expanded = arg;
471 } else {
472 args[i].expanded = dup_list(arg);
473 }
474 expand_list(&args[i].expanded);
475 }
476 }
477 }
478
479 /*
480 * Possibly valid combinations:
481 * - ident + ident -> ident
482 * - ident + number -> ident unless number contains '.', '+' or '-'.
483 * - 'L' + char constant -> wide char constant
484 * - 'L' + string literal -> wide string literal
485 * - number + number -> number
486 * - number + ident -> number
487 * - number + '.' -> number
488 * - number + '+' or '-' -> number, if number used to end on [eEpP].
489 * - '.' + number -> number, if number used to start with a digit.
490 * - special + special -> either special or an error.
491 */
492 static enum token_type combine(struct token *left, struct token *right, char *p)
493 {
494 int len;
495 enum token_type t1 = token_type(left), t2 = token_type(right);
496
497 if (t1 != TOKEN_IDENT && t1 != TOKEN_NUMBER && t1 != TOKEN_SPECIAL)
498 return TOKEN_ERROR;
499
500 if (t1 == TOKEN_IDENT && left->ident == &L_ident) {
501 if (t2 >= TOKEN_CHAR && t2 < TOKEN_WIDE_CHAR)
502 return t2 + TOKEN_WIDE_CHAR - TOKEN_CHAR;
503 if (t2 == TOKEN_STRING)
504 return TOKEN_WIDE_STRING;
505 }
506
507 if (t2 != TOKEN_IDENT && t2 != TOKEN_NUMBER && t2 != TOKEN_SPECIAL)
508 return TOKEN_ERROR;
509
510 strcpy(p, show_token(left));
511 strcat(p, show_token(right));
512 len = strlen(p);
513
514 if (len >= 256)
515 return TOKEN_ERROR;
516
517 if (t1 == TOKEN_IDENT) {
518 if (t2 == TOKEN_SPECIAL)
519 return TOKEN_ERROR;
520 if (t2 == TOKEN_NUMBER && strpbrk(p, "+-."))
521 return TOKEN_ERROR;
522 return TOKEN_IDENT;
523 }
524
525 if (t1 == TOKEN_NUMBER) {
526 if (t2 == TOKEN_SPECIAL) {
527 switch (right->special) {
528 case '.':
529 break;
530 case '+': case '-':
531 if (strchr("eEpP", p[len - 2]))
532 break;
533 default:
534 return TOKEN_ERROR;
535 }
536 }
537 return TOKEN_NUMBER;
538 }
539
540 if (p[0] == '.' && isdigit((unsigned char)p[1]))
541 return TOKEN_NUMBER;
542
543 return TOKEN_SPECIAL;
544 }
545
546 static int merge(struct token *left, struct token *right)
547 {
548 static char buffer[512];
549 enum token_type res = combine(left, right, buffer);
550 int n;
551
552 switch (res) {
553 case TOKEN_IDENT:
554 left->ident = built_in_ident(buffer);
555 left->pos.noexpand = 0;
556 return 1;
557
558 case TOKEN_NUMBER:
559 token_type(left) = TOKEN_NUMBER; /* could be . + num */
560 left->number = xstrdup(buffer);
561 return 1;
562
563 case TOKEN_SPECIAL:
564 if (buffer[2] && buffer[3])
565 break;
566 for (n = SPECIAL_BASE; n < SPECIAL_ARG_SEPARATOR; n++) {
567 if (!memcmp(buffer, combinations[n-SPECIAL_BASE], 3)) {
568 left->special = n;
569 return 1;
570 }
571 }
572 break;
573
574 case TOKEN_WIDE_CHAR:
575 case TOKEN_WIDE_STRING:
576 token_type(left) = res;
577 left->pos.noexpand = 0;
578 left->string = right->string;
579 return 1;
580
581 case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
582 token_type(left) = res;
583 left->pos.noexpand = 0;
584 memcpy(left->embedded, right->embedded, 4);
585 return 1;
586
587 default:
588 ;
589 }
590 sparse_error(left->pos, "'##' failed: concatenation is not a valid token");
591 return 0;
592 }
593
594 static struct token *dup_token(struct token *token, struct position *streampos)
595 {
596 struct token *alloc = alloc_token(streampos);
597 token_type(alloc) = token_type(token);
598 alloc->pos.newline = token->pos.newline;
599 alloc->pos.whitespace = token->pos.whitespace;
600 alloc->number = token->number;
601 alloc->pos.noexpand = token->pos.noexpand;
602 return alloc;
603 }
604
605 static struct token **copy(struct token **where, struct token *list, int *count)
606 {
607 int need_copy = --*count;
608 while (!eof_token(list)) {
609 struct token *token;
610 if (need_copy)
611 token = dup_token(list, &list->pos);
612 else
613 token = list;
614 if (token_type(token) == TOKEN_IDENT && token->ident->tainted)
615 token->pos.noexpand = 1;
616 *where = token;
617 where = &token->next;
618 list = list->next;
619 }
620 *where = &eof_token_entry;
621 return where;
622 }
623
624 static int handle_kludge(struct token **p, struct arg *args)
625 {
626 struct token *t = (*p)->next->next;
627 while (1) {
628 struct arg *v = &args[t->argnum];
629 if (token_type(t->next) != TOKEN_CONCAT) {
630 if (v->arg) {
631 /* ignore the first ## */
632 *p = (*p)->next;
633 return 0;
634 }
635 /* skip the entire thing */
636 *p = t;
637 return 1;
638 }
639 if (v->arg && !eof_token(v->arg))
640 return 0; /* no magic */
641 t = t->next->next;
642 }
643 }
644
645 static struct token **substitute(struct token **list, struct token *body, struct arg *args)
646 {
647 struct position *base_pos = &(*list)->pos;
648 int *count;
649 enum {Normal, Placeholder, Concat} state = Normal;
650
651 for (; !eof_token(body); body = body->next) {
652 struct token *added, *arg;
653 struct token **tail;
654 struct token *t;
655
656 switch (token_type(body)) {
657 case TOKEN_GNU_KLUDGE:
658 /*
659 * GNU kludge: if we had <comma>##<vararg>, behaviour
660 * depends on whether we had enough arguments to have
661 * a vararg. If we did, ## is just ignored. Otherwise
662 * both , and ## are ignored. Worse, there can be
663 * an arbitrary number of ##<arg> in between; if all of
664 * those are empty, we act as if they hadn't been there,
665 * otherwise we act as if the kludge didn't exist.
666 */
667 t = body;
668 if (handle_kludge(&body, args)) {
669 if (state == Concat)
670 state = Normal;
671 else
672 state = Placeholder;
673 continue;
674 }
675 added = dup_token(t, base_pos);
676 token_type(added) = TOKEN_SPECIAL;
677 tail = &added->next;
678 break;
679
680 case TOKEN_STR_ARGUMENT:
681 arg = args[body->argnum].str;
682 count = &args[body->argnum].n_str;
683 goto copy_arg;
684
685 case TOKEN_QUOTED_ARGUMENT:
686 arg = args[body->argnum].arg;
687 count = &args[body->argnum].n_quoted;
688 if (!arg || eof_token(arg)) {
689 if (state == Concat)
690 state = Normal;
691 else
692 state = Placeholder;
693 continue;
694 }
695 goto copy_arg;
696
697 case TOKEN_MACRO_ARGUMENT:
698 arg = args[body->argnum].expanded;
699 count = &args[body->argnum].n_normal;
700 if (eof_token(arg)) {
701 state = Normal;
702 continue;
703 }
704 copy_arg:
705 tail = copy(&added, arg, count);
706 added->pos.newline = body->pos.newline;
707 added->pos.whitespace = body->pos.whitespace;
708 break;
709
710 case TOKEN_CONCAT:
711 if (state == Placeholder)
712 state = Normal;
713 else
714 state = Concat;
715 continue;
716
717 case TOKEN_IDENT:
718 added = dup_token(body, base_pos);
719 if (added->ident->tainted)
720 added->pos.noexpand = 1;
721 tail = &added->next;
722 break;
723
724 default:
725 added = dup_token(body, base_pos);
726 tail = &added->next;
727 break;
728 }
729
730 /*
731 * if we got to doing real concatenation, we already have
732 * added something into the list, so containing_token() is OK.
733 */
734 if (state == Concat && merge(containing_token(list), added)) {
735 *list = added->next;
736 if (tail != &added->next)
737 list = tail;
738 } else {
739 *list = added;
740 list = tail;
741 }
742 state = Normal;
743 }
744 *list = &eof_token_entry;
745 return list;
746 }
747
748 static int expand(struct token **list, struct symbol *sym)
749 {
750 struct token *last;
751 struct token *token = *list;
752 struct ident *expanding = token->ident;
753 struct token **tail;
754 struct token *expansion = sym->expansion;
755 int nargs = sym->arglist ? sym->arglist->count.normal : 0;
756 struct arg args[nargs];
757
758 if (expanding->tainted) {
759 token->pos.noexpand = 1;
760 return 1;
761 }
762
763 if (sym->arglist) {
764 if (!match_op(scan_next(&token->next), '('))
765 return 1;
766 if (!collect_arguments(token->next, sym->arglist, args, token))
767 return 1;
768 expand_arguments(nargs, args);
769 }
770
771 if (sym->expand)
772 return sym->expand(token, args) ? 0 : 1;
773
774 expanding->tainted = 1;
775
776 last = token->next;
777 tail = substitute(list, expansion, args);
778 /*
779 * Note that it won't be eof - at least TOKEN_UNTAINT will be there.
780 * We still can lose the newline flag if the sucker expands to nothing,
781 * but the price of dealing with that is probably too high (we'd need
782 * to collect the flags during scan_next())
783 */
784 (*list)->pos.newline = token->pos.newline;
785 (*list)->pos.whitespace = token->pos.whitespace;
786 *tail = last;
787
788 return 0;
789 }
790
791 static const char *token_name_sequence(struct token *token, int endop, struct token *start)
792 {
793 static char buffer[256];
794 char *ptr = buffer;
795
796 while (!eof_token(token) && !match_op(token, endop)) {
797 int len;
798 const char *val = token->string->data;
799 if (token_type(token) != TOKEN_STRING)
800 val = show_token(token);
801 len = strlen(val);
802 memcpy(ptr, val, len);
803 ptr += len;
804 token = token->next;
805 }
806 *ptr = 0;
807 if (endop && !match_op(token, endop))
808 sparse_error(start->pos, "expected '>' at end of filename");
809 return buffer;
810 }
811
812 static int already_tokenized(const char *path)
813 {
814 int stream, next;
815
816 for (stream = *hash_stream(path); stream >= 0 ; stream = next) {
817 struct stream *s = input_streams + stream;
818
819 next = s->next_stream;
820 if (s->once) {
821 if (strcmp(path, s->name))
822 continue;
823 return 1;
824 }
825 if (s->constant != CONSTANT_FILE_YES)
826 continue;
827 if (strcmp(path, s->name))
828 continue;
829 if (s->protect && !lookup_macro(s->protect))
830 continue;
831 return 1;
832 }
833 return 0;
834 }
835
836 /* Handle include of header files.
837 * The relevant options are made compatible with gcc. The only options that
838 * are not supported is -withprefix and friends.
839 *
840 * Three set of include paths are known:
841 * quote_includepath: Path to search when using #include "file.h"
842 * angle_includepath: Paths to search when using #include <file.h>
843 * isys_includepath: Paths specified with -isystem, come before the
844 * built-in system include paths. Gcc would suppress
845 * warnings from system headers. Here we separate
846 * them from the angle_ ones to keep search ordering.
847 *
848 * sys_includepath: Built-in include paths.
849 * dirafter_includepath Paths added with -dirafter.
850 *
851 * The above is implemented as one array with pointers
852 * +--------------+
853 * quote_includepath ---> | |
854 * +--------------+
855 * | |
856 * +--------------+
857 * angle_includepath ---> | |
858 * +--------------+
859 * isys_includepath ---> | |
860 * +--------------+
861 * sys_includepath ---> | |
862 * +--------------+
863 * dirafter_includepath -> | |
864 * +--------------+
865 *
866 * -I dir insert dir just before isys_includepath and move the rest
867 * -I- makes all dirs specified with -I before to quote dirs only and
868 * angle_includepath is set equal to isys_includepath.
869 * -nostdinc removes all sys dirs by storing NULL in entry pointed
870 * to by * sys_includepath. Note that this will reset all dirs built-in
871 * and added before -nostdinc by -isystem and -idirafter.
872 * -isystem dir adds dir where isys_includepath points adding this dir as
873 * first systemdir
874 * -idirafter dir adds dir to the end of the list
875 */
876
877 static void set_stream_include_path(struct stream *stream)
878 {
879 const char *path = stream->path;
880 if (!path) {
881 const char *p = strrchr(stream->name, '/');
882 path = "";
883 if (p) {
884 int len = p - stream->name + 1;
885 char *m = malloc(len+1);
886 /* This includes the final "/" */
887 memcpy(m, stream->name, len);
888 m[len] = 0;
889 path = m;
890 }
891 stream->path = path;
892 }
893 includepath[0] = path;
894 }
895
896 static int try_include(struct position pos, const char *path, const char *filename, int flen, struct token **where, const char **next_path)
897 {
898 int fd;
899 int plen = strlen(path);
900 static char fullname[PATH_MAX];
901
902 memcpy(fullname, path, plen);
903 if (plen && path[plen-1] != '/') {
904 fullname[plen] = '/';
905 plen++;
906 }
907 memcpy(fullname+plen, filename, flen);
908 if (already_tokenized(fullname))
909 return 1;
910 fd = open(fullname, O_RDONLY);
911 if (fd >= 0) {
912 char *streamname = xmemdup(fullname, plen + flen);
913 *where = tokenize(&pos, streamname, fd, *where, next_path);
914 close(fd);
915 return 1;
916 }
917 return 0;
918 }
919
920 static int do_include_path(const char **pptr, struct token **list, struct token *token, const char *filename, int flen)
921 {
922 const char *path;
923
924 while ((path = *pptr++) != NULL) {
925 if (!try_include(token->pos, path, filename, flen, list, pptr))
926 continue;
927 return 1;
928 }
929 return 0;
930 }
931
932 static int free_preprocessor_line(struct token *token)
933 {
934 while (token_type(token) != TOKEN_EOF) {
935 struct token *free = token;
936 token = token->next;
937 __free_token(free);
938 };
939 return 1;
940 }
941
942 const char *find_include(const char *skip, const char *look_for)
943 {
944 DIR *dp;
945 struct dirent *entry;
946 struct stat statbuf;
947 const char *ret;
948 char cwd[PATH_MAX];
949 static char buf[PATH_MAX + 1];
950
951 dp = opendir(".");
952 if (!dp)
953 return NULL;
954
955 if (!getcwd(cwd, sizeof(cwd)))
956 goto close;
957
958 while ((entry = readdir(dp))) {
959 lstat(entry->d_name, &statbuf);
960
961 if (strcmp(entry->d_name, look_for) == 0) {
962 snprintf(buf, sizeof(buf), "%s/%s", cwd, entry->d_name);
963 closedir(dp);
964 return buf;
965 }
966
967 if (S_ISDIR(statbuf.st_mode)) {
968 /* Found a directory, but ignore . and .. */
969 if (strcmp(".", entry->d_name) == 0 ||
970 strcmp("..", entry->d_name) == 0 ||
971 strcmp(skip, entry->d_name) == 0)
972 continue;
973
974 chdir(entry->d_name);
975 ret = find_include("", look_for);
976 chdir("..");
977 if (ret) {
978 closedir(dp);
979 return ret;
980 }
981 }
982 }
983 close:
984 closedir(dp);
985
986 return NULL;
987 }
988
989 const char *search_dir(const char *stop, const char *look_for)
990 {
991 char cwd[PATH_MAX];
992 int len;
993 const char *ret;
994 int cnt = 0;
995
996 if (!getcwd(cwd, sizeof(cwd)))
997 return NULL;
998
999 len = strlen(cwd);
1000 while (len >= 0) {
1001 ret = find_include(cnt++ ? cwd + len + 1 : "", look_for);
1002 if (ret)
1003 return ret;
1004
1005 if (strcmp(cwd, stop) == 0 ||
1006 strcmp(cwd, "/usr/include") == 0 ||
1007 strcmp(cwd, "/usr/local/include") == 0 ||
1008 strlen(cwd) <= 10 || /* heck... don't search /usr/lib/ */
1009 strcmp(cwd, "/") == 0)
1010 return NULL;
1011
1012 while (--len >= 0) {
1013 if (cwd[len] == '/') {
1014 cwd[len] = '\0';
1015 break;
1016 }
1017 }
1018
1019 chdir("..");
1020 }
1021 return NULL;
1022 }
1023
1024 static void use_best_guess_header_file(struct token *token, const char *filename, struct token **list)
1025 {
1026 char cwd[PATH_MAX];
1027 char dir_part[PATH_MAX];
1028 const char *file_part;
1029 const char *include_name;
1030 static int cnt;
1031 int len;
1032
1033 /* Avoid guessing includes recursively. */
1034 if (cnt++ > 1000)
1035 return;
1036
1037 if (!filename || filename[0] == '\0')
1038 return;
1039
1040 file_part = filename;
1041 while ((filename = strchr(filename, '/'))) {
1042 ++filename;
1043 if (filename[0])
1044 file_part = filename;
1045 }
1046
1047 snprintf(dir_part, sizeof(dir_part), "%s", stream_name(token->pos.stream));
1048 len = strlen(dir_part);
1049 while (--len >= 0) {
1050 if (dir_part[len] == '/') {
1051 dir_part[len] = '\0';
1052 break;
1053 }
1054 }
1055 if (len < 0)
1056 sprintf(dir_part, ".");
1057
1058 if (!getcwd(cwd, sizeof(cwd)))
1059 return;
1060
1061 chdir(dir_part);
1062 include_name = search_dir(cwd, file_part);
1063 chdir(cwd);
1064 if (!include_name)
1065 return;
1066 sparse_error(token->pos, "using '%s'", include_name);
1067
1068 try_include(token->pos, "", include_name, strlen(include_name), list, includepath);
1069 }
1070
1071 static int handle_include_path(struct stream *stream, struct token **list, struct token *token, int how)
1072 {
1073 const char *filename;
1074 struct token *next;
1075 const char **path;
1076 int expect;
1077 int flen;
1078
1079 next = token->next;
1080 expect = '>';
1081 if (!match_op(next, '<')) {
1082 expand_list(&token->next);
1083 expect = 0;
1084 next = token;
1085 if (match_op(token->next, '<')) {
1086 next = token->next;
1087 expect = '>';
1088 }
1089 }
1090
1091 token = next->next;
1092 filename = token_name_sequence(token, expect, token);
1093 flen = strlen(filename) + 1;
1094
1095 /* Absolute path? */
1096 if (filename[0] == '/') {
1097 if (try_include(token->pos, "", filename, flen, list, includepath))
1098 return 0;
1099 goto out;
1100 }
1101
1102 switch (how) {
1103 case 1:
1104 path = stream->next_path;
1105 break;
1106 case 2:
1107 includepath[0] = "";
1108 path = includepath;
1109 break;
1110 default:
1111 /* Dir of input file is first dir to search for quoted includes */
1112 set_stream_include_path(stream);
1113 path = expect ? angle_includepath : quote_includepath;
1114 break;
1115 }
1116 /* Check the standard include paths.. */
1117 if (do_include_path(path, list, token, filename, flen))
1118 return 0;
1119 out:
1120 sparse_error(token->pos, "unable to open '%s'", filename);
1121 use_best_guess_header_file(token, filename, list);
1122 return 0;
1123 }
1124
1125 static int handle_include(struct stream *stream, struct token **list, struct token *token)
1126 {
1127 return handle_include_path(stream, list, token, 0);
1128 }
1129
1130 static int handle_include_next(struct stream *stream, struct token **list, struct token *token)
1131 {
1132 return handle_include_path(stream, list, token, 1);
1133 }
1134
1135 static int handle_argv_include(struct stream *stream, struct token **list, struct token *token)
1136 {
1137 return handle_include_path(stream, list, token, 2);
1138 }
1139
1140 static int token_different(struct token *t1, struct token *t2)
1141 {
1142 int different;
1143
1144 if (token_type(t1) != token_type(t2))
1145 return 1;
1146
1147 switch (token_type(t1)) {
1148 case TOKEN_IDENT:
1149 different = t1->ident != t2->ident;
1150 break;
1151 case TOKEN_ARG_COUNT:
1152 case TOKEN_UNTAINT:
1153 case TOKEN_CONCAT:
1154 case TOKEN_GNU_KLUDGE:
1155 different = 0;
1156 break;
1157 case TOKEN_NUMBER:
1158 different = strcmp(t1->number, t2->number);
1159 break;
1160 case TOKEN_SPECIAL:
1161 different = t1->special != t2->special;
1162 break;
1163 case TOKEN_MACRO_ARGUMENT:
1164 case TOKEN_QUOTED_ARGUMENT:
1165 case TOKEN_STR_ARGUMENT:
1166 different = t1->argnum != t2->argnum;
1167 break;
1168 case TOKEN_CHAR_EMBEDDED_0 ... TOKEN_CHAR_EMBEDDED_3:
1169 case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
1170 different = memcmp(t1->embedded, t2->embedded, 4);
1171 break;
1172 case TOKEN_CHAR:
1173 case TOKEN_WIDE_CHAR:
1174 case TOKEN_STRING:
1175 case TOKEN_WIDE_STRING: {
1176 struct string *s1, *s2;
1177
1178 s1 = t1->string;
1179 s2 = t2->string;
1180 different = 1;
1181 if (s1->length != s2->length)
1182 break;
1183 different = memcmp(s1->data, s2->data, s1->length);
1184 break;
1185 }
1186 default:
1187 different = 1;
1188 break;
1189 }
1190 return different;
1191 }
1192
1193 static int token_list_different(struct token *list1, struct token *list2)
1194 {
1195 for (;;) {
1196 if (list1 == list2)
1197 return 0;
1198 if (!list1 || !list2)
1199 return 1;
1200 if (token_different(list1, list2))
1201 return 1;
1202 list1 = list1->next;
1203 list2 = list2->next;
1204 }
1205 }
1206
1207 static inline void set_arg_count(struct token *token)
1208 {
1209 token_type(token) = TOKEN_ARG_COUNT;
1210 token->count.normal = token->count.quoted =
1211 token->count.str = token->count.vararg = 0;
1212 }
1213
1214 static struct token *parse_arguments(struct token *list)
1215 {
1216 struct token *arg = list->next, *next = list;
1217 struct argcount *count = &list->count;
1218
1219 set_arg_count(list);
1220
1221 if (match_op(arg, ')')) {
1222 next = arg->next;
1223 list->next = &eof_token_entry;
1224 return next;
1225 }
1226
1227 while (token_type(arg) == TOKEN_IDENT) {
1228 if (arg->ident == &__VA_ARGS___ident)
1229 goto Eva_args;
1230 if (!++count->normal)
1231 goto Eargs;
1232 next = arg->next;
1233
1234 if (match_op(next, ',')) {
1235 set_arg_count(next);
1236 arg = next->next;
1237 continue;
1238 }
1239
1240 if (match_op(next, ')')) {
1241 set_arg_count(next);
1242 next = next->next;
1243 arg->next->next = &eof_token_entry;
1244 return next;
1245 }
1246
1247 /* normal cases are finished here */
1248
1249 if (match_op(next, SPECIAL_ELLIPSIS)) {
1250 if (match_op(next->next, ')')) {
1251 set_arg_count(next);
1252 next->count.vararg = 1;
1253 next = next->next;
1254 arg->next->next = &eof_token_entry;
1255 return next->next;
1256 }
1257
1258 arg = next;
1259 goto Enotclosed;
1260 }
1261
1262 if (eof_token(next)) {
1263 goto Enotclosed;
1264 } else {
1265 arg = next;
1266 goto Ebadstuff;
1267 }
1268 }
1269
1270 if (match_op(arg, SPECIAL_ELLIPSIS)) {
1271 next = arg->next;
1272 token_type(arg) = TOKEN_IDENT;
1273 arg->ident = &__VA_ARGS___ident;
1274 if (!match_op(next, ')'))
1275 goto Enotclosed;
1276 if (!++count->normal)
1277 goto Eargs;
1278 set_arg_count(next);
1279 next->count.vararg = 1;
1280 next = next->next;
1281 arg->next->next = &eof_token_entry;
1282 return next;
1283 }
1284
1285 if (eof_token(arg)) {
1286 arg = next;
1287 goto Enotclosed;
1288 }
1289 if (match_op(arg, ','))
1290 goto Emissing;
1291 else
1292 goto Ebadstuff;
1293
1294
1295 Emissing:
1296 sparse_error(arg->pos, "parameter name missing");
1297 return NULL;
1298 Ebadstuff:
1299 sparse_error(arg->pos, "\"%s\" may not appear in macro parameter list",
1300 show_token(arg));
1301 return NULL;
1302 Enotclosed:
1303 sparse_error(arg->pos, "missing ')' in macro parameter list");
1304 return NULL;
1305 Eva_args:
1306 sparse_error(arg->pos, "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
1307 return NULL;
1308 Eargs:
1309 sparse_error(arg->pos, "too many arguments in macro definition");
1310 return NULL;
1311 }
1312
1313 static int try_arg(struct token *token, enum token_type type, struct token *arglist)
1314 {
1315 struct ident *ident = token->ident;
1316 int nr;
1317
1318 if (!arglist || token_type(token) != TOKEN_IDENT)
1319 return 0;
1320
1321 arglist = arglist->next;
1322
1323 for (nr = 0; !eof_token(arglist); nr++, arglist = arglist->next->next) {
1324 if (arglist->ident == ident) {
1325 struct argcount *count = &arglist->next->count;
1326 int n;
1327
1328 token->argnum = nr;
1329 token_type(token) = type;
1330 switch (type) {
1331 case TOKEN_MACRO_ARGUMENT:
1332 n = ++count->normal;
1333 break;
1334 case TOKEN_QUOTED_ARGUMENT:
1335 n = ++count->quoted;
1336 break;
1337 default:
1338 n = ++count->str;
1339 }
1340 if (n)
1341 return count->vararg ? 2 : 1;
1342 /*
1343 * XXX - need saner handling of that
1344 * (>= 1024 instances of argument)
1345 */
1346 token_type(token) = TOKEN_ERROR;
1347 return -1;
1348 }
1349 }
1350 return 0;
1351 }
1352
1353 static struct token *handle_hash(struct token **p, struct token *arglist)
1354 {
1355 struct token *token = *p;
1356 if (arglist) {
1357 struct token *next = token->next;
1358 if (!try_arg(next, TOKEN_STR_ARGUMENT, arglist))
1359 goto Equote;
1360 next->pos.whitespace = token->pos.whitespace;
1361 __free_token(token);
1362 token = *p = next;
1363 } else {
1364 token->pos.noexpand = 1;
1365 }
1366 return token;
1367
1368 Equote:
1369 sparse_error(token->pos, "'#' is not followed by a macro parameter");
1370 return NULL;
1371 }
1372
1373 /* token->next is ## */
1374 static struct token *handle_hashhash(struct token *token, struct token *arglist)
1375 {
1376 struct token *last = token;
1377 struct token *concat;
1378 int state = match_op(token, ',');
1379
1380 try_arg(token, TOKEN_QUOTED_ARGUMENT, arglist);
1381
1382 while (1) {
1383 struct token *t;
1384 int is_arg;
1385
1386 /* eat duplicate ## */
1387 concat = token->next;
1388 while (match_op(t = concat->next, SPECIAL_HASHHASH)) {
1389 token->next = t;
1390 __free_token(concat);
1391 concat = t;
1392 }
1393 token_type(concat) = TOKEN_CONCAT;
1394
1395 if (eof_token(t))
1396 goto Econcat;
1397
1398 if (match_op(t, '#')) {
1399 t = handle_hash(&concat->next, arglist);
1400 if (!t)
1401 return NULL;
1402 }
1403
1404 is_arg = try_arg(t, TOKEN_QUOTED_ARGUMENT, arglist);
1405
1406 if (state == 1 && is_arg) {
1407 state = is_arg;
1408 } else {
1409 last = t;
1410 state = match_op(t, ',');
1411 }
1412
1413 token = t;
1414 if (!match_op(token->next, SPECIAL_HASHHASH))
1415 break;
1416 }
1417 /* handle GNU ,##__VA_ARGS__ kludge, in all its weirdness */
1418 if (state == 2)
1419 token_type(last) = TOKEN_GNU_KLUDGE;
1420 return token;
1421
1422 Econcat:
1423 sparse_error(concat->pos, "'##' cannot appear at the ends of macro expansion");
1424 return NULL;
1425 }
1426
1427 static struct token *parse_expansion(struct token *expansion, struct token *arglist, struct ident *name)
1428 {
1429 struct token *token = expansion;
1430 struct token **p;
1431
1432 if (match_op(token, SPECIAL_HASHHASH))
1433 goto Econcat;
1434
1435 for (p = &expansion; !eof_token(token); p = &token->next, token = *p) {
1436 if (match_op(token, '#')) {
1437 token = handle_hash(p, arglist);
1438 if (!token)
1439 return NULL;
1440 }
1441 if (match_op(token->next, SPECIAL_HASHHASH)) {
1442 token = handle_hashhash(token, arglist);
1443 if (!token)
1444 return NULL;
1445 } else {
1446 try_arg(token, TOKEN_MACRO_ARGUMENT, arglist);
1447 }
1448 switch (token_type(token)) {
1449 case TOKEN_ERROR:
1450 goto Earg;
1451
1452 case TOKEN_STRING:
1453 case TOKEN_WIDE_STRING:
1454 token->string->immutable = 1;
1455 break;
1456 }
1457 }
1458 token = alloc_token(&expansion->pos);
1459 token_type(token) = TOKEN_UNTAINT;
1460 token->ident = name;
1461 token->next = *p;
1462 *p = token;
1463 return expansion;
1464
1465 Econcat:
1466 sparse_error(token->pos, "'##' cannot appear at the ends of macro expansion");
1467 return NULL;
1468 Earg:
1469 sparse_error(token->pos, "too many instances of argument in body");
1470 return NULL;
1471 }
1472
1473 static int do_define(struct position pos, struct token *token, struct ident *name,
1474 struct token *arglist, struct token *expansion, int attr)
1475 {
1476 struct symbol *sym;
1477 int ret = 1;
1478
1479 expansion = parse_expansion(expansion, arglist, name);
1480 if (!expansion)
1481 return 1;
1482
1483 sym = lookup_symbol(name, NS_MACRO | NS_UNDEF);
1484 if (sym) {
1485 int clean;
1486
1487 if (attr < sym->attr)
1488 goto out;
1489
1490 clean = (attr == sym->attr && sym->namespace == NS_MACRO);
1491
1492 if (token_list_different(sym->expansion, expansion) ||
1493 token_list_different(sym->arglist, arglist)) {
1494 ret = 0;
1495 if ((clean && attr == SYM_ATTR_NORMAL)
1496 || sym->used_in == file_scope) {
1497 warning(pos, "preprocessor token %.*s redefined",
1498 name->len, name->name);
1499 info(sym->pos, "this was the original definition");
1500 }
1501 } else if (clean)
1502 goto out;
1503 }
1504
1505 if (!sym || sym->scope != file_scope) {
1506 sym = alloc_symbol(pos, SYM_NODE);
1507 bind_symbol(sym, name, NS_MACRO);
1508 add_ident(&macros, name);
1509 ret = 0;
1510 }
1511
1512 if (!ret) {
1513 sym->expansion = expansion;
1514 sym->arglist = arglist;
1515 if (token) /* Free the "define" token, but not the rest of the line */
1516 __free_token(token);
1517 }
1518
1519 sym->namespace = NS_MACRO;
1520 sym->used_in = NULL;
1521 sym->attr = attr;
1522 out:
1523 return ret;
1524 }
1525
1526 ///
1527 // predefine a macro with a printf-formatted value
1528 // @name: the name of the macro
1529 // @weak: 0/1 for a normal or a weak define
1530 // @fmt: the printf format followed by it's arguments.
1531 //
1532 // The type of the value is automatically infered:
1533 // TOKEN_NUMBER if it starts by a digit, TOKEN_IDENT otherwise.
1534 // If @fmt is null or empty, the macro is defined with an empty definition.
1535 void predefine(const char *name, int weak, const char *fmt, ...)
1536 {
1537 struct ident *ident = built_in_ident(name);
1538 struct token *value = &eof_token_entry;
1539 int attr = weak ? SYM_ATTR_WEAK : SYM_ATTR_NORMAL;
1540
1541 if (fmt && fmt[0]) {
1542 static char buf[256];
1543 va_list ap;
1544
1545 va_start(ap, fmt);
1546 vsnprintf(buf, sizeof(buf), fmt, ap);
1547 va_end(ap);
1548
1549 value = __alloc_token(0);
1550 if (isdigit((unsigned char)buf[0])) {
1551 token_type(value) = TOKEN_NUMBER;
1552 value->number = xstrdup(buf);
1553 } else {
1554 token_type(value) = TOKEN_IDENT;
1555 value->ident = built_in_ident(buf);
1556 }
1557 value->pos.whitespace = 1;
1558 value->next = &eof_token_entry;
1559 }
1560
1561 do_define(value->pos, NULL, ident, NULL, value, attr);
1562 }
1563
1564 ///
1565 // like predefine() but only if one of the non-standard dialect is chosen
1566 void predefine_nostd(const char *name)
1567 {
1568 if ((standard & STANDARD_GNU) || (standard == STANDARD_NONE))
1569 predefine(name, 1, "1");
1570 }
1571
1572 static void predefine_fmt(const char *fmt, int weak, va_list ap)
1573 {
1574 static char buf[256];
1575
1576 vsnprintf(buf, sizeof(buf), fmt, ap);
1577 predefine(buf, weak, "1");
1578 }
1579
1580 void predefine_strong(const char *fmt, ...)
1581 {
1582 va_list ap;
1583
1584 va_start(ap, fmt);
1585 predefine_fmt(fmt, 0, ap);
1586 va_end(ap);
1587 }
1588
1589 void predefine_weak(const char *fmt, ...)
1590 {
1591 va_list ap;
1592
1593 va_start(ap, fmt);
1594 predefine_fmt(fmt, 1, ap);
1595 va_end(ap);
1596 }
1597
1598 static int do_handle_define(struct stream *stream, struct token **line, struct token *token, int attr)
1599 {
1600 struct token *arglist, *expansion;
1601 struct token *left = token->next;
1602 struct ident *name;
1603
1604 if (token_type(left) != TOKEN_IDENT) {
1605 sparse_error(token->pos, "expected identifier to 'define'");
1606 return 1;
1607 }
1608
1609 name = left->ident;
1610
1611 arglist = NULL;
1612 expansion = left->next;
1613 if (!expansion->pos.whitespace) {
1614 if (match_op(expansion, '(')) {
1615 arglist = expansion;
1616 expansion = parse_arguments(expansion);
1617 if (!expansion)
1618 return 1;
1619 } else if (!eof_token(expansion)) {
1620 warning(expansion->pos,
1621 "no whitespace before object-like macro body");
1622 }
1623 }
1624
1625 return do_define(left->pos, token, name, arglist, expansion, attr);
1626 }
1627
1628 static int handle_define(struct stream *stream, struct token **line, struct token *token)
1629 {
1630 return do_handle_define(stream, line, token, SYM_ATTR_NORMAL);
1631 }
1632
1633 static int handle_weak_define(struct stream *stream, struct token **line, struct token *token)
1634 {
1635 return do_handle_define(stream, line, token, SYM_ATTR_WEAK);
1636 }
1637
1638 static int handle_strong_define(struct stream *stream, struct token **line, struct token *token)
1639 {
1640 return do_handle_define(stream, line, token, SYM_ATTR_STRONG);
1641 }
1642
1643 static int do_handle_undef(struct stream *stream, struct token **line, struct token *token, int attr)
1644 {
1645 struct token *left = token->next;
1646 struct symbol *sym;
1647
1648 if (token_type(left) != TOKEN_IDENT) {
1649 sparse_error(token->pos, "expected identifier to 'undef'");
1650 return 1;
1651 }
1652
1653 sym = lookup_symbol(left->ident, NS_MACRO | NS_UNDEF);
1654 if (sym) {
1655 if (attr < sym->attr)
1656 return 1;
1657 if (attr == sym->attr && sym->namespace == NS_UNDEF)
1658 return 1;
1659 } else if (attr <= SYM_ATTR_NORMAL)
1660 return 1;
1661
1662 if (!sym || sym->scope != file_scope) {
1663 sym = alloc_symbol(left->pos, SYM_NODE);
1664 bind_symbol(sym, left->ident, NS_MACRO);
1665 }
1666
1667 sym->namespace = NS_UNDEF;
1668 sym->used_in = NULL;
1669 sym->attr = attr;
1670
1671 return 1;
1672 }
1673
1674 static int handle_undef(struct stream *stream, struct token **line, struct token *token)
1675 {
1676 return do_handle_undef(stream, line, token, SYM_ATTR_NORMAL);
1677 }
1678
1679 static int handle_strong_undef(struct stream *stream, struct token **line, struct token *token)
1680 {
1681 return do_handle_undef(stream, line, token, SYM_ATTR_STRONG);
1682 }
1683
1684 static int preprocessor_if(struct stream *stream, struct token *token, int cond)
1685 {
1686 token_type(token) = false_nesting ? TOKEN_SKIP_GROUPS : TOKEN_IF;
1687 free_preprocessor_line(token->next);
1688 token->next = stream->top_if;
1689 stream->top_if = token;
1690 if (false_nesting || cond != 1)
1691 false_nesting++;
1692 return 0;
1693 }
1694
1695 static int handle_ifdef(struct stream *stream, struct token **line, struct token *token)
1696 {
1697 struct token *next = token->next;
1698 int arg;
1699 if (token_type(next) == TOKEN_IDENT) {
1700 arg = token_defined(next);
1701 } else {
1702 dirty_stream(stream);
1703 if (!false_nesting)
1704 sparse_error(token->pos, "expected preprocessor identifier");
1705 arg = -1;
1706 }
1707 return preprocessor_if(stream, token, arg);
1708 }
1709
1710 static int handle_ifndef(struct stream *stream, struct token **line, struct token *token)
1711 {
1712 struct token *next = token->next;
1713 int arg;
1714 if (token_type(next) == TOKEN_IDENT) {
1715 if (!stream->dirty && !stream->ifndef) {
1716 if (!stream->protect) {
1717 stream->ifndef = token;
1718 stream->protect = next->ident;
1719 } else if (stream->protect == next->ident) {
1720 stream->ifndef = token;
1721 stream->dirty = 1;
1722 }
1723 }
1724 arg = !token_defined(next);
1725 } else {
1726 dirty_stream(stream);
1727 if (!false_nesting)
1728 sparse_error(token->pos, "expected preprocessor identifier");
1729 arg = -1;
1730 }
1731
1732 return preprocessor_if(stream, token, arg);
1733 }
1734
1735 /*
1736 * Expression handling for #if and #elif; it differs from normal expansion
1737 * due to special treatment of "defined".
1738 */
1739 static int expression_value(struct token **where)
1740 {
1741 struct expression *expr;
1742 struct token *p;
1743 struct token **list = where, **beginning = NULL;
1744 long long value;
1745 int state = 0;
1746
1747 while (!eof_token(p = scan_next(list))) {
1748 switch (state) {
1749 case 0:
1750 if (token_type(p) != TOKEN_IDENT)
1751 break;
1752 if (p->ident == &defined_ident) {
1753 state = 1;
1754 beginning = list;
1755 break;
1756 }
1757 if (!expand_one_symbol(list))
1758 continue;
1759 if (token_type(p) != TOKEN_IDENT)
1760 break;
1761 token_type(p) = TOKEN_ZERO_IDENT;
1762 break;
1763 case 1:
1764 if (match_op(p, '(')) {
1765 state = 2;
1766 } else {
1767 state = 0;
1768 replace_with_defined(p);
1769 *beginning = p;
1770 }
1771 break;
1772 case 2:
1773 if (token_type(p) == TOKEN_IDENT)
1774 state = 3;
1775 else
1776 state = 0;
1777 replace_with_defined(p);
1778 *beginning = p;
1779 break;
1780 case 3:
1781 state = 0;
1782 if (!match_op(p, ')'))
1783 sparse_error(p->pos, "missing ')' after \"defined\"");
1784 *list = p->next;
1785 continue;
1786 }
1787 list = &p->next;
1788 }
1789
1790 p = constant_expression(*where, &expr);
1791 if (!eof_token(p))
1792 sparse_error(p->pos, "garbage at end: %s", show_token_sequence(p, 0));
1793 value = get_expression_value(expr);
1794 return value != 0;
1795 }
1796
1797 static int handle_if(struct stream *stream, struct token **line, struct token *token)
1798 {
1799 int value = 0;
1800 if (!false_nesting)
1801 value = expression_value(&token->next);
1802
1803 dirty_stream(stream);
1804 return preprocessor_if(stream, token, value);
1805 }
1806
1807 static int handle_elif(struct stream * stream, struct token **line, struct token *token)
1808 {
1809 struct token *top_if = stream->top_if;
1810 end_group(stream);
1811
1812 if (!top_if) {
1813 nesting_error(stream);
1814 sparse_error(token->pos, "unmatched #elif within stream");
1815 return 1;
1816 }
1817
1818 if (token_type(top_if) == TOKEN_ELSE) {
1819 nesting_error(stream);
1820 sparse_error(token->pos, "#elif after #else");
1821 if (!false_nesting)
1822 false_nesting = 1;
1823 return 1;
1824 }
1825
1826 dirty_stream(stream);
1827 if (token_type(top_if) != TOKEN_IF)
1828 return 1;
1829 if (false_nesting) {
1830 false_nesting = 0;
1831 if (!expression_value(&token->next))
1832 false_nesting = 1;
1833 } else {
1834 false_nesting = 1;
1835 token_type(top_if) = TOKEN_SKIP_GROUPS;
1836 }
1837 return 1;
1838 }
1839
1840 static int handle_else(struct stream *stream, struct token **line, struct token *token)
1841 {
1842 struct token *top_if = stream->top_if;
1843 end_group(stream);
1844
1845 if (!top_if) {
1846 nesting_error(stream);
1847 sparse_error(token->pos, "unmatched #else within stream");
1848 return 1;
1849 }
1850
1851 if (token_type(top_if) == TOKEN_ELSE) {
1852 nesting_error(stream);
1853 sparse_error(token->pos, "#else after #else");
1854 }
1855 if (false_nesting) {
1856 if (token_type(top_if) == TOKEN_IF)
1857 false_nesting = 0;
1858 } else {
1859 false_nesting = 1;
1860 }
1861 token_type(top_if) = TOKEN_ELSE;
1862 return 1;
1863 }
1864
1865 static int handle_endif(struct stream *stream, struct token **line, struct token *token)
1866 {
1867 struct token *top_if = stream->top_if;
1868 end_group(stream);
1869 if (!top_if) {
1870 nesting_error(stream);
1871 sparse_error(token->pos, "unmatched #endif in stream");
1872 return 1;
1873 }
1874 if (false_nesting)
1875 false_nesting--;
1876 stream->top_if = top_if->next;
1877 __free_token(top_if);
1878 return 1;
1879 }
1880
1881 static int handle_warning(struct stream *stream, struct token **line, struct token *token)
1882 {
1883 warning(token->pos, "%s", show_token_sequence(token->next, 0));
1884 return 1;
1885 }
1886
1887 static int handle_error(struct stream *stream, struct token **line, struct token *token)
1888 {
1889 sparse_error(token->pos, "%s", show_token_sequence(token->next, 0));
1890 return 1;
1891 }
1892
1893 static int handle_nostdinc(struct stream *stream, struct token **line, struct token *token)
1894 {
1895 /*
1896 * Do we have any non-system includes?
1897 * Clear them out if so..
1898 */
1899 *sys_includepath = NULL;
1900 return 1;
1901 }
1902
1903 static inline void update_inc_ptrs(const char ***where)
1904 {
1905
1906 if (*where <= dirafter_includepath) {
1907 dirafter_includepath++;
1908 /* If this was the entry that we prepend, don't
1909 * rise the lower entries, even if they are at
1910 * the same level. */
1911 if (where == &dirafter_includepath)
1912 return;
1913 }
1914 if (*where <= sys_includepath) {
1915 sys_includepath++;
1916 if (where == &sys_includepath)
1917 return;
1918 }
1919 if (*where <= isys_includepath) {
1920 isys_includepath++;
1921 if (where == &isys_includepath)
1922 return;
1923 }
1924
1925 /* angle_includepath is actually never updated, since we
1926 * don't suppport -iquote rught now. May change some day. */
1927 if (*where <= angle_includepath) {
1928 angle_includepath++;
1929 if (where == &angle_includepath)
1930 return;
1931 }
1932 }
1933
1934 /* Add a path before 'where' and update the pointers associated with the
1935 * includepath array */
1936 static void add_path_entry(struct token *token, const char *path,
1937 const char ***where)
1938 {
1939 const char **dst;
1940 const char *next;
1941
1942 /* Need one free entry.. */
1943 if (includepath[INCLUDEPATHS-2])
1944 error_die(token->pos, "too many include path entries");
1945
1946 /* check that this is not a duplicate */
1947 dst = includepath;
1948 while (*dst) {
1949 if (strcmp(*dst, path) == 0)
1950 return;
1951 dst++;
1952 }
1953 next = path;
1954 dst = *where;
1955
1956 update_inc_ptrs(where);
1957
1958 /*
1959 * Move them all up starting at dst,
1960 * insert the new entry..
1961 */
1962 do {
1963 const char *tmp = *dst;
1964 *dst = next;
1965 next = tmp;
1966 dst++;
1967 } while (next);
1968 }
1969
1970 static int handle_add_include(struct stream *stream, struct token **line, struct token *token)
1971 {
1972 for (;;) {
1973 token = token->next;
1974 if (eof_token(token))
1975 return 1;
1976 if (token_type(token) != TOKEN_STRING) {
1977 warning(token->pos, "expected path string");
1978 return 1;
1979 }
1980 add_path_entry(token, token->string->data, &isys_includepath);
1981 }
1982 }
1983
1984 static int handle_add_isystem(struct stream *stream, struct token **line, struct token *token)
1985 {
1986 for (;;) {
1987 token = token->next;
1988 if (eof_token(token))
1989 return 1;
1990 if (token_type(token) != TOKEN_STRING) {
1991 sparse_error(token->pos, "expected path string");
1992 return 1;
1993 }
1994 add_path_entry(token, token->string->data, &sys_includepath);
1995 }
1996 }
1997
1998 static int handle_add_system(struct stream *stream, struct token **line, struct token *token)
1999 {
2000 for (;;) {
2001 token = token->next;
2002 if (eof_token(token))
2003 return 1;
2004 if (token_type(token) != TOKEN_STRING) {
2005 sparse_error(token->pos, "expected path string");
2006 return 1;
2007 }
2008 add_path_entry(token, token->string->data, &dirafter_includepath);
2009 }
2010 }
2011
2012 /* Add to end on includepath list - no pointer updates */
2013 static void add_dirafter_entry(struct token *token, const char *path)
2014 {
2015 const char **dst = includepath;
2016
2017 /* Need one free entry.. */
2018 if (includepath[INCLUDEPATHS-2])
2019 error_die(token->pos, "too many include path entries");
2020
2021 /* Add to the end */
2022 while (*dst)
2023 dst++;
2024 *dst = path;
2025 dst++;
2026 *dst = NULL;
2027 }
2028
2029 static int handle_add_dirafter(struct stream *stream, struct token **line, struct token *token)
2030 {
2031 for (;;) {
2032 token = token->next;
2033 if (eof_token(token))
2034 return 1;
2035 if (token_type(token) != TOKEN_STRING) {
2036 sparse_error(token->pos, "expected path string");
2037 return 1;
2038 }
2039 add_dirafter_entry(token, token->string->data);
2040 }
2041 }
2042
2043 static int handle_split_include(struct stream *stream, struct token **line, struct token *token)
2044 {
2045 /*
2046 * -I-
2047 * From info gcc:
2048 * Split the include path. Any directories specified with `-I'
2049 * options before `-I-' are searched only for headers requested with
2050 * `#include "FILE"'; they are not searched for `#include <FILE>'.
2051 * If additional directories are specified with `-I' options after
2052 * the `-I-', those directories are searched for all `#include'
2053 * directives.
2054 * In addition, `-I-' inhibits the use of the directory of the current
2055 * file directory as the first search directory for `#include "FILE"'.
2056 */
2057 quote_includepath = includepath+1;
2058 angle_includepath = sys_includepath;
2059 return 1;
2060 }
2061
2062 /*
2063 * We replace "#pragma xxx" with "__pragma__" in the token
2064 * stream. Just as an example.
2065 *
2066 * We'll just #define that away for now, but the theory here
2067 * is that we can use this to insert arbitrary token sequences
2068 * to turn the pragmas into internal front-end sequences for
2069 * when we actually start caring about them.
2070 *
2071 * So eventually this will turn into some kind of extended
2072 * __attribute__() like thing, except called __pragma__(xxx).
2073 */
2074 static int handle_pragma(struct stream *stream, struct token **line, struct token *token)
2075 {
2076 struct token *next = *line;
2077
2078 if (match_ident(token->next, &once_ident) && eof_token(token->next->next)) {
2079 stream->once = 1;
2080 return 1;
2081 }
2082 token->ident = &pragma_ident;
2083 token->pos.newline = 1;
2084 token->pos.whitespace = 1;
2085 token->pos.pos = 1;
2086 *line = token;
2087 token->next = next;
2088 return 0;
2089 }
2090
2091 /*
2092 * We ignore #line for now.
2093 */
2094 static int handle_line(struct stream *stream, struct token **line, struct token *token)
2095 {
2096 return 1;
2097 }
2098
2099 static int handle_ident(struct stream *stream, struct token **line, struct token *token)
2100 {
2101 return 1;
2102 }
2103
2104 static int handle_nondirective(struct stream *stream, struct token **line, struct token *token)
2105 {
2106 sparse_error(token->pos, "unrecognized preprocessor line '%s'", show_token_sequence(token, 0));
2107 return 1;
2108 }
2109
2110 static bool expand_has_attribute(struct token *token, struct arg *args)
2111 {
2112 struct token *arg = args[0].expanded;
2113 struct symbol *sym;
2114
2115 if (token_type(arg) != TOKEN_IDENT) {
2116 sparse_error(arg->pos, "identifier expected");
2117 return false;
2118 }
2119
2120 sym = lookup_symbol(arg->ident, NS_KEYWORD);
2121 replace_with_bool(token, sym && sym->op && sym->op->attribute);
2122 return true;
2123 }
2124
2125 static bool expand_has_builtin(struct token *token, struct arg *args)
2126 {
2127 struct token *arg = args[0].expanded;
2128 struct symbol *sym;
2129
2130 if (token_type(arg) != TOKEN_IDENT) {
2131 sparse_error(arg->pos, "identifier expected");
2132 return false;
2133 }
2134
2135 sym = lookup_symbol(arg->ident, NS_SYMBOL);
2136 replace_with_bool(token, sym && sym->builtin);
2137 return true;
2138 }
2139
2140 static bool expand_has_extension(struct token *token, struct arg *args)
2141 {
2142 struct token *arg = args[0].expanded;
2143 struct ident *ident;
2144 bool val = false;
2145
2146 if (token_type(arg) != TOKEN_IDENT) {
2147 sparse_error(arg->pos, "identifier expected");
2148 return false;
2149 }
2150
2151 ident = arg->ident;
2152 if (ident == &c_alignas_ident)
2153 val = true;
2154 else if (ident == &c_alignof_ident)
2155 val = true;
2156 else if (ident == &c_generic_selections_ident)
2157 val = true;
2158 else if (ident == &c_static_assert_ident)
2159 val = true;
2160
2161 replace_with_bool(token, val);
2162 return 1;
2163 }
2164
2165 static bool expand_has_feature(struct token *token, struct arg *args)
2166 {
2167 struct token *arg = args[0].expanded;
2168 struct ident *ident;
2169 bool val = false;
2170
2171 if (token_type(arg) != TOKEN_IDENT) {
2172 sparse_error(arg->pos, "identifier expected");
2173 return false;
2174 }
2175
2176 ident = arg->ident;
2177 if (standard >= STANDARD_C11) {
2178 if (ident == &c_alignas_ident)
2179 val = true;
2180 else if (ident == &c_alignof_ident)
2181 val = true;
2182 else if (ident == &c_generic_selections_ident)
2183 val = true;
2184 else if (ident == &c_static_assert_ident)
2185 val = true;
2186 }
2187
2188 replace_with_bool(token, val);
2189 return 1;
2190 }
2191
2192 static void create_arglist(struct symbol *sym, int count)
2193 {
2194 struct token *token;
2195 struct token **next;
2196
2197 if (!count)
2198 return;
2199
2200 token = __alloc_token(0);
2201 token_type(token) = TOKEN_ARG_COUNT;
2202 token->count.normal = count;
2203 sym->arglist = token;
2204 next = &token->next;
2205
2206 while (count--) {
2207 struct token *id, *uses;
2208 id = __alloc_token(0);
2209 token_type(id) = TOKEN_IDENT;
2210 uses = __alloc_token(0);
2211 token_type(uses) = TOKEN_ARG_COUNT;
2212 uses->count.normal = 1;
2213
2214 *next = id;
2215 id->next = uses;
2216 next = &uses->next;
2217 }
2218 *next = &eof_token_entry;
2219 }
2220
2221 static void init_preprocessor(void)
2222 {
2223 int i;
2224 int stream = init_stream(NULL, "preprocessor", -1, includepath);
2225 static struct {
2226 const char *name;
2227 int (*handler)(struct stream *, struct token **, struct token *);
2228 } normal[] = {
2229 { "define", handle_define },
2230 { "weak_define", handle_weak_define },
2231 { "strong_define", handle_strong_define },
2232 { "undef", handle_undef },
2233 { "strong_undef", handle_strong_undef },
2234 { "warning", handle_warning },
2235 { "error", handle_error },
2236 { "include", handle_include },
2237 { "include_next", handle_include_next },
2238 { "pragma", handle_pragma },
2239 { "line", handle_line },
2240 { "ident", handle_ident },
2241
2242 // our internal preprocessor tokens
2243 { "nostdinc", handle_nostdinc },
2244 { "add_include", handle_add_include },
2245 { "add_isystem", handle_add_isystem },
2246 { "add_system", handle_add_system },
2247 { "add_dirafter", handle_add_dirafter },
2248 { "split_include", handle_split_include },
2249 { "argv_include", handle_argv_include },
2250 }, special[] = {
2251 { "ifdef", handle_ifdef },
2252 { "ifndef", handle_ifndef },
2253 { "else", handle_else },
2254 { "endif", handle_endif },
2255 { "if", handle_if },
2256 { "elif", handle_elif },
2257 };
2258 static struct {
2259 const char *name;
2260 void (*expand_simple)(struct token *);
2261 bool (*expand)(struct token *, struct arg *args);
2262 } dynamic[] = {
2263 { "__LINE__", expand_line },
2264 { "__FILE__", expand_file },
2265 { "__BASE_FILE__", expand_basefile },
2266 { "__DATE__", expand_date },
2267 { "__TIME__", expand_time },
2268 { "__COUNTER__", expand_counter },
2269 { "__INCLUDE_LEVEL__", expand_include_level },
2270 { "__has_attribute", NULL, expand_has_attribute },
2271 { "__has_builtin", NULL, expand_has_builtin },
2272 { "__has_extension", NULL, expand_has_extension },
2273 { "__has_feature", NULL, expand_has_feature },
2274 };
2275
2276 for (i = 0; i < ARRAY_SIZE(normal); i++) {
2277 struct symbol *sym;
2278 sym = create_symbol(stream, normal[i].name, SYM_PREPROCESSOR, NS_PREPROCESSOR);
2279 sym->handler = normal[i].handler;
2280 sym->normal = 1;
2281 }
2282 for (i = 0; i < ARRAY_SIZE(special); i++) {
2283 struct symbol *sym;
2284 sym = create_symbol(stream, special[i].name, SYM_PREPROCESSOR, NS_PREPROCESSOR);
2285 sym->handler = special[i].handler;
2286 sym->normal = 0;
2287 }
2288 for (i = 0; i < ARRAY_SIZE(dynamic); i++) {
2289 struct symbol *sym;
2290 sym = create_symbol(stream, dynamic[i].name, SYM_NODE, NS_MACRO);
2291 sym->expand_simple = dynamic[i].expand_simple;
2292 if ((sym->expand = dynamic[i].expand) != NULL)
2293 create_arglist(sym, 1);
2294 }
2295
2296 counter_macro = 0;
2297 }
2298
2299 static void handle_preprocessor_line(struct stream *stream, struct token **line, struct token *start)
2300 {
2301 int (*handler)(struct stream *, struct token **, struct token *);
2302 struct token *token = start->next;
2303 int is_normal = 1;
2304 int is_cond = 0; // is one of {is,ifdef,ifndef,elif,else,endif}
2305
2306 if (eof_token(token))
2307 return;
2308
2309 if (token_type(token) == TOKEN_IDENT) {
2310 struct symbol *sym = lookup_symbol(token->ident, NS_PREPROCESSOR);
2311 if (sym) {
2312 handler = sym->handler;
2313 is_normal = sym->normal;
2314 is_cond = !sym->normal;
2315 } else {
2316 handler = handle_nondirective;
2317 }
2318 } else if (token_type(token) == TOKEN_NUMBER) {
2319 handler = handle_line;
2320 } else {
2321 handler = handle_nondirective;
2322 }
2323
2324 if (is_normal) {
2325 dirty_stream(stream);
2326 if (false_nesting)
2327 goto out;
2328 }
2329
2330 if (expanding) {
2331 if (!is_cond || Wpedantic)
2332 warning(start->pos, "directive in macro's argument list");
2333 }
2334 if (!handler(stream, line, token)) /* all set */
2335 return;
2336
2337 out:
2338 free_preprocessor_line(token);
2339 }
2340
2341 static void preprocessor_line(struct stream *stream, struct token **line)
2342 {
2343 struct token *start = *line, *next;
2344 struct token **tp = &start->next;
2345
2346 for (;;) {
2347 next = *tp;
2348 if (next->pos.newline)
2349 break;
2350 tp = &next->next;
2351 }
2352 *line = next;
2353 *tp = &eof_token_entry;
2354 handle_preprocessor_line(stream, line, start);
2355 }
2356
2357 static void do_preprocess(struct token **list)
2358 {
2359 struct token *next;
2360
2361 while (!eof_token(next = scan_next(list))) {
2362 struct stream *stream = input_streams + next->pos.stream;
2363
2364 if (next->pos.newline && match_op(next, '#')) {
2365 if (!next->pos.noexpand) {
2366 preprocessor_line(stream, list);
2367 __free_token(next); /* Free the '#' token */
2368 continue;
2369 }
2370 }
2371
2372 switch (token_type(next)) {
2373 case TOKEN_STREAMEND:
2374 if (stream->top_if) {
2375 nesting_error(stream);
2376 sparse_error(stream->top_if->pos, "unterminated preprocessor conditional");
2377 stream->top_if = NULL;
2378 false_nesting = 0;
2379 }
2380 if (!stream->dirty)
2381 stream->constant = CONSTANT_FILE_YES;
2382 *list = next->next;
2383 include_level--;
2384 continue;
2385 case TOKEN_STREAMBEGIN:
2386 *list = next->next;
2387 include_level++;
2388 continue;
2389
2390 default:
2391 dirty_stream(stream);
2392 if (false_nesting) {
2393 *list = next->next;
2394 __free_token(next);
2395 continue;
2396 }
2397
2398 if (token_type(next) != TOKEN_IDENT ||
2399 expand_one_symbol(list))
2400 list = &next->next;
2401 }
2402 }
2403 }
2404
2405 void init_include_path(void)
2406 {
2407 FILE *fp;
2408 char path[256];
2409 char arch[32];
2410 char os[32];
2411
2412 fp = popen("/bin/uname -m", "r");
2413 if (!fp)
2414 return;
2415 if (!fgets(arch, sizeof(arch) - 1, fp))
2416 return;
2417 pclose(fp);
2418 if (arch[strlen(arch) - 1] == '\n')
2419 arch[strlen(arch) - 1] = '\0';
2420
2421 fp = popen("/bin/uname -o", "r");
2422 if (!fp)
2423 return;
2424 fgets(os, sizeof(os) - 1, fp);
2425 pclose(fp);
2426
2427 if (strcmp(os, "GNU/Linux\n") != 0)
2428 return;
2429 strcpy(os, "linux-gnu");
2430
2431 snprintf(path, sizeof(path), "/usr/include/%s-%s/", arch, os);
2432 add_pre_buffer("#add_system \"%s/\"\n", path);
2433 }
2434
2435 struct token * preprocess(struct token *token)
2436 {
2437 preprocessing = 1;
2438 init_preprocessor();
2439 do_preprocess(&token);
2440
2441 // Drop all expressions from preprocessing, they're not used any more.
2442 // This is not true when we have multiple files, though ;/
2443 // clear_expression_alloc();
2444 preprocessing = 0;
2445
2446 return token;
2447 }
2448
2449 static int is_VA_ARGS_token(struct token *token)
2450 {
2451 return (token_type(token) == TOKEN_IDENT) &&
2452 (token->ident == &__VA_ARGS___ident);
2453 }
2454
2455 static void dump_macro(struct symbol *sym)
2456 {
2457 int nargs = sym->arglist ? sym->arglist->count.normal : 0;
2458 struct token *args[nargs];
2459 struct token *token;
2460
2461 printf("#define %s", show_ident(sym->ident));
2462 token = sym->arglist;
2463 if (token) {
2464 const char *sep = "";
2465 int narg = 0;
2466 putchar('(');
2467 for (; !eof_token(token); token = token->next) {
2468 if (token_type(token) == TOKEN_ARG_COUNT)
2469 continue;
2470 if (is_VA_ARGS_token(token))
2471 printf("%s...", sep);
2472 else
2473 printf("%s%s", sep, show_token(token));
2474 args[narg++] = token;
2475 sep = ",";
2476 }
2477 putchar(')');
2478 }
2479
2480 token = sym->expansion;
2481 while (token_type(token) != TOKEN_UNTAINT) {
2482 struct token *next = token->next;
2483 if (token->pos.whitespace)
2484 putchar(' ');
2485 switch (token_type(token)) {
2486 case TOKEN_CONCAT:
2487 printf("##");
2488 break;
2489 case TOKEN_STR_ARGUMENT:
2490 printf("#");
2491 /* fall-through */
2492 case TOKEN_QUOTED_ARGUMENT:
2493 case TOKEN_MACRO_ARGUMENT:
2494 token = args[token->argnum];
2495 /* fall-through */
2496 default:
2497 printf("%s", show_token(token));
2498 }
2499 token = next;
2500 }
2501 putchar('\n');
2502 }
2503
2504 void dump_macro_definitions(void)
2505 {
2506 struct ident *name;
2507
2508 FOR_EACH_PTR(macros, name) {
2509 struct symbol *sym = lookup_macro(name);
2510 if (sym)
2511 dump_macro(sym);
2512 } END_FOR_EACH_PTR(name);
2513 }
Static analysis for C