?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
user_data: update a comment
[smatch.git] / pre-process.c
blob065a22392cf5e64e56a78613a42deba160d1cca2
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 #ifndef PATH_MAX
897 #define PATH_MAX 4096 // for Hurd where it's not defined
898 #endif
899
900 static int try_include(const char *path, const char *filename, int flen, struct token **where, const char **next_path)
901 {
902 int fd;
903 int plen = strlen(path);
904 static char fullname[PATH_MAX];
905
906 memcpy(fullname, path, plen);
907 if (plen && path[plen-1] != '/') {
908 fullname[plen] = '/';
909 plen++;
910 }
911 memcpy(fullname+plen, filename, flen);
912 if (already_tokenized(fullname))
913 return 1;
914 fd = open(fullname, O_RDONLY);
915 if (fd >= 0) {
916 char *streamname = xmemdup(fullname, plen + flen);
917 *where = tokenize(streamname, fd, *where, next_path);
918 close(fd);
919 return 1;
920 }
921 return 0;
922 }
923
924 static int do_include_path(const char **pptr, struct token **list, struct token *token, const char *filename, int flen)
925 {
926 const char *path;
927
928 while ((path = *pptr++) != NULL) {
929 if (!try_include(path, filename, flen, list, pptr))
930 continue;
931 return 1;
932 }
933 return 0;
934 }
935
936 static int free_preprocessor_line(struct token *token)
937 {
938 while (token_type(token) != TOKEN_EOF) {
939 struct token *free = token;
940 token = token->next;
941 __free_token(free);
942 };
943 return 1;
944 }
945
946 const char *find_include(const char *skip, const char *look_for)
947 {
948 DIR *dp;
949 struct dirent *entry;
950 struct stat statbuf;
951 const char *ret;
952 char cwd[PATH_MAX];
953 static char buf[PATH_MAX + 1];
954
955 dp = opendir(".");
956 if (!dp)
957 return NULL;
958
959 if (!getcwd(cwd, sizeof(cwd)))
960 goto close;
961
962 while ((entry = readdir(dp))) {
963 lstat(entry->d_name, &statbuf);
964
965 if (strcmp(entry->d_name, look_for) == 0) {
966 snprintf(buf, sizeof(buf), "%s/%s", cwd, entry->d_name);
967 closedir(dp);
968 return buf;
969 }
970
971 if (S_ISDIR(statbuf.st_mode)) {
972 /* Found a directory, but ignore . and .. */
973 if (strcmp(".", entry->d_name) == 0 ||
974 strcmp("..", entry->d_name) == 0 ||
975 strcmp(skip, entry->d_name) == 0)
976 continue;
977
978 chdir(entry->d_name);
979 ret = find_include("", look_for);
980 chdir("..");
981 if (ret) {
982 closedir(dp);
983 return ret;
984 }
985 }
986 }
987 close:
988 closedir(dp);
989
990 return NULL;
991 }
992
993 const char *search_dir(const char *stop, const char *look_for)
994 {
995 char cwd[PATH_MAX];
996 int len;
997 const char *ret;
998 int cnt = 0;
999
1000 if (!getcwd(cwd, sizeof(cwd)))
1001 return NULL;
1002
1003 len = strlen(cwd);
1004 while (len >= 0) {
1005 ret = find_include(cnt++ ? cwd + len + 1 : "", look_for);
1006 if (ret)
1007 return ret;
1008
1009 if (strcmp(cwd, stop) == 0 ||
1010 strcmp(cwd, "/usr/include") == 0 ||
1011 strcmp(cwd, "/usr/local/include") == 0 ||
1012 strlen(cwd) <= 10 || /* heck... don't search /usr/lib/ */
1013 strcmp(cwd, "/") == 0)
1014 return NULL;
1015
1016 while (--len >= 0) {
1017 if (cwd[len] == '/') {
1018 cwd[len] = '\0';
1019 break;
1020 }
1021 }
1022
1023 chdir("..");
1024 }
1025 return NULL;
1026 }
1027
1028 static void use_best_guess_header_file(struct token *token, const char *filename, struct token **list)
1029 {
1030 char cwd[PATH_MAX];
1031 char dir_part[PATH_MAX];
1032 const char *file_part;
1033 const char *include_name;
1034 static int cnt;
1035 int len;
1036
1037 /* Avoid guessing includes recursively. */
1038 if (cnt++ > 1000)
1039 return;
1040
1041 if (!filename || filename[0] == '\0')
1042 return;
1043
1044 file_part = filename;
1045 while ((filename = strchr(filename, '/'))) {
1046 ++filename;
1047 if (filename[0])
1048 file_part = filename;
1049 }
1050
1051 snprintf(dir_part, sizeof(dir_part), "%s", stream_name(token->pos.stream));
1052 len = strlen(dir_part);
1053 while (--len >= 0) {
1054 if (dir_part[len] == '/') {
1055 dir_part[len] = '\0';
1056 break;
1057 }
1058 }
1059 if (len < 0)
1060 sprintf(dir_part, ".");
1061
1062 if (!getcwd(cwd, sizeof(cwd)))
1063 return;
1064
1065 chdir(dir_part);
1066 include_name = search_dir(cwd, file_part);
1067 chdir(cwd);
1068 if (!include_name)
1069 return;
1070 sparse_error(token->pos, "using '%s'", include_name);
1071
1072 try_include("", include_name, strlen(include_name), list, includepath);
1073 }
1074
1075 static int handle_include_path(struct stream *stream, struct token **list, struct token *token, int how)
1076 {
1077 const char *filename;
1078 struct token *next;
1079 const char **path;
1080 int expect;
1081 int flen;
1082
1083 next = token->next;
1084 expect = '>';
1085 if (!match_op(next, '<')) {
1086 expand_list(&token->next);
1087 expect = 0;
1088 next = token;
1089 if (match_op(token->next, '<')) {
1090 next = token->next;
1091 expect = '>';
1092 }
1093 }
1094
1095 token = next->next;
1096 filename = token_name_sequence(token, expect, token);
1097 flen = strlen(filename) + 1;
1098
1099 /* Absolute path? */
1100 if (filename[0] == '/') {
1101 if (try_include("", filename, flen, list, includepath))
1102 return 0;
1103 goto out;
1104 }
1105
1106 switch (how) {
1107 case 1:
1108 path = stream->next_path;
1109 break;
1110 case 2:
1111 includepath[0] = "";
1112 path = includepath;
1113 break;
1114 default:
1115 /* Dir of input file is first dir to search for quoted includes */
1116 set_stream_include_path(stream);
1117 path = expect ? angle_includepath : quote_includepath;
1118 break;
1119 }
1120 /* Check the standard include paths.. */
1121 if (do_include_path(path, list, token, filename, flen))
1122 return 0;
1123 out:
1124 sparse_error(token->pos, "unable to open '%s'", filename);
1125 use_best_guess_header_file(token, filename, list);
1126 return 0;
1127 }
1128
1129 static int handle_include(struct stream *stream, struct token **list, struct token *token)
1130 {
1131 return handle_include_path(stream, list, token, 0);
1132 }
1133
1134 static int handle_include_next(struct stream *stream, struct token **list, struct token *token)
1135 {
1136 return handle_include_path(stream, list, token, 1);
1137 }
1138
1139 static int handle_argv_include(struct stream *stream, struct token **list, struct token *token)
1140 {
1141 return handle_include_path(stream, list, token, 2);
1142 }
1143
1144 static int token_different(struct token *t1, struct token *t2)
1145 {
1146 int different;
1147
1148 if (token_type(t1) != token_type(t2))
1149 return 1;
1150
1151 switch (token_type(t1)) {
1152 case TOKEN_IDENT:
1153 different = t1->ident != t2->ident;
1154 break;
1155 case TOKEN_ARG_COUNT:
1156 case TOKEN_UNTAINT:
1157 case TOKEN_CONCAT:
1158 case TOKEN_GNU_KLUDGE:
1159 different = 0;
1160 break;
1161 case TOKEN_NUMBER:
1162 different = strcmp(t1->number, t2->number);
1163 break;
1164 case TOKEN_SPECIAL:
1165 different = t1->special != t2->special;
1166 break;
1167 case TOKEN_MACRO_ARGUMENT:
1168 case TOKEN_QUOTED_ARGUMENT:
1169 case TOKEN_STR_ARGUMENT:
1170 different = t1->argnum != t2->argnum;
1171 break;
1172 case TOKEN_CHAR_EMBEDDED_0 ... TOKEN_CHAR_EMBEDDED_3:
1173 case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
1174 different = memcmp(t1->embedded, t2->embedded, 4);
1175 break;
1176 case TOKEN_CHAR:
1177 case TOKEN_WIDE_CHAR:
1178 case TOKEN_STRING:
1179 case TOKEN_WIDE_STRING: {
1180 struct string *s1, *s2;
1181
1182 s1 = t1->string;
1183 s2 = t2->string;
1184 different = 1;
1185 if (s1->length != s2->length)
1186 break;
1187 different = memcmp(s1->data, s2->data, s1->length);
1188 break;
1189 }
1190 default:
1191 different = 1;
1192 break;
1193 }
1194 return different;
1195 }
1196
1197 static int token_list_different(struct token *list1, struct token *list2)
1198 {
1199 for (;;) {
1200 if (list1 == list2)
1201 return 0;
1202 if (!list1 || !list2)
1203 return 1;
1204 if (token_different(list1, list2))
1205 return 1;
1206 list1 = list1->next;
1207 list2 = list2->next;
1208 }
1209 }
1210
1211 static inline void set_arg_count(struct token *token)
1212 {
1213 token_type(token) = TOKEN_ARG_COUNT;
1214 token->count.normal = token->count.quoted =
1215 token->count.str = token->count.vararg = 0;
1216 }
1217
1218 static struct token *parse_arguments(struct token *list)
1219 {
1220 struct token *arg = list->next, *next = list;
1221 struct argcount *count = &list->count;
1222
1223 set_arg_count(list);
1224
1225 if (match_op(arg, ')')) {
1226 next = arg->next;
1227 list->next = &eof_token_entry;
1228 return next;
1229 }
1230
1231 while (token_type(arg) == TOKEN_IDENT) {
1232 if (arg->ident == &__VA_ARGS___ident)
1233 goto Eva_args;
1234 if (!++count->normal)
1235 goto Eargs;
1236 next = arg->next;
1237
1238 if (match_op(next, ',')) {
1239 set_arg_count(next);
1240 arg = next->next;
1241 continue;
1242 }
1243
1244 if (match_op(next, ')')) {
1245 set_arg_count(next);
1246 next = next->next;
1247 arg->next->next = &eof_token_entry;
1248 return next;
1249 }
1250
1251 /* normal cases are finished here */
1252
1253 if (match_op(next, SPECIAL_ELLIPSIS)) {
1254 if (match_op(next->next, ')')) {
1255 set_arg_count(next);
1256 next->count.vararg = 1;
1257 next = next->next;
1258 arg->next->next = &eof_token_entry;
1259 return next->next;
1260 }
1261
1262 arg = next;
1263 goto Enotclosed;
1264 }
1265
1266 if (eof_token(next)) {
1267 goto Enotclosed;
1268 } else {
1269 arg = next;
1270 goto Ebadstuff;
1271 }
1272 }
1273
1274 if (match_op(arg, SPECIAL_ELLIPSIS)) {
1275 next = arg->next;
1276 token_type(arg) = TOKEN_IDENT;
1277 arg->ident = &__VA_ARGS___ident;
1278 if (!match_op(next, ')'))
1279 goto Enotclosed;
1280 if (!++count->normal)
1281 goto Eargs;
1282 set_arg_count(next);
1283 next->count.vararg = 1;
1284 next = next->next;
1285 arg->next->next = &eof_token_entry;
1286 return next;
1287 }
1288
1289 if (eof_token(arg)) {
1290 arg = next;
1291 goto Enotclosed;
1292 }
1293 if (match_op(arg, ','))
1294 goto Emissing;
1295 else
1296 goto Ebadstuff;
1297
1298
1299 Emissing:
1300 sparse_error(arg->pos, "parameter name missing");
1301 return NULL;
1302 Ebadstuff:
1303 sparse_error(arg->pos, "\"%s\" may not appear in macro parameter list",
1304 show_token(arg));
1305 return NULL;
1306 Enotclosed:
1307 sparse_error(arg->pos, "missing ')' in macro parameter list");
1308 return NULL;
1309 Eva_args:
1310 sparse_error(arg->pos, "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
1311 return NULL;
1312 Eargs:
1313 sparse_error(arg->pos, "too many arguments in macro definition");
1314 return NULL;
1315 }
1316
1317 static int try_arg(struct token *token, enum token_type type, struct token *arglist)
1318 {
1319 struct ident *ident = token->ident;
1320 int nr;
1321
1322 if (!arglist || token_type(token) != TOKEN_IDENT)
1323 return 0;
1324
1325 arglist = arglist->next;
1326
1327 for (nr = 0; !eof_token(arglist); nr++, arglist = arglist->next->next) {
1328 if (arglist->ident == ident) {
1329 struct argcount *count = &arglist->next->count;
1330 int n;
1331
1332 token->argnum = nr;
1333 token_type(token) = type;
1334 switch (type) {
1335 case TOKEN_MACRO_ARGUMENT:
1336 n = ++count->normal;
1337 break;
1338 case TOKEN_QUOTED_ARGUMENT:
1339 n = ++count->quoted;
1340 break;
1341 default:
1342 n = ++count->str;
1343 }
1344 if (n)
1345 return count->vararg ? 2 : 1;
1346 /*
1347 * XXX - need saner handling of that
1348 * (>= 1024 instances of argument)
1349 */
1350 token_type(token) = TOKEN_ERROR;
1351 return -1;
1352 }
1353 }
1354 return 0;
1355 }
1356
1357 static struct token *handle_hash(struct token **p, struct token *arglist)
1358 {
1359 struct token *token = *p;
1360 if (arglist) {
1361 struct token *next = token->next;
1362 if (!try_arg(next, TOKEN_STR_ARGUMENT, arglist))
1363 goto Equote;
1364 next->pos.whitespace = token->pos.whitespace;
1365 __free_token(token);
1366 token = *p = next;
1367 } else {
1368 token->pos.noexpand = 1;
1369 }
1370 return token;
1371
1372 Equote:
1373 sparse_error(token->pos, "'#' is not followed by a macro parameter");
1374 return NULL;
1375 }
1376
1377 /* token->next is ## */
1378 static struct token *handle_hashhash(struct token *token, struct token *arglist)
1379 {
1380 struct token *last = token;
1381 struct token *concat;
1382 int state = match_op(token, ',');
1383
1384 try_arg(token, TOKEN_QUOTED_ARGUMENT, arglist);
1385
1386 while (1) {
1387 struct token *t;
1388 int is_arg;
1389
1390 /* eat duplicate ## */
1391 concat = token->next;
1392 while (match_op(t = concat->next, SPECIAL_HASHHASH)) {
1393 token->next = t;
1394 __free_token(concat);
1395 concat = t;
1396 }
1397 token_type(concat) = TOKEN_CONCAT;
1398
1399 if (eof_token(t))
1400 goto Econcat;
1401
1402 if (match_op(t, '#')) {
1403 t = handle_hash(&concat->next, arglist);
1404 if (!t)
1405 return NULL;
1406 }
1407
1408 is_arg = try_arg(t, TOKEN_QUOTED_ARGUMENT, arglist);
1409
1410 if (state == 1 && is_arg) {
1411 state = is_arg;
1412 } else {
1413 last = t;
1414 state = match_op(t, ',');
1415 }
1416
1417 token = t;
1418 if (!match_op(token->next, SPECIAL_HASHHASH))
1419 break;
1420 }
1421 /* handle GNU ,##__VA_ARGS__ kludge, in all its weirdness */
1422 if (state == 2)
1423 token_type(last) = TOKEN_GNU_KLUDGE;
1424 return token;
1425
1426 Econcat:
1427 sparse_error(concat->pos, "'##' cannot appear at the ends of macro expansion");
1428 return NULL;
1429 }
1430
1431 static struct token *parse_expansion(struct token *expansion, struct token *arglist, struct ident *name)
1432 {
1433 struct token *token = expansion;
1434 struct token **p;
1435
1436 if (match_op(token, SPECIAL_HASHHASH))
1437 goto Econcat;
1438
1439 for (p = &expansion; !eof_token(token); p = &token->next, token = *p) {
1440 if (match_op(token, '#')) {
1441 token = handle_hash(p, arglist);
1442 if (!token)
1443 return NULL;
1444 }
1445 if (match_op(token->next, SPECIAL_HASHHASH)) {
1446 token = handle_hashhash(token, arglist);
1447 if (!token)
1448 return NULL;
1449 } else {
1450 try_arg(token, TOKEN_MACRO_ARGUMENT, arglist);
1451 }
1452 switch (token_type(token)) {
1453 case TOKEN_ERROR:
1454 goto Earg;
1455
1456 case TOKEN_STRING:
1457 case TOKEN_WIDE_STRING:
1458 token->string->immutable = 1;
1459 break;
1460 }
1461 }
1462 token = alloc_token(&expansion->pos);
1463 token_type(token) = TOKEN_UNTAINT;
1464 token->ident = name;
1465 token->next = *p;
1466 *p = token;
1467 return expansion;
1468
1469 Econcat:
1470 sparse_error(token->pos, "'##' cannot appear at the ends of macro expansion");
1471 return NULL;
1472 Earg:
1473 sparse_error(token->pos, "too many instances of argument in body");
1474 return NULL;
1475 }
1476
1477 static int do_define(struct position pos, struct token *token, struct ident *name,
1478 struct token *arglist, struct token *expansion, int attr)
1479 {
1480 struct symbol *sym;
1481 int ret = 1;
1482
1483 expansion = parse_expansion(expansion, arglist, name);
1484 if (!expansion)
1485 return 1;
1486
1487 sym = lookup_symbol(name, NS_MACRO | NS_UNDEF);
1488 if (sym) {
1489 int clean;
1490
1491 if (attr < sym->attr)
1492 goto out;
1493
1494 clean = (attr == sym->attr && sym->namespace == NS_MACRO);
1495
1496 if (token_list_different(sym->expansion, expansion) ||
1497 token_list_different(sym->arglist, arglist)) {
1498 ret = 0;
1499 if ((clean && attr == SYM_ATTR_NORMAL)
1500 || sym->used_in == file_scope) {
1501 warning(pos, "preprocessor token %.*s redefined",
1502 name->len, name->name);
1503 info(sym->pos, "this was the original definition");
1504 }
1505 } else if (clean)
1506 goto out;
1507 }
1508
1509 if (!sym || sym->scope != file_scope) {
1510 sym = alloc_symbol(pos, SYM_NODE);
1511 bind_symbol(sym, name, NS_MACRO);
1512 add_ident(&macros, name);
1513 ret = 0;
1514 }
1515
1516 if (!ret) {
1517 sym->expansion = expansion;
1518 sym->arglist = arglist;
1519 if (token) /* Free the "define" token, but not the rest of the line */
1520 __free_token(token);
1521 }
1522
1523 sym->namespace = NS_MACRO;
1524 sym->used_in = NULL;
1525 sym->attr = attr;
1526 out:
1527 return ret;
1528 }
1529
1530 ///
1531 // predefine a macro with a printf-formatted value
1532 // @name: the name of the macro
1533 // @weak: 0/1 for a normal or a weak define
1534 // @fmt: the printf format followed by it's arguments.
1535 //
1536 // The type of the value is automatically infered:
1537 // TOKEN_NUMBER if it starts by a digit, TOKEN_IDENT otherwise.
1538 // If @fmt is null or empty, the macro is defined with an empty definition.
1539 void predefine(const char *name, int weak, const char *fmt, ...)
1540 {
1541 struct ident *ident = built_in_ident(name);
1542 struct token *value = &eof_token_entry;
1543 int attr = weak ? SYM_ATTR_WEAK : SYM_ATTR_NORMAL;
1544
1545 if (fmt && fmt[0]) {
1546 static char buf[256];
1547 va_list ap;
1548
1549 va_start(ap, fmt);
1550 vsnprintf(buf, sizeof(buf), fmt, ap);
1551 va_end(ap);
1552
1553 value = __alloc_token(0);
1554 if (isdigit((unsigned char)buf[0])) {
1555 token_type(value) = TOKEN_NUMBER;
1556 value->number = xstrdup(buf);
1557 } else {
1558 token_type(value) = TOKEN_IDENT;
1559 value->ident = built_in_ident(buf);
1560 }
1561 value->pos.whitespace = 1;
1562 value->next = &eof_token_entry;
1563 }
1564
1565 do_define(value->pos, NULL, ident, NULL, value, attr);
1566 }
1567
1568 ///
1569 // like predefine() but only if one of the non-standard dialect is chosen
1570 void predefine_nostd(const char *name)
1571 {
1572 if ((standard & STANDARD_GNU) || (standard == STANDARD_NONE))
1573 predefine(name, 1, "1");
1574 }
1575
1576 static void predefine_fmt(const char *fmt, int weak, va_list ap)
1577 {
1578 static char buf[256];
1579
1580 vsnprintf(buf, sizeof(buf), fmt, ap);
1581 predefine(buf, weak, "1");
1582 }
1583
1584 void predefine_strong(const char *fmt, ...)
1585 {
1586 va_list ap;
1587
1588 va_start(ap, fmt);
1589 predefine_fmt(fmt, 0, ap);
1590 va_end(ap);
1591 }
1592
1593 void predefine_weak(const char *fmt, ...)
1594 {
1595 va_list ap;
1596
1597 va_start(ap, fmt);
1598 predefine_fmt(fmt, 1, ap);
1599 va_end(ap);
1600 }
1601
1602 static int do_handle_define(struct stream *stream, struct token **line, struct token *token, int attr)
1603 {
1604 struct token *arglist, *expansion;
1605 struct token *left = token->next;
1606 struct ident *name;
1607
1608 if (token_type(left) != TOKEN_IDENT) {
1609 sparse_error(token->pos, "expected identifier to 'define'");
1610 return 1;
1611 }
1612
1613 name = left->ident;
1614
1615 arglist = NULL;
1616 expansion = left->next;
1617 if (!expansion->pos.whitespace) {
1618 if (match_op(expansion, '(')) {
1619 arglist = expansion;
1620 expansion = parse_arguments(expansion);
1621 if (!expansion)
1622 return 1;
1623 } else if (!eof_token(expansion)) {
1624 warning(expansion->pos,
1625 "no whitespace before object-like macro body");
1626 }
1627 }
1628
1629 return do_define(left->pos, token, name, arglist, expansion, attr);
1630 }
1631
1632 static int handle_define(struct stream *stream, struct token **line, struct token *token)
1633 {
1634 return do_handle_define(stream, line, token, SYM_ATTR_NORMAL);
1635 }
1636
1637 static int handle_weak_define(struct stream *stream, struct token **line, struct token *token)
1638 {
1639 return do_handle_define(stream, line, token, SYM_ATTR_WEAK);
1640 }
1641
1642 static int handle_strong_define(struct stream *stream, struct token **line, struct token *token)
1643 {
1644 return do_handle_define(stream, line, token, SYM_ATTR_STRONG);
1645 }
1646
1647 static int do_handle_undef(struct stream *stream, struct token **line, struct token *token, int attr)
1648 {
1649 struct token *left = token->next;
1650 struct symbol *sym;
1651
1652 if (token_type(left) != TOKEN_IDENT) {
1653 sparse_error(token->pos, "expected identifier to 'undef'");
1654 return 1;
1655 }
1656
1657 sym = lookup_symbol(left->ident, NS_MACRO | NS_UNDEF);
1658 if (sym) {
1659 if (attr < sym->attr)
1660 return 1;
1661 if (attr == sym->attr && sym->namespace == NS_UNDEF)
1662 return 1;
1663 } else if (attr <= SYM_ATTR_NORMAL)
1664 return 1;
1665
1666 if (!sym || sym->scope != file_scope) {
1667 sym = alloc_symbol(left->pos, SYM_NODE);
1668 bind_symbol(sym, left->ident, NS_MACRO);
1669 }
1670
1671 sym->namespace = NS_UNDEF;
1672 sym->used_in = NULL;
1673 sym->attr = attr;
1674
1675 return 1;
1676 }
1677
1678 static int handle_undef(struct stream *stream, struct token **line, struct token *token)
1679 {
1680 return do_handle_undef(stream, line, token, SYM_ATTR_NORMAL);
1681 }
1682
1683 static int handle_strong_undef(struct stream *stream, struct token **line, struct token *token)
1684 {
1685 return do_handle_undef(stream, line, token, SYM_ATTR_STRONG);
1686 }
1687
1688 static int preprocessor_if(struct stream *stream, struct token *token, int cond)
1689 {
1690 token_type(token) = false_nesting ? TOKEN_SKIP_GROUPS : TOKEN_IF;
1691 free_preprocessor_line(token->next);
1692 token->next = stream->top_if;
1693 stream->top_if = token;
1694 if (false_nesting || cond != 1)
1695 false_nesting++;
1696 return 0;
1697 }
1698
1699 static int handle_ifdef(struct stream *stream, struct token **line, struct token *token)
1700 {
1701 struct token *next = token->next;
1702 int arg;
1703 if (token_type(next) == TOKEN_IDENT) {
1704 arg = token_defined(next);
1705 } else {
1706 dirty_stream(stream);
1707 if (!false_nesting)
1708 sparse_error(token->pos, "expected preprocessor identifier");
1709 arg = -1;
1710 }
1711 return preprocessor_if(stream, token, arg);
1712 }
1713
1714 static int handle_ifndef(struct stream *stream, struct token **line, struct token *token)
1715 {
1716 struct token *next = token->next;
1717 int arg;
1718 if (token_type(next) == TOKEN_IDENT) {
1719 if (!stream->dirty && !stream->ifndef) {
1720 if (!stream->protect) {
1721 stream->ifndef = token;
1722 stream->protect = next->ident;
1723 } else if (stream->protect == next->ident) {
1724 stream->ifndef = token;
1725 stream->dirty = 1;
1726 }
1727 }
1728 arg = !token_defined(next);
1729 } else {
1730 dirty_stream(stream);
1731 if (!false_nesting)
1732 sparse_error(token->pos, "expected preprocessor identifier");
1733 arg = -1;
1734 }
1735
1736 return preprocessor_if(stream, token, arg);
1737 }
1738
1739 /*
1740 * Expression handling for #if and #elif; it differs from normal expansion
1741 * due to special treatment of "defined".
1742 */
1743 static int expression_value(struct token **where)
1744 {
1745 struct expression *expr;
1746 struct token *p;
1747 struct token **list = where, **beginning = NULL;
1748 long long value;
1749 int state = 0;
1750
1751 while (!eof_token(p = scan_next(list))) {
1752 switch (state) {
1753 case 0:
1754 if (token_type(p) != TOKEN_IDENT)
1755 break;
1756 if (p->ident == &defined_ident) {
1757 state = 1;
1758 beginning = list;
1759 break;
1760 }
1761 if (!expand_one_symbol(list))
1762 continue;
1763 if (token_type(p) != TOKEN_IDENT)
1764 break;
1765 token_type(p) = TOKEN_ZERO_IDENT;
1766 break;
1767 case 1:
1768 if (match_op(p, '(')) {
1769 state = 2;
1770 } else {
1771 state = 0;
1772 replace_with_defined(p);
1773 *beginning = p;
1774 }
1775 break;
1776 case 2:
1777 if (token_type(p) == TOKEN_IDENT)
1778 state = 3;
1779 else
1780 state = 0;
1781 replace_with_defined(p);
1782 *beginning = p;
1783 break;
1784 case 3:
1785 state = 0;
1786 if (!match_op(p, ')'))
1787 sparse_error(p->pos, "missing ')' after \"defined\"");
1788 *list = p->next;
1789 continue;
1790 }
1791 list = &p->next;
1792 }
1793
1794 p = constant_expression(*where, &expr);
1795 if (!eof_token(p))
1796 sparse_error(p->pos, "garbage at end: %s", show_token_sequence(p, 0));
1797 value = get_expression_value(expr);
1798 return value != 0;
1799 }
1800
1801 static int handle_if(struct stream *stream, struct token **line, struct token *token)
1802 {
1803 int value = 0;
1804 if (!false_nesting)
1805 value = expression_value(&token->next);
1806
1807 dirty_stream(stream);
1808 return preprocessor_if(stream, token, value);
1809 }
1810
1811 static int handle_elif(struct stream * stream, struct token **line, struct token *token)
1812 {
1813 struct token *top_if = stream->top_if;
1814 end_group(stream);
1815
1816 if (!top_if) {
1817 nesting_error(stream);
1818 sparse_error(token->pos, "unmatched #elif within stream");
1819 return 1;
1820 }
1821
1822 if (token_type(top_if) == TOKEN_ELSE) {
1823 nesting_error(stream);
1824 sparse_error(token->pos, "#elif after #else");
1825 if (!false_nesting)
1826 false_nesting = 1;
1827 return 1;
1828 }
1829
1830 dirty_stream(stream);
1831 if (token_type(top_if) != TOKEN_IF)
1832 return 1;
1833 if (false_nesting) {
1834 false_nesting = 0;
1835 if (!expression_value(&token->next))
1836 false_nesting = 1;
1837 } else {
1838 false_nesting = 1;
1839 token_type(top_if) = TOKEN_SKIP_GROUPS;
1840 }
1841 return 1;
1842 }
1843
1844 static int handle_else(struct stream *stream, struct token **line, struct token *token)
1845 {
1846 struct token *top_if = stream->top_if;
1847 end_group(stream);
1848
1849 if (!top_if) {
1850 nesting_error(stream);
1851 sparse_error(token->pos, "unmatched #else within stream");
1852 return 1;
1853 }
1854
1855 if (token_type(top_if) == TOKEN_ELSE) {
1856 nesting_error(stream);
1857 sparse_error(token->pos, "#else after #else");
1858 }
1859 if (false_nesting) {
1860 if (token_type(top_if) == TOKEN_IF)
1861 false_nesting = 0;
1862 } else {
1863 false_nesting = 1;
1864 }
1865 token_type(top_if) = TOKEN_ELSE;
1866 return 1;
1867 }
1868
1869 static int handle_endif(struct stream *stream, struct token **line, struct token *token)
1870 {
1871 struct token *top_if = stream->top_if;
1872 end_group(stream);
1873 if (!top_if) {
1874 nesting_error(stream);
1875 sparse_error(token->pos, "unmatched #endif in stream");
1876 return 1;
1877 }
1878 if (false_nesting)
1879 false_nesting--;
1880 stream->top_if = top_if->next;
1881 __free_token(top_if);
1882 return 1;
1883 }
1884
1885 static int handle_warning(struct stream *stream, struct token **line, struct token *token)
1886 {
1887 warning(token->pos, "%s", show_token_sequence(token->next, 0));
1888 return 1;
1889 }
1890
1891 static int handle_error(struct stream *stream, struct token **line, struct token *token)
1892 {
1893 sparse_error(token->pos, "%s", show_token_sequence(token->next, 0));
1894 return 1;
1895 }
1896
1897 static int handle_nostdinc(struct stream *stream, struct token **line, struct token *token)
1898 {
1899 /*
1900 * Do we have any non-system includes?
1901 * Clear them out if so..
1902 */
1903 *sys_includepath = NULL;
1904 return 1;
1905 }
1906
1907 static inline void update_inc_ptrs(const char ***where)
1908 {
1909
1910 if (*where <= dirafter_includepath) {
1911 dirafter_includepath++;
1912 /* If this was the entry that we prepend, don't
1913 * rise the lower entries, even if they are at
1914 * the same level. */
1915 if (where == &dirafter_includepath)
1916 return;
1917 }
1918 if (*where <= sys_includepath) {
1919 sys_includepath++;
1920 if (where == &sys_includepath)
1921 return;
1922 }
1923 if (*where <= isys_includepath) {
1924 isys_includepath++;
1925 if (where == &isys_includepath)
1926 return;
1927 }
1928
1929 /* angle_includepath is actually never updated, since we
1930 * don't suppport -iquote rught now. May change some day. */
1931 if (*where <= angle_includepath) {
1932 angle_includepath++;
1933 if (where == &angle_includepath)
1934 return;
1935 }
1936 }
1937
1938 /* Add a path before 'where' and update the pointers associated with the
1939 * includepath array */
1940 static void add_path_entry(struct token *token, const char *path,
1941 const char ***where)
1942 {
1943 const char **dst;
1944 const char *next;
1945
1946 /* Need one free entry.. */
1947 if (includepath[INCLUDEPATHS-2])
1948 error_die(token->pos, "too many include path entries");
1949
1950 /* check that this is not a duplicate */
1951 dst = includepath;
1952 while (*dst) {
1953 if (strcmp(*dst, path) == 0)
1954 return;
1955 dst++;
1956 }
1957 next = path;
1958 dst = *where;
1959
1960 update_inc_ptrs(where);
1961
1962 /*
1963 * Move them all up starting at dst,
1964 * insert the new entry..
1965 */
1966 do {
1967 const char *tmp = *dst;
1968 *dst = next;
1969 next = tmp;
1970 dst++;
1971 } while (next);
1972 }
1973
1974 static int handle_add_include(struct stream *stream, struct token **line, struct token *token)
1975 {
1976 for (;;) {
1977 token = token->next;
1978 if (eof_token(token))
1979 return 1;
1980 if (token_type(token) != TOKEN_STRING) {
1981 warning(token->pos, "expected path string");
1982 return 1;
1983 }
1984 add_path_entry(token, token->string->data, &isys_includepath);
1985 }
1986 }
1987
1988 static int handle_add_isystem(struct stream *stream, struct token **line, struct token *token)
1989 {
1990 for (;;) {
1991 token = token->next;
1992 if (eof_token(token))
1993 return 1;
1994 if (token_type(token) != TOKEN_STRING) {
1995 sparse_error(token->pos, "expected path string");
1996 return 1;
1997 }
1998 add_path_entry(token, token->string->data, &sys_includepath);
1999 }
2000 }
2001
2002 static int handle_add_system(struct stream *stream, struct token **line, struct token *token)
2003 {
2004 for (;;) {
2005 token = token->next;
2006 if (eof_token(token))
2007 return 1;
2008 if (token_type(token) != TOKEN_STRING) {
2009 sparse_error(token->pos, "expected path string");
2010 return 1;
2011 }
2012 add_path_entry(token, token->string->data, &dirafter_includepath);
2013 }
2014 }
2015
2016 /* Add to end on includepath list - no pointer updates */
2017 static void add_dirafter_entry(struct token *token, const char *path)
2018 {
2019 const char **dst = includepath;
2020
2021 /* Need one free entry.. */
2022 if (includepath[INCLUDEPATHS-2])
2023 error_die(token->pos, "too many include path entries");
2024
2025 /* Add to the end */
2026 while (*dst)
2027 dst++;
2028 *dst = path;
2029 dst++;
2030 *dst = NULL;
2031 }
2032
2033 static int handle_add_dirafter(struct stream *stream, struct token **line, struct token *token)
2034 {
2035 for (;;) {
2036 token = token->next;
2037 if (eof_token(token))
2038 return 1;
2039 if (token_type(token) != TOKEN_STRING) {
2040 sparse_error(token->pos, "expected path string");
2041 return 1;
2042 }
2043 add_dirafter_entry(token, token->string->data);
2044 }
2045 }
2046
2047 static int handle_split_include(struct stream *stream, struct token **line, struct token *token)
2048 {
2049 /*
2050 * -I-
2051 * From info gcc:
2052 * Split the include path. Any directories specified with `-I'
2053 * options before `-I-' are searched only for headers requested with
2054 * `#include "FILE"'; they are not searched for `#include <FILE>'.
2055 * If additional directories are specified with `-I' options after
2056 * the `-I-', those directories are searched for all `#include'
2057 * directives.
2058 * In addition, `-I-' inhibits the use of the directory of the current
2059 * file directory as the first search directory for `#include "FILE"'.
2060 */
2061 quote_includepath = includepath+1;
2062 angle_includepath = sys_includepath;
2063 return 1;
2064 }
2065
2066 /*
2067 * We replace "#pragma xxx" with "__pragma__" in the token
2068 * stream. Just as an example.
2069 *
2070 * We'll just #define that away for now, but the theory here
2071 * is that we can use this to insert arbitrary token sequences
2072 * to turn the pragmas into internal front-end sequences for
2073 * when we actually start caring about them.
2074 *
2075 * So eventually this will turn into some kind of extended
2076 * __attribute__() like thing, except called __pragma__(xxx).
2077 */
2078 static int handle_pragma(struct stream *stream, struct token **line, struct token *token)
2079 {
2080 struct token *next = *line;
2081
2082 if (match_ident(token->next, &once_ident) && eof_token(token->next->next)) {
2083 stream->once = 1;
2084 return 1;
2085 }
2086 token->ident = &pragma_ident;
2087 token->pos.newline = 1;
2088 token->pos.whitespace = 1;
2089 token->pos.pos = 1;
2090 *line = token;
2091 token->next = next;
2092 return 0;
2093 }
2094
2095 /*
2096 * We ignore #line for now.
2097 */
2098 static int handle_line(struct stream *stream, struct token **line, struct token *token)
2099 {
2100 return 1;
2101 }
2102
2103 static int handle_ident(struct stream *stream, struct token **line, struct token *token)
2104 {
2105 return 1;
2106 }
2107
2108 static int handle_nondirective(struct stream *stream, struct token **line, struct token *token)
2109 {
2110 sparse_error(token->pos, "unrecognized preprocessor line '%s'", show_token_sequence(token, 0));
2111 return 1;
2112 }
2113
2114 static bool expand_has_attribute(struct token *token, struct arg *args)
2115 {
2116 struct token *arg = args[0].expanded;
2117 struct symbol *sym;
2118
2119 if (token_type(arg) != TOKEN_IDENT) {
2120 sparse_error(arg->pos, "identifier expected");
2121 return false;
2122 }
2123
2124 sym = lookup_symbol(arg->ident, NS_KEYWORD);
2125 replace_with_bool(token, sym && sym->op && sym->op->attribute);
2126 return true;
2127 }
2128
2129 static bool expand_has_builtin(struct token *token, struct arg *args)
2130 {
2131 struct token *arg = args[0].expanded;
2132 struct symbol *sym;
2133
2134 if (token_type(arg) != TOKEN_IDENT) {
2135 sparse_error(arg->pos, "identifier expected");
2136 return false;
2137 }
2138
2139 sym = lookup_symbol(arg->ident, NS_SYMBOL);
2140 replace_with_bool(token, sym && sym->builtin);
2141 return true;
2142 }
2143
2144 static bool expand_has_extension(struct token *token, struct arg *args)
2145 {
2146 struct token *arg = args[0].expanded;
2147 struct ident *ident;
2148 bool val = false;
2149
2150 if (token_type(arg) != TOKEN_IDENT) {
2151 sparse_error(arg->pos, "identifier expected");
2152 return false;
2153 }
2154
2155 ident = arg->ident;
2156 if (ident == &c_alignas_ident)
2157 val = true;
2158 else if (ident == &c_alignof_ident)
2159 val = true;
2160 else if (ident == &c_generic_selections_ident)
2161 val = true;
2162 else if (ident == &c_static_assert_ident)
2163 val = true;
2164
2165 replace_with_bool(token, val);
2166 return 1;
2167 }
2168
2169 static bool expand_has_feature(struct token *token, struct arg *args)
2170 {
2171 struct token *arg = args[0].expanded;
2172 struct ident *ident;
2173 bool val = false;
2174
2175 if (token_type(arg) != TOKEN_IDENT) {
2176 sparse_error(arg->pos, "identifier expected");
2177 return false;
2178 }
2179
2180 ident = arg->ident;
2181 if (standard >= STANDARD_C11) {
2182 if (ident == &c_alignas_ident)
2183 val = true;
2184 else if (ident == &c_alignof_ident)
2185 val = true;
2186 else if (ident == &c_generic_selections_ident)
2187 val = true;
2188 else if (ident == &c_static_assert_ident)
2189 val = true;
2190 }
2191
2192 replace_with_bool(token, val);
2193 return 1;
2194 }
2195
2196 static void create_arglist(struct symbol *sym, int count)
2197 {
2198 struct token *token;
2199 struct token **next;
2200
2201 if (!count)
2202 return;
2203
2204 token = __alloc_token(0);
2205 token_type(token) = TOKEN_ARG_COUNT;
2206 token->count.normal = count;
2207 sym->arglist = token;
2208 next = &token->next;
2209
2210 while (count--) {
2211 struct token *id, *uses;
2212 id = __alloc_token(0);
2213 token_type(id) = TOKEN_IDENT;
2214 uses = __alloc_token(0);
2215 token_type(uses) = TOKEN_ARG_COUNT;
2216 uses->count.normal = 1;
2217
2218 *next = id;
2219 id->next = uses;
2220 next = &uses->next;
2221 }
2222 *next = &eof_token_entry;
2223 }
2224
2225 static void init_preprocessor(void)
2226 {
2227 int i;
2228 int stream = init_stream("preprocessor", -1, includepath);
2229 static struct {
2230 const char *name;
2231 int (*handler)(struct stream *, struct token **, struct token *);
2232 } normal[] = {
2233 { "define", handle_define },
2234 { "weak_define", handle_weak_define },
2235 { "strong_define", handle_strong_define },
2236 { "undef", handle_undef },
2237 { "strong_undef", handle_strong_undef },
2238 { "warning", handle_warning },
2239 { "error", handle_error },
2240 { "include", handle_include },
2241 { "include_next", handle_include_next },
2242 { "pragma", handle_pragma },
2243 { "line", handle_line },
2244 { "ident", handle_ident },
2245
2246 // our internal preprocessor tokens
2247 { "nostdinc", handle_nostdinc },
2248 { "add_include", handle_add_include },
2249 { "add_isystem", handle_add_isystem },
2250 { "add_system", handle_add_system },
2251 { "add_dirafter", handle_add_dirafter },
2252 { "split_include", handle_split_include },
2253 { "argv_include", handle_argv_include },
2254 }, special[] = {
2255 { "ifdef", handle_ifdef },
2256 { "ifndef", handle_ifndef },
2257 { "else", handle_else },
2258 { "endif", handle_endif },
2259 { "if", handle_if },
2260 { "elif", handle_elif },
2261 };
2262 static struct {
2263 const char *name;
2264 void (*expand_simple)(struct token *);
2265 bool (*expand)(struct token *, struct arg *args);
2266 } dynamic[] = {
2267 { "__LINE__", expand_line },
2268 { "__FILE__", expand_file },
2269 { "__BASE_FILE__", expand_basefile },
2270 { "__DATE__", expand_date },
2271 { "__TIME__", expand_time },
2272 { "__COUNTER__", expand_counter },
2273 { "__INCLUDE_LEVEL__", expand_include_level },
2274 { "__has_attribute", NULL, expand_has_attribute },
2275 { "__has_builtin", NULL, expand_has_builtin },
2276 { "__has_extension", NULL, expand_has_extension },
2277 { "__has_feature", NULL, expand_has_feature },
2278 };
2279
2280 for (i = 0; i < ARRAY_SIZE(normal); i++) {
2281 struct symbol *sym;
2282 sym = create_symbol(stream, normal[i].name, SYM_PREPROCESSOR, NS_PREPROCESSOR);
2283 sym->handler = normal[i].handler;
2284 sym->normal = 1;
2285 }
2286 for (i = 0; i < ARRAY_SIZE(special); i++) {
2287 struct symbol *sym;
2288 sym = create_symbol(stream, special[i].name, SYM_PREPROCESSOR, NS_PREPROCESSOR);
2289 sym->handler = special[i].handler;
2290 sym->normal = 0;
2291 }
2292 for (i = 0; i < ARRAY_SIZE(dynamic); i++) {
2293 struct symbol *sym;
2294 sym = create_symbol(stream, dynamic[i].name, SYM_NODE, NS_MACRO);
2295 sym->expand_simple = dynamic[i].expand_simple;
2296 if ((sym->expand = dynamic[i].expand) != NULL)
2297 create_arglist(sym, 1);
2298 }
2299
2300 counter_macro = 0;
2301 }
2302
2303 static void handle_preprocessor_line(struct stream *stream, struct token **line, struct token *start)
2304 {
2305 int (*handler)(struct stream *, struct token **, struct token *);
2306 struct token *token = start->next;
2307 int is_normal = 1;
2308 int is_cond = 0; // is one of {is,ifdef,ifndef,elif,else,endif}
2309
2310 if (eof_token(token))
2311 return;
2312
2313 if (token_type(token) == TOKEN_IDENT) {
2314 struct symbol *sym = lookup_symbol(token->ident, NS_PREPROCESSOR);
2315 if (sym) {
2316 handler = sym->handler;
2317 is_normal = sym->normal;
2318 is_cond = !sym->normal;
2319 } else {
2320 handler = handle_nondirective;
2321 }
2322 } else if (token_type(token) == TOKEN_NUMBER) {
2323 handler = handle_line;
2324 } else {
2325 handler = handle_nondirective;
2326 }
2327
2328 if (is_normal) {
2329 dirty_stream(stream);
2330 if (false_nesting)
2331 goto out;
2332 }
2333
2334 if (expanding) {
2335 if (!is_cond || Wpedantic)
2336 warning(start->pos, "directive in macro's argument list");
2337 }
2338 if (!handler(stream, line, token)) /* all set */
2339 return;
2340
2341 out:
2342 free_preprocessor_line(token);
2343 }
2344
2345 static void preprocessor_line(struct stream *stream, struct token **line)
2346 {
2347 struct token *start = *line, *next;
2348 struct token **tp = &start->next;
2349
2350 for (;;) {
2351 next = *tp;
2352 if (next->pos.newline)
2353 break;
2354 tp = &next->next;
2355 }
2356 *line = next;
2357 *tp = &eof_token_entry;
2358 handle_preprocessor_line(stream, line, start);
2359 }
2360
2361 static void do_preprocess(struct token **list)
2362 {
2363 struct token *next;
2364
2365 while (!eof_token(next = scan_next(list))) {
2366 struct stream *stream = input_streams + next->pos.stream;
2367
2368 if (next->pos.newline && match_op(next, '#')) {
2369 if (!next->pos.noexpand) {
2370 preprocessor_line(stream, list);
2371 __free_token(next); /* Free the '#' token */
2372 continue;
2373 }
2374 }
2375
2376 switch (token_type(next)) {
2377 case TOKEN_STREAMEND:
2378 if (stream->top_if) {
2379 nesting_error(stream);
2380 sparse_error(stream->top_if->pos, "unterminated preprocessor conditional");
2381 stream->top_if = NULL;
2382 false_nesting = 0;
2383 }
2384 if (!stream->dirty)
2385 stream->constant = CONSTANT_FILE_YES;
2386 *list = next->next;
2387 include_level--;
2388 continue;
2389 case TOKEN_STREAMBEGIN:
2390 *list = next->next;
2391 include_level++;
2392 continue;
2393
2394 default:
2395 dirty_stream(stream);
2396 if (false_nesting) {
2397 *list = next->next;
2398 __free_token(next);
2399 continue;
2400 }
2401
2402 if (token_type(next) != TOKEN_IDENT ||
2403 expand_one_symbol(list))
2404 list = &next->next;
2405 }
2406 }
2407 }
2408
2409 void init_include_path(void)
2410 {
2411 FILE *fp;
2412 char path[256];
2413 char arch[32];
2414 char os[32];
2415
2416 fp = popen("/bin/uname -m", "r");
2417 if (!fp)
2418 return;
2419 if (!fgets(arch, sizeof(arch) - 1, fp))
2420 return;
2421 pclose(fp);
2422 if (arch[strlen(arch) - 1] == '\n')
2423 arch[strlen(arch) - 1] = '\0';
2424
2425 fp = popen("/bin/uname -o", "r");
2426 if (!fp)
2427 return;
2428 fgets(os, sizeof(os) - 1, fp);
2429 pclose(fp);
2430
2431 if (strcmp(os, "GNU/Linux\n") != 0)
2432 return;
2433 strcpy(os, "linux-gnu");
2434
2435 snprintf(path, sizeof(path), "/usr/include/%s-%s/", arch, os);
2436 add_pre_buffer("#add_system \"%s/\"\n", path);
2437 }
2438
2439 struct token * preprocess(struct token *token)
2440 {
2441 preprocessing = 1;
2442 init_preprocessor();
2443 do_preprocess(&token);
2444
2445 // Drop all expressions from preprocessing, they're not used any more.
2446 // This is not true when we have multiple files, though ;/
2447 // clear_expression_alloc();
2448 preprocessing = 0;
2449
2450 return token;
2451 }
2452
2453 static int is_VA_ARGS_token(struct token *token)
2454 {
2455 return (token_type(token) == TOKEN_IDENT) &&
2456 (token->ident == &__VA_ARGS___ident);
2457 }
2458
2459 static void dump_macro(struct symbol *sym)
2460 {
2461 int nargs = sym->arglist ? sym->arglist->count.normal : 0;
2462 struct token *args[nargs];
2463 struct token *token;
2464
2465 printf("#define %s", show_ident(sym->ident));
2466 token = sym->arglist;
2467 if (token) {
2468 const char *sep = "";
2469 int narg = 0;
2470 putchar('(');
2471 for (; !eof_token(token); token = token->next) {
2472 if (token_type(token) == TOKEN_ARG_COUNT)
2473 continue;
2474 if (is_VA_ARGS_token(token))
2475 printf("%s...", sep);
2476 else
2477 printf("%s%s", sep, show_token(token));
2478 args[narg++] = token;
2479 sep = ",";
2480 }
2481 putchar(')');
2482 }
2483
2484 token = sym->expansion;
2485 while (token_type(token) != TOKEN_UNTAINT) {
2486 struct token *next = token->next;
2487 if (token->pos.whitespace)
2488 putchar(' ');
2489 switch (token_type(token)) {
2490 case TOKEN_CONCAT:
2491 printf("##");
2492 break;
2493 case TOKEN_STR_ARGUMENT:
2494 printf("#");
2495 /* fall-through */
2496 case TOKEN_QUOTED_ARGUMENT:
2497 case TOKEN_MACRO_ARGUMENT:
2498 token = args[token->argnum];
2499 /* fall-through */
2500 default:
2501 printf("%s", show_token(token));
2502 }
2503 token = next;
2504 }
2505 putchar('\n');
2506 }
2507
2508 void dump_macro_definitions(void)
2509 {
2510 struct ident *name;
2511
2512 FOR_EACH_PTR(macros, name) {
2513 struct symbol *sym = lookup_macro(name);
2514 if (sym)
2515 dump_macro(sym);
2516 } END_FOR_EACH_PTR(name);
2517 }
Static analysis for C