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 
54 #define INCLUDEPATHS 300
55 const char *includepath[INCLUDEPATHS+1] = {
56 	"",
57 	"/usr/include",
58 	"/usr/local/include",
59 	NULL
60 };
61 
62 static const char **quote_includepath = includepath;
63 static const char **angle_includepath = includepath + 1;
64 static const char **isys_includepath   = includepath + 1;
65 static const char **sys_includepath   = includepath + 1;
66 static const char **dirafter_includepath = includepath + 3;
67 
68 #define dirty_stream(stream)				\
69 	do {						\
70 		if (!stream->dirty) {			\
71 			stream->dirty = 1;		\
72 			if (!stream->ifndef)		\
73 				stream->protect = NULL;	\
74 		}					\
75 	} while(0)
76 
77 #define end_group(stream)					\
78 	do {							\
79 		if (stream->ifndef == stream->top_if) {		\
80 			stream->ifndef = NULL;			\
81 			if (!stream->dirty)			\
82 				stream->protect = NULL;		\
83 			else if (stream->protect)		\
84 				stream->dirty = 0;		\
85 		}						\
86 	} while(0)
87 
88 #define nesting_error(stream)		\
89 	do {				\
90 		stream->dirty = 1;	\
91 		stream->ifndef = NULL;	\
92 		stream->protect = NULL;	\
93 	} while(0)
94 
alloc_token(struct position * pos)95 static struct token *alloc_token(struct position *pos)
96 {
97 	struct token *token = __alloc_token(0);
98 
99 	token->pos.stream = pos->stream;
100 	token->pos.line = pos->line;
101 	token->pos.pos = pos->pos;
102 	token->pos.whitespace = 1;
103 	return token;
104 }
105 
106 /* Expand symbol 'sym' at '*list' */
107 static int expand(struct token **, struct symbol *);
108 
replace_with_string(struct token * token,const char * str)109 static void replace_with_string(struct token *token, const char *str)
110 {
111 	int size = strlen(str) + 1;
112 	struct string *s = __alloc_string(size);
113 
114 	s->length = size;
115 	memcpy(s->data, str, size);
116 	token_type(token) = TOKEN_STRING;
117 	token->string = s;
118 }
119 
replace_with_integer(struct token * token,unsigned int val)120 static void replace_with_integer(struct token *token, unsigned int val)
121 {
122 	char *buf = __alloc_bytes(11);
123 	sprintf(buf, "%u", val);
124 	token_type(token) = TOKEN_NUMBER;
125 	token->number = buf;
126 }
127 
lookup_macro(struct ident * ident)128 static struct symbol *lookup_macro(struct ident *ident)
129 {
130 	struct symbol *sym = lookup_symbol(ident, NS_MACRO | NS_UNDEF);
131 	if (sym && sym->namespace != NS_MACRO)
132 		sym = NULL;
133 	return sym;
134 }
135 
token_defined(struct token * token)136 static int token_defined(struct token *token)
137 {
138 	if (token_type(token) == TOKEN_IDENT) {
139 		struct symbol *sym = lookup_macro(token->ident);
140 		if (sym) {
141 			sym->used_in = file_scope;
142 			return 1;
143 		}
144 		return 0;
145 	}
146 
147 	sparse_error(token->pos, "expected preprocessor identifier");
148 	return 0;
149 }
150 
replace_with_bool(struct token * token,bool val)151 static void replace_with_bool(struct token *token, bool val)
152 {
153 	static const char *string[] = { "0", "1" };
154 
155 	token_type(token) = TOKEN_NUMBER;
156 	token->number = string[val];
157 }
158 
replace_with_defined(struct token * token)159 static void replace_with_defined(struct token *token)
160 {
161 	replace_with_bool(token, token_defined(token));
162 }
163 
replace_with_has_builtin(struct token * token)164 static void replace_with_has_builtin(struct token *token)
165 {
166 	struct symbol *sym = lookup_symbol(token->ident, NS_SYMBOL);
167 	replace_with_bool(token, sym && sym->builtin);
168 }
169 
replace_with_has_attribute(struct token * token)170 static void replace_with_has_attribute(struct token *token)
171 {
172 	struct symbol *sym = lookup_symbol(token->ident, NS_KEYWORD);
173 	replace_with_bool(token, sym && sym->op && sym->op->attribute);
174 }
175 
expand_line(struct token * token)176 static void expand_line(struct token *token)
177 {
178 	replace_with_integer(token, token->pos.line);
179 }
180 
expand_file(struct token * token)181 static void expand_file(struct token *token)
182 {
183 	replace_with_string(token, stream_name(token->pos.stream));
184 }
185 
expand_basefile(struct token * token)186 static void expand_basefile(struct token *token)
187 {
188 	replace_with_string(token, base_filename);
189 }
190 
191 static time_t t = 0;
expand_date(struct token * token)192 static void expand_date(struct token *token)
193 {
194 	static char buffer[12]; /* __DATE__: 3 + ' ' + 2 + ' ' + 4 + '\0' */
195 
196 	if (!t)
197 		time(&t);
198 	strftime(buffer, 12, "%b %e %Y", localtime(&t));
199 	replace_with_string(token, buffer);
200 }
201 
expand_time(struct token * token)202 static void expand_time(struct token *token)
203 {
204 	static char buffer[9]; /* __TIME__: 2 + ':' + 2 + ':' + 2 + '\0' */
205 
206 	if (!t)
207 		time(&t);
208 	strftime(buffer, 9, "%T", localtime(&t));
209 	replace_with_string(token, buffer);
210 }
211 
expand_counter(struct token * token)212 static void expand_counter(struct token *token)
213 {
214 	replace_with_integer(token, counter_macro++);
215 }
216 
expand_include_level(struct token * token)217 static void expand_include_level(struct token *token)
218 {
219 	replace_with_integer(token, include_level - 1);
220 }
221 
expand_one_symbol(struct token ** list)222 static int expand_one_symbol(struct token **list)
223 {
224 	struct token *token = *list;
225 	struct symbol *sym;
226 
227 	if (token->pos.noexpand)
228 		return 1;
229 
230 	sym = lookup_macro(token->ident);
231 	if (!sym)
232 		return 1;
233 	store_macro_pos(token);
234 	if (sym->expander) {
235 		sym->expander(token);
236 		return 1;
237 	} else {
238 		sym->used_in = file_scope;
239 		return expand(list, sym);
240 	}
241 }
242 
scan_next(struct token ** where)243 static inline struct token *scan_next(struct token **where)
244 {
245 	struct token *token = *where;
246 	if (token_type(token) != TOKEN_UNTAINT)
247 		return token;
248 	do {
249 		token->ident->tainted = 0;
250 		token = token->next;
251 	} while (token_type(token) == TOKEN_UNTAINT);
252 	*where = token;
253 	return token;
254 }
255 
expand_list(struct token ** list)256 static void expand_list(struct token **list)
257 {
258 	struct token *next;
259 	while (!eof_token(next = scan_next(list))) {
260 		if (token_type(next) != TOKEN_IDENT || expand_one_symbol(list))
261 			list = &next->next;
262 	}
263 }
264 
265 static void preprocessor_line(struct stream *stream, struct token **line);
266 
collect_arg(struct token * prev,int vararg,struct position * pos,int count)267 static struct token *collect_arg(struct token *prev, int vararg, struct position *pos, int count)
268 {
269 	struct stream *stream = input_streams + prev->pos.stream;
270 	struct token **p = &prev->next;
271 	struct token *next;
272 	int nesting = 0;
273 
274 	while (!eof_token(next = scan_next(p))) {
275 		if (next->pos.newline && match_op(next, '#')) {
276 			if (!next->pos.noexpand) {
277 				sparse_error(next->pos,
278 					     "directive in argument list");
279 				preprocessor_line(stream, p);
280 				__free_token(next);	/* Free the '#' token */
281 				continue;
282 			}
283 		}
284 		switch (token_type(next)) {
285 		case TOKEN_STREAMEND:
286 		case TOKEN_STREAMBEGIN:
287 			*p = &eof_token_entry;
288 			return next;
289 		case TOKEN_STRING:
290 		case TOKEN_WIDE_STRING:
291 			if (count > 1)
292 				next->string->immutable = 1;
293 			break;
294 		}
295 		if (false_nesting) {
296 			*p = next->next;
297 			__free_token(next);
298 			continue;
299 		}
300 		if (match_op(next, '(')) {
301 			nesting++;
302 		} else if (match_op(next, ')')) {
303 			if (!nesting--)
304 				break;
305 		} else if (match_op(next, ',') && !nesting && !vararg) {
306 			break;
307 		}
308 		next->pos.stream = pos->stream;
309 		next->pos.line = pos->line;
310 		next->pos.pos = pos->pos;
311 		p = &next->next;
312 	}
313 	*p = &eof_token_entry;
314 	return next;
315 }
316 
317 /*
318  * We store arglist as <counter> [arg1] <number of uses for arg1> ... eof
319  */
320 
321 struct arg {
322 	struct token *arg;
323 	struct token *expanded;
324 	struct token *str;
325 	int n_normal;
326 	int n_quoted;
327 	int n_str;
328 };
329 
collect_arguments(struct token * start,struct token * arglist,struct arg * args,struct token * what)330 static int collect_arguments(struct token *start, struct token *arglist, struct arg *args, struct token *what)
331 {
332 	int wanted = arglist->count.normal;
333 	struct token *next = NULL;
334 	int count = 0;
335 
336 	arglist = arglist->next;	/* skip counter */
337 
338 	if (!wanted) {
339 		next = collect_arg(start, 0, &what->pos, 0);
340 		if (eof_token(next))
341 			goto Eclosing;
342 		if (!eof_token(start->next) || !match_op(next, ')')) {
343 			count++;
344 			goto Emany;
345 		}
346 	} else {
347 		for (count = 0; count < wanted; count++) {
348 			struct argcount *p = &arglist->next->count;
349 			next = collect_arg(start, p->vararg, &what->pos, p->normal);
350 			if (eof_token(next))
351 				goto Eclosing;
352 			if (p->vararg && wanted == 1 && eof_token(start->next))
353 				break;
354 			arglist = arglist->next->next;
355 			args[count].arg = start->next;
356 			args[count].n_normal = p->normal;
357 			args[count].n_quoted = p->quoted;
358 			args[count].n_str = p->str;
359 			if (match_op(next, ')')) {
360 				count++;
361 				break;
362 			}
363 			start = next;
364 		}
365 		if (count == wanted && !match_op(next, ')'))
366 			goto Emany;
367 		if (count == wanted - 1) {
368 			struct argcount *p = &arglist->next->count;
369 			if (!p->vararg)
370 				goto Efew;
371 			args[count].arg = NULL;
372 			args[count].n_normal = p->normal;
373 			args[count].n_quoted = p->quoted;
374 			args[count].n_str = p->str;
375 		}
376 		if (count < wanted - 1)
377 			goto Efew;
378 	}
379 	what->next = next->next;
380 	return 1;
381 
382 Efew:
383 	sparse_error(what->pos, "macro \"%s\" requires %d arguments, but only %d given",
384 		show_token(what), wanted, count);
385 	goto out;
386 Emany:
387 	while (match_op(next, ',')) {
388 		next = collect_arg(next, 0, &what->pos, 0);
389 		count++;
390 	}
391 	if (eof_token(next))
392 		goto Eclosing;
393 	sparse_error(what->pos, "macro \"%s\" passed %d arguments, but takes just %d",
394 		show_token(what), count, wanted);
395 	goto out;
396 Eclosing:
397 	sparse_error(what->pos, "unterminated argument list invoking macro \"%s\"",
398 		show_token(what));
399 out:
400 	what->next = next->next;
401 	return 0;
402 }
403 
dup_list(struct token * list)404 static struct token *dup_list(struct token *list)
405 {
406 	struct token *res = NULL;
407 	struct token **p = &res;
408 
409 	while (!eof_token(list)) {
410 		struct token *newtok = __alloc_token(0);
411 		*newtok = *list;
412 		*p = newtok;
413 		p = &newtok->next;
414 		list = list->next;
415 	}
416 	return res;
417 }
418 
show_token_sequence(struct token * token,int quote)419 static const char *show_token_sequence(struct token *token, int quote)
420 {
421 	static char buffer[MAX_STRING];
422 	char *ptr = buffer;
423 	int whitespace = 0;
424 
425 	if (!token && !quote)
426 		return "<none>";
427 	while (!eof_token(token)) {
428 		const char *val = quote ? quote_token(token) : show_token(token);
429 		int len = strlen(val);
430 
431 		if (ptr + whitespace + len >= buffer + sizeof(buffer)) {
432 			sparse_error(token->pos, "too long token expansion");
433 			break;
434 		}
435 
436 		if (whitespace)
437 			*ptr++ = ' ';
438 		memcpy(ptr, val, len);
439 		ptr += len;
440 		token = token->next;
441 		whitespace = token->pos.whitespace;
442 	}
443 	*ptr = 0;
444 	return buffer;
445 }
446 
stringify(struct token * arg)447 static struct token *stringify(struct token *arg)
448 {
449 	const char *s = show_token_sequence(arg, 1);
450 	int size = strlen(s)+1;
451 	struct token *token = __alloc_token(0);
452 	struct string *string = __alloc_string(size);
453 
454 	memcpy(string->data, s, size);
455 	string->length = size;
456 	token->pos = arg->pos;
457 	token_type(token) = TOKEN_STRING;
458 	token->string = string;
459 	token->next = &eof_token_entry;
460 	return token;
461 }
462 
expand_arguments(int count,struct arg * args)463 static void expand_arguments(int count, struct arg *args)
464 {
465 	int i;
466 	for (i = 0; i < count; i++) {
467 		struct token *arg = args[i].arg;
468 		if (!arg)
469 			arg = &eof_token_entry;
470 		if (args[i].n_str)
471 			args[i].str = stringify(arg);
472 		if (args[i].n_normal) {
473 			if (!args[i].n_quoted) {
474 				args[i].expanded = arg;
475 				args[i].arg = NULL;
476 			} else if (eof_token(arg)) {
477 				args[i].expanded = arg;
478 			} else {
479 				args[i].expanded = dup_list(arg);
480 			}
481 			expand_list(&args[i].expanded);
482 		}
483 	}
484 }
485 
486 /*
487  * Possibly valid combinations:
488  *  - ident + ident -> ident
489  *  - ident + number -> ident unless number contains '.', '+' or '-'.
490  *  - 'L' + char constant -> wide char constant
491  *  - 'L' + string literal -> wide string literal
492  *  - number + number -> number
493  *  - number + ident -> number
494  *  - number + '.' -> number
495  *  - number + '+' or '-' -> number, if number used to end on [eEpP].
496  *  - '.' + number -> number, if number used to start with a digit.
497  *  - special + special -> either special or an error.
498  */
combine(struct token * left,struct token * right,char * p)499 static enum token_type combine(struct token *left, struct token *right, char *p)
500 {
501 	int len;
502 	enum token_type t1 = token_type(left), t2 = token_type(right);
503 
504 	if (t1 != TOKEN_IDENT && t1 != TOKEN_NUMBER && t1 != TOKEN_SPECIAL)
505 		return TOKEN_ERROR;
506 
507 	if (t1 == TOKEN_IDENT && left->ident == &L_ident) {
508 		if (t2 >= TOKEN_CHAR && t2 < TOKEN_WIDE_CHAR)
509 			return t2 + TOKEN_WIDE_CHAR - TOKEN_CHAR;
510 		if (t2 == TOKEN_STRING)
511 			return TOKEN_WIDE_STRING;
512 	}
513 
514 	if (t2 != TOKEN_IDENT && t2 != TOKEN_NUMBER && t2 != TOKEN_SPECIAL)
515 		return TOKEN_ERROR;
516 
517 	strcpy(p, show_token(left));
518 	strcat(p, show_token(right));
519 	len = strlen(p);
520 
521 	if (len >= 256)
522 		return TOKEN_ERROR;
523 
524 	if (t1 == TOKEN_IDENT) {
525 		if (t2 == TOKEN_SPECIAL)
526 			return TOKEN_ERROR;
527 		if (t2 == TOKEN_NUMBER && strpbrk(p, "+-."))
528 			return TOKEN_ERROR;
529 		return TOKEN_IDENT;
530 	}
531 
532 	if (t1 == TOKEN_NUMBER) {
533 		if (t2 == TOKEN_SPECIAL) {
534 			switch (right->special) {
535 			case '.':
536 				break;
537 			case '+': case '-':
538 				if (strchr("eEpP", p[len - 2]))
539 					break;
540 			default:
541 				return TOKEN_ERROR;
542 			}
543 		}
544 		return TOKEN_NUMBER;
545 	}
546 
547 	if (p[0] == '.' && isdigit((unsigned char)p[1]))
548 		return TOKEN_NUMBER;
549 
550 	return TOKEN_SPECIAL;
551 }
552 
merge(struct token * left,struct token * right)553 static int merge(struct token *left, struct token *right)
554 {
555 	static char buffer[512];
556 	enum token_type res = combine(left, right, buffer);
557 	int n;
558 
559 	switch (res) {
560 	case TOKEN_IDENT:
561 		left->ident = built_in_ident(buffer);
562 		left->pos.noexpand = 0;
563 		return 1;
564 
565 	case TOKEN_NUMBER:
566 		token_type(left) = TOKEN_NUMBER;	/* could be . + num */
567 		left->number = xstrdup(buffer);
568 		return 1;
569 
570 	case TOKEN_SPECIAL:
571 		if (buffer[2] && buffer[3])
572 			break;
573 		for (n = SPECIAL_BASE; n < SPECIAL_ARG_SEPARATOR; n++) {
574 			if (!memcmp(buffer, combinations[n-SPECIAL_BASE], 3)) {
575 				left->special = n;
576 				return 1;
577 			}
578 		}
579 		break;
580 
581 	case TOKEN_WIDE_CHAR:
582 	case TOKEN_WIDE_STRING:
583 		token_type(left) = res;
584 		left->pos.noexpand = 0;
585 		left->string = right->string;
586 		return 1;
587 
588 	case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
589 		token_type(left) = res;
590 		left->pos.noexpand = 0;
591 		memcpy(left->embedded, right->embedded, 4);
592 		return 1;
593 
594 	default:
595 		;
596 	}
597 	sparse_error(left->pos, "'##' failed: concatenation is not a valid token");
598 	return 0;
599 }
600 
dup_token(struct token * token,struct position * streampos)601 static struct token *dup_token(struct token *token, struct position *streampos)
602 {
603 	struct token *alloc = alloc_token(streampos);
604 	token_type(alloc) = token_type(token);
605 	alloc->pos.newline = token->pos.newline;
606 	alloc->pos.whitespace = token->pos.whitespace;
607 	alloc->number = token->number;
608 	alloc->pos.noexpand = token->pos.noexpand;
609 	return alloc;
610 }
611 
copy(struct token ** where,struct token * list,int * count)612 static struct token **copy(struct token **where, struct token *list, int *count)
613 {
614 	int need_copy = --*count;
615 	while (!eof_token(list)) {
616 		struct token *token;
617 		if (need_copy)
618 			token = dup_token(list, &list->pos);
619 		else
620 			token = list;
621 		if (token_type(token) == TOKEN_IDENT && token->ident->tainted)
622 			token->pos.noexpand = 1;
623 		*where = token;
624 		where = &token->next;
625 		list = list->next;
626 	}
627 	*where = &eof_token_entry;
628 	return where;
629 }
630 
handle_kludge(struct token ** p,struct arg * args)631 static int handle_kludge(struct token **p, struct arg *args)
632 {
633 	struct token *t = (*p)->next->next;
634 	while (1) {
635 		struct arg *v = &args[t->argnum];
636 		if (token_type(t->next) != TOKEN_CONCAT) {
637 			if (v->arg) {
638 				/* ignore the first ## */
639 				*p = (*p)->next;
640 				return 0;
641 			}
642 			/* skip the entire thing */
643 			*p = t;
644 			return 1;
645 		}
646 		if (v->arg && !eof_token(v->arg))
647 			return 0; /* no magic */
648 		t = t->next->next;
649 	}
650 }
651 
substitute(struct token ** list,struct token * body,struct arg * args)652 static struct token **substitute(struct token **list, struct token *body, struct arg *args)
653 {
654 	struct position *base_pos = &(*list)->pos;
655 	int *count;
656 	enum {Normal, Placeholder, Concat} state = Normal;
657 
658 	for (; !eof_token(body); body = body->next) {
659 		struct token *added, *arg;
660 		struct token **tail;
661 		struct token *t;
662 
663 		switch (token_type(body)) {
664 		case TOKEN_GNU_KLUDGE:
665 			/*
666 			 * GNU kludge: if we had <comma>##<vararg>, behaviour
667 			 * depends on whether we had enough arguments to have
668 			 * a vararg.  If we did, ## is just ignored.  Otherwise
669 			 * both , and ## are ignored.  Worse, there can be
670 			 * an arbitrary number of ##<arg> in between; if all of
671 			 * those are empty, we act as if they hadn't been there,
672 			 * otherwise we act as if the kludge didn't exist.
673 			 */
674 			t = body;
675 			if (handle_kludge(&body, args)) {
676 				if (state == Concat)
677 					state = Normal;
678 				else
679 					state = Placeholder;
680 				continue;
681 			}
682 			added = dup_token(t, base_pos);
683 			token_type(added) = TOKEN_SPECIAL;
684 			tail = &added->next;
685 			break;
686 
687 		case TOKEN_STR_ARGUMENT:
688 			arg = args[body->argnum].str;
689 			count = &args[body->argnum].n_str;
690 			goto copy_arg;
691 
692 		case TOKEN_QUOTED_ARGUMENT:
693 			arg = args[body->argnum].arg;
694 			count = &args[body->argnum].n_quoted;
695 			if (!arg || eof_token(arg)) {
696 				if (state == Concat)
697 					state = Normal;
698 				else
699 					state = Placeholder;
700 				continue;
701 			}
702 			goto copy_arg;
703 
704 		case TOKEN_MACRO_ARGUMENT:
705 			arg = args[body->argnum].expanded;
706 			count = &args[body->argnum].n_normal;
707 			if (eof_token(arg)) {
708 				state = Normal;
709 				continue;
710 			}
711 		copy_arg:
712 			tail = copy(&added, arg, count);
713 			added->pos.newline = body->pos.newline;
714 			added->pos.whitespace = body->pos.whitespace;
715 			break;
716 
717 		case TOKEN_CONCAT:
718 			if (state == Placeholder)
719 				state = Normal;
720 			else
721 				state = Concat;
722 			continue;
723 
724 		case TOKEN_IDENT:
725 			added = dup_token(body, base_pos);
726 			if (added->ident->tainted)
727 				added->pos.noexpand = 1;
728 			tail = &added->next;
729 			break;
730 
731 		default:
732 			added = dup_token(body, base_pos);
733 			tail = &added->next;
734 			break;
735 		}
736 
737 		/*
738 		 * if we got to doing real concatenation, we already have
739 		 * added something into the list, so containing_token() is OK.
740 		 */
741 		if (state == Concat && merge(containing_token(list), added)) {
742 			*list = added->next;
743 			if (tail != &added->next)
744 				list = tail;
745 		} else {
746 			*list = added;
747 			list = tail;
748 		}
749 		state = Normal;
750 	}
751 	*list = &eof_token_entry;
752 	return list;
753 }
754 
expand(struct token ** list,struct symbol * sym)755 static int expand(struct token **list, struct symbol *sym)
756 {
757 	struct token *last;
758 	struct token *token = *list;
759 	struct ident *expanding = token->ident;
760 	struct token **tail;
761 	int nargs = sym->arglist ? sym->arglist->count.normal : 0;
762 	struct arg args[nargs];
763 
764 	if (expanding->tainted) {
765 		token->pos.noexpand = 1;
766 		return 1;
767 	}
768 
769 	if (sym->arglist) {
770 		if (!match_op(scan_next(&token->next), '('))
771 			return 1;
772 		if (!collect_arguments(token->next, sym->arglist, args, token))
773 			return 1;
774 		expand_arguments(nargs, args);
775 	}
776 
777 	expanding->tainted = 1;
778 
779 	last = token->next;
780 	tail = substitute(list, sym->expansion, args);
781 	/*
782 	 * Note that it won't be eof - at least TOKEN_UNTAINT will be there.
783 	 * We still can lose the newline flag if the sucker expands to nothing,
784 	 * but the price of dealing with that is probably too high (we'd need
785 	 * to collect the flags during scan_next())
786 	 */
787 	(*list)->pos.newline = token->pos.newline;
788 	(*list)->pos.whitespace = token->pos.whitespace;
789 	*tail = last;
790 
791 	return 0;
792 }
793 
token_name_sequence(struct token * token,int endop,struct token * start)794 static const char *token_name_sequence(struct token *token, int endop, struct token *start)
795 {
796 	static char buffer[256];
797 	char *ptr = buffer;
798 
799 	while (!eof_token(token) && !match_op(token, endop)) {
800 		int len;
801 		const char *val = token->string->data;
802 		if (token_type(token) != TOKEN_STRING)
803 			val = show_token(token);
804 		len = strlen(val);
805 		memcpy(ptr, val, len);
806 		ptr += len;
807 		token = token->next;
808 	}
809 	*ptr = 0;
810 	if (endop && !match_op(token, endop))
811 		sparse_error(start->pos, "expected '>' at end of filename");
812 	return buffer;
813 }
814 
already_tokenized(const char * path)815 static int already_tokenized(const char *path)
816 {
817 	int stream, next;
818 
819 	for (stream = *hash_stream(path); stream >= 0 ; stream = next) {
820 		struct stream *s = input_streams + stream;
821 
822 		next = s->next_stream;
823 		if (s->once) {
824 			if (strcmp(path, s->name))
825 				continue;
826 			return 1;
827 		}
828 		if (s->constant != CONSTANT_FILE_YES)
829 			continue;
830 		if (strcmp(path, s->name))
831 			continue;
832 		if (s->protect && !lookup_macro(s->protect))
833 			continue;
834 		return 1;
835 	}
836 	return 0;
837 }
838 
839 /* Handle include of header files.
840  * The relevant options are made compatible with gcc. The only options that
841  * are not supported is -withprefix and friends.
842  *
843  * Three set of include paths are known:
844  * quote_includepath:	Path to search when using #include "file.h"
845  * angle_includepath:	Paths to search when using #include <file.h>
846  * isys_includepath:	Paths specified with -isystem, come before the
847  *			built-in system include paths. Gcc would suppress
848  *			warnings from system headers. Here we separate
849  *			them from the angle_ ones to keep search ordering.
850  *
851  * sys_includepath:	Built-in include paths.
852  * dirafter_includepath Paths added with -dirafter.
853  *
854  * The above is implemented as one array with pointers
855  *                         +--------------+
856  * quote_includepath --->  |              |
857  *                         +--------------+
858  *                         |              |
859  *                         +--------------+
860  * angle_includepath --->  |              |
861  *                         +--------------+
862  * isys_includepath  --->  |              |
863  *                         +--------------+
864  * sys_includepath   --->  |              |
865  *                         +--------------+
866  * dirafter_includepath -> |              |
867  *                         +--------------+
868  *
869  * -I dir insert dir just before isys_includepath and move the rest
870  * -I- makes all dirs specified with -I before to quote dirs only and
871  *   angle_includepath is set equal to isys_includepath.
872  * -nostdinc removes all sys dirs by storing NULL in entry pointed
873  *   to by * sys_includepath. Note that this will reset all dirs built-in
874  *   and added before -nostdinc by -isystem and -idirafter.
875  * -isystem dir adds dir where isys_includepath points adding this dir as
876  *   first systemdir
877  * -idirafter dir adds dir to the end of the list
878  */
879 
set_stream_include_path(struct stream * stream)880 static void set_stream_include_path(struct stream *stream)
881 {
882 	const char *path = stream->path;
883 	if (!path) {
884 		const char *p = strrchr(stream->name, '/');
885 		path = "";
886 		if (p) {
887 			int len = p - stream->name + 1;
888 			char *m = malloc(len+1);
889 			/* This includes the final "/" */
890 			memcpy(m, stream->name, len);
891 			m[len] = 0;
892 			path = m;
893 		}
894 		stream->path = path;
895 	}
896 	includepath[0] = path;
897 }
898 
899 #ifndef PATH_MAX
900 #define PATH_MAX 4096	// for Hurd where it's not defined
901 #endif
902 
try_include(const char * path,const char * filename,int flen,struct token ** where,const char ** next_path)903 static int try_include(const char *path, const char *filename, int flen, struct token **where, const char **next_path)
904 {
905 	int fd;
906 	int plen = strlen(path);
907 	static char fullname[PATH_MAX];
908 
909 	memcpy(fullname, path, plen);
910 	if (plen && path[plen-1] != '/') {
911 		fullname[plen] = '/';
912 		plen++;
913 	}
914 	memcpy(fullname+plen, filename, flen);
915 	if (already_tokenized(fullname))
916 		return 1;
917 	fd = open(fullname, O_RDONLY);
918 	if (fd >= 0) {
919 		char *streamname = xmemdup(fullname, plen + flen);
920 		*where = tokenize(streamname, fd, *where, next_path);
921 		close(fd);
922 		return 1;
923 	}
924 	return 0;
925 }
926 
do_include_path(const char ** pptr,struct token ** list,struct token * token,const char * filename,int flen)927 static int do_include_path(const char **pptr, struct token **list, struct token *token, const char *filename, int flen)
928 {
929 	const char *path;
930 
931 	while ((path = *pptr++) != NULL) {
932 		if (!try_include(path, filename, flen, list, pptr))
933 			continue;
934 		return 1;
935 	}
936 	return 0;
937 }
938 
free_preprocessor_line(struct token * token)939 static int free_preprocessor_line(struct token *token)
940 {
941 	while (token_type(token) != TOKEN_EOF) {
942 		struct token *free = token;
943 		token = token->next;
944 		__free_token(free);
945 	};
946 	return 1;
947 }
948 
find_include(const char * skip,const char * look_for)949 const char *find_include(const char *skip, const char *look_for)
950 {
951 	DIR *dp;
952 	struct dirent *entry;
953 	struct stat statbuf;
954 	const char *ret;
955 	char cwd[PATH_MAX];
956 	static char buf[PATH_MAX + 1];
957 
958 	dp = opendir(".");
959 	if (!dp)
960 		return NULL;
961 
962 	if (!getcwd(cwd, sizeof(cwd)))
963 		goto close;
964 
965 	while ((entry = readdir(dp))) {
966 		lstat(entry->d_name, &statbuf);
967 
968 		if (strcmp(entry->d_name, look_for) == 0) {
969 			snprintf(buf, sizeof(buf), "%s/%s", cwd, entry->d_name);
970 			closedir(dp);
971 			return buf;
972 		}
973 
974 		if (S_ISDIR(statbuf.st_mode)) {
975 			/* Found a directory, but ignore . and .. */
976 			if (strcmp(".", entry->d_name) == 0 ||
977 			    strcmp("..", entry->d_name) == 0 ||
978 			    strcmp(skip, entry->d_name) == 0)
979 				continue;
980 
981 			chdir(entry->d_name);
982 			ret = find_include("", look_for);
983 			chdir("..");
984 			if (ret) {
985 				closedir(dp);
986 				return ret;
987 			}
988 		}
989 	}
990 close:
991 	closedir(dp);
992 
993 	return NULL;
994 }
995 
search_dir(const char * stop,const char * look_for)996 const char *search_dir(const char *stop, const char *look_for)
997 {
998 	char cwd[PATH_MAX];
999 	int len;
1000 	const char *ret;
1001 	int cnt = 0;
1002 
1003 	if (!getcwd(cwd, sizeof(cwd)))
1004 		return NULL;
1005 
1006 	len = strlen(cwd);
1007 	while (len >= 0) {
1008 		ret = find_include(cnt++ ? cwd + len + 1 : "", look_for);
1009 		if (ret)
1010 			return ret;
1011 
1012 		if (strcmp(cwd, stop) == 0 ||
1013 		    strcmp(cwd, "/usr/include") == 0 ||
1014 		    strcmp(cwd, "/usr/local/include") == 0 ||
1015 		    strlen(cwd) <= 10 ||  /* heck...  don't search /usr/lib/ */
1016 		    strcmp(cwd, "/") == 0)
1017 			return NULL;
1018 
1019 		while (--len >= 0) {
1020 			if (cwd[len] == '/') {
1021 				cwd[len] = '\0';
1022 				break;
1023 			}
1024 		}
1025 
1026 		chdir("..");
1027 	}
1028 	return NULL;
1029 }
1030 
use_best_guess_header_file(struct token * token,const char * filename,struct token ** list)1031 static void use_best_guess_header_file(struct token *token, const char *filename, struct token **list)
1032 {
1033 	char cwd[PATH_MAX];
1034 	char dir_part[PATH_MAX];
1035 	const char *file_part;
1036 	const char *include_name;
1037 	static int cnt;
1038 	int len;
1039 
1040 	/* Avoid guessing includes recursively. */
1041 	if (cnt++ > 1000)
1042 		return;
1043 
1044 	if (!filename || filename[0] == '\0')
1045 		return;
1046 
1047 	file_part = filename;
1048 	while ((filename = strchr(filename, '/'))) {
1049 		++filename;
1050 		if (filename[0])
1051 			file_part = filename;
1052 	}
1053 
1054 	snprintf(dir_part, sizeof(dir_part), "%s", stream_name(token->pos.stream));
1055 	len = strlen(dir_part);
1056 	while (--len >= 0) {
1057 		if (dir_part[len] == '/') {
1058 			dir_part[len] = '\0';
1059 			break;
1060 		}
1061 	}
1062 	if (len < 0)
1063 		sprintf(dir_part, ".");
1064 
1065 	if (!getcwd(cwd, sizeof(cwd)))
1066 		return;
1067 
1068 	chdir(dir_part);
1069 	include_name = search_dir(cwd, file_part);
1070 	chdir(cwd);
1071 	if (!include_name)
1072 		return;
1073 	sparse_error(token->pos, "using '%s'", include_name);
1074 
1075 	try_include("", include_name, strlen(include_name), list, includepath);
1076 }
1077 
handle_include_path(struct stream * stream,struct token ** list,struct token * token,int how)1078 static int handle_include_path(struct stream *stream, struct token **list, struct token *token, int how)
1079 {
1080 	const char *filename;
1081 	struct token *next;
1082 	const char **path;
1083 	int expect;
1084 	int flen;
1085 
1086 	next = token->next;
1087 	expect = '>';
1088 	if (!match_op(next, '<')) {
1089 		expand_list(&token->next);
1090 		expect = 0;
1091 		next = token;
1092 		if (match_op(token->next, '<')) {
1093 			next = token->next;
1094 			expect = '>';
1095 		}
1096 	}
1097 
1098 	token = next->next;
1099 	filename = token_name_sequence(token, expect, token);
1100 	flen = strlen(filename) + 1;
1101 
1102 	/* Absolute path? */
1103 	if (filename[0] == '/') {
1104 		if (try_include("", filename, flen, list, includepath))
1105 			return 0;
1106 		goto out;
1107 	}
1108 
1109 	switch (how) {
1110 	case 1:
1111 		path = stream->next_path;
1112 		break;
1113 	case 2:
1114 		includepath[0] = "";
1115 		path = includepath;
1116 		break;
1117 	default:
1118 		/* Dir of input file is first dir to search for quoted includes */
1119 		set_stream_include_path(stream);
1120 		path = expect ? angle_includepath : quote_includepath;
1121 		break;
1122 	}
1123 	/* Check the standard include paths.. */
1124 	if (do_include_path(path, list, token, filename, flen))
1125 		return 0;
1126 out:
1127 	sparse_error(token->pos, "unable to open '%s'", filename);
1128 	use_best_guess_header_file(token, filename, list);
1129 	return 0;
1130 }
1131 
handle_include(struct stream * stream,struct token ** list,struct token * token)1132 static int handle_include(struct stream *stream, struct token **list, struct token *token)
1133 {
1134 	return handle_include_path(stream, list, token, 0);
1135 }
1136 
handle_include_next(struct stream * stream,struct token ** list,struct token * token)1137 static int handle_include_next(struct stream *stream, struct token **list, struct token *token)
1138 {
1139 	return handle_include_path(stream, list, token, 1);
1140 }
1141 
handle_argv_include(struct stream * stream,struct token ** list,struct token * token)1142 static int handle_argv_include(struct stream *stream, struct token **list, struct token *token)
1143 {
1144 	return handle_include_path(stream, list, token, 2);
1145 }
1146 
token_different(struct token * t1,struct token * t2)1147 static int token_different(struct token *t1, struct token *t2)
1148 {
1149 	int different;
1150 
1151 	if (token_type(t1) != token_type(t2))
1152 		return 1;
1153 
1154 	switch (token_type(t1)) {
1155 	case TOKEN_IDENT:
1156 		different = t1->ident != t2->ident;
1157 		break;
1158 	case TOKEN_ARG_COUNT:
1159 	case TOKEN_UNTAINT:
1160 	case TOKEN_CONCAT:
1161 	case TOKEN_GNU_KLUDGE:
1162 		different = 0;
1163 		break;
1164 	case TOKEN_NUMBER:
1165 		different = strcmp(t1->number, t2->number);
1166 		break;
1167 	case TOKEN_SPECIAL:
1168 		different = t1->special != t2->special;
1169 		break;
1170 	case TOKEN_MACRO_ARGUMENT:
1171 	case TOKEN_QUOTED_ARGUMENT:
1172 	case TOKEN_STR_ARGUMENT:
1173 		different = t1->argnum != t2->argnum;
1174 		break;
1175 	case TOKEN_CHAR_EMBEDDED_0 ... TOKEN_CHAR_EMBEDDED_3:
1176 	case TOKEN_WIDE_CHAR_EMBEDDED_0 ... TOKEN_WIDE_CHAR_EMBEDDED_3:
1177 		different = memcmp(t1->embedded, t2->embedded, 4);
1178 		break;
1179 	case TOKEN_CHAR:
1180 	case TOKEN_WIDE_CHAR:
1181 	case TOKEN_STRING:
1182 	case TOKEN_WIDE_STRING: {
1183 		struct string *s1, *s2;
1184 
1185 		s1 = t1->string;
1186 		s2 = t2->string;
1187 		different = 1;
1188 		if (s1->length != s2->length)
1189 			break;
1190 		different = memcmp(s1->data, s2->data, s1->length);
1191 		break;
1192 	}
1193 	default:
1194 		different = 1;
1195 		break;
1196 	}
1197 	return different;
1198 }
1199 
token_list_different(struct token * list1,struct token * list2)1200 static int token_list_different(struct token *list1, struct token *list2)
1201 {
1202 	for (;;) {
1203 		if (list1 == list2)
1204 			return 0;
1205 		if (!list1 || !list2)
1206 			return 1;
1207 		if (token_different(list1, list2))
1208 			return 1;
1209 		list1 = list1->next;
1210 		list2 = list2->next;
1211 	}
1212 }
1213 
set_arg_count(struct token * token)1214 static inline void set_arg_count(struct token *token)
1215 {
1216 	token_type(token) = TOKEN_ARG_COUNT;
1217 	token->count.normal = token->count.quoted =
1218 	token->count.str = token->count.vararg = 0;
1219 }
1220 
parse_arguments(struct token * list)1221 static struct token *parse_arguments(struct token *list)
1222 {
1223 	struct token *arg = list->next, *next = list;
1224 	struct argcount *count = &list->count;
1225 
1226 	set_arg_count(list);
1227 
1228 	if (match_op(arg, ')')) {
1229 		next = arg->next;
1230 		list->next = &eof_token_entry;
1231 		return next;
1232 	}
1233 
1234 	while (token_type(arg) == TOKEN_IDENT) {
1235 		if (arg->ident == &__VA_ARGS___ident)
1236 			goto Eva_args;
1237 		if (!++count->normal)
1238 			goto Eargs;
1239 		next = arg->next;
1240 
1241 		if (match_op(next, ',')) {
1242 			set_arg_count(next);
1243 			arg = next->next;
1244 			continue;
1245 		}
1246 
1247 		if (match_op(next, ')')) {
1248 			set_arg_count(next);
1249 			next = next->next;
1250 			arg->next->next = &eof_token_entry;
1251 			return next;
1252 		}
1253 
1254 		/* normal cases are finished here */
1255 
1256 		if (match_op(next, SPECIAL_ELLIPSIS)) {
1257 			if (match_op(next->next, ')')) {
1258 				set_arg_count(next);
1259 				next->count.vararg = 1;
1260 				next = next->next;
1261 				arg->next->next = &eof_token_entry;
1262 				return next->next;
1263 			}
1264 
1265 			arg = next;
1266 			goto Enotclosed;
1267 		}
1268 
1269 		if (eof_token(next)) {
1270 			goto Enotclosed;
1271 		} else {
1272 			arg = next;
1273 			goto Ebadstuff;
1274 		}
1275 	}
1276 
1277 	if (match_op(arg, SPECIAL_ELLIPSIS)) {
1278 		next = arg->next;
1279 		token_type(arg) = TOKEN_IDENT;
1280 		arg->ident = &__VA_ARGS___ident;
1281 		if (!match_op(next, ')'))
1282 			goto Enotclosed;
1283 		if (!++count->normal)
1284 			goto Eargs;
1285 		set_arg_count(next);
1286 		next->count.vararg = 1;
1287 		next = next->next;
1288 		arg->next->next = &eof_token_entry;
1289 		return next;
1290 	}
1291 
1292 	if (eof_token(arg)) {
1293 		arg = next;
1294 		goto Enotclosed;
1295 	}
1296 	if (match_op(arg, ','))
1297 		goto Emissing;
1298 	else
1299 		goto Ebadstuff;
1300 
1301 
1302 Emissing:
1303 	sparse_error(arg->pos, "parameter name missing");
1304 	return NULL;
1305 Ebadstuff:
1306 	sparse_error(arg->pos, "\"%s\" may not appear in macro parameter list",
1307 		show_token(arg));
1308 	return NULL;
1309 Enotclosed:
1310 	sparse_error(arg->pos, "missing ')' in macro parameter list");
1311 	return NULL;
1312 Eva_args:
1313 	sparse_error(arg->pos, "__VA_ARGS__ can only appear in the expansion of a C99 variadic macro");
1314 	return NULL;
1315 Eargs:
1316 	sparse_error(arg->pos, "too many arguments in macro definition");
1317 	return NULL;
1318 }
1319 
try_arg(struct token * token,enum token_type type,struct token * arglist)1320 static int try_arg(struct token *token, enum token_type type, struct token *arglist)
1321 {
1322 	struct ident *ident = token->ident;
1323 	int nr;
1324 
1325 	if (!arglist || token_type(token) != TOKEN_IDENT)
1326 		return 0;
1327 
1328 	arglist = arglist->next;
1329 
1330 	for (nr = 0; !eof_token(arglist); nr++, arglist = arglist->next->next) {
1331 		if (arglist->ident == ident) {
1332 			struct argcount *count = &arglist->next->count;
1333 			int n;
1334 
1335 			token->argnum = nr;
1336 			token_type(token) = type;
1337 			switch (type) {
1338 			case TOKEN_MACRO_ARGUMENT:
1339 				n = ++count->normal;
1340 				break;
1341 			case TOKEN_QUOTED_ARGUMENT:
1342 				n = ++count->quoted;
1343 				break;
1344 			default:
1345 				n = ++count->str;
1346 			}
1347 			if (n)
1348 				return count->vararg ? 2 : 1;
1349 			/*
1350 			 * XXX - need saner handling of that
1351 			 * (>= 1024 instances of argument)
1352 			 */
1353 			token_type(token) = TOKEN_ERROR;
1354 			return -1;
1355 		}
1356 	}
1357 	return 0;
1358 }
1359 
handle_hash(struct token ** p,struct token * arglist)1360 static struct token *handle_hash(struct token **p, struct token *arglist)
1361 {
1362 	struct token *token = *p;
1363 	if (arglist) {
1364 		struct token *next = token->next;
1365 		if (!try_arg(next, TOKEN_STR_ARGUMENT, arglist))
1366 			goto Equote;
1367 		next->pos.whitespace = token->pos.whitespace;
1368 		__free_token(token);
1369 		token = *p = next;
1370 	} else {
1371 		token->pos.noexpand = 1;
1372 	}
1373 	return token;
1374 
1375 Equote:
1376 	sparse_error(token->pos, "'#' is not followed by a macro parameter");
1377 	return NULL;
1378 }
1379 
1380 /* token->next is ## */
handle_hashhash(struct token * token,struct token * arglist)1381 static struct token *handle_hashhash(struct token *token, struct token *arglist)
1382 {
1383 	struct token *last = token;
1384 	struct token *concat;
1385 	int state = match_op(token, ',');
1386 
1387 	try_arg(token, TOKEN_QUOTED_ARGUMENT, arglist);
1388 
1389 	while (1) {
1390 		struct token *t;
1391 		int is_arg;
1392 
1393 		/* eat duplicate ## */
1394 		concat = token->next;
1395 		while (match_op(t = concat->next, SPECIAL_HASHHASH)) {
1396 			token->next = t;
1397 			__free_token(concat);
1398 			concat = t;
1399 		}
1400 		token_type(concat) = TOKEN_CONCAT;
1401 
1402 		if (eof_token(t))
1403 			goto Econcat;
1404 
1405 		if (match_op(t, '#')) {
1406 			t = handle_hash(&concat->next, arglist);
1407 			if (!t)
1408 				return NULL;
1409 		}
1410 
1411 		is_arg = try_arg(t, TOKEN_QUOTED_ARGUMENT, arglist);
1412 
1413 		if (state == 1 && is_arg) {
1414 			state = is_arg;
1415 		} else {
1416 			last = t;
1417 			state = match_op(t, ',');
1418 		}
1419 
1420 		token = t;
1421 		if (!match_op(token->next, SPECIAL_HASHHASH))
1422 			break;
1423 	}
1424 	/* handle GNU ,##__VA_ARGS__ kludge, in all its weirdness */
1425 	if (state == 2)
1426 		token_type(last) = TOKEN_GNU_KLUDGE;
1427 	return token;
1428 
1429 Econcat:
1430 	sparse_error(concat->pos, "'##' cannot appear at the ends of macro expansion");
1431 	return NULL;
1432 }
1433 
parse_expansion(struct token * expansion,struct token * arglist,struct ident * name)1434 static struct token *parse_expansion(struct token *expansion, struct token *arglist, struct ident *name)
1435 {
1436 	struct token *token = expansion;
1437 	struct token **p;
1438 
1439 	if (match_op(token, SPECIAL_HASHHASH))
1440 		goto Econcat;
1441 
1442 	for (p = &expansion; !eof_token(token); p = &token->next, token = *p) {
1443 		if (match_op(token, '#')) {
1444 			token = handle_hash(p, arglist);
1445 			if (!token)
1446 				return NULL;
1447 		}
1448 		if (match_op(token->next, SPECIAL_HASHHASH)) {
1449 			token = handle_hashhash(token, arglist);
1450 			if (!token)
1451 				return NULL;
1452 		} else {
1453 			try_arg(token, TOKEN_MACRO_ARGUMENT, arglist);
1454 		}
1455 		switch (token_type(token)) {
1456 		case TOKEN_ERROR:
1457 			goto Earg;
1458 
1459 		case TOKEN_STRING:
1460 		case TOKEN_WIDE_STRING:
1461 			token->string->immutable = 1;
1462 			break;
1463 		}
1464 	}
1465 	token = alloc_token(&expansion->pos);
1466 	token_type(token) = TOKEN_UNTAINT;
1467 	token->ident = name;
1468 	token->next = *p;
1469 	*p = token;
1470 	return expansion;
1471 
1472 Econcat:
1473 	sparse_error(token->pos, "'##' cannot appear at the ends of macro expansion");
1474 	return NULL;
1475 Earg:
1476 	sparse_error(token->pos, "too many instances of argument in body");
1477 	return NULL;
1478 }
1479 
do_define(struct position pos,struct token * token,struct ident * name,struct token * arglist,struct token * expansion,int attr)1480 static int do_define(struct position pos, struct token *token, struct ident *name,
1481 		     struct token *arglist, struct token *expansion, int attr)
1482 {
1483 	struct symbol *sym;
1484 	int ret = 1;
1485 
1486 	expansion = parse_expansion(expansion, arglist, name);
1487 	if (!expansion)
1488 		return 1;
1489 
1490 	sym = lookup_symbol(name, NS_MACRO | NS_UNDEF);
1491 	if (sym) {
1492 		int clean;
1493 
1494 		if (attr < sym->attr)
1495 			goto out;
1496 
1497 		clean = (attr == sym->attr && sym->namespace == NS_MACRO);
1498 
1499 		if (token_list_different(sym->expansion, expansion) ||
1500 		    token_list_different(sym->arglist, arglist)) {
1501 			ret = 0;
1502 			if ((clean && attr == SYM_ATTR_NORMAL)
1503 					|| sym->used_in == file_scope) {
1504 				warning(pos, "preprocessor token %.*s redefined",
1505 						name->len, name->name);
1506 				info(sym->pos, "this was the original definition");
1507 			}
1508 		} else if (clean)
1509 			goto out;
1510 	}
1511 
1512 	if (!sym || sym->scope != file_scope) {
1513 		sym = alloc_symbol(pos, SYM_NODE);
1514 		bind_symbol(sym, name, NS_MACRO);
1515 		add_ident(&macros, name);
1516 		ret = 0;
1517 	}
1518 
1519 	if (!ret) {
1520 		sym->expansion = expansion;
1521 		sym->arglist = arglist;
1522 		if (token) /* Free the "define" token, but not the rest of the line */
1523 			__free_token(token);
1524 	}
1525 
1526 	sym->namespace = NS_MACRO;
1527 	sym->used_in = NULL;
1528 	sym->attr = attr;
1529 out:
1530 	return ret;
1531 }
1532 
1533 ///
1534 // predefine a macro with a printf-formatted value
1535 // @name: the name of the macro
1536 // @weak: 0/1 for a normal or a weak define
1537 // @fmt: the printf format followed by it's arguments.
1538 //
1539 // The type of the value is automatically infered:
1540 // TOKEN_NUMBER if it starts by a digit, TOKEN_IDENT otherwise.
1541 // If @fmt is null or empty, the macro is defined with an empty definition.
predefine(const char * name,int weak,const char * fmt,...)1542 void predefine(const char *name, int weak, const char *fmt, ...)
1543 {
1544 	struct ident *ident = built_in_ident(name);
1545 	struct token *value = &eof_token_entry;
1546 	int attr = weak ? SYM_ATTR_WEAK : SYM_ATTR_NORMAL;
1547 
1548 	if (fmt && fmt[0]) {
1549 		static char buf[256];
1550 		va_list ap;
1551 
1552 		va_start(ap, fmt);
1553 		vsnprintf(buf, sizeof(buf), fmt, ap);
1554 		va_end(ap);
1555 
1556 		value = __alloc_token(0);
1557 		if (isdigit(buf[0])) {
1558 			token_type(value) = TOKEN_NUMBER;
1559 			value->number = xstrdup(buf);
1560 		} else {
1561 			token_type(value) = TOKEN_IDENT;
1562 			value->ident = built_in_ident(buf);
1563 		}
1564 		value->pos.whitespace = 1;
1565 		value->next = &eof_token_entry;
1566 	}
1567 
1568 	do_define(value->pos, NULL, ident, NULL, value, attr);
1569 }
1570 
1571 ///
1572 // like predefine() but only if one of the non-standard dialect is chosen
predefine_nostd(const char * name)1573 void predefine_nostd(const char *name)
1574 {
1575 	if ((standard & STANDARD_GNU) || (standard == STANDARD_NONE))
1576 		predefine(name, 1, "1");
1577 }
1578 
do_handle_define(struct stream * stream,struct token ** line,struct token * token,int attr)1579 static int do_handle_define(struct stream *stream, struct token **line, struct token *token, int attr)
1580 {
1581 	struct token *arglist, *expansion;
1582 	struct token *left = token->next;
1583 	struct ident *name;
1584 
1585 	if (token_type(left) != TOKEN_IDENT) {
1586 		sparse_error(token->pos, "expected identifier to 'define'");
1587 		return 1;
1588 	}
1589 
1590 	name = left->ident;
1591 
1592 	arglist = NULL;
1593 	expansion = left->next;
1594 	if (!expansion->pos.whitespace) {
1595 		if (match_op(expansion, '(')) {
1596 			arglist = expansion;
1597 			expansion = parse_arguments(expansion);
1598 			if (!expansion)
1599 				return 1;
1600 		} else if (!eof_token(expansion)) {
1601 			warning(expansion->pos,
1602 				"no whitespace before object-like macro body");
1603 		}
1604 	}
1605 
1606 	return do_define(left->pos, token, name, arglist, expansion, attr);
1607 }
1608 
handle_define(struct stream * stream,struct token ** line,struct token * token)1609 static int handle_define(struct stream *stream, struct token **line, struct token *token)
1610 {
1611 	return do_handle_define(stream, line, token, SYM_ATTR_NORMAL);
1612 }
1613 
handle_weak_define(struct stream * stream,struct token ** line,struct token * token)1614 static int handle_weak_define(struct stream *stream, struct token **line, struct token *token)
1615 {
1616 	return do_handle_define(stream, line, token, SYM_ATTR_WEAK);
1617 }
1618 
handle_strong_define(struct stream * stream,struct token ** line,struct token * token)1619 static int handle_strong_define(struct stream *stream, struct token **line, struct token *token)
1620 {
1621 	return do_handle_define(stream, line, token, SYM_ATTR_STRONG);
1622 }
1623 
do_handle_undef(struct stream * stream,struct token ** line,struct token * token,int attr)1624 static int do_handle_undef(struct stream *stream, struct token **line, struct token *token, int attr)
1625 {
1626 	struct token *left = token->next;
1627 	struct symbol *sym;
1628 
1629 	if (token_type(left) != TOKEN_IDENT) {
1630 		sparse_error(token->pos, "expected identifier to 'undef'");
1631 		return 1;
1632 	}
1633 
1634 	sym = lookup_symbol(left->ident, NS_MACRO | NS_UNDEF);
1635 	if (sym) {
1636 		if (attr < sym->attr)
1637 			return 1;
1638 		if (attr == sym->attr && sym->namespace == NS_UNDEF)
1639 			return 1;
1640 	} else if (attr <= SYM_ATTR_NORMAL)
1641 		return 1;
1642 
1643 	if (!sym || sym->scope != file_scope) {
1644 		sym = alloc_symbol(left->pos, SYM_NODE);
1645 		bind_symbol(sym, left->ident, NS_MACRO);
1646 	}
1647 
1648 	sym->namespace = NS_UNDEF;
1649 	sym->used_in = NULL;
1650 	sym->attr = attr;
1651 
1652 	return 1;
1653 }
1654 
handle_undef(struct stream * stream,struct token ** line,struct token * token)1655 static int handle_undef(struct stream *stream, struct token **line, struct token *token)
1656 {
1657 	return do_handle_undef(stream, line, token, SYM_ATTR_NORMAL);
1658 }
1659 
handle_strong_undef(struct stream * stream,struct token ** line,struct token * token)1660 static int handle_strong_undef(struct stream *stream, struct token **line, struct token *token)
1661 {
1662 	return do_handle_undef(stream, line, token, SYM_ATTR_STRONG);
1663 }
1664 
preprocessor_if(struct stream * stream,struct token * token,int cond)1665 static int preprocessor_if(struct stream *stream, struct token *token, int cond)
1666 {
1667 	token_type(token) = false_nesting ? TOKEN_SKIP_GROUPS : TOKEN_IF;
1668 	free_preprocessor_line(token->next);
1669 	token->next = stream->top_if;
1670 	stream->top_if = token;
1671 	if (false_nesting || cond != 1)
1672 		false_nesting++;
1673 	return 0;
1674 }
1675 
handle_ifdef(struct stream * stream,struct token ** line,struct token * token)1676 static int handle_ifdef(struct stream *stream, struct token **line, struct token *token)
1677 {
1678 	struct token *next = token->next;
1679 	int arg;
1680 	if (token_type(next) == TOKEN_IDENT) {
1681 		arg = token_defined(next);
1682 	} else {
1683 		dirty_stream(stream);
1684 		if (!false_nesting)
1685 			sparse_error(token->pos, "expected preprocessor identifier");
1686 		arg = -1;
1687 	}
1688 	return preprocessor_if(stream, token, arg);
1689 }
1690 
handle_ifndef(struct stream * stream,struct token ** line,struct token * token)1691 static int handle_ifndef(struct stream *stream, struct token **line, struct token *token)
1692 {
1693 	struct token *next = token->next;
1694 	int arg;
1695 	if (token_type(next) == TOKEN_IDENT) {
1696 		if (!stream->dirty && !stream->ifndef) {
1697 			if (!stream->protect) {
1698 				stream->ifndef = token;
1699 				stream->protect = next->ident;
1700 			} else if (stream->protect == next->ident) {
1701 				stream->ifndef = token;
1702 				stream->dirty = 1;
1703 			}
1704 		}
1705 		arg = !token_defined(next);
1706 	} else {
1707 		dirty_stream(stream);
1708 		if (!false_nesting)
1709 			sparse_error(token->pos, "expected preprocessor identifier");
1710 		arg = -1;
1711 	}
1712 
1713 	return preprocessor_if(stream, token, arg);
1714 }
1715 
1716 static const char *show_token_sequence(struct token *token, int quote);
1717 
1718 /*
1719  * Expression handling for #if and #elif; it differs from normal expansion
1720  * due to special treatment of "defined".
1721  */
expression_value(struct token ** where)1722 static int expression_value(struct token **where)
1723 {
1724 	struct expression *expr;
1725 	struct token *p;
1726 	struct token **list = where, **beginning = NULL;
1727 	long long value;
1728 	int state = 0;
1729 
1730 	while (!eof_token(p = scan_next(list))) {
1731 		switch (state) {
1732 		case 0:
1733 			if (token_type(p) != TOKEN_IDENT)
1734 				break;
1735 			if (p->ident == &defined_ident) {
1736 				state = 1;
1737 				beginning = list;
1738 				break;
1739 			} else if (p->ident == &__has_builtin_ident) {
1740 				state = 4;
1741 				beginning = list;
1742 				break;
1743 			} else if (p->ident == &__has_attribute_ident) {
1744 				state = 6;
1745 				beginning = list;
1746 				break;
1747 			}
1748 			if (!expand_one_symbol(list))
1749 				continue;
1750 			if (token_type(p) != TOKEN_IDENT)
1751 				break;
1752 			token_type(p) = TOKEN_ZERO_IDENT;
1753 			break;
1754 		case 1:
1755 			if (match_op(p, '(')) {
1756 				state = 2;
1757 			} else {
1758 				state = 0;
1759 				replace_with_defined(p);
1760 				*beginning = p;
1761 			}
1762 			break;
1763 		case 2:
1764 			if (token_type(p) == TOKEN_IDENT)
1765 				state = 3;
1766 			else
1767 				state = 0;
1768 			replace_with_defined(p);
1769 			*beginning = p;
1770 			break;
1771 		case 3:
1772 			state = 0;
1773 			if (!match_op(p, ')'))
1774 				sparse_error(p->pos, "missing ')' after \"defined\"");
1775 			*list = p->next;
1776 			continue;
1777 
1778 		// __has_builtin(x) or __has_attribute(x)
1779 		case 4: case 6:
1780 			if (match_op(p, '(')) {
1781 				state++;
1782 			} else {
1783 				sparse_error(p->pos, "missing '(' after \"__has_%s\"",
1784 					state == 4 ? "builtin" : "attribute");
1785 				state = 0;
1786 			}
1787 			*beginning = p;
1788 			break;
1789 		case 5: case 7:
1790 			if (token_type(p) != TOKEN_IDENT) {
1791 				sparse_error(p->pos, "identifier expected");
1792 				state = 0;
1793 				break;
1794 			}
1795 			if (!match_op(p->next, ')'))
1796 				sparse_error(p->pos, "missing ')' after \"__has_%s\"",
1797 					state == 5 ? "builtin" : "attribute");
1798 			if (state == 5)
1799 				replace_with_has_builtin(p);
1800 			else
1801 				replace_with_has_attribute(p);
1802 			state = 8;
1803 			*beginning = p;
1804 			break;
1805 		case 8:
1806 			state = 0;
1807 			*list = p->next;
1808 			continue;
1809 		}
1810 		list = &p->next;
1811 	}
1812 
1813 	p = constant_expression(*where, &expr);
1814 	if (!eof_token(p))
1815 		sparse_error(p->pos, "garbage at end: %s", show_token_sequence(p, 0));
1816 	value = get_expression_value(expr);
1817 	return value != 0;
1818 }
1819 
handle_if(struct stream * stream,struct token ** line,struct token * token)1820 static int handle_if(struct stream *stream, struct token **line, struct token *token)
1821 {
1822 	int value = 0;
1823 	if (!false_nesting)
1824 		value = expression_value(&token->next);
1825 
1826 	dirty_stream(stream);
1827 	return preprocessor_if(stream, token, value);
1828 }
1829 
handle_elif(struct stream * stream,struct token ** line,struct token * token)1830 static int handle_elif(struct stream * stream, struct token **line, struct token *token)
1831 {
1832 	struct token *top_if = stream->top_if;
1833 	end_group(stream);
1834 
1835 	if (!top_if) {
1836 		nesting_error(stream);
1837 		sparse_error(token->pos, "unmatched #elif within stream");
1838 		return 1;
1839 	}
1840 
1841 	if (token_type(top_if) == TOKEN_ELSE) {
1842 		nesting_error(stream);
1843 		sparse_error(token->pos, "#elif after #else");
1844 		if (!false_nesting)
1845 			false_nesting = 1;
1846 		return 1;
1847 	}
1848 
1849 	dirty_stream(stream);
1850 	if (token_type(top_if) != TOKEN_IF)
1851 		return 1;
1852 	if (false_nesting) {
1853 		false_nesting = 0;
1854 		if (!expression_value(&token->next))
1855 			false_nesting = 1;
1856 	} else {
1857 		false_nesting = 1;
1858 		token_type(top_if) = TOKEN_SKIP_GROUPS;
1859 	}
1860 	return 1;
1861 }
1862 
handle_else(struct stream * stream,struct token ** line,struct token * token)1863 static int handle_else(struct stream *stream, struct token **line, struct token *token)
1864 {
1865 	struct token *top_if = stream->top_if;
1866 	end_group(stream);
1867 
1868 	if (!top_if) {
1869 		nesting_error(stream);
1870 		sparse_error(token->pos, "unmatched #else within stream");
1871 		return 1;
1872 	}
1873 
1874 	if (token_type(top_if) == TOKEN_ELSE) {
1875 		nesting_error(stream);
1876 		sparse_error(token->pos, "#else after #else");
1877 	}
1878 	if (false_nesting) {
1879 		if (token_type(top_if) == TOKEN_IF)
1880 			false_nesting = 0;
1881 	} else {
1882 		false_nesting = 1;
1883 	}
1884 	token_type(top_if) = TOKEN_ELSE;
1885 	return 1;
1886 }
1887 
handle_endif(struct stream * stream,struct token ** line,struct token * token)1888 static int handle_endif(struct stream *stream, struct token **line, struct token *token)
1889 {
1890 	struct token *top_if = stream->top_if;
1891 	end_group(stream);
1892 	if (!top_if) {
1893 		nesting_error(stream);
1894 		sparse_error(token->pos, "unmatched #endif in stream");
1895 		return 1;
1896 	}
1897 	if (false_nesting)
1898 		false_nesting--;
1899 	stream->top_if = top_if->next;
1900 	__free_token(top_if);
1901 	return 1;
1902 }
1903 
handle_warning(struct stream * stream,struct token ** line,struct token * token)1904 static int handle_warning(struct stream *stream, struct token **line, struct token *token)
1905 {
1906 	warning(token->pos, "%s", show_token_sequence(token->next, 0));
1907 	return 1;
1908 }
1909 
handle_error(struct stream * stream,struct token ** line,struct token * token)1910 static int handle_error(struct stream *stream, struct token **line, struct token *token)
1911 {
1912 	sparse_error(token->pos, "%s", show_token_sequence(token->next, 0));
1913 	return 1;
1914 }
1915 
handle_nostdinc(struct stream * stream,struct token ** line,struct token * token)1916 static int handle_nostdinc(struct stream *stream, struct token **line, struct token *token)
1917 {
1918 	/*
1919 	 * Do we have any non-system includes?
1920 	 * Clear them out if so..
1921 	 */
1922 	*sys_includepath = NULL;
1923 	return 1;
1924 }
1925 
update_inc_ptrs(const char *** where)1926 static inline void update_inc_ptrs(const char ***where)
1927 {
1928 
1929 	if (*where <= dirafter_includepath) {
1930 		dirafter_includepath++;
1931 		/* If this was the entry that we prepend, don't
1932 		 * rise the lower entries, even if they are at
1933 		 * the same level. */
1934 		if (where == &dirafter_includepath)
1935 			return;
1936 	}
1937 	if (*where <= sys_includepath) {
1938 		sys_includepath++;
1939 		if (where == &sys_includepath)
1940 			return;
1941 	}
1942 	if (*where <= isys_includepath) {
1943 		isys_includepath++;
1944 		if (where == &isys_includepath)
1945 			return;
1946 	}
1947 
1948 	/* angle_includepath is actually never updated, since we
1949 	 * don't suppport -iquote rught now. May change some day. */
1950 	if (*where <= angle_includepath) {
1951 		angle_includepath++;
1952 		if (where == &angle_includepath)
1953 			return;
1954 	}
1955 }
1956 
1957 /* Add a path before 'where' and update the pointers associated with the
1958  * includepath array */
add_path_entry(struct token * token,const char * path,const char *** where)1959 static void add_path_entry(struct token *token, const char *path,
1960 	const char ***where)
1961 {
1962 	const char **dst;
1963 	const char *next;
1964 
1965 	/* Need one free entry.. */
1966 	if (includepath[INCLUDEPATHS-2])
1967 		error_die(token->pos, "too many include path entries");
1968 
1969 	/* check that this is not a duplicate */
1970 	dst = includepath;
1971 	while (*dst) {
1972 		if (strcmp(*dst, path) == 0)
1973 			return;
1974 		dst++;
1975 	}
1976 	next = path;
1977 	dst = *where;
1978 
1979 	update_inc_ptrs(where);
1980 
1981 	/*
1982 	 * Move them all up starting at dst,
1983 	 * insert the new entry..
1984 	 */
1985 	do {
1986 		const char *tmp = *dst;
1987 		*dst = next;
1988 		next = tmp;
1989 		dst++;
1990 	} while (next);
1991 }
1992 
handle_add_include(struct stream * stream,struct token ** line,struct token * token)1993 static int handle_add_include(struct stream *stream, struct token **line, struct token *token)
1994 {
1995 	for (;;) {
1996 		token = token->next;
1997 		if (eof_token(token))
1998 			return 1;
1999 		if (token_type(token) != TOKEN_STRING) {
2000 			warning(token->pos, "expected path string");
2001 			return 1;
2002 		}
2003 		add_path_entry(token, token->string->data, &isys_includepath);
2004 	}
2005 }
2006 
handle_add_isystem(struct stream * stream,struct token ** line,struct token * token)2007 static int handle_add_isystem(struct stream *stream, struct token **line, struct token *token)
2008 {
2009 	for (;;) {
2010 		token = token->next;
2011 		if (eof_token(token))
2012 			return 1;
2013 		if (token_type(token) != TOKEN_STRING) {
2014 			sparse_error(token->pos, "expected path string");
2015 			return 1;
2016 		}
2017 		add_path_entry(token, token->string->data, &sys_includepath);
2018 	}
2019 }
2020 
handle_add_system(struct stream * stream,struct token ** line,struct token * token)2021 static int handle_add_system(struct stream *stream, struct token **line, struct token *token)
2022 {
2023 	for (;;) {
2024 		token = token->next;
2025 		if (eof_token(token))
2026 			return 1;
2027 		if (token_type(token) != TOKEN_STRING) {
2028 			sparse_error(token->pos, "expected path string");
2029 			return 1;
2030 		}
2031 		add_path_entry(token, token->string->data, &dirafter_includepath);
2032 	}
2033 }
2034 
2035 /* Add to end on includepath list - no pointer updates */
add_dirafter_entry(struct token * token,const char * path)2036 static void add_dirafter_entry(struct token *token, const char *path)
2037 {
2038 	const char **dst = includepath;
2039 
2040 	/* Need one free entry.. */
2041 	if (includepath[INCLUDEPATHS-2])
2042 		error_die(token->pos, "too many include path entries");
2043 
2044 	/* Add to the end */
2045 	while (*dst)
2046 		dst++;
2047 	*dst = path;
2048 	dst++;
2049 	*dst = NULL;
2050 }
2051 
handle_add_dirafter(struct stream * stream,struct token ** line,struct token * token)2052 static int handle_add_dirafter(struct stream *stream, struct token **line, struct token *token)
2053 {
2054 	for (;;) {
2055 		token = token->next;
2056 		if (eof_token(token))
2057 			return 1;
2058 		if (token_type(token) != TOKEN_STRING) {
2059 			sparse_error(token->pos, "expected path string");
2060 			return 1;
2061 		}
2062 		add_dirafter_entry(token, token->string->data);
2063 	}
2064 }
2065 
handle_split_include(struct stream * stream,struct token ** line,struct token * token)2066 static int handle_split_include(struct stream *stream, struct token **line, struct token *token)
2067 {
2068 	/*
2069 	 * -I-
2070 	 *  From info gcc:
2071 	 *  Split the include path.  Any directories specified with `-I'
2072 	 *  options before `-I-' are searched only for headers requested with
2073 	 *  `#include "FILE"'; they are not searched for `#include <FILE>'.
2074 	 *  If additional directories are specified with `-I' options after
2075 	 *  the `-I-', those directories are searched for all `#include'
2076 	 *  directives.
2077 	 *  In addition, `-I-' inhibits the use of the directory of the current
2078 	 *  file directory as the first search directory for `#include "FILE"'.
2079 	 */
2080 	quote_includepath = includepath+1;
2081 	angle_includepath = sys_includepath;
2082 	return 1;
2083 }
2084 
2085 /*
2086  * We replace "#pragma xxx" with "__pragma__" in the token
2087  * stream. Just as an example.
2088  *
2089  * We'll just #define that away for now, but the theory here
2090  * is that we can use this to insert arbitrary token sequences
2091  * to turn the pragmas into internal front-end sequences for
2092  * when we actually start caring about them.
2093  *
2094  * So eventually this will turn into some kind of extended
2095  * __attribute__() like thing, except called __pragma__(xxx).
2096  */
handle_pragma(struct stream * stream,struct token ** line,struct token * token)2097 static int handle_pragma(struct stream *stream, struct token **line, struct token *token)
2098 {
2099 	struct token *next = *line;
2100 
2101 	if (match_ident(token->next, &once_ident) && eof_token(token->next->next)) {
2102 		stream->once = 1;
2103 		return 1;
2104 	}
2105 	token->ident = &pragma_ident;
2106 	token->pos.newline = 1;
2107 	token->pos.whitespace = 1;
2108 	token->pos.pos = 1;
2109 	*line = token;
2110 	token->next = next;
2111 	return 0;
2112 }
2113 
2114 /*
2115  * We ignore #line for now.
2116  */
handle_line(struct stream * stream,struct token ** line,struct token * token)2117 static int handle_line(struct stream *stream, struct token **line, struct token *token)
2118 {
2119 	return 1;
2120 }
2121 
handle_ident(struct stream * stream,struct token ** line,struct token * token)2122 static int handle_ident(struct stream *stream, struct token **line, struct token *token)
2123 {
2124 	return 1;
2125 }
2126 
handle_nondirective(struct stream * stream,struct token ** line,struct token * token)2127 static int handle_nondirective(struct stream *stream, struct token **line, struct token *token)
2128 {
2129 	sparse_error(token->pos, "unrecognized preprocessor line '%s'", show_token_sequence(token, 0));
2130 	return 1;
2131 }
2132 
2133 
init_preprocessor(void)2134 static void init_preprocessor(void)
2135 {
2136 	int i;
2137 	int stream = init_stream("preprocessor", -1, includepath);
2138 	static struct {
2139 		const char *name;
2140 		int (*handler)(struct stream *, struct token **, struct token *);
2141 	} normal[] = {
2142 		{ "define",		handle_define },
2143 		{ "weak_define",	handle_weak_define },
2144 		{ "strong_define",	handle_strong_define },
2145 		{ "undef",		handle_undef },
2146 		{ "strong_undef",	handle_strong_undef },
2147 		{ "warning",		handle_warning },
2148 		{ "error",		handle_error },
2149 		{ "include",		handle_include },
2150 		{ "include_next",	handle_include_next },
2151 		{ "pragma",		handle_pragma },
2152 		{ "line",		handle_line },
2153 		{ "ident",		handle_ident },
2154 
2155 		// our internal preprocessor tokens
2156 		{ "nostdinc",	   handle_nostdinc },
2157 		{ "add_include",   handle_add_include },
2158 		{ "add_isystem",   handle_add_isystem },
2159 		{ "add_system",    handle_add_system },
2160 		{ "add_dirafter",  handle_add_dirafter },
2161 		{ "split_include", handle_split_include },
2162 		{ "argv_include",  handle_argv_include },
2163 	}, special[] = {
2164 		{ "ifdef",	handle_ifdef },
2165 		{ "ifndef",	handle_ifndef },
2166 		{ "else",	handle_else },
2167 		{ "endif",	handle_endif },
2168 		{ "if",		handle_if },
2169 		{ "elif",	handle_elif },
2170 	};
2171 	static struct {
2172 		const char *name;
2173 		void (*expander)(struct token *);
2174 	} dynamic[] = {
2175 		{ "__LINE__",		expand_line },
2176 		{ "__FILE__",		expand_file },
2177 		{ "__BASE_FILE__",	expand_basefile },
2178 		{ "__DATE__",		expand_date },
2179 		{ "__TIME__",		expand_time },
2180 		{ "__COUNTER__",	expand_counter },
2181 		{ "__INCLUDE_LEVEL__",	expand_include_level },
2182 	};
2183 
2184 	for (i = 0; i < ARRAY_SIZE(normal); i++) {
2185 		struct symbol *sym;
2186 		sym = create_symbol(stream, normal[i].name, SYM_PREPROCESSOR, NS_PREPROCESSOR);
2187 		sym->handler = normal[i].handler;
2188 		sym->normal = 1;
2189 	}
2190 	for (i = 0; i < ARRAY_SIZE(special); i++) {
2191 		struct symbol *sym;
2192 		sym = create_symbol(stream, special[i].name, SYM_PREPROCESSOR, NS_PREPROCESSOR);
2193 		sym->handler = special[i].handler;
2194 		sym->normal = 0;
2195 	}
2196 	for (i = 0; i < ARRAY_SIZE(dynamic); i++) {
2197 		struct symbol *sym;
2198 		sym = create_symbol(stream, dynamic[i].name, SYM_NODE, NS_MACRO);
2199 		sym->expander = dynamic[i].expander;
2200 	}
2201 
2202 	counter_macro = 0;
2203 }
2204 
handle_preprocessor_line(struct stream * stream,struct token ** line,struct token * start)2205 static void handle_preprocessor_line(struct stream *stream, struct token **line, struct token *start)
2206 {
2207 	int (*handler)(struct stream *, struct token **, struct token *);
2208 	struct token *token = start->next;
2209 	int is_normal = 1;
2210 
2211 	if (eof_token(token))
2212 		return;
2213 
2214 	if (token_type(token) == TOKEN_IDENT) {
2215 		struct symbol *sym = lookup_symbol(token->ident, NS_PREPROCESSOR);
2216 		if (sym) {
2217 			handler = sym->handler;
2218 			is_normal = sym->normal;
2219 		} else {
2220 			handler = handle_nondirective;
2221 		}
2222 	} else if (token_type(token) == TOKEN_NUMBER) {
2223 		handler = handle_line;
2224 	} else {
2225 		handler = handle_nondirective;
2226 	}
2227 
2228 	if (is_normal) {
2229 		dirty_stream(stream);
2230 		if (false_nesting)
2231 			goto out;
2232 	}
2233 	if (!handler(stream, line, token))	/* all set */
2234 		return;
2235 
2236 out:
2237 	free_preprocessor_line(token);
2238 }
2239 
preprocessor_line(struct stream * stream,struct token ** line)2240 static void preprocessor_line(struct stream *stream, struct token **line)
2241 {
2242 	struct token *start = *line, *next;
2243 	struct token **tp = &start->next;
2244 
2245 	for (;;) {
2246 		next = *tp;
2247 		if (next->pos.newline)
2248 			break;
2249 		tp = &next->next;
2250 	}
2251 	*line = next;
2252 	*tp = &eof_token_entry;
2253 	handle_preprocessor_line(stream, line, start);
2254 }
2255 
do_preprocess(struct token ** list)2256 static void do_preprocess(struct token **list)
2257 {
2258 	struct token *next;
2259 
2260 	while (!eof_token(next = scan_next(list))) {
2261 		struct stream *stream = input_streams + next->pos.stream;
2262 
2263 		if (next->pos.newline && match_op(next, '#')) {
2264 			if (!next->pos.noexpand) {
2265 				preprocessor_line(stream, list);
2266 				__free_token(next);	/* Free the '#' token */
2267 				continue;
2268 			}
2269 		}
2270 
2271 		switch (token_type(next)) {
2272 		case TOKEN_STREAMEND:
2273 			if (stream->top_if) {
2274 				nesting_error(stream);
2275 				sparse_error(stream->top_if->pos, "unterminated preprocessor conditional");
2276 				stream->top_if = NULL;
2277 				false_nesting = 0;
2278 			}
2279 			if (!stream->dirty)
2280 				stream->constant = CONSTANT_FILE_YES;
2281 			*list = next->next;
2282 			include_level--;
2283 			continue;
2284 		case TOKEN_STREAMBEGIN:
2285 			*list = next->next;
2286 			include_level++;
2287 			continue;
2288 
2289 		default:
2290 			dirty_stream(stream);
2291 			if (false_nesting) {
2292 				*list = next->next;
2293 				__free_token(next);
2294 				continue;
2295 			}
2296 
2297 			if (token_type(next) != TOKEN_IDENT ||
2298 			    expand_one_symbol(list))
2299 				list = &next->next;
2300 		}
2301 	}
2302 }
2303 
init_include_path(void)2304 void init_include_path(void)
2305 {
2306 	FILE *fp;
2307 	char path[256];
2308 	char arch[32];
2309 	char os[32];
2310 
2311 	fp = popen("/bin/uname -m", "r");
2312 	if (!fp)
2313 		return;
2314 	if (!fgets(arch, sizeof(arch) - 1, fp))
2315 		return;
2316 	pclose(fp);
2317 	if (arch[strlen(arch) - 1] == '\n')
2318 		arch[strlen(arch) - 1] = '\0';
2319 
2320 	fp = popen("/bin/uname -o", "r");
2321 	if (!fp)
2322 		return;
2323 	fgets(os, sizeof(os) - 1, fp);
2324 	pclose(fp);
2325 
2326 	if (strcmp(os, "GNU/Linux\n") != 0)
2327 		return;
2328 	strcpy(os, "linux-gnu");
2329 
2330 	snprintf(path, sizeof(path), "/usr/include/%s-%s/", arch, os);
2331 	add_pre_buffer("#add_system \"%s/\"\n", path);
2332 }
2333 
preprocess(struct token * token)2334 struct token * preprocess(struct token *token)
2335 {
2336 	preprocessing = 1;
2337 	init_preprocessor();
2338 	do_preprocess(&token);
2339 
2340 	// Drop all expressions from preprocessing, they're not used any more.
2341 	// This is not true when we have multiple files, though ;/
2342 	// clear_expression_alloc();
2343 	preprocessing = 0;
2344 
2345 	return token;
2346 }
2347 
is_VA_ARGS_token(struct token * token)2348 static int is_VA_ARGS_token(struct token *token)
2349 {
2350 	return (token_type(token) == TOKEN_IDENT) &&
2351 		(token->ident == &__VA_ARGS___ident);
2352 }
2353 
dump_macro(struct symbol * sym)2354 static void dump_macro(struct symbol *sym)
2355 {
2356 	int nargs = sym->arglist ? sym->arglist->count.normal : 0;
2357 	struct token *args[nargs];
2358 	struct token *token;
2359 
2360 	printf("#define %s", show_ident(sym->ident));
2361 	token = sym->arglist;
2362 	if (token) {
2363 		const char *sep = "";
2364 		int narg = 0;
2365 		putchar('(');
2366 		for (; !eof_token(token); token = token->next) {
2367 			if (token_type(token) == TOKEN_ARG_COUNT)
2368 				continue;
2369 			if (is_VA_ARGS_token(token))
2370 				printf("%s...", sep);
2371 			else
2372 				printf("%s%s", sep, show_token(token));
2373 			args[narg++] = token;
2374 			sep = ",";
2375 		}
2376 		putchar(')');
2377 	}
2378 
2379 	token = sym->expansion;
2380 	while (token_type(token) != TOKEN_UNTAINT) {
2381 		struct token *next = token->next;
2382 		if (token->pos.whitespace)
2383 			putchar(' ');
2384 		switch (token_type(token)) {
2385 		case TOKEN_CONCAT:
2386 			printf("##");
2387 			break;
2388 		case TOKEN_STR_ARGUMENT:
2389 			printf("#");
2390 			/* fall-through */
2391 		case TOKEN_QUOTED_ARGUMENT:
2392 		case TOKEN_MACRO_ARGUMENT:
2393 			token = args[token->argnum];
2394 			/* fall-through */
2395 		default:
2396 			printf("%s", show_token(token));
2397 		}
2398 		token = next;
2399 	}
2400 	putchar('\n');
2401 }
2402 
dump_macro_definitions(void)2403 void dump_macro_definitions(void)
2404 {
2405 	struct ident *name;
2406 
2407 	FOR_EACH_PTR(macros, name) {
2408 		struct symbol *sym = lookup_macro(name);
2409 		if (sym)
2410 			dump_macro(sym);
2411 	} END_FOR_EACH_PTR(name);
2412 }
2413