Filter-out warning about softfloat in tests2
[tinycc.git] / tccpp.c
1 /*
2 * TCC - Tiny C Compiler
3 *
4 * Copyright (c) 2001-2004 Fabrice Bellard
5 *
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19 */
20
21 #include "tcc.h"
22
23 /********************************************************/
24 /* global variables */
25
26 ST_DATA int tok_flags;
27 /* additional informations about token */
28 #define TOK_FLAG_BOL 0x0001 /* beginning of line before */
29 #define TOK_FLAG_BOF 0x0002 /* beginning of file before */
30 #define TOK_FLAG_ENDIF 0x0004 /* a endif was found matching starting #ifdef */
31 #define TOK_FLAG_EOF 0x0008 /* end of file */
32
33 ST_DATA int parse_flags;
34 #define PARSE_FLAG_PREPROCESS 0x0001 /* activate preprocessing */
35 #define PARSE_FLAG_TOK_NUM 0x0002 /* return numbers instead of TOK_PPNUM */
36 #define PARSE_FLAG_LINEFEED 0x0004 /* line feed is returned as a
37 token. line feed is also
38 returned at eof */
39 #define PARSE_FLAG_ASM_COMMENTS 0x0008 /* '#' can be used for line comment */
40 #define PARSE_FLAG_SPACES 0x0010 /* next() returns space tokens (for -E) */
41
42 ST_DATA struct BufferedFile *file;
43 ST_DATA int ch, tok;
44 ST_DATA CValue tokc;
45 ST_DATA const int *macro_ptr;
46 ST_DATA CString tokcstr; /* current parsed string, if any */
47
48 /* display benchmark infos */
49 ST_DATA int total_lines;
50 ST_DATA int total_bytes;
51 ST_DATA int tok_ident;
52 ST_DATA TokenSym **table_ident;
53
54 /* ------------------------------------------------------------------------- */
55
56 static int *macro_ptr_allocated;
57 static const int *unget_saved_macro_ptr;
58 static int unget_saved_buffer[TOK_MAX_SIZE + 1];
59 static int unget_buffer_enabled;
60 static TokenSym *hash_ident[TOK_HASH_SIZE];
61 static char token_buf[STRING_MAX_SIZE + 1];
62 /* true if isid(c) || isnum(c) */
63 static unsigned char isidnum_table[256-CH_EOF];
64
65 static const char tcc_keywords[] =
66 #define DEF(id, str) str "\0"
67 #include "tcctok.h"
68 #undef DEF
69 ;
70
71 /* WARNING: the content of this string encodes token numbers */
72 static const unsigned char tok_two_chars[] =
73 /* outdated -- gr
74 "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
75 "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
76 */{
77 '<','=', TOK_LE,
78 '>','=', TOK_GE,
79 '!','=', TOK_NE,
80 '&','&', TOK_LAND,
81 '|','|', TOK_LOR,
82 '+','+', TOK_INC,
83 '-','-', TOK_DEC,
84 '=','=', TOK_EQ,
85 '<','<', TOK_SHL,
86 '>','>', TOK_SAR,
87 '+','=', TOK_A_ADD,
88 '-','=', TOK_A_SUB,
89 '*','=', TOK_A_MUL,
90 '/','=', TOK_A_DIV,
91 '%','=', TOK_A_MOD,
92 '&','=', TOK_A_AND,
93 '^','=', TOK_A_XOR,
94 '|','=', TOK_A_OR,
95 '-','>', TOK_ARROW,
96 '.','.', 0xa8, // C++ token ?
97 '#','#', TOK_TWOSHARPS,
98 0
99 };
100
101 struct macro_level {
102 struct macro_level *prev;
103 const int *p;
104 };
105
106 static void next_nomacro_spc(void);
107 static void macro_subst(
108 TokenString *tok_str,
109 Sym **nested_list,
110 const int *macro_str,
111 struct macro_level **can_read_stream
112 );
113
114 ST_FUNC void skip(int c)
115 {
116 if (tok != c)
117 tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
118 next();
119 }
120
121 ST_FUNC void expect(const char *msg)
122 {
123 tcc_error("%s expected", msg);
124 }
125
126 /* ------------------------------------------------------------------------- */
127 /* CString handling */
128 static void cstr_realloc(CString *cstr, int new_size)
129 {
130 int size;
131 void *data;
132
133 size = cstr->size_allocated;
134 if (size == 0)
135 size = 8; /* no need to allocate a too small first string */
136 while (size < new_size)
137 size = size * 2;
138 data = tcc_realloc(cstr->data_allocated, size);
139 cstr->data_allocated = data;
140 cstr->size_allocated = size;
141 cstr->data = data;
142 }
143
144 /* add a byte */
145 ST_FUNC void cstr_ccat(CString *cstr, int ch)
146 {
147 int size;
148 size = cstr->size + 1;
149 if (size > cstr->size_allocated)
150 cstr_realloc(cstr, size);
151 ((unsigned char *)cstr->data)[size - 1] = ch;
152 cstr->size = size;
153 }
154
155 ST_FUNC void cstr_cat(CString *cstr, const char *str)
156 {
157 int c;
158 for(;;) {
159 c = *str;
160 if (c == '\0')
161 break;
162 cstr_ccat(cstr, c);
163 str++;
164 }
165 }
166
167 /* add a wide char */
168 ST_FUNC void cstr_wccat(CString *cstr, int ch)
169 {
170 int size;
171 size = cstr->size + sizeof(nwchar_t);
172 if (size > cstr->size_allocated)
173 cstr_realloc(cstr, size);
174 *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
175 cstr->size = size;
176 }
177
178 ST_FUNC void cstr_new(CString *cstr)
179 {
180 memset(cstr, 0, sizeof(CString));
181 }
182
183 /* free string and reset it to NULL */
184 ST_FUNC void cstr_free(CString *cstr)
185 {
186 tcc_free(cstr->data_allocated);
187 cstr_new(cstr);
188 }
189
190 /* reset string to empty */
191 ST_FUNC void cstr_reset(CString *cstr)
192 {
193 cstr->size = 0;
194 }
195
196 /* XXX: unicode ? */
197 static void add_char(CString *cstr, int c)
198 {
199 if (c == '\'' || c == '\"' || c == '\\') {
200 /* XXX: could be more precise if char or string */
201 cstr_ccat(cstr, '\\');
202 }
203 if (c >= 32 && c <= 126) {
204 cstr_ccat(cstr, c);
205 } else {
206 cstr_ccat(cstr, '\\');
207 if (c == '\n') {
208 cstr_ccat(cstr, 'n');
209 } else {
210 cstr_ccat(cstr, '0' + ((c >> 6) & 7));
211 cstr_ccat(cstr, '0' + ((c >> 3) & 7));
212 cstr_ccat(cstr, '0' + (c & 7));
213 }
214 }
215 }
216
217 /* ------------------------------------------------------------------------- */
218 /* allocate a new token */
219 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
220 {
221 TokenSym *ts, **ptable;
222 int i;
223
224 if (tok_ident >= SYM_FIRST_ANOM)
225 tcc_error("memory full (symbols)");
226
227 /* expand token table if needed */
228 i = tok_ident - TOK_IDENT;
229 if ((i % TOK_ALLOC_INCR) == 0) {
230 ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
231 table_ident = ptable;
232 }
233
234 ts = tcc_malloc(sizeof(TokenSym) + len);
235 table_ident[i] = ts;
236 ts->tok = tok_ident++;
237 ts->sym_define = NULL;
238 ts->sym_label = NULL;
239 ts->sym_struct = NULL;
240 ts->sym_identifier = NULL;
241 ts->len = len;
242 ts->hash_next = NULL;
243 memcpy(ts->str, str, len);
244 ts->str[len] = '\0';
245 *pts = ts;
246 return ts;
247 }
248
249 #define TOK_HASH_INIT 1
250 #define TOK_HASH_FUNC(h, c) ((h) * 263 + (c))
251
252 /* find a token and add it if not found */
253 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
254 {
255 TokenSym *ts, **pts;
256 int i;
257 unsigned int h;
258
259 h = TOK_HASH_INIT;
260 for(i=0;i<len;i++)
261 h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
262 h &= (TOK_HASH_SIZE - 1);
263
264 pts = &hash_ident[h];
265 for(;;) {
266 ts = *pts;
267 if (!ts)
268 break;
269 if (ts->len == len && !memcmp(ts->str, str, len))
270 return ts;
271 pts = &(ts->hash_next);
272 }
273 return tok_alloc_new(pts, str, len);
274 }
275
276 /* XXX: buffer overflow */
277 /* XXX: float tokens */
278 ST_FUNC char *get_tok_str(int v, CValue *cv)
279 {
280 static char buf[STRING_MAX_SIZE + 1];
281 static CString cstr_buf;
282 CString *cstr;
283 char *p;
284 int i, len;
285
286 /* NOTE: to go faster, we give a fixed buffer for small strings */
287 cstr_reset(&cstr_buf);
288 cstr_buf.data = buf;
289 cstr_buf.size_allocated = sizeof(buf);
290 p = buf;
291
292 /* just an explanation, should never happen:
293 if (v <= TOK_LINENUM && v >= TOK_CINT && cv == NULL)
294 tcc_error("internal error: get_tok_str"); */
295
296 switch(v) {
297 case TOK_CINT:
298 case TOK_CUINT:
299 /* XXX: not quite exact, but only useful for testing */
300 sprintf(p, "%u", cv->ui);
301 break;
302 case TOK_CLLONG:
303 case TOK_CULLONG:
304 /* XXX: not quite exact, but only useful for testing */
305 #ifdef _WIN32
306 sprintf(p, "%u", (unsigned)cv->ull);
307 #else
308 sprintf(p, "%llu", cv->ull);
309 #endif
310 break;
311 case TOK_LCHAR:
312 cstr_ccat(&cstr_buf, 'L');
313 case TOK_CCHAR:
314 cstr_ccat(&cstr_buf, '\'');
315 add_char(&cstr_buf, cv->i);
316 cstr_ccat(&cstr_buf, '\'');
317 cstr_ccat(&cstr_buf, '\0');
318 break;
319 case TOK_PPNUM:
320 cstr = cv->cstr;
321 len = cstr->size - 1;
322 for(i=0;i<len;i++)
323 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
324 cstr_ccat(&cstr_buf, '\0');
325 break;
326 case TOK_LSTR:
327 cstr_ccat(&cstr_buf, 'L');
328 case TOK_STR:
329 cstr = cv->cstr;
330 cstr_ccat(&cstr_buf, '\"');
331 if (v == TOK_STR) {
332 len = cstr->size - 1;
333 for(i=0;i<len;i++)
334 add_char(&cstr_buf, ((unsigned char *)cstr->data)[i]);
335 } else {
336 len = (cstr->size / sizeof(nwchar_t)) - 1;
337 for(i=0;i<len;i++)
338 add_char(&cstr_buf, ((nwchar_t *)cstr->data)[i]);
339 }
340 cstr_ccat(&cstr_buf, '\"');
341 cstr_ccat(&cstr_buf, '\0');
342 break;
343
344 case TOK_CFLOAT:
345 case TOK_CDOUBLE:
346 case TOK_CLDOUBLE:
347 case TOK_LINENUM:
348 return NULL; /* should not happen */
349
350 /* above tokens have value, the ones below don't */
351
352 case TOK_LT:
353 v = '<';
354 goto addv;
355 case TOK_GT:
356 v = '>';
357 goto addv;
358 case TOK_DOTS:
359 return strcpy(p, "...");
360 case TOK_A_SHL:
361 return strcpy(p, "<<=");
362 case TOK_A_SAR:
363 return strcpy(p, ">>=");
364 default:
365 if (v < TOK_IDENT) {
366 /* search in two bytes table */
367 const unsigned char *q = tok_two_chars;
368 while (*q) {
369 if (q[2] == v) {
370 *p++ = q[0];
371 *p++ = q[1];
372 *p = '\0';
373 return buf;
374 }
375 q += 3;
376 }
377 addv:
378 *p++ = v;
379 *p = '\0';
380 } else if (v < tok_ident) {
381 return table_ident[v - TOK_IDENT]->str;
382 } else if (v >= SYM_FIRST_ANOM) {
383 /* special name for anonymous symbol */
384 sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
385 } else {
386 /* should never happen */
387 return NULL;
388 }
389 break;
390 }
391 return cstr_buf.data;
392 }
393
394 /* fill input buffer and peek next char */
395 static int tcc_peekc_slow(BufferedFile *bf)
396 {
397 int len;
398 /* only tries to read if really end of buffer */
399 if (bf->buf_ptr >= bf->buf_end) {
400 if (bf->fd != -1) {
401 #if defined(PARSE_DEBUG)
402 len = 8;
403 #else
404 len = IO_BUF_SIZE;
405 #endif
406 len = read(bf->fd, bf->buffer, len);
407 if (len < 0)
408 len = 0;
409 } else {
410 len = 0;
411 }
412 total_bytes += len;
413 bf->buf_ptr = bf->buffer;
414 bf->buf_end = bf->buffer + len;
415 *bf->buf_end = CH_EOB;
416 }
417 if (bf->buf_ptr < bf->buf_end) {
418 return bf->buf_ptr[0];
419 } else {
420 bf->buf_ptr = bf->buf_end;
421 return CH_EOF;
422 }
423 }
424
425 /* return the current character, handling end of block if necessary
426 (but not stray) */
427 ST_FUNC int handle_eob(void)
428 {
429 return tcc_peekc_slow(file);
430 }
431
432 /* read next char from current input file and handle end of input buffer */
433 ST_INLN void inp(void)
434 {
435 ch = *(++(file->buf_ptr));
436 /* end of buffer/file handling */
437 if (ch == CH_EOB)
438 ch = handle_eob();
439 }
440
441 /* handle '\[\r]\n' */
442 static int handle_stray_noerror(void)
443 {
444 while (ch == '\\') {
445 inp();
446 if (ch == '\n') {
447 file->line_num++;
448 inp();
449 } else if (ch == '\r') {
450 inp();
451 if (ch != '\n')
452 goto fail;
453 file->line_num++;
454 inp();
455 } else {
456 fail:
457 return 1;
458 }
459 }
460 return 0;
461 }
462
463 static void handle_stray(void)
464 {
465 if (handle_stray_noerror())
466 tcc_error("stray '\\' in program");
467 }
468
469 /* skip the stray and handle the \\n case. Output an error if
470 incorrect char after the stray */
471 static int handle_stray1(uint8_t *p)
472 {
473 int c;
474
475 if (p >= file->buf_end) {
476 file->buf_ptr = p;
477 c = handle_eob();
478 p = file->buf_ptr;
479 if (c == '\\')
480 goto parse_stray;
481 } else {
482 parse_stray:
483 file->buf_ptr = p;
484 ch = *p;
485 handle_stray();
486 p = file->buf_ptr;
487 c = *p;
488 }
489 return c;
490 }
491
492 /* handle just the EOB case, but not stray */
493 #define PEEKC_EOB(c, p)\
494 {\
495 p++;\
496 c = *p;\
497 if (c == '\\') {\
498 file->buf_ptr = p;\
499 c = handle_eob();\
500 p = file->buf_ptr;\
501 }\
502 }
503
504 /* handle the complicated stray case */
505 #define PEEKC(c, p)\
506 {\
507 p++;\
508 c = *p;\
509 if (c == '\\') {\
510 c = handle_stray1(p);\
511 p = file->buf_ptr;\
512 }\
513 }
514
515 /* input with '\[\r]\n' handling. Note that this function cannot
516 handle other characters after '\', so you cannot call it inside
517 strings or comments */
518 ST_FUNC void minp(void)
519 {
520 inp();
521 if (ch == '\\')
522 handle_stray();
523 }
524
525
526 /* single line C++ comments */
527 static uint8_t *parse_line_comment(uint8_t *p)
528 {
529 int c;
530
531 p++;
532 for(;;) {
533 c = *p;
534 redo:
535 if (c == '\n' || c == CH_EOF) {
536 break;
537 } else if (c == '\\') {
538 file->buf_ptr = p;
539 c = handle_eob();
540 p = file->buf_ptr;
541 if (c == '\\') {
542 PEEKC_EOB(c, p);
543 if (c == '\n') {
544 file->line_num++;
545 PEEKC_EOB(c, p);
546 } else if (c == '\r') {
547 PEEKC_EOB(c, p);
548 if (c == '\n') {
549 file->line_num++;
550 PEEKC_EOB(c, p);
551 }
552 }
553 } else {
554 goto redo;
555 }
556 } else {
557 p++;
558 }
559 }
560 return p;
561 }
562
563 /* C comments */
564 ST_FUNC uint8_t *parse_comment(uint8_t *p)
565 {
566 int c;
567
568 p++;
569 for(;;) {
570 /* fast skip loop */
571 for(;;) {
572 c = *p;
573 if (c == '\n' || c == '*' || c == '\\')
574 break;
575 p++;
576 c = *p;
577 if (c == '\n' || c == '*' || c == '\\')
578 break;
579 p++;
580 }
581 /* now we can handle all the cases */
582 if (c == '\n') {
583 file->line_num++;
584 p++;
585 } else if (c == '*') {
586 p++;
587 for(;;) {
588 c = *p;
589 if (c == '*') {
590 p++;
591 } else if (c == '/') {
592 goto end_of_comment;
593 } else if (c == '\\') {
594 file->buf_ptr = p;
595 c = handle_eob();
596 p = file->buf_ptr;
597 if (c == '\\') {
598 /* skip '\[\r]\n', otherwise just skip the stray */
599 while (c == '\\') {
600 PEEKC_EOB(c, p);
601 if (c == '\n') {
602 file->line_num++;
603 PEEKC_EOB(c, p);
604 } else if (c == '\r') {
605 PEEKC_EOB(c, p);
606 if (c == '\n') {
607 file->line_num++;
608 PEEKC_EOB(c, p);
609 }
610 } else {
611 goto after_star;
612 }
613 }
614 }
615 } else {
616 break;
617 }
618 }
619 after_star: ;
620 } else {
621 /* stray, eob or eof */
622 file->buf_ptr = p;
623 c = handle_eob();
624 p = file->buf_ptr;
625 if (c == CH_EOF) {
626 tcc_error("unexpected end of file in comment");
627 } else if (c == '\\') {
628 p++;
629 }
630 }
631 }
632 end_of_comment:
633 p++;
634 return p;
635 }
636
637 #define cinp minp
638
639 static inline void skip_spaces(void)
640 {
641 while (is_space(ch))
642 cinp();
643 }
644
645 static inline int check_space(int t, int *spc)
646 {
647 if (is_space(t)) {
648 if (*spc)
649 return 1;
650 *spc = 1;
651 } else
652 *spc = 0;
653 return 0;
654 }
655
656 /* parse a string without interpreting escapes */
657 static uint8_t *parse_pp_string(uint8_t *p,
658 int sep, CString *str)
659 {
660 int c;
661 p++;
662 for(;;) {
663 c = *p;
664 if (c == sep) {
665 break;
666 } else if (c == '\\') {
667 file->buf_ptr = p;
668 c = handle_eob();
669 p = file->buf_ptr;
670 if (c == CH_EOF) {
671 unterminated_string:
672 /* XXX: indicate line number of start of string */
673 tcc_error("missing terminating %c character", sep);
674 } else if (c == '\\') {
675 /* escape : just skip \[\r]\n */
676 PEEKC_EOB(c, p);
677 if (c == '\n') {
678 file->line_num++;
679 p++;
680 } else if (c == '\r') {
681 PEEKC_EOB(c, p);
682 if (c != '\n')
683 expect("'\n' after '\r'");
684 file->line_num++;
685 p++;
686 } else if (c == CH_EOF) {
687 goto unterminated_string;
688 } else {
689 if (str) {
690 cstr_ccat(str, '\\');
691 cstr_ccat(str, c);
692 }
693 p++;
694 }
695 }
696 } else if (c == '\n') {
697 file->line_num++;
698 goto add_char;
699 } else if (c == '\r') {
700 PEEKC_EOB(c, p);
701 if (c != '\n') {
702 if (str)
703 cstr_ccat(str, '\r');
704 } else {
705 file->line_num++;
706 goto add_char;
707 }
708 } else {
709 add_char:
710 if (str)
711 cstr_ccat(str, c);
712 p++;
713 }
714 }
715 p++;
716 return p;
717 }
718
719 /* skip block of text until #else, #elif or #endif. skip also pairs of
720 #if/#endif */
721 static void preprocess_skip(void)
722 {
723 int a, start_of_line, c, in_warn_or_error;
724 uint8_t *p;
725
726 p = file->buf_ptr;
727 a = 0;
728 redo_start:
729 start_of_line = 1;
730 in_warn_or_error = 0;
731 for(;;) {
732 redo_no_start:
733 c = *p;
734 switch(c) {
735 case ' ':
736 case '\t':
737 case '\f':
738 case '\v':
739 case '\r':
740 p++;
741 goto redo_no_start;
742 case '\n':
743 file->line_num++;
744 p++;
745 goto redo_start;
746 case '\\':
747 file->buf_ptr = p;
748 c = handle_eob();
749 if (c == CH_EOF) {
750 expect("#endif");
751 } else if (c == '\\') {
752 ch = file->buf_ptr[0];
753 handle_stray_noerror();
754 }
755 p = file->buf_ptr;
756 goto redo_no_start;
757 /* skip strings */
758 case '\"':
759 case '\'':
760 if (in_warn_or_error)
761 goto _default;
762 p = parse_pp_string(p, c, NULL);
763 break;
764 /* skip comments */
765 case '/':
766 if (in_warn_or_error)
767 goto _default;
768 file->buf_ptr = p;
769 ch = *p;
770 minp();
771 p = file->buf_ptr;
772 if (ch == '*') {
773 p = parse_comment(p);
774 } else if (ch == '/') {
775 p = parse_line_comment(p);
776 }
777 break;
778 case '#':
779 p++;
780 if (start_of_line) {
781 file->buf_ptr = p;
782 next_nomacro();
783 p = file->buf_ptr;
784 if (a == 0 &&
785 (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
786 goto the_end;
787 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
788 a++;
789 else if (tok == TOK_ENDIF)
790 a--;
791 else if( tok == TOK_ERROR || tok == TOK_WARNING)
792 in_warn_or_error = 1;
793 else if (tok == TOK_LINEFEED)
794 goto redo_start;
795 }
796 break;
797 _default:
798 default:
799 p++;
800 break;
801 }
802 start_of_line = 0;
803 }
804 the_end: ;
805 file->buf_ptr = p;
806 }
807
808 /* ParseState handling */
809
810 /* XXX: currently, no include file info is stored. Thus, we cannot display
811 accurate messages if the function or data definition spans multiple
812 files */
813
814 /* save current parse state in 's' */
815 ST_FUNC void save_parse_state(ParseState *s)
816 {
817 s->line_num = file->line_num;
818 s->macro_ptr = macro_ptr;
819 s->tok = tok;
820 s->tokc = tokc;
821 }
822
823 /* restore parse state from 's' */
824 ST_FUNC void restore_parse_state(ParseState *s)
825 {
826 file->line_num = s->line_num;
827 macro_ptr = s->macro_ptr;
828 tok = s->tok;
829 tokc = s->tokc;
830 }
831
832 /* return the number of additional 'ints' necessary to store the
833 token */
834 static inline int tok_ext_size(int t)
835 {
836 switch(t) {
837 /* 4 bytes */
838 case TOK_CINT:
839 case TOK_CUINT:
840 case TOK_CCHAR:
841 case TOK_LCHAR:
842 case TOK_CFLOAT:
843 case TOK_LINENUM:
844 return 1;
845 case TOK_STR:
846 case TOK_LSTR:
847 case TOK_PPNUM:
848 tcc_error("unsupported token");
849 return 1;
850 case TOK_CDOUBLE:
851 case TOK_CLLONG:
852 case TOK_CULLONG:
853 return 2;
854 case TOK_CLDOUBLE:
855 return LDOUBLE_SIZE / 4;
856 default:
857 return 0;
858 }
859 }
860
861 /* token string handling */
862
863 ST_INLN void tok_str_new(TokenString *s)
864 {
865 s->str = NULL;
866 s->len = 0;
867 s->allocated_len = 0;
868 s->last_line_num = -1;
869 }
870
871 ST_FUNC void tok_str_free(int *str)
872 {
873 tcc_free(str);
874 }
875
876 static int *tok_str_realloc(TokenString *s)
877 {
878 int *str, len;
879
880 if (s->allocated_len == 0) {
881 len = 8;
882 } else {
883 len = s->allocated_len * 2;
884 }
885 str = tcc_realloc(s->str, len * sizeof(int));
886 s->allocated_len = len;
887 s->str = str;
888 return str;
889 }
890
891 ST_FUNC void tok_str_add(TokenString *s, int t)
892 {
893 int len, *str;
894
895 len = s->len;
896 str = s->str;
897 if (len >= s->allocated_len)
898 str = tok_str_realloc(s);
899 str[len++] = t;
900 s->len = len;
901 }
902
903 static void tok_str_add2(TokenString *s, int t, CValue *cv)
904 {
905 int len, *str;
906
907 len = s->len;
908 str = s->str;
909
910 /* allocate space for worst case */
911 if (len + TOK_MAX_SIZE > s->allocated_len)
912 str = tok_str_realloc(s);
913 str[len++] = t;
914 switch(t) {
915 case TOK_CINT:
916 case TOK_CUINT:
917 case TOK_CCHAR:
918 case TOK_LCHAR:
919 case TOK_CFLOAT:
920 case TOK_LINENUM:
921 str[len++] = cv->tab[0];
922 break;
923 case TOK_PPNUM:
924 case TOK_STR:
925 case TOK_LSTR:
926 {
927 int nb_words;
928 CString *cstr;
929
930 nb_words = (sizeof(CString) + cv->cstr->size + 3) >> 2;
931 while ((len + nb_words) > s->allocated_len)
932 str = tok_str_realloc(s);
933 cstr = (CString *)(str + len);
934 cstr->data = NULL;
935 cstr->size = cv->cstr->size;
936 cstr->data_allocated = NULL;
937 cstr->size_allocated = cstr->size;
938 memcpy((char *)cstr + sizeof(CString),
939 cv->cstr->data, cstr->size);
940 len += nb_words;
941 }
942 break;
943 case TOK_CDOUBLE:
944 case TOK_CLLONG:
945 case TOK_CULLONG:
946 #if LDOUBLE_SIZE == 8
947 case TOK_CLDOUBLE:
948 #endif
949 str[len++] = cv->tab[0];
950 str[len++] = cv->tab[1];
951 break;
952 #if LDOUBLE_SIZE == 12
953 case TOK_CLDOUBLE:
954 str[len++] = cv->tab[0];
955 str[len++] = cv->tab[1];
956 str[len++] = cv->tab[2];
957 #elif LDOUBLE_SIZE == 16
958 case TOK_CLDOUBLE:
959 str[len++] = cv->tab[0];
960 str[len++] = cv->tab[1];
961 str[len++] = cv->tab[2];
962 str[len++] = cv->tab[3];
963 #elif LDOUBLE_SIZE != 8
964 #error add long double size support
965 #endif
966 break;
967 default:
968 break;
969 }
970 s->len = len;
971 }
972
973 /* add the current parse token in token string 's' */
974 ST_FUNC void tok_str_add_tok(TokenString *s)
975 {
976 CValue cval;
977
978 /* save line number info */
979 if (file->line_num != s->last_line_num) {
980 s->last_line_num = file->line_num;
981 cval.i = s->last_line_num;
982 tok_str_add2(s, TOK_LINENUM, &cval);
983 }
984 tok_str_add2(s, tok, &tokc);
985 }
986
987 /* get a token from an integer array and increment pointer
988 accordingly. we code it as a macro to avoid pointer aliasing. */
989 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
990 {
991 const int *p = *pp;
992 int n, *tab;
993
994 tab = cv->tab;
995 switch(*t = *p++) {
996 case TOK_CINT:
997 case TOK_CUINT:
998 case TOK_CCHAR:
999 case TOK_LCHAR:
1000 case TOK_CFLOAT:
1001 case TOK_LINENUM:
1002 tab[0] = *p++;
1003 break;
1004 case TOK_STR:
1005 case TOK_LSTR:
1006 case TOK_PPNUM:
1007 cv->cstr = (CString *)p;
1008 cv->cstr->data = (char *)p + sizeof(CString);
1009 p += (sizeof(CString) + cv->cstr->size + 3) >> 2;
1010 break;
1011 case TOK_CDOUBLE:
1012 case TOK_CLLONG:
1013 case TOK_CULLONG:
1014 n = 2;
1015 goto copy;
1016 case TOK_CLDOUBLE:
1017 #if LDOUBLE_SIZE == 16
1018 n = 4;
1019 #elif LDOUBLE_SIZE == 12
1020 n = 3;
1021 #elif LDOUBLE_SIZE == 8
1022 n = 2;
1023 #else
1024 # error add long double size support
1025 #endif
1026 copy:
1027 do
1028 *tab++ = *p++;
1029 while (--n);
1030 break;
1031 default:
1032 break;
1033 }
1034 *pp = p;
1035 }
1036
1037 static int macro_is_equal(const int *a, const int *b)
1038 {
1039 char buf[STRING_MAX_SIZE + 1];
1040 CValue cv;
1041 int t;
1042 while (*a && *b) {
1043 TOK_GET(&t, &a, &cv);
1044 pstrcpy(buf, sizeof buf, get_tok_str(t, &cv));
1045 TOK_GET(&t, &b, &cv);
1046 if (strcmp(buf, get_tok_str(t, &cv)))
1047 return 0;
1048 }
1049 return !(*a || *b);
1050 }
1051
1052 /* defines handling */
1053 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1054 {
1055 Sym *s;
1056
1057 s = define_find(v);
1058 if (s && !macro_is_equal(s->d, str))
1059 tcc_warning("%s redefined", get_tok_str(v, NULL));
1060
1061 s = sym_push2(&define_stack, v, macro_type, 0);
1062 s->d = str;
1063 s->next = first_arg;
1064 table_ident[v - TOK_IDENT]->sym_define = s;
1065 }
1066
1067 /* undefined a define symbol. Its name is just set to zero */
1068 ST_FUNC void define_undef(Sym *s)
1069 {
1070 int v;
1071 v = s->v;
1072 if (v >= TOK_IDENT && v < tok_ident)
1073 table_ident[v - TOK_IDENT]->sym_define = NULL;
1074 s->v = 0;
1075 }
1076
1077 ST_INLN Sym *define_find(int v)
1078 {
1079 v -= TOK_IDENT;
1080 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1081 return NULL;
1082 return table_ident[v]->sym_define;
1083 }
1084
1085 /* free define stack until top reaches 'b' */
1086 ST_FUNC void free_defines(Sym *b)
1087 {
1088 Sym *top, *top1;
1089 int v;
1090
1091 top = define_stack;
1092 while (top != b) {
1093 top1 = top->prev;
1094 /* do not free args or predefined defines */
1095 if (top->d)
1096 tok_str_free(top->d);
1097 v = top->v;
1098 if (v >= TOK_IDENT && v < tok_ident)
1099 table_ident[v - TOK_IDENT]->sym_define = NULL;
1100 sym_free(top);
1101 top = top1;
1102 }
1103 define_stack = b;
1104 }
1105
1106 /* label lookup */
1107 ST_FUNC Sym *label_find(int v)
1108 {
1109 v -= TOK_IDENT;
1110 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1111 return NULL;
1112 return table_ident[v]->sym_label;
1113 }
1114
1115 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1116 {
1117 Sym *s, **ps;
1118 s = sym_push2(ptop, v, 0, 0);
1119 s->r = flags;
1120 ps = &table_ident[v - TOK_IDENT]->sym_label;
1121 if (ptop == &global_label_stack) {
1122 /* modify the top most local identifier, so that
1123 sym_identifier will point to 's' when popped */
1124 while (*ps != NULL)
1125 ps = &(*ps)->prev_tok;
1126 }
1127 s->prev_tok = *ps;
1128 *ps = s;
1129 return s;
1130 }
1131
1132 /* pop labels until element last is reached. Look if any labels are
1133 undefined. Define symbols if '&&label' was used. */
1134 ST_FUNC void label_pop(Sym **ptop, Sym *slast)
1135 {
1136 Sym *s, *s1;
1137 for(s = *ptop; s != slast; s = s1) {
1138 s1 = s->prev;
1139 if (s->r == LABEL_DECLARED) {
1140 tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1141 } else if (s->r == LABEL_FORWARD) {
1142 tcc_error("label '%s' used but not defined",
1143 get_tok_str(s->v, NULL));
1144 } else {
1145 if (s->c) {
1146 /* define corresponding symbol. A size of
1147 1 is put. */
1148 put_extern_sym(s, cur_text_section, s->jnext, 1);
1149 }
1150 }
1151 /* remove label */
1152 table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1153 sym_free(s);
1154 }
1155 *ptop = slast;
1156 }
1157
1158 /* eval an expression for #if/#elif */
1159 static int expr_preprocess(void)
1160 {
1161 int c, t;
1162 TokenString str;
1163
1164 tok_str_new(&str);
1165 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1166 next(); /* do macro subst */
1167 if (tok == TOK_DEFINED) {
1168 next_nomacro();
1169 t = tok;
1170 if (t == '(')
1171 next_nomacro();
1172 c = define_find(tok) != 0;
1173 if (t == '(')
1174 next_nomacro();
1175 tok = TOK_CINT;
1176 tokc.i = c;
1177 } else if (tok >= TOK_IDENT) {
1178 /* if undefined macro */
1179 tok = TOK_CINT;
1180 tokc.i = 0;
1181 }
1182 tok_str_add_tok(&str);
1183 }
1184 tok_str_add(&str, -1); /* simulate end of file */
1185 tok_str_add(&str, 0);
1186 /* now evaluate C constant expression */
1187 macro_ptr = str.str;
1188 next();
1189 c = expr_const();
1190 macro_ptr = NULL;
1191 tok_str_free(str.str);
1192 return c != 0;
1193 }
1194
1195 #if defined(PARSE_DEBUG) || defined(PP_DEBUG)
1196 static void tok_print(int *str)
1197 {
1198 int t;
1199 CValue cval;
1200
1201 printf("<");
1202 while (1) {
1203 TOK_GET(&t, &str, &cval);
1204 if (!t)
1205 break;
1206 printf("%s", get_tok_str(t, &cval));
1207 }
1208 printf(">\n");
1209 }
1210 #endif
1211
1212 /* parse after #define */
1213 ST_FUNC void parse_define(void)
1214 {
1215 Sym *s, *first, **ps;
1216 int v, t, varg, is_vaargs, spc, ptok, macro_list_start;
1217 TokenString str;
1218
1219 v = tok;
1220 if (v < TOK_IDENT)
1221 tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1222 /* XXX: should check if same macro (ANSI) */
1223 first = NULL;
1224 t = MACRO_OBJ;
1225 /* '(' must be just after macro definition for MACRO_FUNC */
1226 next_nomacro_spc();
1227 if (tok == '(') {
1228 next_nomacro();
1229 ps = &first;
1230 while (tok != ')') {
1231 varg = tok;
1232 next_nomacro();
1233 is_vaargs = 0;
1234 if (varg == TOK_DOTS) {
1235 varg = TOK___VA_ARGS__;
1236 is_vaargs = 1;
1237 } else if (tok == TOK_DOTS && gnu_ext) {
1238 is_vaargs = 1;
1239 next_nomacro();
1240 }
1241 if (varg < TOK_IDENT)
1242 tcc_error( "\'%s\' may not appear in parameter list", get_tok_str(varg, NULL));
1243 s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1244 *ps = s;
1245 ps = &s->next;
1246 if (tok != ',')
1247 continue;
1248 next_nomacro();
1249 }
1250 next_nomacro_spc();
1251 t = MACRO_FUNC;
1252 }
1253 tok_str_new(&str);
1254 spc = 2;
1255 /* EOF testing necessary for '-D' handling */
1256 ptok = 0;
1257 macro_list_start = 1;
1258 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1259 if (!macro_list_start && spc == 2 && tok == TOK_TWOSHARPS)
1260 tcc_error("'##' invalid at start of macro");
1261 ptok = tok;
1262 /* remove spaces around ## and after '#' */
1263 if (TOK_TWOSHARPS == tok) {
1264 if (1 == spc)
1265 --str.len;
1266 spc = 2;
1267 } else if ('#' == tok) {
1268 spc = 2;
1269 } else if (check_space(tok, &spc)) {
1270 goto skip;
1271 }
1272 tok_str_add2(&str, tok, &tokc);
1273 skip:
1274 next_nomacro_spc();
1275 macro_list_start = 0;
1276 }
1277 if (ptok == TOK_TWOSHARPS)
1278 tcc_error("'##' invalid at end of macro");
1279 if (spc == 1)
1280 --str.len; /* remove trailing space */
1281 tok_str_add(&str, 0);
1282 #ifdef PP_DEBUG
1283 printf("define %s %d: ", get_tok_str(v, NULL), t);
1284 tok_print(str.str);
1285 #endif
1286 define_push(v, t, str.str, first);
1287 }
1288
1289 static inline int hash_cached_include(const char *filename)
1290 {
1291 const unsigned char *s;
1292 unsigned int h;
1293
1294 h = TOK_HASH_INIT;
1295 s = (unsigned char *) filename;
1296 while (*s) {
1297 h = TOK_HASH_FUNC(h, *s);
1298 s++;
1299 }
1300 h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1301 return h;
1302 }
1303
1304 static CachedInclude *search_cached_include(TCCState *s1, const char *filename)
1305 {
1306 CachedInclude *e;
1307 int i, h;
1308 h = hash_cached_include(filename);
1309 i = s1->cached_includes_hash[h];
1310 for(;;) {
1311 if (i == 0)
1312 break;
1313 e = s1->cached_includes[i - 1];
1314 if (0 == PATHCMP(e->filename, filename))
1315 return e;
1316 i = e->hash_next;
1317 }
1318 return NULL;
1319 }
1320
1321 static inline void add_cached_include(TCCState *s1, const char *filename, int ifndef_macro)
1322 {
1323 CachedInclude *e;
1324 int h;
1325
1326 if (search_cached_include(s1, filename))
1327 return;
1328 #ifdef INC_DEBUG
1329 printf("adding cached '%s' %s\n", filename, get_tok_str(ifndef_macro, NULL));
1330 #endif
1331 e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1332 strcpy(e->filename, filename);
1333 e->ifndef_macro = ifndef_macro;
1334 dynarray_add((void ***)&s1->cached_includes, &s1->nb_cached_includes, e);
1335 /* add in hash table */
1336 h = hash_cached_include(filename);
1337 e->hash_next = s1->cached_includes_hash[h];
1338 s1->cached_includes_hash[h] = s1->nb_cached_includes;
1339 }
1340
1341 static void pragma_parse(TCCState *s1)
1342 {
1343 int val;
1344
1345 next();
1346 if (tok == TOK_pack) {
1347 /*
1348 This may be:
1349 #pragma pack(1) // set
1350 #pragma pack() // reset to default
1351 #pragma pack(push,1) // push & set
1352 #pragma pack(pop) // restore previous
1353 */
1354 next();
1355 skip('(');
1356 if (tok == TOK_ASM_pop) {
1357 next();
1358 if (s1->pack_stack_ptr <= s1->pack_stack) {
1359 stk_error:
1360 tcc_error("out of pack stack");
1361 }
1362 s1->pack_stack_ptr--;
1363 } else {
1364 val = 0;
1365 if (tok != ')') {
1366 if (tok == TOK_ASM_push) {
1367 next();
1368 if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1369 goto stk_error;
1370 s1->pack_stack_ptr++;
1371 skip(',');
1372 }
1373 if (tok != TOK_CINT) {
1374 pack_error:
1375 tcc_error("invalid pack pragma");
1376 }
1377 val = tokc.i;
1378 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1379 goto pack_error;
1380 next();
1381 }
1382 *s1->pack_stack_ptr = val;
1383 skip(')');
1384 }
1385 }
1386 }
1387
1388 /* is_bof is true if first non space token at beginning of file */
1389 ST_FUNC void preprocess(int is_bof)
1390 {
1391 TCCState *s1 = tcc_state;
1392 int i, c, n, saved_parse_flags;
1393 char buf[1024], *q;
1394 Sym *s;
1395
1396 saved_parse_flags = parse_flags;
1397 parse_flags = PARSE_FLAG_PREPROCESS | PARSE_FLAG_TOK_NUM |
1398 PARSE_FLAG_LINEFEED;
1399 next_nomacro();
1400 redo:
1401 switch(tok) {
1402 case TOK_DEFINE:
1403 next_nomacro();
1404 parse_define();
1405 break;
1406 case TOK_UNDEF:
1407 next_nomacro();
1408 s = define_find(tok);
1409 /* undefine symbol by putting an invalid name */
1410 if (s)
1411 define_undef(s);
1412 break;
1413 case TOK_INCLUDE:
1414 case TOK_INCLUDE_NEXT:
1415 ch = file->buf_ptr[0];
1416 /* XXX: incorrect if comments : use next_nomacro with a special mode */
1417 skip_spaces();
1418 if (ch == '<') {
1419 c = '>';
1420 goto read_name;
1421 } else if (ch == '\"') {
1422 c = ch;
1423 read_name:
1424 inp();
1425 q = buf;
1426 while (ch != c && ch != '\n' && ch != CH_EOF) {
1427 if ((q - buf) < sizeof(buf) - 1)
1428 *q++ = ch;
1429 if (ch == '\\') {
1430 if (handle_stray_noerror() == 0)
1431 --q;
1432 } else
1433 inp();
1434 }
1435 *q = '\0';
1436 minp();
1437 #if 0
1438 /* eat all spaces and comments after include */
1439 /* XXX: slightly incorrect */
1440 while (ch1 != '\n' && ch1 != CH_EOF)
1441 inp();
1442 #endif
1443 } else {
1444 /* computed #include : either we have only strings or
1445 we have anything enclosed in '<>' */
1446 next();
1447 buf[0] = '\0';
1448 if (tok == TOK_STR) {
1449 while (tok != TOK_LINEFEED) {
1450 if (tok != TOK_STR) {
1451 include_syntax:
1452 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1453 }
1454 pstrcat(buf, sizeof(buf), (char *)tokc.cstr->data);
1455 next();
1456 }
1457 c = '\"';
1458 } else {
1459 int len;
1460 while (tok != TOK_LINEFEED) {
1461 pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1462 next();
1463 }
1464 len = strlen(buf);
1465 /* check syntax and remove '<>' */
1466 if (len < 2 || buf[0] != '<' || buf[len - 1] != '>')
1467 goto include_syntax;
1468 memmove(buf, buf + 1, len - 2);
1469 buf[len - 2] = '\0';
1470 c = '>';
1471 }
1472 }
1473
1474 if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1475 tcc_error("#include recursion too deep");
1476 /* store current file in stack, but increment stack later below */
1477 *s1->include_stack_ptr = file;
1478
1479 n = s1->nb_include_paths + s1->nb_sysinclude_paths;
1480 for (i = -2; i < n; ++i) {
1481 char buf1[sizeof file->filename];
1482 CachedInclude *e;
1483 BufferedFile **f;
1484 const char *path;
1485
1486 if (i == -2) {
1487 /* check absolute include path */
1488 if (!IS_ABSPATH(buf))
1489 continue;
1490 buf1[0] = 0;
1491 i = n; /* force end loop */
1492
1493 } else if (i == -1) {
1494 /* search in current dir if "header.h" */
1495 if (c != '\"')
1496 continue;
1497 path = file->filename;
1498 pstrncpy(buf1, path, tcc_basename(path) - path);
1499
1500 } else {
1501 /* search in all the include paths */
1502 if (i < s1->nb_include_paths)
1503 path = s1->include_paths[i];
1504 else
1505 path = s1->sysinclude_paths[i - s1->nb_include_paths];
1506 pstrcpy(buf1, sizeof(buf1), path);
1507 pstrcat(buf1, sizeof(buf1), "/");
1508 }
1509
1510 pstrcat(buf1, sizeof(buf1), buf);
1511
1512 if (tok == TOK_INCLUDE_NEXT)
1513 for (f = s1->include_stack_ptr; f >= s1->include_stack; --f)
1514 if (0 == PATHCMP((*f)->filename, buf1)) {
1515 #ifdef INC_DEBUG
1516 printf("%s: #include_next skipping %s\n", file->filename, buf1);
1517 #endif
1518 goto include_trynext;
1519 }
1520
1521 e = search_cached_include(s1, buf1);
1522 if (e && define_find(e->ifndef_macro)) {
1523 /* no need to parse the include because the 'ifndef macro'
1524 is defined */
1525 #ifdef INC_DEBUG
1526 printf("%s: skipping cached %s\n", file->filename, buf1);
1527 #endif
1528 goto include_done;
1529 }
1530
1531 if (tcc_open(s1, buf1) < 0)
1532 include_trynext:
1533 continue;
1534
1535 #ifdef INC_DEBUG
1536 printf("%s: including %s\n", file->prev->filename, file->filename);
1537 #endif
1538 /* update target deps */
1539 dynarray_add((void ***)&s1->target_deps, &s1->nb_target_deps,
1540 tcc_strdup(buf1));
1541 /* push current file in stack */
1542 ++s1->include_stack_ptr;
1543 /* add include file debug info */
1544 if (s1->do_debug)
1545 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1546 tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1547 ch = file->buf_ptr[0];
1548 goto the_end;
1549 }
1550 tcc_error("include file '%s' not found", buf);
1551 include_done:
1552 break;
1553 case TOK_IFNDEF:
1554 c = 1;
1555 goto do_ifdef;
1556 case TOK_IF:
1557 c = expr_preprocess();
1558 goto do_if;
1559 case TOK_IFDEF:
1560 c = 0;
1561 do_ifdef:
1562 next_nomacro();
1563 if (tok < TOK_IDENT)
1564 tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1565 if (is_bof) {
1566 if (c) {
1567 #ifdef INC_DEBUG
1568 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1569 #endif
1570 file->ifndef_macro = tok;
1571 }
1572 }
1573 c = (define_find(tok) != 0) ^ c;
1574 do_if:
1575 if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1576 tcc_error("memory full (ifdef)");
1577 *s1->ifdef_stack_ptr++ = c;
1578 goto test_skip;
1579 case TOK_ELSE:
1580 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1581 tcc_error("#else without matching #if");
1582 if (s1->ifdef_stack_ptr[-1] & 2)
1583 tcc_error("#else after #else");
1584 c = (s1->ifdef_stack_ptr[-1] ^= 3);
1585 goto test_else;
1586 case TOK_ELIF:
1587 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1588 tcc_error("#elif without matching #if");
1589 c = s1->ifdef_stack_ptr[-1];
1590 if (c > 1)
1591 tcc_error("#elif after #else");
1592 /* last #if/#elif expression was true: we skip */
1593 if (c == 1)
1594 goto skip;
1595 c = expr_preprocess();
1596 s1->ifdef_stack_ptr[-1] = c;
1597 test_else:
1598 if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1599 file->ifndef_macro = 0;
1600 test_skip:
1601 if (!(c & 1)) {
1602 skip:
1603 preprocess_skip();
1604 is_bof = 0;
1605 goto redo;
1606 }
1607 break;
1608 case TOK_ENDIF:
1609 if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1610 tcc_error("#endif without matching #if");
1611 s1->ifdef_stack_ptr--;
1612 /* '#ifndef macro' was at the start of file. Now we check if
1613 an '#endif' is exactly at the end of file */
1614 if (file->ifndef_macro &&
1615 s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1616 file->ifndef_macro_saved = file->ifndef_macro;
1617 /* need to set to zero to avoid false matches if another
1618 #ifndef at middle of file */
1619 file->ifndef_macro = 0;
1620 while (tok != TOK_LINEFEED)
1621 next_nomacro();
1622 tok_flags |= TOK_FLAG_ENDIF;
1623 goto the_end;
1624 }
1625 break;
1626 case TOK_LINE:
1627 next();
1628 if (tok != TOK_CINT)
1629 tcc_error("#line");
1630 file->line_num = tokc.i - 1; /* the line number will be incremented after */
1631 next();
1632 if (tok != TOK_LINEFEED) {
1633 if (tok != TOK_STR)
1634 tcc_error("#line");
1635 pstrcpy(file->filename, sizeof(file->filename),
1636 (char *)tokc.cstr->data);
1637 }
1638 break;
1639 case TOK_ERROR:
1640 case TOK_WARNING:
1641 c = tok;
1642 ch = file->buf_ptr[0];
1643 skip_spaces();
1644 q = buf;
1645 while (ch != '\n' && ch != CH_EOF) {
1646 if ((q - buf) < sizeof(buf) - 1)
1647 *q++ = ch;
1648 if (ch == '\\') {
1649 if (handle_stray_noerror() == 0)
1650 --q;
1651 } else
1652 inp();
1653 }
1654 *q = '\0';
1655 if (c == TOK_ERROR)
1656 tcc_error("#error %s", buf);
1657 else
1658 tcc_warning("#warning %s", buf);
1659 break;
1660 case TOK_PRAGMA:
1661 pragma_parse(s1);
1662 break;
1663 default:
1664 if (tok == TOK_LINEFEED || tok == '!' || tok == TOK_PPNUM) {
1665 /* '!' is ignored to allow C scripts. numbers are ignored
1666 to emulate cpp behaviour */
1667 } else {
1668 if (!(saved_parse_flags & PARSE_FLAG_ASM_COMMENTS))
1669 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
1670 else {
1671 /* this is a gas line comment in an 'S' file. */
1672 file->buf_ptr = parse_line_comment(file->buf_ptr);
1673 goto the_end;
1674 }
1675 }
1676 break;
1677 }
1678 /* ignore other preprocess commands or #! for C scripts */
1679 while (tok != TOK_LINEFEED)
1680 next_nomacro();
1681 the_end:
1682 parse_flags = saved_parse_flags;
1683 }
1684
1685 /* evaluate escape codes in a string. */
1686 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
1687 {
1688 int c, n;
1689 const uint8_t *p;
1690
1691 p = buf;
1692 for(;;) {
1693 c = *p;
1694 if (c == '\0')
1695 break;
1696 if (c == '\\') {
1697 p++;
1698 /* escape */
1699 c = *p;
1700 switch(c) {
1701 case '0': case '1': case '2': case '3':
1702 case '4': case '5': case '6': case '7':
1703 /* at most three octal digits */
1704 n = c - '0';
1705 p++;
1706 c = *p;
1707 if (isoct(c)) {
1708 n = n * 8 + c - '0';
1709 p++;
1710 c = *p;
1711 if (isoct(c)) {
1712 n = n * 8 + c - '0';
1713 p++;
1714 }
1715 }
1716 c = n;
1717 goto add_char_nonext;
1718 case 'x':
1719 case 'u':
1720 case 'U':
1721 p++;
1722 n = 0;
1723 for(;;) {
1724 c = *p;
1725 if (c >= 'a' && c <= 'f')
1726 c = c - 'a' + 10;
1727 else if (c >= 'A' && c <= 'F')
1728 c = c - 'A' + 10;
1729 else if (isnum(c))
1730 c = c - '0';
1731 else
1732 break;
1733 n = n * 16 + c;
1734 p++;
1735 }
1736 c = n;
1737 goto add_char_nonext;
1738 case 'a':
1739 c = '\a';
1740 break;
1741 case 'b':
1742 c = '\b';
1743 break;
1744 case 'f':
1745 c = '\f';
1746 break;
1747 case 'n':
1748 c = '\n';
1749 break;
1750 case 'r':
1751 c = '\r';
1752 break;
1753 case 't':
1754 c = '\t';
1755 break;
1756 case 'v':
1757 c = '\v';
1758 break;
1759 case 'e':
1760 if (!gnu_ext)
1761 goto invalid_escape;
1762 c = 27;
1763 break;
1764 case '\'':
1765 case '\"':
1766 case '\\':
1767 case '?':
1768 break;
1769 default:
1770 invalid_escape:
1771 if (c >= '!' && c <= '~')
1772 tcc_warning("unknown escape sequence: \'\\%c\'", c);
1773 else
1774 tcc_warning("unknown escape sequence: \'\\x%x\'", c);
1775 break;
1776 }
1777 }
1778 p++;
1779 add_char_nonext:
1780 if (!is_long)
1781 cstr_ccat(outstr, c);
1782 else
1783 cstr_wccat(outstr, c);
1784 }
1785 /* add a trailing '\0' */
1786 if (!is_long)
1787 cstr_ccat(outstr, '\0');
1788 else
1789 cstr_wccat(outstr, '\0');
1790 }
1791
1792 /* we use 64 bit numbers */
1793 #define BN_SIZE 2
1794
1795 /* bn = (bn << shift) | or_val */
1796 static void bn_lshift(unsigned int *bn, int shift, int or_val)
1797 {
1798 int i;
1799 unsigned int v;
1800 for(i=0;i<BN_SIZE;i++) {
1801 v = bn[i];
1802 bn[i] = (v << shift) | or_val;
1803 or_val = v >> (32 - shift);
1804 }
1805 }
1806
1807 static void bn_zero(unsigned int *bn)
1808 {
1809 int i;
1810 for(i=0;i<BN_SIZE;i++) {
1811 bn[i] = 0;
1812 }
1813 }
1814
1815 /* parse number in null terminated string 'p' and return it in the
1816 current token */
1817 static void parse_number(const char *p)
1818 {
1819 int b, t, shift, frac_bits, s, exp_val, ch;
1820 char *q;
1821 unsigned int bn[BN_SIZE];
1822 double d;
1823
1824 /* number */
1825 q = token_buf;
1826 ch = *p++;
1827 t = ch;
1828 ch = *p++;
1829 *q++ = t;
1830 b = 10;
1831 if (t == '.') {
1832 goto float_frac_parse;
1833 } else if (t == '0') {
1834 if (ch == 'x' || ch == 'X') {
1835 q--;
1836 ch = *p++;
1837 b = 16;
1838 } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
1839 q--;
1840 ch = *p++;
1841 b = 2;
1842 }
1843 }
1844 /* parse all digits. cannot check octal numbers at this stage
1845 because of floating point constants */
1846 while (1) {
1847 if (ch >= 'a' && ch <= 'f')
1848 t = ch - 'a' + 10;
1849 else if (ch >= 'A' && ch <= 'F')
1850 t = ch - 'A' + 10;
1851 else if (isnum(ch))
1852 t = ch - '0';
1853 else
1854 break;
1855 if (t >= b)
1856 break;
1857 if (q >= token_buf + STRING_MAX_SIZE) {
1858 num_too_long:
1859 tcc_error("number too long");
1860 }
1861 *q++ = ch;
1862 ch = *p++;
1863 }
1864 if (ch == '.' ||
1865 ((ch == 'e' || ch == 'E') && b == 10) ||
1866 ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
1867 if (b != 10) {
1868 /* NOTE: strtox should support that for hexa numbers, but
1869 non ISOC99 libcs do not support it, so we prefer to do
1870 it by hand */
1871 /* hexadecimal or binary floats */
1872 /* XXX: handle overflows */
1873 *q = '\0';
1874 if (b == 16)
1875 shift = 4;
1876 else
1877 shift = 2;
1878 bn_zero(bn);
1879 q = token_buf;
1880 while (1) {
1881 t = *q++;
1882 if (t == '\0') {
1883 break;
1884 } else if (t >= 'a') {
1885 t = t - 'a' + 10;
1886 } else if (t >= 'A') {
1887 t = t - 'A' + 10;
1888 } else {
1889 t = t - '0';
1890 }
1891 bn_lshift(bn, shift, t);
1892 }
1893 frac_bits = 0;
1894 if (ch == '.') {
1895 ch = *p++;
1896 while (1) {
1897 t = ch;
1898 if (t >= 'a' && t <= 'f') {
1899 t = t - 'a' + 10;
1900 } else if (t >= 'A' && t <= 'F') {
1901 t = t - 'A' + 10;
1902 } else if (t >= '0' && t <= '9') {
1903 t = t - '0';
1904 } else {
1905 break;
1906 }
1907 if (t >= b)
1908 tcc_error("invalid digit");
1909 bn_lshift(bn, shift, t);
1910 frac_bits += shift;
1911 ch = *p++;
1912 }
1913 }
1914 if (ch != 'p' && ch != 'P')
1915 expect("exponent");
1916 ch = *p++;
1917 s = 1;
1918 exp_val = 0;
1919 if (ch == '+') {
1920 ch = *p++;
1921 } else if (ch == '-') {
1922 s = -1;
1923 ch = *p++;
1924 }
1925 if (ch < '0' || ch > '9')
1926 expect("exponent digits");
1927 while (ch >= '0' && ch <= '9') {
1928 exp_val = exp_val * 10 + ch - '0';
1929 ch = *p++;
1930 }
1931 exp_val = exp_val * s;
1932
1933 /* now we can generate the number */
1934 /* XXX: should patch directly float number */
1935 d = (double)bn[1] * 4294967296.0 + (double)bn[0];
1936 d = ldexp(d, exp_val - frac_bits);
1937 t = toup(ch);
1938 if (t == 'F') {
1939 ch = *p++;
1940 tok = TOK_CFLOAT;
1941 /* float : should handle overflow */
1942 tokc.f = (float)d;
1943 } else if (t == 'L') {
1944 ch = *p++;
1945 #ifdef TCC_TARGET_PE
1946 tok = TOK_CDOUBLE;
1947 tokc.d = d;
1948 #else
1949 tok = TOK_CLDOUBLE;
1950 /* XXX: not large enough */
1951 tokc.ld = (long double)d;
1952 #endif
1953 } else {
1954 tok = TOK_CDOUBLE;
1955 tokc.d = d;
1956 }
1957 } else {
1958 /* decimal floats */
1959 if (ch == '.') {
1960 if (q >= token_buf + STRING_MAX_SIZE)
1961 goto num_too_long;
1962 *q++ = ch;
1963 ch = *p++;
1964 float_frac_parse:
1965 while (ch >= '0' && ch <= '9') {
1966 if (q >= token_buf + STRING_MAX_SIZE)
1967 goto num_too_long;
1968 *q++ = ch;
1969 ch = *p++;
1970 }
1971 }
1972 if (ch == 'e' || ch == 'E') {
1973 if (q >= token_buf + STRING_MAX_SIZE)
1974 goto num_too_long;
1975 *q++ = ch;
1976 ch = *p++;
1977 if (ch == '-' || ch == '+') {
1978 if (q >= token_buf + STRING_MAX_SIZE)
1979 goto num_too_long;
1980 *q++ = ch;
1981 ch = *p++;
1982 }
1983 if (ch < '0' || ch > '9')
1984 expect("exponent digits");
1985 while (ch >= '0' && ch <= '9') {
1986 if (q >= token_buf + STRING_MAX_SIZE)
1987 goto num_too_long;
1988 *q++ = ch;
1989 ch = *p++;
1990 }
1991 }
1992 *q = '\0';
1993 t = toup(ch);
1994 errno = 0;
1995 if (t == 'F') {
1996 ch = *p++;
1997 tok = TOK_CFLOAT;
1998 tokc.f = strtof(token_buf, NULL);
1999 } else if (t == 'L') {
2000 ch = *p++;
2001 #ifdef TCC_TARGET_PE
2002 tok = TOK_CDOUBLE;
2003 tokc.d = strtod(token_buf, NULL);
2004 #else
2005 tok = TOK_CLDOUBLE;
2006 tokc.ld = strtold(token_buf, NULL);
2007 #endif
2008 } else {
2009 tok = TOK_CDOUBLE;
2010 tokc.d = strtod(token_buf, NULL);
2011 }
2012 }
2013 } else {
2014 unsigned long long n, n1;
2015 int lcount, ucount;
2016
2017 /* integer number */
2018 *q = '\0';
2019 q = token_buf;
2020 if (b == 10 && *q == '0') {
2021 b = 8;
2022 q++;
2023 }
2024 n = 0;
2025 while(1) {
2026 t = *q++;
2027 /* no need for checks except for base 10 / 8 errors */
2028 if (t == '\0') {
2029 break;
2030 } else if (t >= 'a') {
2031 t = t - 'a' + 10;
2032 } else if (t >= 'A') {
2033 t = t - 'A' + 10;
2034 } else {
2035 t = t - '0';
2036 if (t >= b)
2037 tcc_error("invalid digit");
2038 }
2039 n1 = n;
2040 n = n * b + t;
2041 /* detect overflow */
2042 /* XXX: this test is not reliable */
2043 if (n < n1)
2044 tcc_error("integer constant overflow");
2045 }
2046
2047 /* XXX: not exactly ANSI compliant */
2048 if ((n & 0xffffffff00000000LL) != 0) {
2049 if ((n >> 63) != 0)
2050 tok = TOK_CULLONG;
2051 else
2052 tok = TOK_CLLONG;
2053 } else if (n > 0x7fffffff) {
2054 tok = TOK_CUINT;
2055 } else {
2056 tok = TOK_CINT;
2057 }
2058 lcount = 0;
2059 ucount = 0;
2060 for(;;) {
2061 t = toup(ch);
2062 if (t == 'L') {
2063 if (lcount >= 2)
2064 tcc_error("three 'l's in integer constant");
2065 lcount++;
2066 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2067 if (lcount == 2) {
2068 #endif
2069 if (tok == TOK_CINT)
2070 tok = TOK_CLLONG;
2071 else if (tok == TOK_CUINT)
2072 tok = TOK_CULLONG;
2073 #if !defined TCC_TARGET_X86_64 || defined TCC_TARGET_PE
2074 }
2075 #endif
2076 ch = *p++;
2077 } else if (t == 'U') {
2078 if (ucount >= 1)
2079 tcc_error("two 'u's in integer constant");
2080 ucount++;
2081 if (tok == TOK_CINT)
2082 tok = TOK_CUINT;
2083 else if (tok == TOK_CLLONG)
2084 tok = TOK_CULLONG;
2085 ch = *p++;
2086 } else {
2087 break;
2088 }
2089 }
2090 if (tok == TOK_CINT || tok == TOK_CUINT)
2091 tokc.ui = n;
2092 else
2093 tokc.ull = n;
2094 }
2095 if (ch)
2096 tcc_error("invalid number\n");
2097 }
2098
2099
2100 #define PARSE2(c1, tok1, c2, tok2) \
2101 case c1: \
2102 PEEKC(c, p); \
2103 if (c == c2) { \
2104 p++; \
2105 tok = tok2; \
2106 } else { \
2107 tok = tok1; \
2108 } \
2109 break;
2110
2111 /* return next token without macro substitution */
2112 static inline void next_nomacro1(void)
2113 {
2114 int t, c, is_long;
2115 TokenSym *ts;
2116 uint8_t *p, *p1;
2117 unsigned int h;
2118
2119 p = file->buf_ptr;
2120 redo_no_start:
2121 c = *p;
2122 switch(c) {
2123 case ' ':
2124 case '\t':
2125 tok = c;
2126 p++;
2127 goto keep_tok_flags;
2128 case '\f':
2129 case '\v':
2130 case '\r':
2131 p++;
2132 goto redo_no_start;
2133 case '\\':
2134 /* first look if it is in fact an end of buffer */
2135 if (p >= file->buf_end) {
2136 file->buf_ptr = p;
2137 handle_eob();
2138 p = file->buf_ptr;
2139 if (p >= file->buf_end)
2140 goto parse_eof;
2141 else
2142 goto redo_no_start;
2143 } else {
2144 file->buf_ptr = p;
2145 ch = *p;
2146 handle_stray();
2147 p = file->buf_ptr;
2148 goto redo_no_start;
2149 }
2150 parse_eof:
2151 {
2152 TCCState *s1 = tcc_state;
2153 if ((parse_flags & PARSE_FLAG_LINEFEED)
2154 && !(tok_flags & TOK_FLAG_EOF)) {
2155 tok_flags |= TOK_FLAG_EOF;
2156 tok = TOK_LINEFEED;
2157 goto keep_tok_flags;
2158 } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2159 tok = TOK_EOF;
2160 } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2161 tcc_error("missing #endif");
2162 } else if (s1->include_stack_ptr == s1->include_stack) {
2163 /* no include left : end of file. */
2164 tok = TOK_EOF;
2165 } else {
2166 tok_flags &= ~TOK_FLAG_EOF;
2167 /* pop include file */
2168
2169 /* test if previous '#endif' was after a #ifdef at
2170 start of file */
2171 if (tok_flags & TOK_FLAG_ENDIF) {
2172 #ifdef INC_DEBUG
2173 printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2174 #endif
2175 add_cached_include(s1, file->filename, file->ifndef_macro_saved);
2176 tok_flags &= ~TOK_FLAG_ENDIF;
2177 }
2178
2179 /* add end of include file debug info */
2180 if (tcc_state->do_debug) {
2181 put_stabd(N_EINCL, 0, 0);
2182 }
2183 /* pop include stack */
2184 tcc_close();
2185 s1->include_stack_ptr--;
2186 p = file->buf_ptr;
2187 goto redo_no_start;
2188 }
2189 }
2190 break;
2191
2192 case '\n':
2193 file->line_num++;
2194 tok_flags |= TOK_FLAG_BOL;
2195 p++;
2196 maybe_newline:
2197 if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2198 goto redo_no_start;
2199 tok = TOK_LINEFEED;
2200 goto keep_tok_flags;
2201
2202 case '#':
2203 /* XXX: simplify */
2204 PEEKC(c, p);
2205 if ((tok_flags & TOK_FLAG_BOL) &&
2206 (parse_flags & PARSE_FLAG_PREPROCESS)) {
2207 file->buf_ptr = p;
2208 preprocess(tok_flags & TOK_FLAG_BOF);
2209 p = file->buf_ptr;
2210 goto maybe_newline;
2211 } else {
2212 if (c == '#') {
2213 p++;
2214 tok = TOK_TWOSHARPS;
2215 } else {
2216 if (parse_flags & PARSE_FLAG_ASM_COMMENTS) {
2217 p = parse_line_comment(p - 1);
2218 goto redo_no_start;
2219 } else {
2220 tok = '#';
2221 }
2222 }
2223 }
2224 break;
2225
2226 case 'a': case 'b': case 'c': case 'd':
2227 case 'e': case 'f': case 'g': case 'h':
2228 case 'i': case 'j': case 'k': case 'l':
2229 case 'm': case 'n': case 'o': case 'p':
2230 case 'q': case 'r': case 's': case 't':
2231 case 'u': case 'v': case 'w': case 'x':
2232 case 'y': case 'z':
2233 case 'A': case 'B': case 'C': case 'D':
2234 case 'E': case 'F': case 'G': case 'H':
2235 case 'I': case 'J': case 'K':
2236 case 'M': case 'N': case 'O': case 'P':
2237 case 'Q': case 'R': case 'S': case 'T':
2238 case 'U': case 'V': case 'W': case 'X':
2239 case 'Y': case 'Z':
2240 case '_':
2241 parse_ident_fast:
2242 p1 = p;
2243 h = TOK_HASH_INIT;
2244 h = TOK_HASH_FUNC(h, c);
2245 p++;
2246 for(;;) {
2247 c = *p;
2248 if (!isidnum_table[c-CH_EOF])
2249 break;
2250 h = TOK_HASH_FUNC(h, c);
2251 p++;
2252 }
2253 if (c != '\\') {
2254 TokenSym **pts;
2255 int len;
2256
2257 /* fast case : no stray found, so we have the full token
2258 and we have already hashed it */
2259 len = p - p1;
2260 h &= (TOK_HASH_SIZE - 1);
2261 pts = &hash_ident[h];
2262 for(;;) {
2263 ts = *pts;
2264 if (!ts)
2265 break;
2266 if (ts->len == len && !memcmp(ts->str, p1, len))
2267 goto token_found;
2268 pts = &(ts->hash_next);
2269 }
2270 ts = tok_alloc_new(pts, (char *) p1, len);
2271 token_found: ;
2272 } else {
2273 /* slower case */
2274 cstr_reset(&tokcstr);
2275
2276 while (p1 < p) {
2277 cstr_ccat(&tokcstr, *p1);
2278 p1++;
2279 }
2280 p--;
2281 PEEKC(c, p);
2282 parse_ident_slow:
2283 while (isidnum_table[c-CH_EOF]) {
2284 cstr_ccat(&tokcstr, c);
2285 PEEKC(c, p);
2286 }
2287 ts = tok_alloc(tokcstr.data, tokcstr.size);
2288 }
2289 tok = ts->tok;
2290 break;
2291 case 'L':
2292 t = p[1];
2293 if (t != '\\' && t != '\'' && t != '\"') {
2294 /* fast case */
2295 goto parse_ident_fast;
2296 } else {
2297 PEEKC(c, p);
2298 if (c == '\'' || c == '\"') {
2299 is_long = 1;
2300 goto str_const;
2301 } else {
2302 cstr_reset(&tokcstr);
2303 cstr_ccat(&tokcstr, 'L');
2304 goto parse_ident_slow;
2305 }
2306 }
2307 break;
2308 case '0': case '1': case '2': case '3':
2309 case '4': case '5': case '6': case '7':
2310 case '8': case '9':
2311
2312 cstr_reset(&tokcstr);
2313 /* after the first digit, accept digits, alpha, '.' or sign if
2314 prefixed by 'eEpP' */
2315 parse_num:
2316 for(;;) {
2317 t = c;
2318 cstr_ccat(&tokcstr, c);
2319 PEEKC(c, p);
2320 if (!(isnum(c) || isid(c) || c == '.' ||
2321 ((c == '+' || c == '-') &&
2322 (t == 'e' || t == 'E' || t == 'p' || t == 'P'))))
2323 break;
2324 }
2325 /* We add a trailing '\0' to ease parsing */
2326 cstr_ccat(&tokcstr, '\0');
2327 tokc.cstr = &tokcstr;
2328 tok = TOK_PPNUM;
2329 break;
2330 case '.':
2331 /* special dot handling because it can also start a number */
2332 PEEKC(c, p);
2333 if (isnum(c)) {
2334 cstr_reset(&tokcstr);
2335 cstr_ccat(&tokcstr, '.');
2336 goto parse_num;
2337 } else if (c == '.') {
2338 PEEKC(c, p);
2339 if (c != '.')
2340 expect("'.'");
2341 PEEKC(c, p);
2342 tok = TOK_DOTS;
2343 } else {
2344 tok = '.';
2345 }
2346 break;
2347 case '\'':
2348 case '\"':
2349 is_long = 0;
2350 str_const:
2351 {
2352 CString str;
2353 int sep;
2354
2355 sep = c;
2356
2357 /* parse the string */
2358 cstr_new(&str);
2359 p = parse_pp_string(p, sep, &str);
2360 cstr_ccat(&str, '\0');
2361
2362 /* eval the escape (should be done as TOK_PPNUM) */
2363 cstr_reset(&tokcstr);
2364 parse_escape_string(&tokcstr, str.data, is_long);
2365 cstr_free(&str);
2366
2367 if (sep == '\'') {
2368 int char_size;
2369 /* XXX: make it portable */
2370 if (!is_long)
2371 char_size = 1;
2372 else
2373 char_size = sizeof(nwchar_t);
2374 if (tokcstr.size <= char_size)
2375 tcc_error("empty character constant");
2376 if (tokcstr.size > 2 * char_size)
2377 tcc_warning("multi-character character constant");
2378 if (!is_long) {
2379 tokc.i = *(int8_t *)tokcstr.data;
2380 tok = TOK_CCHAR;
2381 } else {
2382 tokc.i = *(nwchar_t *)tokcstr.data;
2383 tok = TOK_LCHAR;
2384 }
2385 } else {
2386 tokc.cstr = &tokcstr;
2387 if (!is_long)
2388 tok = TOK_STR;
2389 else
2390 tok = TOK_LSTR;
2391 }
2392 }
2393 break;
2394
2395 case '<':
2396 PEEKC(c, p);
2397 if (c == '=') {
2398 p++;
2399 tok = TOK_LE;
2400 } else if (c == '<') {
2401 PEEKC(c, p);
2402 if (c == '=') {
2403 p++;
2404 tok = TOK_A_SHL;
2405 } else {
2406 tok = TOK_SHL;
2407 }
2408 } else {
2409 tok = TOK_LT;
2410 }
2411 break;
2412
2413 case '>':
2414 PEEKC(c, p);
2415 if (c == '=') {
2416 p++;
2417 tok = TOK_GE;
2418 } else if (c == '>') {
2419 PEEKC(c, p);
2420 if (c == '=') {
2421 p++;
2422 tok = TOK_A_SAR;
2423 } else {
2424 tok = TOK_SAR;
2425 }
2426 } else {
2427 tok = TOK_GT;
2428 }
2429 break;
2430
2431 case '&':
2432 PEEKC(c, p);
2433 if (c == '&') {
2434 p++;
2435 tok = TOK_LAND;
2436 } else if (c == '=') {
2437 p++;
2438 tok = TOK_A_AND;
2439 } else {
2440 tok = '&';
2441 }
2442 break;
2443
2444 case '|':
2445 PEEKC(c, p);
2446 if (c == '|') {
2447 p++;
2448 tok = TOK_LOR;
2449 } else if (c == '=') {
2450 p++;
2451 tok = TOK_A_OR;
2452 } else {
2453 tok = '|';
2454 }
2455 break;
2456
2457 case '+':
2458 PEEKC(c, p);
2459 if (c == '+') {
2460 p++;
2461 tok = TOK_INC;
2462 } else if (c == '=') {
2463 p++;
2464 tok = TOK_A_ADD;
2465 } else {
2466 tok = '+';
2467 }
2468 break;
2469
2470 case '-':
2471 PEEKC(c, p);
2472 if (c == '-') {
2473 p++;
2474 tok = TOK_DEC;
2475 } else if (c == '=') {
2476 p++;
2477 tok = TOK_A_SUB;
2478 } else if (c == '>') {
2479 p++;
2480 tok = TOK_ARROW;
2481 } else {
2482 tok = '-';
2483 }
2484 break;
2485
2486 PARSE2('!', '!', '=', TOK_NE)
2487 PARSE2('=', '=', '=', TOK_EQ)
2488 PARSE2('*', '*', '=', TOK_A_MUL)
2489 PARSE2('%', '%', '=', TOK_A_MOD)
2490 PARSE2('^', '^', '=', TOK_A_XOR)
2491
2492 /* comments or operator */
2493 case '/':
2494 PEEKC(c, p);
2495 if (c == '*') {
2496 p = parse_comment(p);
2497 /* comments replaced by a blank */
2498 tok = ' ';
2499 goto keep_tok_flags;
2500 } else if (c == '/') {
2501 p = parse_line_comment(p);
2502 tok = ' ';
2503 goto keep_tok_flags;
2504 } else if (c == '=') {
2505 p++;
2506 tok = TOK_A_DIV;
2507 } else {
2508 tok = '/';
2509 }
2510 break;
2511
2512 /* simple tokens */
2513 case '(':
2514 case ')':
2515 case '[':
2516 case ']':
2517 case '{':
2518 case '}':
2519 case ',':
2520 case ';':
2521 case ':':
2522 case '?':
2523 case '~':
2524 case '$': /* only used in assembler */
2525 case '@': /* dito */
2526 tok = c;
2527 p++;
2528 break;
2529 default:
2530 tcc_error("unrecognized character \\x%02x", c);
2531 break;
2532 }
2533 tok_flags = 0;
2534 keep_tok_flags:
2535 file->buf_ptr = p;
2536 #if defined(PARSE_DEBUG)
2537 printf("token = %s\n", get_tok_str(tok, &tokc));
2538 #endif
2539 }
2540
2541 /* return next token without macro substitution. Can read input from
2542 macro_ptr buffer */
2543 static void next_nomacro_spc(void)
2544 {
2545 if (macro_ptr) {
2546 redo:
2547 tok = *macro_ptr;
2548 if (tok) {
2549 TOK_GET(&tok, &macro_ptr, &tokc);
2550 if (tok == TOK_LINENUM) {
2551 file->line_num = tokc.i;
2552 goto redo;
2553 }
2554 }
2555 } else {
2556 next_nomacro1();
2557 }
2558 }
2559
2560 ST_FUNC void next_nomacro(void)
2561 {
2562 do {
2563 next_nomacro_spc();
2564 } while (is_space(tok));
2565 }
2566
2567 /* substitute arguments in replacement lists in macro_str by the values in
2568 args (field d) and return allocated string */
2569 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2570 {
2571 int last_tok, t, spc;
2572 const int *st;
2573 Sym *s;
2574 CValue cval;
2575 TokenString str;
2576 CString cstr;
2577
2578 tok_str_new(&str);
2579 last_tok = 0;
2580 while(1) {
2581 TOK_GET(&t, &macro_str, &cval);
2582 if (!t)
2583 break;
2584 if (t == '#') {
2585 /* stringize */
2586 TOK_GET(&t, &macro_str, &cval);
2587 if (!t)
2588 break;
2589 s = sym_find2(args, t);
2590 if (s) {
2591 cstr_new(&cstr);
2592 st = s->d;
2593 spc = 0;
2594 while (*st) {
2595 TOK_GET(&t, &st, &cval);
2596 if (!check_space(t, &spc))
2597 cstr_cat(&cstr, get_tok_str(t, &cval));
2598 }
2599 cstr.size -= spc;
2600 cstr_ccat(&cstr, '\0');
2601 #ifdef PP_DEBUG
2602 printf("stringize: %s\n", (char *)cstr.data);
2603 #endif
2604 /* add string */
2605 cval.cstr = &cstr;
2606 tok_str_add2(&str, TOK_STR, &cval);
2607 cstr_free(&cstr);
2608 } else {
2609 tok_str_add2(&str, t, &cval);
2610 }
2611 } else if (t >= TOK_IDENT) {
2612 s = sym_find2(args, t);
2613 if (s) {
2614 st = s->d;
2615 /* if '##' is present before or after, no arg substitution */
2616 if (*macro_str == TOK_TWOSHARPS || last_tok == TOK_TWOSHARPS) {
2617 /* special case for var arg macros : ## eats the
2618 ',' if empty VA_ARGS variable. */
2619 /* XXX: test of the ',' is not 100%
2620 reliable. should fix it to avoid security
2621 problems */
2622 if (gnu_ext && s->type.t &&
2623 last_tok == TOK_TWOSHARPS &&
2624 str.len >= 2 && str.str[str.len - 2] == ',') {
2625 if (*st == TOK_PLCHLDR) {
2626 /* suppress ',' '##' */
2627 str.len -= 2;
2628 } else {
2629 /* suppress '##' and add variable */
2630 str.len--;
2631 goto add_var;
2632 }
2633 } else {
2634 int t1;
2635 add_var:
2636 for(;;) {
2637 TOK_GET(&t1, &st, &cval);
2638 if (!t1)
2639 break;
2640 tok_str_add2(&str, t1, &cval);
2641 }
2642 }
2643 } else {
2644 /* NOTE: the stream cannot be read when macro
2645 substituing an argument */
2646 macro_subst(&str, nested_list, st, NULL);
2647 }
2648 } else {
2649 tok_str_add(&str, t);
2650 }
2651 } else {
2652 tok_str_add2(&str, t, &cval);
2653 }
2654 last_tok = t;
2655 }
2656 tok_str_add(&str, 0);
2657 return str.str;
2658 }
2659
2660 static char const ab_month_name[12][4] =
2661 {
2662 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
2663 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
2664 };
2665
2666 /* do macro substitution of current token with macro 's' and add
2667 result to (tok_str,tok_len). 'nested_list' is the list of all
2668 macros we got inside to avoid recursing. Return non zero if no
2669 substitution needs to be done */
2670 static int macro_subst_tok(TokenString *tok_str,
2671 Sym **nested_list, Sym *s, struct macro_level **can_read_stream)
2672 {
2673 Sym *args, *sa, *sa1;
2674 int mstr_allocated, parlevel, *mstr, t, t1, spc;
2675 const int *p;
2676 TokenString str;
2677 char *cstrval;
2678 CValue cval;
2679 CString cstr;
2680 char buf[32];
2681
2682 /* if symbol is a macro, prepare substitution */
2683 /* special macros */
2684 if (tok == TOK___LINE__) {
2685 snprintf(buf, sizeof(buf), "%d", file->line_num);
2686 cstrval = buf;
2687 t1 = TOK_PPNUM;
2688 goto add_cstr1;
2689 } else if (tok == TOK___FILE__) {
2690 cstrval = file->filename;
2691 goto add_cstr;
2692 } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
2693 time_t ti;
2694 struct tm *tm;
2695
2696 time(&ti);
2697 tm = localtime(&ti);
2698 if (tok == TOK___DATE__) {
2699 snprintf(buf, sizeof(buf), "%s %2d %d",
2700 ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
2701 } else {
2702 snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
2703 tm->tm_hour, tm->tm_min, tm->tm_sec);
2704 }
2705 cstrval = buf;
2706 add_cstr:
2707 t1 = TOK_STR;
2708 add_cstr1:
2709 cstr_new(&cstr);
2710 cstr_cat(&cstr, cstrval);
2711 cstr_ccat(&cstr, '\0');
2712 cval.cstr = &cstr;
2713 tok_str_add2(tok_str, t1, &cval);
2714 cstr_free(&cstr);
2715 } else {
2716 mstr = s->d;
2717 mstr_allocated = 0;
2718 if (s->type.t == MACRO_FUNC) {
2719 /* NOTE: we do not use next_nomacro to avoid eating the
2720 next token. XXX: find better solution */
2721 redo:
2722 if (macro_ptr) {
2723 p = macro_ptr;
2724 while (is_space(t = *p) || TOK_LINEFEED == t)
2725 ++p;
2726 if (t == 0 && can_read_stream) {
2727 /* end of macro stream: we must look at the token
2728 after in the file */
2729 struct macro_level *ml = *can_read_stream;
2730 macro_ptr = NULL;
2731 if (ml)
2732 {
2733 macro_ptr = ml->p;
2734 ml->p = NULL;
2735 *can_read_stream = ml -> prev;
2736 }
2737 /* also, end of scope for nested defined symbol */
2738 (*nested_list)->v = -1;
2739 goto redo;
2740 }
2741 } else {
2742 ch = file->buf_ptr[0];
2743 while (is_space(ch) || ch == '\n' || ch == '/')
2744 {
2745 if (ch == '/')
2746 {
2747 int c;
2748 uint8_t *p = file->buf_ptr;
2749 PEEKC(c, p);
2750 if (c == '*') {
2751 p = parse_comment(p);
2752 file->buf_ptr = p - 1;
2753 } else if (c == '/') {
2754 p = parse_line_comment(p);
2755 file->buf_ptr = p - 1;
2756 } else
2757 break;
2758 }
2759 cinp();
2760 }
2761 t = ch;
2762 }
2763 if (t != '(') /* no macro subst */
2764 return -1;
2765
2766 /* argument macro */
2767 next_nomacro();
2768 next_nomacro();
2769 args = NULL;
2770 sa = s->next;
2771 /* NOTE: empty args are allowed, except if no args */
2772 for(;;) {
2773 /* handle '()' case */
2774 if (!args && !sa && tok == ')')
2775 break;
2776 if (!sa)
2777 tcc_error("macro '%s' used with too many args",
2778 get_tok_str(s->v, 0));
2779 tok_str_new(&str);
2780 parlevel = spc = 0;
2781 /* NOTE: non zero sa->t indicates VA_ARGS */
2782 while ((parlevel > 0 ||
2783 (tok != ')' &&
2784 (tok != ',' || sa->type.t))) &&
2785 tok != -1) {
2786 if (tok == '(')
2787 parlevel++;
2788 else if (tok == ')')
2789 parlevel--;
2790 if (tok == TOK_LINEFEED)
2791 tok = ' ';
2792 if (!check_space(tok, &spc))
2793 tok_str_add2(&str, tok, &tokc);
2794 next_nomacro_spc();
2795 }
2796 if (!str.len)
2797 tok_str_add(&str, TOK_PLCHLDR);
2798 str.len -= spc;
2799 tok_str_add(&str, 0);
2800 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
2801 sa1->d = str.str;
2802 sa = sa->next;
2803 if (tok == ')') {
2804 /* special case for gcc var args: add an empty
2805 var arg argument if it is omitted */
2806 if (sa && sa->type.t && gnu_ext)
2807 continue;
2808 else
2809 break;
2810 }
2811 if (tok != ',')
2812 expect(",");
2813 next_nomacro();
2814 }
2815 if (sa) {
2816 tcc_error("macro '%s' used with too few args",
2817 get_tok_str(s->v, 0));
2818 }
2819
2820 /* now subst each arg */
2821 mstr = macro_arg_subst(nested_list, mstr, args);
2822 /* free memory */
2823 sa = args;
2824 while (sa) {
2825 sa1 = sa->prev;
2826 tok_str_free(sa->d);
2827 sym_free(sa);
2828 sa = sa1;
2829 }
2830 mstr_allocated = 1;
2831 }
2832 sym_push2(nested_list, s->v, 0, 0);
2833 macro_subst(tok_str, nested_list, mstr, can_read_stream);
2834 /* pop nested defined symbol */
2835 sa1 = *nested_list;
2836 *nested_list = sa1->prev;
2837 sym_free(sa1);
2838 if (mstr_allocated)
2839 tok_str_free(mstr);
2840 }
2841 return 0;
2842 }
2843
2844 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
2845 return the resulting string (which must be freed). */
2846 static inline int *macro_twosharps(const int *macro_str)
2847 {
2848 const int *ptr;
2849 int t;
2850 TokenString macro_str1;
2851 CString cstr;
2852 int n, start_of_nosubsts;
2853
2854 /* we search the first '##' */
2855 for(ptr = macro_str;;) {
2856 CValue cval;
2857 TOK_GET(&t, &ptr, &cval);
2858 if (t == TOK_TWOSHARPS)
2859 break;
2860 /* nothing more to do if end of string */
2861 if (t == 0)
2862 return NULL;
2863 }
2864
2865 /* we saw '##', so we need more processing to handle it */
2866 start_of_nosubsts = -1;
2867 tok_str_new(&macro_str1);
2868 for(ptr = macro_str;;) {
2869 TOK_GET(&tok, &ptr, &tokc);
2870 if (tok == 0)
2871 break;
2872 if (tok == TOK_TWOSHARPS)
2873 continue;
2874 if (tok == TOK_NOSUBST && start_of_nosubsts < 0)
2875 start_of_nosubsts = macro_str1.len;
2876 while (*ptr == TOK_TWOSHARPS) {
2877 /* given 'a##b', remove nosubsts preceding 'a' */
2878 if (start_of_nosubsts >= 0)
2879 macro_str1.len = start_of_nosubsts;
2880 /* given 'a##b', skip '##' */
2881 t = *++ptr;
2882 /* given 'a##b', remove nosubsts preceding 'b' */
2883 while (t == TOK_NOSUBST)
2884 t = *++ptr;
2885 if (t && t != TOK_TWOSHARPS) {
2886 CValue cval;
2887 TOK_GET(&t, &ptr, &cval);
2888 /* We concatenate the two tokens */
2889 cstr_new(&cstr);
2890 if (tok != TOK_PLCHLDR)
2891 cstr_cat(&cstr, get_tok_str(tok, &tokc));
2892 n = cstr.size;
2893 if (t != TOK_PLCHLDR || tok == TOK_PLCHLDR)
2894 cstr_cat(&cstr, get_tok_str(t, &cval));
2895 cstr_ccat(&cstr, '\0');
2896
2897 tcc_open_bf(tcc_state, ":paste:", cstr.size);
2898 memcpy(file->buffer, cstr.data, cstr.size);
2899 for (;;) {
2900 next_nomacro1();
2901 if (0 == *file->buf_ptr)
2902 break;
2903 tok_str_add2(&macro_str1, tok, &tokc);
2904 tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid preprocessing token",
2905 n, cstr.data, (char*)cstr.data + n);
2906 }
2907 tcc_close();
2908 cstr_free(&cstr);
2909 }
2910 }
2911 if (tok != TOK_NOSUBST) {
2912 tok_str_add2(&macro_str1, tok, &tokc);
2913 tok = ' ';
2914 start_of_nosubsts = -1;
2915 }
2916 tok_str_add2(&macro_str1, tok, &tokc);
2917 }
2918 tok_str_add(&macro_str1, 0);
2919 return macro_str1.str;
2920 }
2921
2922
2923 /* do macro substitution of macro_str and add result to
2924 (tok_str,tok_len). 'nested_list' is the list of all macros we got
2925 inside to avoid recursing. */
2926 static void macro_subst(TokenString *tok_str, Sym **nested_list,
2927 const int *macro_str, struct macro_level ** can_read_stream)
2928 {
2929 Sym *s;
2930 int *macro_str1;
2931 const int *ptr;
2932 int t, ret, spc;
2933 CValue cval;
2934 struct macro_level ml;
2935 int force_blank;
2936
2937 /* first scan for '##' operator handling */
2938 ptr = macro_str;
2939 macro_str1 = macro_twosharps(ptr);
2940
2941 if (macro_str1)
2942 ptr = macro_str1;
2943 spc = 0;
2944 force_blank = 0;
2945
2946 while (1) {
2947 /* NOTE: ptr == NULL can only happen if tokens are read from
2948 file stream due to a macro function call */
2949 if (ptr == NULL)
2950 break;
2951 TOK_GET(&t, &ptr, &cval);
2952 if (t == 0)
2953 break;
2954 if (t == TOK_NOSUBST) {
2955 /* following token has already been subst'd. just copy it on */
2956 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2957 TOK_GET(&t, &ptr, &cval);
2958 goto no_subst;
2959 }
2960 s = define_find(t);
2961 if (s != NULL) {
2962 /* if nested substitution, do nothing */
2963 if (sym_find2(*nested_list, t)) {
2964 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
2965 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
2966 goto no_subst;
2967 }
2968 ml.p = macro_ptr;
2969 if (can_read_stream)
2970 ml.prev = *can_read_stream, *can_read_stream = &ml;
2971 macro_ptr = (int *)ptr;
2972 tok = t;
2973 ret = macro_subst_tok(tok_str, nested_list, s, can_read_stream);
2974 ptr = (int *)macro_ptr;
2975 macro_ptr = ml.p;
2976 if (can_read_stream && *can_read_stream == &ml)
2977 *can_read_stream = ml.prev;
2978 if (ret != 0)
2979 goto no_subst;
2980 if (parse_flags & PARSE_FLAG_SPACES)
2981 force_blank = 1;
2982 } else {
2983 no_subst:
2984 if (force_blank) {
2985 tok_str_add(tok_str, ' ');
2986 spc = 1;
2987 force_blank = 0;
2988 }
2989 if (!check_space(t, &spc))
2990 tok_str_add2(tok_str, t, &cval);
2991 }
2992 }
2993 if (macro_str1)
2994 tok_str_free(macro_str1);
2995 }
2996
2997 /* return next token with macro substitution */
2998 ST_FUNC void next(void)
2999 {
3000 Sym *nested_list, *s;
3001 TokenString str;
3002 struct macro_level *ml;
3003
3004 redo:
3005 if (parse_flags & PARSE_FLAG_SPACES)
3006 next_nomacro_spc();
3007 else
3008 next_nomacro();
3009 if (!macro_ptr) {
3010 /* if not reading from macro substituted string, then try
3011 to substitute macros */
3012 if (tok >= TOK_IDENT &&
3013 (parse_flags & PARSE_FLAG_PREPROCESS)) {
3014 s = define_find(tok);
3015 if (s) {
3016 /* we have a macro: we try to substitute */
3017 tok_str_new(&str);
3018 nested_list = NULL;
3019 ml = NULL;
3020 if (macro_subst_tok(&str, &nested_list, s, &ml) == 0) {
3021 /* substitution done, NOTE: maybe empty */
3022 tok_str_add(&str, 0);
3023 macro_ptr = str.str;
3024 macro_ptr_allocated = str.str;
3025 goto redo;
3026 }
3027 }
3028 }
3029 } else {
3030 if (tok == 0) {
3031 /* end of macro or end of unget buffer */
3032 if (unget_buffer_enabled) {
3033 macro_ptr = unget_saved_macro_ptr;
3034 unget_buffer_enabled = 0;
3035 } else {
3036 /* end of macro string: free it */
3037 tok_str_free(macro_ptr_allocated);
3038 macro_ptr_allocated = NULL;
3039 macro_ptr = NULL;
3040 }
3041 goto redo;
3042 } else if (tok == TOK_NOSUBST) {
3043 /* discard preprocessor's nosubst markers */
3044 goto redo;
3045 }
3046 }
3047
3048 /* convert preprocessor tokens into C tokens */
3049 if (tok == TOK_PPNUM &&
3050 (parse_flags & PARSE_FLAG_TOK_NUM)) {
3051 parse_number((char *)tokc.cstr->data);
3052 }
3053 }
3054
3055 /* push back current token and set current token to 'last_tok'. Only
3056 identifier case handled for labels. */
3057 ST_INLN void unget_tok(int last_tok)
3058 {
3059 int i, n;
3060 int *q;
3061 if (unget_buffer_enabled)
3062 {
3063 /* assert(macro_ptr == unget_saved_buffer + 1);
3064 assert(*macro_ptr == 0); */
3065 }
3066 else
3067 {
3068 unget_saved_macro_ptr = macro_ptr;
3069 unget_buffer_enabled = 1;
3070 }
3071 q = unget_saved_buffer;
3072 macro_ptr = q;
3073 *q++ = tok;
3074 n = tok_ext_size(tok) - 1;
3075 for(i=0;i<n;i++)
3076 *q++ = tokc.tab[i];
3077 *q = 0; /* end of token string */
3078 tok = last_tok;
3079 }
3080
3081
3082 /* better than nothing, but needs extension to handle '-E' option
3083 correctly too */
3084 ST_FUNC void preprocess_init(TCCState *s1)
3085 {
3086 s1->include_stack_ptr = s1->include_stack;
3087 /* XXX: move that before to avoid having to initialize
3088 file->ifdef_stack_ptr ? */
3089 s1->ifdef_stack_ptr = s1->ifdef_stack;
3090 file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3091
3092 vtop = vstack - 1;
3093 s1->pack_stack[0] = 0;
3094 s1->pack_stack_ptr = s1->pack_stack;
3095 }
3096
3097 ST_FUNC void preprocess_new(void)
3098 {
3099 int i, c;
3100 const char *p, *r;
3101
3102 /* init isid table */
3103 for(i=CH_EOF;i<256;i++)
3104 isidnum_table[i-CH_EOF] = isid(i) || isnum(i);
3105
3106 /* add all tokens */
3107 table_ident = NULL;
3108 memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3109
3110 tok_ident = TOK_IDENT;
3111 p = tcc_keywords;
3112 while (*p) {
3113 r = p;
3114 for(;;) {
3115 c = *r++;
3116 if (c == '\0')
3117 break;
3118 }
3119 tok_alloc(p, r - p - 1);
3120 p = r;
3121 }
3122 }
3123
3124 /* Preprocess the current file */
3125 ST_FUNC int tcc_preprocess(TCCState *s1)
3126 {
3127 Sym *define_start;
3128
3129 BufferedFile *file_ref, **iptr, **iptr_new;
3130 int token_seen, line_ref, d;
3131 const char *s;
3132
3133 preprocess_init(s1);
3134 define_start = define_stack;
3135 ch = file->buf_ptr[0];
3136 tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3137 parse_flags = PARSE_FLAG_ASM_COMMENTS | PARSE_FLAG_PREPROCESS |
3138 PARSE_FLAG_LINEFEED | PARSE_FLAG_SPACES;
3139 token_seen = 0;
3140 line_ref = 0;
3141 file_ref = NULL;
3142 iptr = s1->include_stack_ptr;
3143
3144 for (;;) {
3145 next();
3146 if (tok == TOK_EOF) {
3147 break;
3148 } else if (file != file_ref) {
3149 goto print_line;
3150 } else if (tok == TOK_LINEFEED) {
3151 if (!token_seen)
3152 continue;
3153 ++line_ref;
3154 token_seen = 0;
3155 } else if (!token_seen) {
3156 d = file->line_num - line_ref;
3157 if (file != file_ref || d < 0 || d >= 8) {
3158 print_line:
3159 iptr_new = s1->include_stack_ptr;
3160 s = iptr_new > iptr ? " 1"
3161 : iptr_new < iptr ? " 2"
3162 : iptr_new > s1->include_stack ? " 3"
3163 : ""
3164 ;
3165 iptr = iptr_new;
3166 fprintf(s1->ppfp, "# %d \"%s\"%s\n", file->line_num, file->filename, s);
3167 } else {
3168 while (d)
3169 fputs("\n", s1->ppfp), --d;
3170 }
3171 line_ref = (file_ref = file)->line_num;
3172 token_seen = tok != TOK_LINEFEED;
3173 if (!token_seen)
3174 continue;
3175 }
3176 fputs(get_tok_str(tok, &tokc), s1->ppfp);
3177 }
3178 free_defines(define_start);
3179 return 0;
3180 }