?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Revert "use int for ssize_t, (u)intptr_t instead of long in stddef.h"
[tinycc.git] / tccpp.c
blob4d5169e98ded91251bb7f0ce8440c563bae36d98
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 ST_DATA int parse_flags;
28
29 ST_DATA struct BufferedFile *file;
30 ST_DATA int ch, tok;
31 ST_DATA CValue tokc;
32 ST_DATA const int *macro_ptr;
33 ST_DATA CString tokcstr; /* current parsed string, if any */
34
35 /* display benchmark infos */
36 ST_DATA int total_lines;
37 ST_DATA int total_bytes;
38 ST_DATA int tok_ident;
39 ST_DATA TokenSym **table_ident;
40
41 /* ------------------------------------------------------------------------- */
42
43 static TokenSym *hash_ident[TOK_HASH_SIZE];
44 static char token_buf[STRING_MAX_SIZE + 1];
45 static CString cstr_buf;
46 static CString macro_equal_buf;
47 static TokenString tokstr_buf;
48 static unsigned char isidnum_table[256 - CH_EOF];
49 static int pp_debug_tok, pp_debug_symv;
50 static int pp_once;
51 static int pp_expr;
52 static int pp_counter;
53 static void tok_print(const char *msg, const int *str);
54
55 static struct TinyAlloc *toksym_alloc;
56 static struct TinyAlloc *tokstr_alloc;
57 static struct TinyAlloc *cstr_alloc;
58
59 static TokenString *macro_stack;
60
61 static const char tcc_keywords[] =
62 #define DEF(id, str) str "\0"
63 #include "tcctok.h"
64 #undef DEF
65 ;
66
67 /* WARNING: the content of this string encodes token numbers */
68 static const unsigned char tok_two_chars[] =
69 /* outdated -- gr
70 "<=\236>=\235!=\225&&\240||\241++\244--\242==\224<<\1>>\2+=\253"
71 "-=\255*=\252/=\257%=\245&=\246^=\336|=\374->\313..\250##\266";
72 */{
73 '<','=', TOK_LE,
74 '>','=', TOK_GE,
75 '!','=', TOK_NE,
76 '&','&', TOK_LAND,
77 '|','|', TOK_LOR,
78 '+','+', TOK_INC,
79 '-','-', TOK_DEC,
80 '=','=', TOK_EQ,
81 '<','<', TOK_SHL,
82 '>','>', TOK_SAR,
83 '+','=', TOK_A_ADD,
84 '-','=', TOK_A_SUB,
85 '*','=', TOK_A_MUL,
86 '/','=', TOK_A_DIV,
87 '%','=', TOK_A_MOD,
88 '&','=', TOK_A_AND,
89 '^','=', TOK_A_XOR,
90 '|','=', TOK_A_OR,
91 '-','>', TOK_ARROW,
92 '.','.', 0xa8, // C++ token ?
93 '#','#', TOK_TWOSHARPS,
94 0
95 };
96
97 static void next_nomacro_spc(void);
98
99 ST_FUNC void skip(int c)
100 {
101 if (tok != c)
102 tcc_error("'%c' expected (got \"%s\")", c, get_tok_str(tok, &tokc));
103 next();
104 }
105
106 ST_FUNC void expect(const char *msg)
107 {
108 tcc_error("%s expected", msg);
109 }
110
111 /* ------------------------------------------------------------------------- */
112 /* Custom allocator for tiny objects */
113
114 #define USE_TAL
115
116 #ifndef USE_TAL
117 #define tal_free(al, p) tcc_free(p)
118 #define tal_realloc(al, p, size) tcc_realloc(p, size)
119 #define tal_new(a,b,c)
120 #define tal_delete(a)
121 #else
122 #if !defined(MEM_DEBUG)
123 #define tal_free(al, p) tal_free_impl(al, p)
124 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size)
125 #define TAL_DEBUG_PARAMS
126 #else
127 #define TAL_DEBUG 1
128 //#define TAL_INFO 1 /* collect and dump allocators stats */
129 #define tal_free(al, p) tal_free_impl(al, p, __FILE__, __LINE__)
130 #define tal_realloc(al, p, size) tal_realloc_impl(&al, p, size, __FILE__, __LINE__)
131 #define TAL_DEBUG_PARAMS , const char *file, int line
132 #define TAL_DEBUG_FILE_LEN 40
133 #endif
134
135 #define TOKSYM_TAL_SIZE (768 * 1024) /* allocator for tiny TokenSym in table_ident */
136 #define TOKSTR_TAL_SIZE (768 * 1024) /* allocator for tiny TokenString instances */
137 #define CSTR_TAL_SIZE (256 * 1024) /* allocator for tiny CString instances */
138 #define TOKSYM_TAL_LIMIT 256 /* prefer unique limits to distinguish allocators debug msgs */
139 #define TOKSTR_TAL_LIMIT 128 /* 32 * sizeof(int) */
140 #define CSTR_TAL_LIMIT 1024
141
142 typedef struct TinyAlloc {
143 unsigned limit;
144 unsigned size;
145 uint8_t *buffer;
146 uint8_t *p;
147 unsigned nb_allocs;
148 struct TinyAlloc *next, *top;
149 #ifdef TAL_INFO
150 unsigned nb_peak;
151 unsigned nb_total;
152 unsigned nb_missed;
153 uint8_t *peak_p;
154 #endif
155 } TinyAlloc;
156
157 typedef struct tal_header_t {
158 unsigned size;
159 #ifdef TAL_DEBUG
160 int line_num; /* negative line_num used for double free check */
161 char file_name[TAL_DEBUG_FILE_LEN + 1];
162 #endif
163 } tal_header_t;
164
165 /* ------------------------------------------------------------------------- */
166
167 static TinyAlloc *tal_new(TinyAlloc **pal, unsigned limit, unsigned size)
168 {
169 TinyAlloc *al = tcc_mallocz(sizeof(TinyAlloc));
170 al->p = al->buffer = tcc_malloc(size);
171 al->limit = limit;
172 al->size = size;
173 if (pal) *pal = al;
174 return al;
175 }
176
177 static void tal_delete(TinyAlloc *al)
178 {
179 TinyAlloc *next;
180
181 tail_call:
182 if (!al)
183 return;
184 #ifdef TAL_INFO
185 fprintf(stderr, "limit=%5d, size=%5g MB, nb_peak=%6d, nb_total=%8d, nb_missed=%6d, usage=%5.1f%%\n",
186 al->limit, al->size / 1024.0 / 1024.0, al->nb_peak, al->nb_total, al->nb_missed,
187 (al->peak_p - al->buffer) * 100.0 / al->size);
188 #endif
189 #ifdef TAL_DEBUG
190 if (al->nb_allocs > 0) {
191 uint8_t *p;
192 fprintf(stderr, "TAL_DEBUG: memory leak %d chunk(s) (limit= %d)\n",
193 al->nb_allocs, al->limit);
194 p = al->buffer;
195 while (p < al->p) {
196 tal_header_t *header = (tal_header_t *)p;
197 if (header->line_num > 0) {
198 fprintf(stderr, "%s:%d: chunk of %d bytes leaked\n",
199 header->file_name, header->line_num, header->size);
200 }
201 p += header->size + sizeof(tal_header_t);
202 }
203 #if MEM_DEBUG-0 == 2
204 exit(2);
205 #endif
206 }
207 #endif
208 next = al->next;
209 tcc_free(al->buffer);
210 tcc_free(al);
211 al = next;
212 goto tail_call;
213 }
214
215 static void tal_free_impl(TinyAlloc *al, void *p TAL_DEBUG_PARAMS)
216 {
217 if (!p)
218 return;
219 tail_call:
220 if (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size) {
221 #ifdef TAL_DEBUG
222 tal_header_t *header = (((tal_header_t *)p) - 1);
223 if (header->line_num < 0) {
224 fprintf(stderr, "%s:%d: TAL_DEBUG: double frees chunk from\n",
225 file, line);
226 fprintf(stderr, "%s:%d: %d bytes\n",
227 header->file_name, (int)-header->line_num, (int)header->size);
228 } else
229 header->line_num = -header->line_num;
230 #endif
231 al->nb_allocs--;
232 if (!al->nb_allocs)
233 al->p = al->buffer;
234 } else if (al->next) {
235 al = al->next;
236 goto tail_call;
237 }
238 else
239 tcc_free(p);
240 }
241
242 static void *tal_realloc_impl(TinyAlloc **pal, void *p, unsigned size TAL_DEBUG_PARAMS)
243 {
244 tal_header_t *header;
245 void *ret;
246 int is_own;
247 unsigned adj_size = (size + 3) & -4;
248 TinyAlloc *al = *pal;
249
250 tail_call:
251 is_own = (al->buffer <= (uint8_t *)p && (uint8_t *)p < al->buffer + al->size);
252 if ((!p || is_own) && size <= al->limit) {
253 if (al->p + adj_size + sizeof(tal_header_t) < al->buffer + al->size) {
254 header = (tal_header_t *)al->p;
255 header->size = adj_size;
256 #ifdef TAL_DEBUG
257 { int ofs = strlen(file) - TAL_DEBUG_FILE_LEN;
258 strncpy(header->file_name, file + (ofs > 0 ? ofs : 0), TAL_DEBUG_FILE_LEN);
259 header->file_name[TAL_DEBUG_FILE_LEN] = 0;
260 header->line_num = line; }
261 #endif
262 ret = al->p + sizeof(tal_header_t);
263 al->p += adj_size + sizeof(tal_header_t);
264 if (is_own) {
265 header = (((tal_header_t *)p) - 1);
266 memcpy(ret, p, header->size);
267 #ifdef TAL_DEBUG
268 header->line_num = -header->line_num;
269 #endif
270 } else {
271 al->nb_allocs++;
272 }
273 #ifdef TAL_INFO
274 if (al->nb_peak < al->nb_allocs)
275 al->nb_peak = al->nb_allocs;
276 if (al->peak_p < al->p)
277 al->peak_p = al->p;
278 al->nb_total++;
279 #endif
280 return ret;
281 } else if (is_own) {
282 al->nb_allocs--;
283 ret = tal_realloc(*pal, 0, size);
284 header = (((tal_header_t *)p) - 1);
285 memcpy(ret, p, header->size);
286 #ifdef TAL_DEBUG
287 header->line_num = -header->line_num;
288 #endif
289 return ret;
290 }
291 if (al->next) {
292 al = al->next;
293 } else {
294 TinyAlloc *bottom = al, *next = al->top ? al->top : al;
295
296 al = tal_new(pal, next->limit, next->size * 2);
297 al->next = next;
298 bottom->top = al;
299 }
300 goto tail_call;
301 }
302 if (is_own) {
303 al->nb_allocs--;
304 ret = tcc_malloc(size);
305 header = (((tal_header_t *)p) - 1);
306 memcpy(ret, p, header->size);
307 #ifdef TAL_DEBUG
308 header->line_num = -header->line_num;
309 #endif
310 } else if (al->next) {
311 al = al->next;
312 goto tail_call;
313 } else
314 ret = tcc_realloc(p, size);
315 #ifdef TAL_INFO
316 al->nb_missed++;
317 #endif
318 return ret;
319 }
320
321 #endif /* USE_TAL */
322
323 /* ------------------------------------------------------------------------- */
324 /* CString handling */
325 static void cstr_realloc(CString *cstr, int new_size)
326 {
327 int size;
328
329 size = cstr->size_allocated;
330 if (size < 8)
331 size = 8; /* no need to allocate a too small first string */
332 while (size < new_size)
333 size = size * 2;
334 cstr->data = tal_realloc(cstr_alloc, cstr->data, size);
335 cstr->size_allocated = size;
336 }
337
338 /* add a byte */
339 ST_INLN void cstr_ccat(CString *cstr, int ch)
340 {
341 int size;
342 size = cstr->size + 1;
343 if (size > cstr->size_allocated)
344 cstr_realloc(cstr, size);
345 ((unsigned char *)cstr->data)[size - 1] = ch;
346 cstr->size = size;
347 }
348
349 ST_FUNC void cstr_cat(CString *cstr, const char *str, int len)
350 {
351 int size;
352 if (len <= 0)
353 len = strlen(str) + 1 + len;
354 size = cstr->size + len;
355 if (size > cstr->size_allocated)
356 cstr_realloc(cstr, size);
357 memmove(((unsigned char *)cstr->data) + cstr->size, str, len);
358 cstr->size = size;
359 }
360
361 /* add a wide char */
362 ST_FUNC void cstr_wccat(CString *cstr, int ch)
363 {
364 int size;
365 size = cstr->size + sizeof(nwchar_t);
366 if (size > cstr->size_allocated)
367 cstr_realloc(cstr, size);
368 *(nwchar_t *)(((unsigned char *)cstr->data) + size - sizeof(nwchar_t)) = ch;
369 cstr->size = size;
370 }
371
372 ST_FUNC void cstr_new(CString *cstr)
373 {
374 memset(cstr, 0, sizeof(CString));
375 }
376
377 /* free string and reset it to NULL */
378 ST_FUNC void cstr_free(CString *cstr)
379 {
380 tal_free(cstr_alloc, cstr->data);
381 cstr_new(cstr);
382 }
383
384 /* reset string to empty */
385 ST_FUNC void cstr_reset(CString *cstr)
386 {
387 cstr->size = 0;
388 }
389
390 /* XXX: unicode ? */
391 static void add_char(CString *cstr, int c)
392 {
393 if (c == '\'' || c == '\"' || c == '\\') {
394 /* XXX: could be more precise if char or string */
395 cstr_ccat(cstr, '\\');
396 }
397 if (c >= 32 && c <= 126) {
398 cstr_ccat(cstr, c);
399 } else {
400 cstr_ccat(cstr, '\\');
401 if (c == '\n') {
402 cstr_ccat(cstr, 'n');
403 } else {
404 cstr_ccat(cstr, '0' + ((c >> 6) & 7));
405 cstr_ccat(cstr, '0' + ((c >> 3) & 7));
406 cstr_ccat(cstr, '0' + (c & 7));
407 }
408 }
409 }
410
411 /* ------------------------------------------------------------------------- */
412 /* allocate a new token */
413 static TokenSym *tok_alloc_new(TokenSym **pts, const char *str, int len)
414 {
415 TokenSym *ts, **ptable;
416 int i;
417
418 if (tok_ident >= SYM_FIRST_ANOM)
419 tcc_error("memory full (symbols)");
420
421 /* expand token table if needed */
422 i = tok_ident - TOK_IDENT;
423 if ((i % TOK_ALLOC_INCR) == 0) {
424 ptable = tcc_realloc(table_ident, (i + TOK_ALLOC_INCR) * sizeof(TokenSym *));
425 table_ident = ptable;
426 }
427
428 ts = tal_realloc(toksym_alloc, 0, sizeof(TokenSym) + len);
429 table_ident[i] = ts;
430 ts->tok = tok_ident++;
431 ts->sym_define = NULL;
432 ts->sym_label = NULL;
433 ts->sym_struct = NULL;
434 ts->sym_identifier = NULL;
435 ts->len = len;
436 ts->hash_next = NULL;
437 memcpy(ts->str, str, len);
438 ts->str[len] = '\0';
439 *pts = ts;
440 return ts;
441 }
442
443 #define TOK_HASH_INIT 1
444 #define TOK_HASH_FUNC(h, c) ((h) + ((h) << 5) + ((h) >> 27) + (c))
445
446
447 /* find a token and add it if not found */
448 ST_FUNC TokenSym *tok_alloc(const char *str, int len)
449 {
450 TokenSym *ts, **pts;
451 int i;
452 unsigned int h;
453
454 h = TOK_HASH_INIT;
455 for(i=0;i<len;i++)
456 h = TOK_HASH_FUNC(h, ((unsigned char *)str)[i]);
457 h &= (TOK_HASH_SIZE - 1);
458
459 pts = &hash_ident[h];
460 for(;;) {
461 ts = *pts;
462 if (!ts)
463 break;
464 if (ts->len == len && !memcmp(ts->str, str, len))
465 return ts;
466 pts = &(ts->hash_next);
467 }
468 return tok_alloc_new(pts, str, len);
469 }
470
471 /* XXX: buffer overflow */
472 /* XXX: float tokens */
473 ST_FUNC const char *get_tok_str(int v, CValue *cv)
474 {
475 char *p;
476 int i, len;
477
478 cstr_reset(&cstr_buf);
479 p = cstr_buf.data;
480
481 switch(v) {
482 case TOK_CINT:
483 case TOK_CUINT:
484 case TOK_CLONG:
485 case TOK_CULONG:
486 case TOK_CLLONG:
487 case TOK_CULLONG:
488 /* XXX: not quite exact, but only useful for testing */
489 #ifdef _WIN32
490 sprintf(p, "%u", (unsigned)cv->i);
491 #else
492 sprintf(p, "%llu", (unsigned long long)cv->i);
493 #endif
494 break;
495 case TOK_LCHAR:
496 cstr_ccat(&cstr_buf, 'L');
497 case TOK_CCHAR:
498 cstr_ccat(&cstr_buf, '\'');
499 add_char(&cstr_buf, cv->i);
500 cstr_ccat(&cstr_buf, '\'');
501 cstr_ccat(&cstr_buf, '\0');
502 break;
503 case TOK_PPNUM:
504 case TOK_PPSTR:
505 return (char*)cv->str.data;
506 case TOK_LSTR:
507 cstr_ccat(&cstr_buf, 'L');
508 case TOK_STR:
509 cstr_ccat(&cstr_buf, '\"');
510 if (v == TOK_STR) {
511 len = cv->str.size - 1;
512 for(i=0;i<len;i++)
513 add_char(&cstr_buf, ((unsigned char *)cv->str.data)[i]);
514 } else {
515 len = (cv->str.size / sizeof(nwchar_t)) - 1;
516 for(i=0;i<len;i++)
517 add_char(&cstr_buf, ((nwchar_t *)cv->str.data)[i]);
518 }
519 cstr_ccat(&cstr_buf, '\"');
520 cstr_ccat(&cstr_buf, '\0');
521 break;
522
523 case TOK_CFLOAT:
524 cstr_cat(&cstr_buf, "<float>", 0);
525 break;
526 case TOK_CDOUBLE:
527 cstr_cat(&cstr_buf, "<double>", 0);
528 break;
529 case TOK_CLDOUBLE:
530 cstr_cat(&cstr_buf, "<long double>", 0);
531 break;
532 case TOK_LINENUM:
533 cstr_cat(&cstr_buf, "<linenumber>", 0);
534 break;
535
536 /* above tokens have value, the ones below don't */
537 case TOK_LT:
538 v = '<';
539 goto addv;
540 case TOK_GT:
541 v = '>';
542 goto addv;
543 case TOK_DOTS:
544 return strcpy(p, "...");
545 case TOK_A_SHL:
546 return strcpy(p, "<<=");
547 case TOK_A_SAR:
548 return strcpy(p, ">>=");
549 case TOK_EOF:
550 return strcpy(p, "<eof>");
551 default:
552 if (v < TOK_IDENT) {
553 /* search in two bytes table */
554 const unsigned char *q = tok_two_chars;
555 while (*q) {
556 if (q[2] == v) {
557 *p++ = q[0];
558 *p++ = q[1];
559 *p = '\0';
560 return cstr_buf.data;
561 }
562 q += 3;
563 }
564 if (v >= 127) {
565 sprintf(cstr_buf.data, "<%02x>", v);
566 return cstr_buf.data;
567 }
568 addv:
569 *p++ = v;
570 *p = '\0';
571 } else if (v < tok_ident) {
572 return table_ident[v - TOK_IDENT]->str;
573 } else if (v >= SYM_FIRST_ANOM) {
574 /* special name for anonymous symbol */
575 sprintf(p, "L.%u", v - SYM_FIRST_ANOM);
576 } else {
577 /* should never happen */
578 return NULL;
579 }
580 break;
581 }
582 return cstr_buf.data;
583 }
584
585 /* return the current character, handling end of block if necessary
586 (but not stray) */
587 ST_FUNC int handle_eob(void)
588 {
589 BufferedFile *bf = file;
590 int len;
591
592 /* only tries to read if really end of buffer */
593 if (bf->buf_ptr >= bf->buf_end) {
594 if (bf->fd != -1) {
595 #if defined(PARSE_DEBUG)
596 len = 1;
597 #else
598 len = IO_BUF_SIZE;
599 #endif
600 len = read(bf->fd, bf->buffer, len);
601 if (len < 0)
602 len = 0;
603 } else {
604 len = 0;
605 }
606 total_bytes += len;
607 bf->buf_ptr = bf->buffer;
608 bf->buf_end = bf->buffer + len;
609 *bf->buf_end = CH_EOB;
610 }
611 if (bf->buf_ptr < bf->buf_end) {
612 return bf->buf_ptr[0];
613 } else {
614 bf->buf_ptr = bf->buf_end;
615 return CH_EOF;
616 }
617 }
618
619 /* read next char from current input file and handle end of input buffer */
620 ST_INLN void inp(void)
621 {
622 ch = *(++(file->buf_ptr));
623 /* end of buffer/file handling */
624 if (ch == CH_EOB)
625 ch = handle_eob();
626 }
627
628 /* handle '\[\r]\n' */
629 static int handle_stray_noerror(void)
630 {
631 while (ch == '\\') {
632 inp();
633 if (ch == '\n') {
634 file->line_num++;
635 inp();
636 } else if (ch == '\r') {
637 inp();
638 if (ch != '\n')
639 goto fail;
640 file->line_num++;
641 inp();
642 } else {
643 fail:
644 return 1;
645 }
646 }
647 return 0;
648 }
649
650 static void handle_stray(void)
651 {
652 if (handle_stray_noerror())
653 tcc_error("stray '\\' in program");
654 }
655
656 /* skip the stray and handle the \\n case. Output an error if
657 incorrect char after the stray */
658 static int handle_stray1(uint8_t *p)
659 {
660 int c;
661
662 file->buf_ptr = p;
663 if (p >= file->buf_end) {
664 c = handle_eob();
665 if (c != '\\')
666 return c;
667 p = file->buf_ptr;
668 }
669 ch = *p;
670 if (handle_stray_noerror()) {
671 if (!(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
672 tcc_error("stray '\\' in program");
673 *--file->buf_ptr = '\\';
674 }
675 p = file->buf_ptr;
676 c = *p;
677 return c;
678 }
679
680 /* handle just the EOB case, but not stray */
681 #define PEEKC_EOB(c, p)\
682 {\
683 p++;\
684 c = *p;\
685 if (c == '\\') {\
686 file->buf_ptr = p;\
687 c = handle_eob();\
688 p = file->buf_ptr;\
689 }\
690 }
691
692 /* handle the complicated stray case */
693 #define PEEKC(c, p)\
694 {\
695 p++;\
696 c = *p;\
697 if (c == '\\') {\
698 c = handle_stray1(p);\
699 p = file->buf_ptr;\
700 }\
701 }
702
703 /* input with '\[\r]\n' handling. Note that this function cannot
704 handle other characters after '\', so you cannot call it inside
705 strings or comments */
706 ST_FUNC void minp(void)
707 {
708 inp();
709 if (ch == '\\')
710 handle_stray();
711 }
712
713 /* single line C++ comments */
714 static uint8_t *parse_line_comment(uint8_t *p)
715 {
716 int c;
717
718 p++;
719 for(;;) {
720 c = *p;
721 redo:
722 if (c == '\n' || c == CH_EOF) {
723 break;
724 } else if (c == '\\') {
725 file->buf_ptr = p;
726 c = handle_eob();
727 p = file->buf_ptr;
728 if (c == '\\') {
729 PEEKC_EOB(c, p);
730 if (c == '\n') {
731 file->line_num++;
732 PEEKC_EOB(c, p);
733 } else if (c == '\r') {
734 PEEKC_EOB(c, p);
735 if (c == '\n') {
736 file->line_num++;
737 PEEKC_EOB(c, p);
738 }
739 }
740 } else {
741 goto redo;
742 }
743 } else {
744 p++;
745 }
746 }
747 return p;
748 }
749
750 /* C comments */
751 ST_FUNC uint8_t *parse_comment(uint8_t *p)
752 {
753 int c;
754
755 p++;
756 for(;;) {
757 /* fast skip loop */
758 for(;;) {
759 c = *p;
760 if (c == '\n' || c == '*' || c == '\\')
761 break;
762 p++;
763 c = *p;
764 if (c == '\n' || c == '*' || c == '\\')
765 break;
766 p++;
767 }
768 /* now we can handle all the cases */
769 if (c == '\n') {
770 file->line_num++;
771 p++;
772 } else if (c == '*') {
773 p++;
774 for(;;) {
775 c = *p;
776 if (c == '*') {
777 p++;
778 } else if (c == '/') {
779 goto end_of_comment;
780 } else if (c == '\\') {
781 file->buf_ptr = p;
782 c = handle_eob();
783 p = file->buf_ptr;
784 if (c == CH_EOF)
785 tcc_error("unexpected end of file in comment");
786 if (c == '\\') {
787 /* skip '\[\r]\n', otherwise just skip the stray */
788 while (c == '\\') {
789 PEEKC_EOB(c, p);
790 if (c == '\n') {
791 file->line_num++;
792 PEEKC_EOB(c, p);
793 } else if (c == '\r') {
794 PEEKC_EOB(c, p);
795 if (c == '\n') {
796 file->line_num++;
797 PEEKC_EOB(c, p);
798 }
799 } else {
800 goto after_star;
801 }
802 }
803 }
804 } else {
805 break;
806 }
807 }
808 after_star: ;
809 } else {
810 /* stray, eob or eof */
811 file->buf_ptr = p;
812 c = handle_eob();
813 p = file->buf_ptr;
814 if (c == CH_EOF) {
815 tcc_error("unexpected end of file in comment");
816 } else if (c == '\\') {
817 p++;
818 }
819 }
820 }
821 end_of_comment:
822 p++;
823 return p;
824 }
825
826 ST_FUNC int set_idnum(int c, int val)
827 {
828 int prev = isidnum_table[c - CH_EOF];
829 isidnum_table[c - CH_EOF] = val;
830 return prev;
831 }
832
833 #define cinp minp
834
835 static inline void skip_spaces(void)
836 {
837 while (isidnum_table[ch - CH_EOF] & IS_SPC)
838 cinp();
839 }
840
841 static inline int check_space(int t, int *spc)
842 {
843 if (t < 256 && (isidnum_table[t - CH_EOF] & IS_SPC)) {
844 if (*spc)
845 return 1;
846 *spc = 1;
847 } else
848 *spc = 0;
849 return 0;
850 }
851
852 /* parse a string without interpreting escapes */
853 static uint8_t *parse_pp_string(uint8_t *p,
854 int sep, CString *str)
855 {
856 int c;
857 p++;
858 for(;;) {
859 c = *p;
860 if (c == sep) {
861 break;
862 } else if (c == '\\') {
863 file->buf_ptr = p;
864 c = handle_eob();
865 p = file->buf_ptr;
866 if (c == CH_EOF) {
867 unterminated_string:
868 /* XXX: indicate line number of start of string */
869 tcc_error("missing terminating %c character", sep);
870 } else if (c == '\\') {
871 /* escape : just skip \[\r]\n */
872 PEEKC_EOB(c, p);
873 if (c == '\n') {
874 file->line_num++;
875 p++;
876 } else if (c == '\r') {
877 PEEKC_EOB(c, p);
878 if (c != '\n')
879 expect("'\n' after '\r'");
880 file->line_num++;
881 p++;
882 } else if (c == CH_EOF) {
883 goto unterminated_string;
884 } else {
885 if (str) {
886 cstr_ccat(str, '\\');
887 cstr_ccat(str, c);
888 }
889 p++;
890 }
891 }
892 } else if (c == '\n') {
893 file->line_num++;
894 goto add_char;
895 } else if (c == '\r') {
896 PEEKC_EOB(c, p);
897 if (c != '\n') {
898 if (str)
899 cstr_ccat(str, '\r');
900 } else {
901 file->line_num++;
902 goto add_char;
903 }
904 } else {
905 add_char:
906 if (str)
907 cstr_ccat(str, c);
908 p++;
909 }
910 }
911 p++;
912 return p;
913 }
914
915 /* skip block of text until #else, #elif or #endif. skip also pairs of
916 #if/#endif */
917 static void preprocess_skip(void)
918 {
919 int a, start_of_line, c, in_warn_or_error;
920 uint8_t *p;
921
922 p = file->buf_ptr;
923 a = 0;
924 redo_start:
925 start_of_line = 1;
926 in_warn_or_error = 0;
927 for(;;) {
928 redo_no_start:
929 c = *p;
930 switch(c) {
931 case ' ':
932 case '\t':
933 case '\f':
934 case '\v':
935 case '\r':
936 p++;
937 goto redo_no_start;
938 case '\n':
939 file->line_num++;
940 p++;
941 goto redo_start;
942 case '\\':
943 file->buf_ptr = p;
944 c = handle_eob();
945 if (c == CH_EOF) {
946 expect("#endif");
947 } else if (c == '\\') {
948 ch = file->buf_ptr[0];
949 handle_stray_noerror();
950 }
951 p = file->buf_ptr;
952 goto redo_no_start;
953 /* skip strings */
954 case '\"':
955 case '\'':
956 if (in_warn_or_error)
957 goto _default;
958 p = parse_pp_string(p, c, NULL);
959 break;
960 /* skip comments */
961 case '/':
962 if (in_warn_or_error)
963 goto _default;
964 file->buf_ptr = p;
965 ch = *p;
966 minp();
967 p = file->buf_ptr;
968 if (ch == '*') {
969 p = parse_comment(p);
970 } else if (ch == '/') {
971 p = parse_line_comment(p);
972 }
973 break;
974 case '#':
975 p++;
976 if (start_of_line) {
977 file->buf_ptr = p;
978 next_nomacro();
979 p = file->buf_ptr;
980 if (a == 0 &&
981 (tok == TOK_ELSE || tok == TOK_ELIF || tok == TOK_ENDIF))
982 goto the_end;
983 if (tok == TOK_IF || tok == TOK_IFDEF || tok == TOK_IFNDEF)
984 a++;
985 else if (tok == TOK_ENDIF)
986 a--;
987 else if( tok == TOK_ERROR || tok == TOK_WARNING)
988 in_warn_or_error = 1;
989 else if (tok == TOK_LINEFEED)
990 goto redo_start;
991 else if (parse_flags & PARSE_FLAG_ASM_FILE)
992 p = parse_line_comment(p - 1);
993 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
994 p = parse_line_comment(p - 1);
995 break;
996 _default:
997 default:
998 p++;
999 break;
1000 }
1001 start_of_line = 0;
1002 }
1003 the_end: ;
1004 file->buf_ptr = p;
1005 }
1006
1007 #if 0
1008 /* return the number of additional 'ints' necessary to store the
1009 token */
1010 static inline int tok_size(const int *p)
1011 {
1012 switch(*p) {
1013 /* 4 bytes */
1014 case TOK_CINT:
1015 case TOK_CUINT:
1016 case TOK_CCHAR:
1017 case TOK_LCHAR:
1018 case TOK_CFLOAT:
1019 case TOK_LINENUM:
1020 #ifndef TCC_LONG_ARE_64_BIT
1021 case TOK_CLONG;
1022 case TOK_CULONG;
1023 #endif
1024 return 1 + 1;
1025 case TOK_STR:
1026 case TOK_LSTR:
1027 case TOK_PPNUM:
1028 case TOK_PPSTR:
1029 return 1 + ((sizeof(CString) + ((CString *)(p+1))->size + 3) >> 2);
1030 case TOK_CDOUBLE:
1031 case TOK_CLLONG:
1032 case TOK_CULLONG:
1033 #ifdef TCC_LONG_ARE_64_BIT
1034 case TOK_CLONG;
1035 case TOK_CULONG;
1036 #endif
1037 return 1 + 2;
1038 case TOK_CLDOUBLE:
1039 return 1 + LDOUBLE_SIZE / 4;
1040 default:
1041 return 1 + 0;
1042 }
1043 }
1044 #endif
1045
1046 /* token string handling */
1047 ST_INLN void tok_str_new(TokenString *s)
1048 {
1049 s->str = NULL;
1050 s->len = s->lastlen = 0;
1051 s->allocated_len = 0;
1052 s->last_line_num = -1;
1053 }
1054
1055 ST_FUNC TokenString *tok_str_alloc(void)
1056 {
1057 TokenString *str = tal_realloc(tokstr_alloc, 0, sizeof *str);
1058 tok_str_new(str);
1059 return str;
1060 }
1061
1062 ST_FUNC int *tok_str_dup(TokenString *s)
1063 {
1064 int *str;
1065
1066 str = tal_realloc(tokstr_alloc, 0, s->len * sizeof(int));
1067 memcpy(str, s->str, s->len * sizeof(int));
1068 return str;
1069 }
1070
1071 ST_FUNC void tok_str_free_str(int *str)
1072 {
1073 tal_free(tokstr_alloc, str);
1074 }
1075
1076 ST_FUNC void tok_str_free(TokenString *str)
1077 {
1078 tok_str_free_str(str->str);
1079 tal_free(tokstr_alloc, str);
1080 }
1081
1082 ST_FUNC int *tok_str_realloc(TokenString *s, int new_size)
1083 {
1084 int *str, size;
1085
1086 size = s->allocated_len;
1087 if (size < 16)
1088 size = 16;
1089 while (size < new_size)
1090 size = size * 2;
1091 if (size > s->allocated_len) {
1092 str = tal_realloc(tokstr_alloc, s->str, size * sizeof(int));
1093 s->allocated_len = size;
1094 s->str = str;
1095 }
1096 return s->str;
1097 }
1098
1099 ST_FUNC void tok_str_add(TokenString *s, int t)
1100 {
1101 int len, *str;
1102
1103 len = s->len;
1104 str = s->str;
1105 if (len >= s->allocated_len)
1106 str = tok_str_realloc(s, len + 1);
1107 str[len++] = t;
1108 s->len = len;
1109 }
1110
1111 ST_FUNC void begin_macro(TokenString *str, int alloc)
1112 {
1113 str->alloc = alloc;
1114 str->prev = macro_stack;
1115 str->prev_ptr = macro_ptr;
1116 str->save_line_num = file->line_num;
1117 macro_ptr = str->str;
1118 macro_stack = str;
1119 }
1120
1121 ST_FUNC void end_macro(void)
1122 {
1123 TokenString *str = macro_stack;
1124 macro_stack = str->prev;
1125 macro_ptr = str->prev_ptr;
1126 file->line_num = str->save_line_num;
1127 if (str->alloc == 2) {
1128 str->alloc = 3; /* just mark as finished */
1129 } else {
1130 tok_str_free(str);
1131 }
1132 }
1133
1134 static void tok_str_add2(TokenString *s, int t, CValue *cv)
1135 {
1136 int len, *str;
1137
1138 len = s->lastlen = s->len;
1139 str = s->str;
1140
1141 /* allocate space for worst case */
1142 if (len + TOK_MAX_SIZE >= s->allocated_len)
1143 str = tok_str_realloc(s, len + TOK_MAX_SIZE + 1);
1144 str[len++] = t;
1145 switch(t) {
1146 case TOK_CINT:
1147 case TOK_CUINT:
1148 case TOK_CCHAR:
1149 case TOK_LCHAR:
1150 case TOK_CFLOAT:
1151 case TOK_LINENUM:
1152 #ifndef TCC_LONG_ARE_64_BIT
1153 case TOK_CLONG:
1154 case TOK_CULONG:
1155 #endif
1156 str[len++] = cv->tab[0];
1157 break;
1158 case TOK_PPNUM:
1159 case TOK_PPSTR:
1160 case TOK_STR:
1161 case TOK_LSTR:
1162 {
1163 /* Insert the string into the int array. */
1164 size_t nb_words =
1165 1 + (cv->str.size + sizeof(int) - 1) / sizeof(int);
1166 if (len + nb_words >= s->allocated_len)
1167 str = tok_str_realloc(s, len + nb_words + 1);
1168 str[len] = cv->str.size;
1169 memcpy(&str[len + 1], cv->str.data, cv->str.size);
1170 len += nb_words;
1171 }
1172 break;
1173 case TOK_CDOUBLE:
1174 case TOK_CLLONG:
1175 case TOK_CULLONG:
1176 #ifdef TCC_LONG_ARE_64_BIT
1177 case TOK_CLONG:
1178 case TOK_CULONG:
1179 #endif
1180 #if LDOUBLE_SIZE == 8
1181 case TOK_CLDOUBLE:
1182 #endif
1183 str[len++] = cv->tab[0];
1184 str[len++] = cv->tab[1];
1185 break;
1186 #if LDOUBLE_SIZE == 12
1187 case TOK_CLDOUBLE:
1188 str[len++] = cv->tab[0];
1189 str[len++] = cv->tab[1];
1190 str[len++] = cv->tab[2];
1191 #elif LDOUBLE_SIZE == 16
1192 case TOK_CLDOUBLE:
1193 str[len++] = cv->tab[0];
1194 str[len++] = cv->tab[1];
1195 str[len++] = cv->tab[2];
1196 str[len++] = cv->tab[3];
1197 #elif LDOUBLE_SIZE != 8
1198 #error add long double size support
1199 #endif
1200 break;
1201 default:
1202 break;
1203 }
1204 s->len = len;
1205 }
1206
1207 /* add the current parse token in token string 's' */
1208 ST_FUNC void tok_str_add_tok(TokenString *s)
1209 {
1210 CValue cval;
1211
1212 /* save line number info */
1213 if (file->line_num != s->last_line_num) {
1214 s->last_line_num = file->line_num;
1215 cval.i = s->last_line_num;
1216 tok_str_add2(s, TOK_LINENUM, &cval);
1217 }
1218 tok_str_add2(s, tok, &tokc);
1219 }
1220
1221 /* get a token from an integer array and increment pointer
1222 accordingly. we code it as a macro to avoid pointer aliasing. */
1223 static inline void TOK_GET(int *t, const int **pp, CValue *cv)
1224 {
1225 const int *p = *pp;
1226 int n, *tab;
1227
1228 tab = cv->tab;
1229 switch(*t = *p++) {
1230 case TOK_CINT:
1231 case TOK_CUINT:
1232 case TOK_CCHAR:
1233 case TOK_LCHAR:
1234 case TOK_LINENUM:
1235 #ifndef TCC_LONG_ARE_64_BIT
1236 case TOK_CLONG:
1237 case TOK_CULONG:
1238 #endif
1239 tab[0] = *p++;
1240 cv->i = (*t == TOK_CUINT) ? (unsigned)cv->i : (int)cv->i;
1241 break;
1242 case TOK_CFLOAT:
1243 tab[0] = *p++;
1244 break;
1245 case TOK_STR:
1246 case TOK_LSTR:
1247 case TOK_PPNUM:
1248 case TOK_PPSTR:
1249 cv->str.size = *p++;
1250 cv->str.data = p;
1251 p += (cv->str.size + sizeof(int) - 1) / sizeof(int);
1252 break;
1253 case TOK_CDOUBLE:
1254 case TOK_CLLONG:
1255 case TOK_CULLONG:
1256 #ifdef TCC_LONG_ARE_64_BIT
1257 case TOK_CLONG:
1258 case TOK_CULONG:
1259 #endif
1260 n = 2;
1261 goto copy;
1262 case TOK_CLDOUBLE:
1263 #if LDOUBLE_SIZE == 16
1264 n = 4;
1265 #elif LDOUBLE_SIZE == 12
1266 n = 3;
1267 #elif LDOUBLE_SIZE == 8
1268 n = 2;
1269 #else
1270 # error add long double size support
1271 #endif
1272 copy:
1273 do
1274 *tab++ = *p++;
1275 while (--n);
1276 break;
1277 default:
1278 break;
1279 }
1280 *pp = p;
1281 }
1282
1283 static int macro_is_equal(const int *a, const int *b)
1284 {
1285 CValue cv;
1286 int t;
1287
1288 if (!a || !b)
1289 return 1;
1290
1291 while (*a && *b) {
1292 /* first time preallocate macro_equal_buf, next time only reset position to start */
1293 cstr_reset(&macro_equal_buf);
1294 TOK_GET(&t, &a, &cv);
1295 cstr_cat(&macro_equal_buf, get_tok_str(t, &cv), 0);
1296 TOK_GET(&t, &b, &cv);
1297 if (strcmp(macro_equal_buf.data, get_tok_str(t, &cv)))
1298 return 0;
1299 }
1300 return !(*a || *b);
1301 }
1302
1303 /* defines handling */
1304 ST_INLN void define_push(int v, int macro_type, int *str, Sym *first_arg)
1305 {
1306 Sym *s, *o;
1307
1308 o = define_find(v);
1309 s = sym_push2(&define_stack, v, macro_type, 0);
1310 s->d = str;
1311 s->next = first_arg;
1312 table_ident[v - TOK_IDENT]->sym_define = s;
1313
1314 if (o && !macro_is_equal(o->d, s->d))
1315 tcc_warning("%s redefined", get_tok_str(v, NULL));
1316 }
1317
1318 /* undefined a define symbol. Its name is just set to zero */
1319 ST_FUNC void define_undef(Sym *s)
1320 {
1321 int v = s->v;
1322 if (v >= TOK_IDENT && v < tok_ident)
1323 table_ident[v - TOK_IDENT]->sym_define = NULL;
1324 }
1325
1326 ST_INLN Sym *define_find(int v)
1327 {
1328 v -= TOK_IDENT;
1329 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1330 return NULL;
1331 return table_ident[v]->sym_define;
1332 }
1333
1334 /* free define stack until top reaches 'b' */
1335 ST_FUNC void free_defines(Sym *b)
1336 {
1337 while (define_stack != b) {
1338 Sym *top = define_stack;
1339 define_stack = top->prev;
1340 tok_str_free_str(top->d);
1341 define_undef(top);
1342 sym_free(top);
1343 }
1344
1345 /* restore remaining (-D or predefined) symbols if they were
1346 #undef'd in the file */
1347 while (b) {
1348 int v = b->v;
1349 if (v >= TOK_IDENT && v < tok_ident) {
1350 Sym **d = &table_ident[v - TOK_IDENT]->sym_define;
1351 if (!*d)
1352 *d = b;
1353 }
1354 b = b->prev;
1355 }
1356 }
1357
1358 /* label lookup */
1359 ST_FUNC Sym *label_find(int v)
1360 {
1361 v -= TOK_IDENT;
1362 if ((unsigned)v >= (unsigned)(tok_ident - TOK_IDENT))
1363 return NULL;
1364 return table_ident[v]->sym_label;
1365 }
1366
1367 ST_FUNC Sym *label_push(Sym **ptop, int v, int flags)
1368 {
1369 Sym *s, **ps;
1370 s = sym_push2(ptop, v, 0, 0);
1371 s->r = flags;
1372 ps = &table_ident[v - TOK_IDENT]->sym_label;
1373 if (ptop == &global_label_stack) {
1374 /* modify the top most local identifier, so that
1375 sym_identifier will point to 's' when popped */
1376 while (*ps != NULL)
1377 ps = &(*ps)->prev_tok;
1378 }
1379 s->prev_tok = *ps;
1380 *ps = s;
1381 return s;
1382 }
1383
1384 /* pop labels until element last is reached. Look if any labels are
1385 undefined. Define symbols if '&&label' was used. */
1386 ST_FUNC void label_pop(Sym **ptop, Sym *slast, int keep)
1387 {
1388 Sym *s, *s1;
1389 for(s = *ptop; s != slast; s = s1) {
1390 s1 = s->prev;
1391 if (s->r == LABEL_DECLARED) {
1392 tcc_warning("label '%s' declared but not used", get_tok_str(s->v, NULL));
1393 } else if (s->r == LABEL_FORWARD) {
1394 tcc_error("label '%s' used but not defined",
1395 get_tok_str(s->v, NULL));
1396 } else {
1397 if (s->c) {
1398 /* define corresponding symbol. A size of
1399 1 is put. */
1400 put_extern_sym(s, cur_text_section, s->jnext, 1);
1401 }
1402 }
1403 /* remove label */
1404 table_ident[s->v - TOK_IDENT]->sym_label = s->prev_tok;
1405 if (!keep)
1406 sym_free(s);
1407 }
1408 if (!keep)
1409 *ptop = slast;
1410 }
1411
1412 /* fake the nth "#if defined test_..." for tcc -dt -run */
1413 static void maybe_run_test(TCCState *s)
1414 {
1415 const char *p;
1416 if (s->include_stack_ptr != s->include_stack)
1417 return;
1418 p = get_tok_str(tok, NULL);
1419 if (0 != memcmp(p, "test_", 5))
1420 return;
1421 if (0 != --s->run_test)
1422 return;
1423 fprintf(s->ppfp, "\n[%s]\n" + !(s->dflag & 32), p), fflush(s->ppfp);
1424 define_push(tok, MACRO_OBJ, NULL, NULL);
1425 }
1426
1427 /* eval an expression for #if/#elif */
1428 static int expr_preprocess(void)
1429 {
1430 int c, t;
1431 TokenString *str;
1432
1433 str = tok_str_alloc();
1434 pp_expr = 1;
1435 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1436 next(); /* do macro subst */
1437 if (tok == TOK_DEFINED) {
1438 next_nomacro();
1439 t = tok;
1440 if (t == '(')
1441 next_nomacro();
1442 if (tok < TOK_IDENT)
1443 expect("identifier");
1444 if (tcc_state->run_test)
1445 maybe_run_test(tcc_state);
1446 c = define_find(tok) != 0;
1447 if (t == '(') {
1448 next_nomacro();
1449 if (tok != ')')
1450 expect("')'");
1451 }
1452 tok = TOK_CINT;
1453 tokc.i = c;
1454 } else if (tok >= TOK_IDENT) {
1455 /* if undefined macro */
1456 tok = TOK_CINT;
1457 tokc.i = 0;
1458 }
1459 tok_str_add_tok(str);
1460 }
1461 pp_expr = 0;
1462 tok_str_add(str, -1); /* simulate end of file */
1463 tok_str_add(str, 0);
1464 /* now evaluate C constant expression */
1465 begin_macro(str, 1);
1466 next();
1467 c = expr_const();
1468 end_macro();
1469 return c != 0;
1470 }
1471
1472
1473 /* parse after #define */
1474 ST_FUNC void parse_define(void)
1475 {
1476 Sym *s, *first, **ps;
1477 int v, t, varg, is_vaargs, spc;
1478 int saved_parse_flags = parse_flags;
1479
1480 v = tok;
1481 if (v < TOK_IDENT || v == TOK_DEFINED)
1482 tcc_error("invalid macro name '%s'", get_tok_str(tok, &tokc));
1483 /* XXX: should check if same macro (ANSI) */
1484 first = NULL;
1485 t = MACRO_OBJ;
1486 /* We have to parse the whole define as if not in asm mode, in particular
1487 no line comment with '#' must be ignored. Also for function
1488 macros the argument list must be parsed without '.' being an ID
1489 character. */
1490 parse_flags = ((parse_flags & ~PARSE_FLAG_ASM_FILE) | PARSE_FLAG_SPACES);
1491 /* '(' must be just after macro definition for MACRO_FUNC */
1492 next_nomacro_spc();
1493 if (tok == '(') {
1494 int dotid = set_idnum('.', 0);
1495 next_nomacro();
1496 ps = &first;
1497 if (tok != ')') for (;;) {
1498 varg = tok;
1499 next_nomacro();
1500 is_vaargs = 0;
1501 if (varg == TOK_DOTS) {
1502 varg = TOK___VA_ARGS__;
1503 is_vaargs = 1;
1504 } else if (tok == TOK_DOTS && gnu_ext) {
1505 is_vaargs = 1;
1506 next_nomacro();
1507 }
1508 if (varg < TOK_IDENT)
1509 bad_list:
1510 tcc_error("bad macro parameter list");
1511 s = sym_push2(&define_stack, varg | SYM_FIELD, is_vaargs, 0);
1512 *ps = s;
1513 ps = &s->next;
1514 if (tok == ')')
1515 break;
1516 if (tok != ',' || is_vaargs)
1517 goto bad_list;
1518 next_nomacro();
1519 }
1520 next_nomacro_spc();
1521 t = MACRO_FUNC;
1522 set_idnum('.', dotid);
1523 }
1524
1525 tokstr_buf.len = 0;
1526 spc = 2;
1527 parse_flags |= PARSE_FLAG_ACCEPT_STRAYS | PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED;
1528 /* The body of a macro definition should be parsed such that identifiers
1529 are parsed like the file mode determines (i.e. with '.' being an
1530 ID character in asm mode). But '#' should be retained instead of
1531 regarded as line comment leader, so still don't set ASM_FILE
1532 in parse_flags. */
1533 while (tok != TOK_LINEFEED && tok != TOK_EOF) {
1534 /* remove spaces around ## and after '#' */
1535 if (TOK_TWOSHARPS == tok) {
1536 if (2 == spc)
1537 goto bad_twosharp;
1538 if (1 == spc)
1539 --tokstr_buf.len;
1540 spc = 3;
1541 tok = TOK_PPJOIN;
1542 } else if ('#' == tok) {
1543 spc = 4;
1544 } else if (check_space(tok, &spc)) {
1545 goto skip;
1546 }
1547 tok_str_add2(&tokstr_buf, tok, &tokc);
1548 skip:
1549 next_nomacro_spc();
1550 }
1551
1552 parse_flags = saved_parse_flags;
1553 if (spc == 1)
1554 --tokstr_buf.len; /* remove trailing space */
1555 tok_str_add(&tokstr_buf, 0);
1556 if (3 == spc)
1557 bad_twosharp:
1558 tcc_error("'##' cannot appear at either end of macro");
1559 define_push(v, t, tok_str_dup(&tokstr_buf), first);
1560 }
1561
1562 static CachedInclude *search_cached_include(TCCState *s1, const char *filename, int add)
1563 {
1564 const unsigned char *s;
1565 unsigned int h;
1566 CachedInclude *e;
1567 int i;
1568
1569 h = TOK_HASH_INIT;
1570 s = (unsigned char *) filename;
1571 while (*s) {
1572 #ifdef _WIN32
1573 h = TOK_HASH_FUNC(h, toup(*s));
1574 #else
1575 h = TOK_HASH_FUNC(h, *s);
1576 #endif
1577 s++;
1578 }
1579 h &= (CACHED_INCLUDES_HASH_SIZE - 1);
1580
1581 i = s1->cached_includes_hash[h];
1582 for(;;) {
1583 if (i == 0)
1584 break;
1585 e = s1->cached_includes[i - 1];
1586 if (0 == PATHCMP(e->filename, filename))
1587 return e;
1588 i = e->hash_next;
1589 }
1590 if (!add)
1591 return NULL;
1592
1593 e = tcc_malloc(sizeof(CachedInclude) + strlen(filename));
1594 strcpy(e->filename, filename);
1595 e->ifndef_macro = e->once = 0;
1596 dynarray_add(&s1->cached_includes, &s1->nb_cached_includes, e);
1597 /* add in hash table */
1598 e->hash_next = s1->cached_includes_hash[h];
1599 s1->cached_includes_hash[h] = s1->nb_cached_includes;
1600 #ifdef INC_DEBUG
1601 printf("adding cached '%s'\n", filename);
1602 #endif
1603 return e;
1604 }
1605
1606 static void pragma_parse(TCCState *s1)
1607 {
1608 next_nomacro();
1609 if (tok == TOK_push_macro || tok == TOK_pop_macro) {
1610 int t = tok, v;
1611 Sym *s;
1612
1613 if (next(), tok != '(')
1614 goto pragma_err;
1615 if (next(), tok != TOK_STR)
1616 goto pragma_err;
1617 v = tok_alloc(tokc.str.data, tokc.str.size - 1)->tok;
1618 if (next(), tok != ')')
1619 goto pragma_err;
1620 if (t == TOK_push_macro) {
1621 while (NULL == (s = define_find(v)))
1622 define_push(v, 0, NULL, NULL);
1623 s->type.ref = s; /* set push boundary */
1624 } else {
1625 for (s = define_stack; s; s = s->prev)
1626 if (s->v == v && s->type.ref == s) {
1627 s->type.ref = NULL;
1628 break;
1629 }
1630 }
1631 if (s)
1632 table_ident[v - TOK_IDENT]->sym_define = s->d ? s : NULL;
1633 else
1634 tcc_warning("unbalanced #pragma pop_macro");
1635 pp_debug_tok = t, pp_debug_symv = v;
1636
1637 } else if (tok == TOK_once) {
1638 search_cached_include(s1, file->filename, 1)->once = pp_once;
1639
1640 } else if (s1->output_type == TCC_OUTPUT_PREPROCESS) {
1641 /* tcc -E: keep pragmas below unchanged */
1642 unget_tok(' ');
1643 unget_tok(TOK_PRAGMA);
1644 unget_tok('#');
1645 unget_tok(TOK_LINEFEED);
1646
1647 } else if (tok == TOK_pack) {
1648 /* This may be:
1649 #pragma pack(1) // set
1650 #pragma pack() // reset to default
1651 #pragma pack(push,1) // push & set
1652 #pragma pack(pop) // restore previous */
1653 next();
1654 skip('(');
1655 if (tok == TOK_ASM_pop) {
1656 next();
1657 if (s1->pack_stack_ptr <= s1->pack_stack) {
1658 stk_error:
1659 tcc_error("out of pack stack");
1660 }
1661 s1->pack_stack_ptr--;
1662 } else {
1663 int val = 0;
1664 if (tok != ')') {
1665 if (tok == TOK_ASM_push) {
1666 next();
1667 if (s1->pack_stack_ptr >= s1->pack_stack + PACK_STACK_SIZE - 1)
1668 goto stk_error;
1669 s1->pack_stack_ptr++;
1670 skip(',');
1671 }
1672 if (tok != TOK_CINT)
1673 goto pragma_err;
1674 val = tokc.i;
1675 if (val < 1 || val > 16 || (val & (val - 1)) != 0)
1676 goto pragma_err;
1677 next();
1678 }
1679 *s1->pack_stack_ptr = val;
1680 }
1681 if (tok != ')')
1682 goto pragma_err;
1683
1684 } else if (tok == TOK_comment) {
1685 char *p; int t;
1686 next();
1687 skip('(');
1688 t = tok;
1689 next();
1690 skip(',');
1691 if (tok != TOK_STR)
1692 goto pragma_err;
1693 p = tcc_strdup((char *)tokc.str.data);
1694 next();
1695 if (tok != ')')
1696 goto pragma_err;
1697 if (t == TOK_lib) {
1698 dynarray_add(&s1->pragma_libs, &s1->nb_pragma_libs, p);
1699 } else {
1700 if (t == TOK_option)
1701 tcc_set_options(s1, p);
1702 tcc_free(p);
1703 }
1704
1705 } else if (s1->warn_unsupported) {
1706 tcc_warning("#pragma %s is ignored", get_tok_str(tok, &tokc));
1707 }
1708 return;
1709
1710 pragma_err:
1711 tcc_error("malformed #pragma directive");
1712 return;
1713 }
1714
1715 /* is_bof is true if first non space token at beginning of file */
1716 ST_FUNC void preprocess(int is_bof)
1717 {
1718 TCCState *s1 = tcc_state;
1719 int i, c, n, saved_parse_flags;
1720 char buf[1024], *q;
1721 Sym *s;
1722
1723 saved_parse_flags = parse_flags;
1724 parse_flags = PARSE_FLAG_PREPROCESS
1725 | PARSE_FLAG_TOK_NUM
1726 | PARSE_FLAG_TOK_STR
1727 | PARSE_FLAG_LINEFEED
1728 | (parse_flags & PARSE_FLAG_ASM_FILE)
1729 ;
1730
1731 next_nomacro();
1732 redo:
1733 switch(tok) {
1734 case TOK_DEFINE:
1735 pp_debug_tok = tok;
1736 next_nomacro();
1737 pp_debug_symv = tok;
1738 parse_define();
1739 break;
1740 case TOK_UNDEF:
1741 pp_debug_tok = tok;
1742 next_nomacro();
1743 pp_debug_symv = tok;
1744 s = define_find(tok);
1745 /* undefine symbol by putting an invalid name */
1746 if (s)
1747 define_undef(s);
1748 break;
1749 case TOK_INCLUDE:
1750 case TOK_INCLUDE_NEXT:
1751 ch = file->buf_ptr[0];
1752 /* XXX: incorrect if comments : use next_nomacro with a special mode */
1753 skip_spaces();
1754 if (ch == '<') {
1755 c = '>';
1756 goto read_name;
1757 } else if (ch == '\"') {
1758 c = ch;
1759 read_name:
1760 inp();
1761 q = buf;
1762 while (ch != c && ch != '\n' && ch != CH_EOF) {
1763 if ((q - buf) < sizeof(buf) - 1)
1764 *q++ = ch;
1765 if (ch == '\\') {
1766 if (handle_stray_noerror() == 0)
1767 --q;
1768 } else
1769 inp();
1770 }
1771 *q = '\0';
1772 minp();
1773 #if 0
1774 /* eat all spaces and comments after include */
1775 /* XXX: slightly incorrect */
1776 while (ch1 != '\n' && ch1 != CH_EOF)
1777 inp();
1778 #endif
1779 } else {
1780 int len;
1781 /* computed #include : concatenate everything up to linefeed,
1782 the result must be one of the two accepted forms.
1783 Don't convert pp-tokens to tokens here. */
1784 parse_flags = (PARSE_FLAG_PREPROCESS
1785 | PARSE_FLAG_LINEFEED
1786 | (parse_flags & PARSE_FLAG_ASM_FILE));
1787 next();
1788 buf[0] = '\0';
1789 while (tok != TOK_LINEFEED) {
1790 pstrcat(buf, sizeof(buf), get_tok_str(tok, &tokc));
1791 next();
1792 }
1793 len = strlen(buf);
1794 /* check syntax and remove '<>|""' */
1795 if ((len < 2 || ((buf[0] != '"' || buf[len-1] != '"') &&
1796 (buf[0] != '<' || buf[len-1] != '>'))))
1797 tcc_error("'#include' expects \"FILENAME\" or <FILENAME>");
1798 c = buf[len-1];
1799 memmove(buf, buf + 1, len - 2);
1800 buf[len - 2] = '\0';
1801 }
1802
1803 if (s1->include_stack_ptr >= s1->include_stack + INCLUDE_STACK_SIZE)
1804 tcc_error("#include recursion too deep");
1805 /* store current file in stack, but increment stack later below */
1806 *s1->include_stack_ptr = file;
1807 i = tok == TOK_INCLUDE_NEXT ? file->include_next_index : 0;
1808 n = 2 + s1->nb_include_paths + s1->nb_sysinclude_paths;
1809 for (; i < n; ++i) {
1810 char buf1[sizeof file->filename];
1811 CachedInclude *e;
1812 const char *path;
1813
1814 if (i == 0) {
1815 /* check absolute include path */
1816 if (!IS_ABSPATH(buf))
1817 continue;
1818 buf1[0] = 0;
1819
1820 } else if (i == 1) {
1821 /* search in file's dir if "header.h" */
1822 if (c != '\"')
1823 continue;
1824 /* https://savannah.nongnu.org/bugs/index.php?50847 */
1825 path = file->true_filename;
1826 pstrncpy(buf1, path, tcc_basename(path) - path);
1827
1828 } else {
1829 /* search in all the include paths */
1830 int j = i - 2, k = j - s1->nb_include_paths;
1831 path = k < 0 ? s1->include_paths[j] : s1->sysinclude_paths[k];
1832 pstrcpy(buf1, sizeof(buf1), path);
1833 pstrcat(buf1, sizeof(buf1), "/");
1834 }
1835
1836 pstrcat(buf1, sizeof(buf1), buf);
1837 e = search_cached_include(s1, buf1, 0);
1838 if (e && (define_find(e->ifndef_macro) || e->once == pp_once)) {
1839 /* no need to parse the include because the 'ifndef macro'
1840 is defined (or had #pragma once) */
1841 #ifdef INC_DEBUG
1842 printf("%s: skipping cached %s\n", file->filename, buf1);
1843 #endif
1844 goto include_done;
1845 }
1846
1847 if (tcc_open(s1, buf1) < 0)
1848 continue;
1849
1850 file->include_next_index = i + 1;
1851 #ifdef INC_DEBUG
1852 printf("%s: including %s\n", file->prev->filename, file->filename);
1853 #endif
1854 /* update target deps */
1855 dynarray_add(&s1->target_deps, &s1->nb_target_deps,
1856 tcc_strdup(buf1));
1857 /* push current file in stack */
1858 ++s1->include_stack_ptr;
1859 /* add include file debug info */
1860 if (s1->do_debug)
1861 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1862 tok_flags |= TOK_FLAG_BOF | TOK_FLAG_BOL;
1863 ch = file->buf_ptr[0];
1864 goto the_end;
1865 }
1866 tcc_error("include file '%s' not found", buf);
1867 include_done:
1868 break;
1869 case TOK_IFNDEF:
1870 c = 1;
1871 goto do_ifdef;
1872 case TOK_IF:
1873 c = expr_preprocess();
1874 goto do_if;
1875 case TOK_IFDEF:
1876 c = 0;
1877 do_ifdef:
1878 next_nomacro();
1879 if (tok < TOK_IDENT)
1880 tcc_error("invalid argument for '#if%sdef'", c ? "n" : "");
1881 if (is_bof) {
1882 if (c) {
1883 #ifdef INC_DEBUG
1884 printf("#ifndef %s\n", get_tok_str(tok, NULL));
1885 #endif
1886 file->ifndef_macro = tok;
1887 }
1888 }
1889 c = (define_find(tok) != 0) ^ c;
1890 do_if:
1891 if (s1->ifdef_stack_ptr >= s1->ifdef_stack + IFDEF_STACK_SIZE)
1892 tcc_error("memory full (ifdef)");
1893 *s1->ifdef_stack_ptr++ = c;
1894 goto test_skip;
1895 case TOK_ELSE:
1896 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1897 tcc_error("#else without matching #if");
1898 if (s1->ifdef_stack_ptr[-1] & 2)
1899 tcc_error("#else after #else");
1900 c = (s1->ifdef_stack_ptr[-1] ^= 3);
1901 goto test_else;
1902 case TOK_ELIF:
1903 if (s1->ifdef_stack_ptr == s1->ifdef_stack)
1904 tcc_error("#elif without matching #if");
1905 c = s1->ifdef_stack_ptr[-1];
1906 if (c > 1)
1907 tcc_error("#elif after #else");
1908 /* last #if/#elif expression was true: we skip */
1909 if (c == 1) {
1910 c = 0;
1911 } else {
1912 c = expr_preprocess();
1913 s1->ifdef_stack_ptr[-1] = c;
1914 }
1915 test_else:
1916 if (s1->ifdef_stack_ptr == file->ifdef_stack_ptr + 1)
1917 file->ifndef_macro = 0;
1918 test_skip:
1919 if (!(c & 1)) {
1920 preprocess_skip();
1921 is_bof = 0;
1922 goto redo;
1923 }
1924 break;
1925 case TOK_ENDIF:
1926 if (s1->ifdef_stack_ptr <= file->ifdef_stack_ptr)
1927 tcc_error("#endif without matching #if");
1928 s1->ifdef_stack_ptr--;
1929 /* '#ifndef macro' was at the start of file. Now we check if
1930 an '#endif' is exactly at the end of file */
1931 if (file->ifndef_macro &&
1932 s1->ifdef_stack_ptr == file->ifdef_stack_ptr) {
1933 file->ifndef_macro_saved = file->ifndef_macro;
1934 /* need to set to zero to avoid false matches if another
1935 #ifndef at middle of file */
1936 file->ifndef_macro = 0;
1937 while (tok != TOK_LINEFEED)
1938 next_nomacro();
1939 tok_flags |= TOK_FLAG_ENDIF;
1940 goto the_end;
1941 }
1942 break;
1943 case TOK_PPNUM:
1944 n = strtoul((char*)tokc.str.data, &q, 10);
1945 goto _line_num;
1946 case TOK_LINE:
1947 next();
1948 if (tok != TOK_CINT)
1949 _line_err:
1950 tcc_error("wrong #line format");
1951 n = tokc.i;
1952 _line_num:
1953 next();
1954 if (tok != TOK_LINEFEED) {
1955 if (tok == TOK_STR) {
1956 if (file->true_filename == file->filename)
1957 file->true_filename = tcc_strdup(file->filename);
1958 pstrcpy(file->filename, sizeof(file->filename), (char *)tokc.str.data);
1959 } else if (parse_flags & PARSE_FLAG_ASM_FILE)
1960 break;
1961 else
1962 goto _line_err;
1963 --n;
1964 }
1965 if (file->fd > 0)
1966 total_lines += file->line_num - n;
1967 file->line_num = n;
1968 if (s1->do_debug)
1969 put_stabs(file->filename, N_BINCL, 0, 0, 0);
1970 break;
1971 case TOK_ERROR:
1972 case TOK_WARNING:
1973 c = tok;
1974 ch = file->buf_ptr[0];
1975 skip_spaces();
1976 q = buf;
1977 while (ch != '\n' && ch != CH_EOF) {
1978 if ((q - buf) < sizeof(buf) - 1)
1979 *q++ = ch;
1980 if (ch == '\\') {
1981 if (handle_stray_noerror() == 0)
1982 --q;
1983 } else
1984 inp();
1985 }
1986 *q = '\0';
1987 if (c == TOK_ERROR)
1988 tcc_error("#error %s", buf);
1989 else
1990 tcc_warning("#warning %s", buf);
1991 break;
1992 case TOK_PRAGMA:
1993 pragma_parse(s1);
1994 break;
1995 case TOK_LINEFEED:
1996 goto the_end;
1997 default:
1998 /* ignore gas line comment in an 'S' file. */
1999 if (saved_parse_flags & PARSE_FLAG_ASM_FILE)
2000 goto ignore;
2001 if (tok == '!' && is_bof)
2002 /* '!' is ignored at beginning to allow C scripts. */
2003 goto ignore;
2004 tcc_warning("Ignoring unknown preprocessing directive #%s", get_tok_str(tok, &tokc));
2005 ignore:
2006 file->buf_ptr = parse_line_comment(file->buf_ptr - 1);
2007 goto the_end;
2008 }
2009 /* ignore other preprocess commands or #! for C scripts */
2010 while (tok != TOK_LINEFEED)
2011 next_nomacro();
2012 the_end:
2013 parse_flags = saved_parse_flags;
2014 }
2015
2016 /* evaluate escape codes in a string. */
2017 static void parse_escape_string(CString *outstr, const uint8_t *buf, int is_long)
2018 {
2019 int c, n;
2020 const uint8_t *p;
2021
2022 p = buf;
2023 for(;;) {
2024 c = *p;
2025 if (c == '\0')
2026 break;
2027 if (c == '\\') {
2028 p++;
2029 /* escape */
2030 c = *p;
2031 switch(c) {
2032 case '0': case '1': case '2': case '3':
2033 case '4': case '5': case '6': case '7':
2034 /* at most three octal digits */
2035 n = c - '0';
2036 p++;
2037 c = *p;
2038 if (isoct(c)) {
2039 n = n * 8 + c - '0';
2040 p++;
2041 c = *p;
2042 if (isoct(c)) {
2043 n = n * 8 + c - '0';
2044 p++;
2045 }
2046 }
2047 c = n;
2048 goto add_char_nonext;
2049 case 'x':
2050 case 'u':
2051 case 'U':
2052 p++;
2053 n = 0;
2054 for(;;) {
2055 c = *p;
2056 if (c >= 'a' && c <= 'f')
2057 c = c - 'a' + 10;
2058 else if (c >= 'A' && c <= 'F')
2059 c = c - 'A' + 10;
2060 else if (isnum(c))
2061 c = c - '0';
2062 else
2063 break;
2064 n = n * 16 + c;
2065 p++;
2066 }
2067 c = n;
2068 goto add_char_nonext;
2069 case 'a':
2070 c = '\a';
2071 break;
2072 case 'b':
2073 c = '\b';
2074 break;
2075 case 'f':
2076 c = '\f';
2077 break;
2078 case 'n':
2079 c = '\n';
2080 break;
2081 case 'r':
2082 c = '\r';
2083 break;
2084 case 't':
2085 c = '\t';
2086 break;
2087 case 'v':
2088 c = '\v';
2089 break;
2090 case 'e':
2091 if (!gnu_ext)
2092 goto invalid_escape;
2093 c = 27;
2094 break;
2095 case '\'':
2096 case '\"':
2097 case '\\':
2098 case '?':
2099 break;
2100 default:
2101 invalid_escape:
2102 if (c >= '!' && c <= '~')
2103 tcc_warning("unknown escape sequence: \'\\%c\'", c);
2104 else
2105 tcc_warning("unknown escape sequence: \'\\x%x\'", c);
2106 break;
2107 }
2108 }
2109 p++;
2110 add_char_nonext:
2111 if (!is_long)
2112 cstr_ccat(outstr, c);
2113 else
2114 cstr_wccat(outstr, c);
2115 }
2116 /* add a trailing '\0' */
2117 if (!is_long)
2118 cstr_ccat(outstr, '\0');
2119 else
2120 cstr_wccat(outstr, '\0');
2121 }
2122
2123 static void parse_string(const char *s, int len)
2124 {
2125 uint8_t buf[1000], *p = buf;
2126 int is_long, sep;
2127
2128 if ((is_long = *s == 'L'))
2129 ++s, --len;
2130 sep = *s++;
2131 len -= 2;
2132 if (len >= sizeof buf)
2133 p = tcc_malloc(len + 1);
2134 memcpy(p, s, len);
2135 p[len] = 0;
2136
2137 cstr_reset(&tokcstr);
2138 parse_escape_string(&tokcstr, p, is_long);
2139 if (p != buf)
2140 tcc_free(p);
2141
2142 if (sep == '\'') {
2143 int char_size;
2144 /* XXX: make it portable */
2145 if (!is_long)
2146 char_size = 1;
2147 else
2148 char_size = sizeof(nwchar_t);
2149 if (tokcstr.size <= char_size)
2150 tcc_error("empty character constant");
2151 if (tokcstr.size > 2 * char_size)
2152 tcc_warning("multi-character character constant");
2153 if (!is_long) {
2154 tokc.i = *(int8_t *)tokcstr.data;
2155 tok = TOK_CCHAR;
2156 } else {
2157 tokc.i = *(nwchar_t *)tokcstr.data;
2158 tok = TOK_LCHAR;
2159 }
2160 } else {
2161 tokc.str.size = tokcstr.size;
2162 tokc.str.data = tokcstr.data;
2163 if (!is_long)
2164 tok = TOK_STR;
2165 else
2166 tok = TOK_LSTR;
2167 }
2168 }
2169
2170 /* we use 64 bit numbers */
2171 #define BN_SIZE 2
2172
2173 /* bn = (bn << shift) | or_val */
2174 static void bn_lshift(unsigned int *bn, int shift, int or_val)
2175 {
2176 int i;
2177 unsigned int v;
2178 for(i=0;i<BN_SIZE;i++) {
2179 v = bn[i];
2180 bn[i] = (v << shift) | or_val;
2181 or_val = v >> (32 - shift);
2182 }
2183 }
2184
2185 static void bn_zero(unsigned int *bn)
2186 {
2187 int i;
2188 for(i=0;i<BN_SIZE;i++) {
2189 bn[i] = 0;
2190 }
2191 }
2192
2193 /* parse number in null terminated string 'p' and return it in the
2194 current token */
2195 static void parse_number(const char *p)
2196 {
2197 int b, t, shift, frac_bits, s, exp_val, ch;
2198 char *q;
2199 unsigned int bn[BN_SIZE];
2200 double d;
2201
2202 /* number */
2203 q = token_buf;
2204 ch = *p++;
2205 t = ch;
2206 ch = *p++;
2207 *q++ = t;
2208 b = 10;
2209 if (t == '.') {
2210 goto float_frac_parse;
2211 } else if (t == '0') {
2212 if (ch == 'x' || ch == 'X') {
2213 q--;
2214 ch = *p++;
2215 b = 16;
2216 } else if (tcc_ext && (ch == 'b' || ch == 'B')) {
2217 q--;
2218 ch = *p++;
2219 b = 2;
2220 }
2221 }
2222 /* parse all digits. cannot check octal numbers at this stage
2223 because of floating point constants */
2224 while (1) {
2225 if (ch >= 'a' && ch <= 'f')
2226 t = ch - 'a' + 10;
2227 else if (ch >= 'A' && ch <= 'F')
2228 t = ch - 'A' + 10;
2229 else if (isnum(ch))
2230 t = ch - '0';
2231 else
2232 break;
2233 if (t >= b)
2234 break;
2235 if (q >= token_buf + STRING_MAX_SIZE) {
2236 num_too_long:
2237 tcc_error("number too long");
2238 }
2239 *q++ = ch;
2240 ch = *p++;
2241 }
2242 if (ch == '.' ||
2243 ((ch == 'e' || ch == 'E') && b == 10) ||
2244 ((ch == 'p' || ch == 'P') && (b == 16 || b == 2))) {
2245 if (b != 10) {
2246 /* NOTE: strtox should support that for hexa numbers, but
2247 non ISOC99 libcs do not support it, so we prefer to do
2248 it by hand */
2249 /* hexadecimal or binary floats */
2250 /* XXX: handle overflows */
2251 *q = '\0';
2252 if (b == 16)
2253 shift = 4;
2254 else
2255 shift = 1;
2256 bn_zero(bn);
2257 q = token_buf;
2258 while (1) {
2259 t = *q++;
2260 if (t == '\0') {
2261 break;
2262 } else if (t >= 'a') {
2263 t = t - 'a' + 10;
2264 } else if (t >= 'A') {
2265 t = t - 'A' + 10;
2266 } else {
2267 t = t - '0';
2268 }
2269 bn_lshift(bn, shift, t);
2270 }
2271 frac_bits = 0;
2272 if (ch == '.') {
2273 ch = *p++;
2274 while (1) {
2275 t = ch;
2276 if (t >= 'a' && t <= 'f') {
2277 t = t - 'a' + 10;
2278 } else if (t >= 'A' && t <= 'F') {
2279 t = t - 'A' + 10;
2280 } else if (t >= '0' && t <= '9') {
2281 t = t - '0';
2282 } else {
2283 break;
2284 }
2285 if (t >= b)
2286 tcc_error("invalid digit");
2287 bn_lshift(bn, shift, t);
2288 frac_bits += shift;
2289 ch = *p++;
2290 }
2291 }
2292 if (ch != 'p' && ch != 'P')
2293 expect("exponent");
2294 ch = *p++;
2295 s = 1;
2296 exp_val = 0;
2297 if (ch == '+') {
2298 ch = *p++;
2299 } else if (ch == '-') {
2300 s = -1;
2301 ch = *p++;
2302 }
2303 if (ch < '0' || ch > '9')
2304 expect("exponent digits");
2305 while (ch >= '0' && ch <= '9') {
2306 exp_val = exp_val * 10 + ch - '0';
2307 ch = *p++;
2308 }
2309 exp_val = exp_val * s;
2310
2311 /* now we can generate the number */
2312 /* XXX: should patch directly float number */
2313 d = (double)bn[1] * 4294967296.0 + (double)bn[0];
2314 d = ldexp(d, exp_val - frac_bits);
2315 t = toup(ch);
2316 if (t == 'F') {
2317 ch = *p++;
2318 tok = TOK_CFLOAT;
2319 /* float : should handle overflow */
2320 tokc.f = (float)d;
2321 } else if (t == 'L') {
2322 ch = *p++;
2323 #ifdef TCC_TARGET_PE
2324 tok = TOK_CDOUBLE;
2325 tokc.d = d;
2326 #else
2327 tok = TOK_CLDOUBLE;
2328 /* XXX: not large enough */
2329 tokc.ld = (long double)d;
2330 #endif
2331 } else {
2332 tok = TOK_CDOUBLE;
2333 tokc.d = d;
2334 }
2335 } else {
2336 /* decimal floats */
2337 if (ch == '.') {
2338 if (q >= token_buf + STRING_MAX_SIZE)
2339 goto num_too_long;
2340 *q++ = ch;
2341 ch = *p++;
2342 float_frac_parse:
2343 while (ch >= '0' && ch <= '9') {
2344 if (q >= token_buf + STRING_MAX_SIZE)
2345 goto num_too_long;
2346 *q++ = ch;
2347 ch = *p++;
2348 }
2349 }
2350 if (ch == 'e' || ch == 'E') {
2351 if (q >= token_buf + STRING_MAX_SIZE)
2352 goto num_too_long;
2353 *q++ = ch;
2354 ch = *p++;
2355 if (ch == '-' || ch == '+') {
2356 if (q >= token_buf + STRING_MAX_SIZE)
2357 goto num_too_long;
2358 *q++ = ch;
2359 ch = *p++;
2360 }
2361 if (ch < '0' || ch > '9')
2362 expect("exponent digits");
2363 while (ch >= '0' && ch <= '9') {
2364 if (q >= token_buf + STRING_MAX_SIZE)
2365 goto num_too_long;
2366 *q++ = ch;
2367 ch = *p++;
2368 }
2369 }
2370 *q = '\0';
2371 t = toup(ch);
2372 errno = 0;
2373 if (t == 'F') {
2374 ch = *p++;
2375 tok = TOK_CFLOAT;
2376 tokc.f = strtof(token_buf, NULL);
2377 } else if (t == 'L') {
2378 ch = *p++;
2379 #ifdef TCC_TARGET_PE
2380 tok = TOK_CDOUBLE;
2381 tokc.d = strtod(token_buf, NULL);
2382 #else
2383 tok = TOK_CLDOUBLE;
2384 tokc.ld = strtold(token_buf, NULL);
2385 #endif
2386 } else {
2387 tok = TOK_CDOUBLE;
2388 tokc.d = strtod(token_buf, NULL);
2389 }
2390 }
2391 } else {
2392 unsigned long long n, n1;
2393 int lcount, ucount, must_64bit;
2394 const char *p1;
2395
2396 /* integer number */
2397 *q = '\0';
2398 q = token_buf;
2399 if (b == 10 && *q == '0') {
2400 b = 8;
2401 q++;
2402 }
2403 n = 0;
2404 while(1) {
2405 t = *q++;
2406 /* no need for checks except for base 10 / 8 errors */
2407 if (t == '\0')
2408 break;
2409 else if (t >= 'a')
2410 t = t - 'a' + 10;
2411 else if (t >= 'A')
2412 t = t - 'A' + 10;
2413 else
2414 t = t - '0';
2415 if (t >= b)
2416 tcc_error("invalid digit");
2417 n1 = n;
2418 n = n * b + t;
2419 /* detect overflow */
2420 /* XXX: this test is not reliable */
2421 if (n < n1)
2422 tcc_error("integer constant overflow");
2423 }
2424
2425 /* Determine the characteristics (unsigned and/or 64bit) the type of
2426 the constant must have according to the constant suffix(es) */
2427 lcount = ucount = must_64bit = 0;
2428 p1 = p;
2429 for(;;) {
2430 t = toup(ch);
2431 if (t == 'L') {
2432 if (lcount >= 2)
2433 tcc_error("three 'l's in integer constant");
2434 if (lcount && *(p - 1) != ch)
2435 tcc_error("incorrect integer suffix: %s", p1);
2436 lcount++;
2437 if (lcount == 2)
2438 must_64bit = 1;
2439 ch = *p++;
2440 } else if (t == 'U') {
2441 if (ucount >= 1)
2442 tcc_error("two 'u's in integer constant");
2443 ucount++;
2444 ch = *p++;
2445 } else {
2446 break;
2447 }
2448 }
2449
2450 /* Whether 64 bits are needed to hold the constant's value */
2451 if (n & 0xffffffff00000000LL || must_64bit) {
2452 tok = TOK_CLLONG;
2453 n1 = n >> 32;
2454 } else if (lcount) {
2455 #ifdef TCC_LONG_ARE_64_BIT
2456 n1 = n >> 32;
2457 #else
2458 n1 = n;
2459 #endif
2460 tok = TOK_CLONG;
2461 } else {
2462 tok = TOK_CINT;
2463 n1 = n;
2464 }
2465
2466 /* Whether type must be unsigned to hold the constant's value */
2467 if (ucount || ((n1 >> 31) && (b != 10))) {
2468 if (tok == TOK_CLLONG)
2469 tok = TOK_CULLONG;
2470 else if (tok == TOK_CLONG)
2471 tok = TOK_CULONG;
2472 else
2473 tok = TOK_CUINT;
2474 /* If decimal and no unsigned suffix, bump to 64 bits or throw error */
2475 } else if (n1 >> 31) {
2476 if (tok == TOK_CINT)
2477 tok = TOK_CLLONG;
2478 else
2479 tcc_error("integer constant overflow");
2480 }
2481
2482 tokc.i = n;
2483 }
2484 if (ch)
2485 tcc_error("invalid number\n");
2486 }
2487
2488
2489 #define PARSE2(c1, tok1, c2, tok2) \
2490 case c1: \
2491 PEEKC(c, p); \
2492 if (c == c2) { \
2493 p++; \
2494 tok = tok2; \
2495 } else { \
2496 tok = tok1; \
2497 } \
2498 break;
2499
2500 /* return next token without macro substitution */
2501 static inline void next_nomacro1(void)
2502 {
2503 int t, c, is_long, len;
2504 TokenSym *ts;
2505 uint8_t *p, *p1;
2506 unsigned int h;
2507
2508 p = file->buf_ptr;
2509 redo_no_start:
2510 c = *p;
2511 switch(c) {
2512 case ' ':
2513 case '\t':
2514 tok = c;
2515 p++;
2516 if (parse_flags & PARSE_FLAG_SPACES)
2517 goto keep_tok_flags;
2518 while (isidnum_table[*p - CH_EOF] & IS_SPC)
2519 ++p;
2520 goto redo_no_start;
2521 case '\f':
2522 case '\v':
2523 case '\r':
2524 p++;
2525 goto redo_no_start;
2526 case '\\':
2527 /* first look if it is in fact an end of buffer */
2528 c = handle_stray1(p);
2529 p = file->buf_ptr;
2530 if (c == '\\')
2531 goto parse_simple;
2532 if (c != CH_EOF)
2533 goto redo_no_start;
2534 {
2535 TCCState *s1 = tcc_state;
2536 if ((parse_flags & PARSE_FLAG_LINEFEED)
2537 && !(tok_flags & TOK_FLAG_EOF)) {
2538 tok_flags |= TOK_FLAG_EOF;
2539 tok = TOK_LINEFEED;
2540 goto keep_tok_flags;
2541 } else if (!(parse_flags & PARSE_FLAG_PREPROCESS)) {
2542 tok = TOK_EOF;
2543 } else if (s1->ifdef_stack_ptr != file->ifdef_stack_ptr) {
2544 tcc_error("missing #endif");
2545 } else if (s1->include_stack_ptr == s1->include_stack) {
2546 /* no include left : end of file. */
2547 tok = TOK_EOF;
2548 } else {
2549 tok_flags &= ~TOK_FLAG_EOF;
2550 /* pop include file */
2551
2552 /* test if previous '#endif' was after a #ifdef at
2553 start of file */
2554 if (tok_flags & TOK_FLAG_ENDIF) {
2555 #ifdef INC_DEBUG
2556 printf("#endif %s\n", get_tok_str(file->ifndef_macro_saved, NULL));
2557 #endif
2558 search_cached_include(s1, file->filename, 1)
2559 ->ifndef_macro = file->ifndef_macro_saved;
2560 tok_flags &= ~TOK_FLAG_ENDIF;
2561 }
2562
2563 /* add end of include file debug info */
2564 if (tcc_state->do_debug) {
2565 put_stabd(N_EINCL, 0, 0);
2566 }
2567 /* pop include stack */
2568 tcc_close();
2569 s1->include_stack_ptr--;
2570 p = file->buf_ptr;
2571 goto redo_no_start;
2572 }
2573 }
2574 break;
2575
2576 case '\n':
2577 file->line_num++;
2578 tok_flags |= TOK_FLAG_BOL;
2579 p++;
2580 maybe_newline:
2581 if (0 == (parse_flags & PARSE_FLAG_LINEFEED))
2582 goto redo_no_start;
2583 tok = TOK_LINEFEED;
2584 goto keep_tok_flags;
2585
2586 case '#':
2587 /* XXX: simplify */
2588 PEEKC(c, p);
2589 if ((tok_flags & TOK_FLAG_BOL) &&
2590 (parse_flags & PARSE_FLAG_PREPROCESS)) {
2591 file->buf_ptr = p;
2592 preprocess(tok_flags & TOK_FLAG_BOF);
2593 p = file->buf_ptr;
2594 goto maybe_newline;
2595 } else {
2596 if (c == '#') {
2597 p++;
2598 tok = TOK_TWOSHARPS;
2599 } else {
2600 if (parse_flags & PARSE_FLAG_ASM_FILE) {
2601 p = parse_line_comment(p - 1);
2602 goto redo_no_start;
2603 } else {
2604 tok = '#';
2605 }
2606 }
2607 }
2608 break;
2609
2610 /* dollar is allowed to start identifiers when not parsing asm */
2611 case '$':
2612 if (!(isidnum_table[c - CH_EOF] & IS_ID)
2613 || (parse_flags & PARSE_FLAG_ASM_FILE))
2614 goto parse_simple;
2615
2616 case 'a': case 'b': case 'c': case 'd':
2617 case 'e': case 'f': case 'g': case 'h':
2618 case 'i': case 'j': case 'k': case 'l':
2619 case 'm': case 'n': case 'o': case 'p':
2620 case 'q': case 'r': case 's': case 't':
2621 case 'u': case 'v': case 'w': case 'x':
2622 case 'y': case 'z':
2623 case 'A': case 'B': case 'C': case 'D':
2624 case 'E': case 'F': case 'G': case 'H':
2625 case 'I': case 'J': case 'K':
2626 case 'M': case 'N': case 'O': case 'P':
2627 case 'Q': case 'R': case 'S': case 'T':
2628 case 'U': case 'V': case 'W': case 'X':
2629 case 'Y': case 'Z':
2630 case '_':
2631 parse_ident_fast:
2632 p1 = p;
2633 h = TOK_HASH_INIT;
2634 h = TOK_HASH_FUNC(h, c);
2635 while (c = *++p, isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2636 h = TOK_HASH_FUNC(h, c);
2637 len = p - p1;
2638 if (c != '\\') {
2639 TokenSym **pts;
2640
2641 /* fast case : no stray found, so we have the full token
2642 and we have already hashed it */
2643 h &= (TOK_HASH_SIZE - 1);
2644 pts = &hash_ident[h];
2645 for(;;) {
2646 ts = *pts;
2647 if (!ts)
2648 break;
2649 if (ts->len == len && !memcmp(ts->str, p1, len))
2650 goto token_found;
2651 pts = &(ts->hash_next);
2652 }
2653 ts = tok_alloc_new(pts, (char *) p1, len);
2654 token_found: ;
2655 } else {
2656 /* slower case */
2657 cstr_reset(&tokcstr);
2658 cstr_cat(&tokcstr, (char *) p1, len);
2659 p--;
2660 PEEKC(c, p);
2661 parse_ident_slow:
2662 while (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2663 {
2664 cstr_ccat(&tokcstr, c);
2665 PEEKC(c, p);
2666 }
2667 ts = tok_alloc(tokcstr.data, tokcstr.size);
2668 }
2669 tok = ts->tok;
2670 break;
2671 case 'L':
2672 t = p[1];
2673 if (t != '\\' && t != '\'' && t != '\"') {
2674 /* fast case */
2675 goto parse_ident_fast;
2676 } else {
2677 PEEKC(c, p);
2678 if (c == '\'' || c == '\"') {
2679 is_long = 1;
2680 goto str_const;
2681 } else {
2682 cstr_reset(&tokcstr);
2683 cstr_ccat(&tokcstr, 'L');
2684 goto parse_ident_slow;
2685 }
2686 }
2687 break;
2688
2689 case '0': case '1': case '2': case '3':
2690 case '4': case '5': case '6': case '7':
2691 case '8': case '9':
2692 t = c;
2693 PEEKC(c, p);
2694 /* after the first digit, accept digits, alpha, '.' or sign if
2695 prefixed by 'eEpP' */
2696 parse_num:
2697 cstr_reset(&tokcstr);
2698 for(;;) {
2699 cstr_ccat(&tokcstr, t);
2700 if (!((isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))
2701 || c == '.'
2702 || ((c == '+' || c == '-')
2703 && (((t == 'e' || t == 'E')
2704 && !(parse_flags & PARSE_FLAG_ASM_FILE
2705 /* 0xe+1 is 3 tokens in asm */
2706 && ((char*)tokcstr.data)[0] == '0'
2707 && toup(((char*)tokcstr.data)[1]) == 'X'))
2708 || t == 'p' || t == 'P'))))
2709 break;
2710 t = c;
2711 PEEKC(c, p);
2712 }
2713 /* We add a trailing '\0' to ease parsing */
2714 cstr_ccat(&tokcstr, '\0');
2715 tokc.str.size = tokcstr.size;
2716 tokc.str.data = tokcstr.data;
2717 tok = TOK_PPNUM;
2718 break;
2719
2720 case '.':
2721 /* special dot handling because it can also start a number */
2722 PEEKC(c, p);
2723 if (isnum(c)) {
2724 t = '.';
2725 goto parse_num;
2726 } else if ((isidnum_table['.' - CH_EOF] & IS_ID)
2727 && (isidnum_table[c - CH_EOF] & (IS_ID|IS_NUM))) {
2728 *--p = c = '.';
2729 goto parse_ident_fast;
2730 } else if (c == '.') {
2731 PEEKC(c, p);
2732 if (c == '.') {
2733 p++;
2734 tok = TOK_DOTS;
2735 } else {
2736 *--p = '.'; /* may underflow into file->unget[] */
2737 tok = '.';
2738 }
2739 } else {
2740 tok = '.';
2741 }
2742 break;
2743 case '\'':
2744 case '\"':
2745 is_long = 0;
2746 str_const:
2747 cstr_reset(&tokcstr);
2748 if (is_long)
2749 cstr_ccat(&tokcstr, 'L');
2750 cstr_ccat(&tokcstr, c);
2751 p = parse_pp_string(p, c, &tokcstr);
2752 cstr_ccat(&tokcstr, c);
2753 cstr_ccat(&tokcstr, '\0');
2754 tokc.str.size = tokcstr.size;
2755 tokc.str.data = tokcstr.data;
2756 tok = TOK_PPSTR;
2757 break;
2758
2759 case '<':
2760 PEEKC(c, p);
2761 if (c == '=') {
2762 p++;
2763 tok = TOK_LE;
2764 } else if (c == '<') {
2765 PEEKC(c, p);
2766 if (c == '=') {
2767 p++;
2768 tok = TOK_A_SHL;
2769 } else {
2770 tok = TOK_SHL;
2771 }
2772 } else {
2773 tok = TOK_LT;
2774 }
2775 break;
2776 case '>':
2777 PEEKC(c, p);
2778 if (c == '=') {
2779 p++;
2780 tok = TOK_GE;
2781 } else if (c == '>') {
2782 PEEKC(c, p);
2783 if (c == '=') {
2784 p++;
2785 tok = TOK_A_SAR;
2786 } else {
2787 tok = TOK_SAR;
2788 }
2789 } else {
2790 tok = TOK_GT;
2791 }
2792 break;
2793
2794 case '&':
2795 PEEKC(c, p);
2796 if (c == '&') {
2797 p++;
2798 tok = TOK_LAND;
2799 } else if (c == '=') {
2800 p++;
2801 tok = TOK_A_AND;
2802 } else {
2803 tok = '&';
2804 }
2805 break;
2806
2807 case '|':
2808 PEEKC(c, p);
2809 if (c == '|') {
2810 p++;
2811 tok = TOK_LOR;
2812 } else if (c == '=') {
2813 p++;
2814 tok = TOK_A_OR;
2815 } else {
2816 tok = '|';
2817 }
2818 break;
2819
2820 case '+':
2821 PEEKC(c, p);
2822 if (c == '+') {
2823 p++;
2824 tok = TOK_INC;
2825 } else if (c == '=') {
2826 p++;
2827 tok = TOK_A_ADD;
2828 } else {
2829 tok = '+';
2830 }
2831 break;
2832
2833 case '-':
2834 PEEKC(c, p);
2835 if (c == '-') {
2836 p++;
2837 tok = TOK_DEC;
2838 } else if (c == '=') {
2839 p++;
2840 tok = TOK_A_SUB;
2841 } else if (c == '>') {
2842 p++;
2843 tok = TOK_ARROW;
2844 } else {
2845 tok = '-';
2846 }
2847 break;
2848
2849 PARSE2('!', '!', '=', TOK_NE)
2850 PARSE2('=', '=', '=', TOK_EQ)
2851 PARSE2('*', '*', '=', TOK_A_MUL)
2852 PARSE2('%', '%', '=', TOK_A_MOD)
2853 PARSE2('^', '^', '=', TOK_A_XOR)
2854
2855 /* comments or operator */
2856 case '/':
2857 PEEKC(c, p);
2858 if (c == '*') {
2859 p = parse_comment(p);
2860 /* comments replaced by a blank */
2861 tok = ' ';
2862 goto keep_tok_flags;
2863 } else if (c == '/') {
2864 p = parse_line_comment(p);
2865 tok = ' ';
2866 goto keep_tok_flags;
2867 } else if (c == '=') {
2868 p++;
2869 tok = TOK_A_DIV;
2870 } else {
2871 tok = '/';
2872 }
2873 break;
2874
2875 /* simple tokens */
2876 case '(':
2877 case ')':
2878 case '[':
2879 case ']':
2880 case '{':
2881 case '}':
2882 case ',':
2883 case ';':
2884 case ':':
2885 case '?':
2886 case '~':
2887 case '@': /* only used in assembler */
2888 parse_simple:
2889 tok = c;
2890 p++;
2891 break;
2892 default:
2893 if (c >= 0x80 && c <= 0xFF) /* utf8 identifiers */
2894 goto parse_ident_fast;
2895 if (parse_flags & PARSE_FLAG_ASM_FILE)
2896 goto parse_simple;
2897 tcc_error("unrecognized character \\x%02x", c);
2898 break;
2899 }
2900 tok_flags = 0;
2901 keep_tok_flags:
2902 file->buf_ptr = p;
2903 #if defined(PARSE_DEBUG)
2904 printf("token = %s\n", get_tok_str(tok, &tokc));
2905 #endif
2906 }
2907
2908 /* return next token without macro substitution. Can read input from
2909 macro_ptr buffer */
2910 static void next_nomacro_spc(void)
2911 {
2912 if (macro_ptr) {
2913 redo:
2914 tok = *macro_ptr;
2915 if (tok) {
2916 TOK_GET(&tok, &macro_ptr, &tokc);
2917 if (tok == TOK_LINENUM) {
2918 file->line_num = tokc.i;
2919 goto redo;
2920 }
2921 }
2922 } else {
2923 next_nomacro1();
2924 }
2925 //printf("token = %s\n", get_tok_str(tok, &tokc));
2926 }
2927
2928 ST_FUNC void next_nomacro(void)
2929 {
2930 do {
2931 next_nomacro_spc();
2932 } while (tok < 256 && (isidnum_table[tok - CH_EOF] & IS_SPC));
2933 }
2934
2935
2936 static void macro_subst(
2937 TokenString *tok_str,
2938 Sym **nested_list,
2939 const int *macro_str
2940 );
2941
2942 /* substitute arguments in replacement lists in macro_str by the values in
2943 args (field d) and return allocated string */
2944 static int *macro_arg_subst(Sym **nested_list, const int *macro_str, Sym *args)
2945 {
2946 int t, t0, t1, spc;
2947 const int *st;
2948 Sym *s;
2949 CValue cval;
2950 TokenString str;
2951 CString cstr;
2952
2953 tok_str_new(&str);
2954 t0 = t1 = 0;
2955 while(1) {
2956 TOK_GET(&t, &macro_str, &cval);
2957 if (!t)
2958 break;
2959 if (t == '#') {
2960 /* stringize */
2961 TOK_GET(&t, &macro_str, &cval);
2962 if (!t)
2963 goto bad_stringy;
2964 s = sym_find2(args, t);
2965 if (s) {
2966 cstr_new(&cstr);
2967 cstr_ccat(&cstr, '\"');
2968 st = s->d;
2969 spc = 0;
2970 while (*st >= 0) {
2971 TOK_GET(&t, &st, &cval);
2972 if (t != TOK_PLCHLDR
2973 && t != TOK_NOSUBST
2974 && 0 == check_space(t, &spc)) {
2975 const char *s = get_tok_str(t, &cval);
2976 while (*s) {
2977 if (t == TOK_PPSTR && *s != '\'')
2978 add_char(&cstr, *s);
2979 else
2980 cstr_ccat(&cstr, *s);
2981 ++s;
2982 }
2983 }
2984 }
2985 cstr.size -= spc;
2986 cstr_ccat(&cstr, '\"');
2987 cstr_ccat(&cstr, '\0');
2988 #ifdef PP_DEBUG
2989 printf("\nstringize: <%s>\n", (char *)cstr.data);
2990 #endif
2991 /* add string */
2992 cval.str.size = cstr.size;
2993 cval.str.data = cstr.data;
2994 tok_str_add2(&str, TOK_PPSTR, &cval);
2995 cstr_free(&cstr);
2996 } else {
2997 bad_stringy:
2998 expect("macro parameter after '#'");
2999 }
3000 } else if (t >= TOK_IDENT) {
3001 s = sym_find2(args, t);
3002 if (s) {
3003 int l0 = str.len;
3004 st = s->d;
3005 /* if '##' is present before or after, no arg substitution */
3006 if (*macro_str == TOK_PPJOIN || t1 == TOK_PPJOIN) {
3007 /* special case for var arg macros : ## eats the ','
3008 if empty VA_ARGS variable. */
3009 if (t1 == TOK_PPJOIN && t0 == ',' && gnu_ext && s->type.t) {
3010 if (*st <= 0) {
3011 /* suppress ',' '##' */
3012 str.len -= 2;
3013 } else {
3014 /* suppress '##' and add variable */
3015 str.len--;
3016 goto add_var;
3017 }
3018 }
3019 } else {
3020 add_var:
3021 if (!s->next) {
3022 /* Expand arguments tokens and store them. In most
3023 cases we could also re-expand each argument if
3024 used multiple times, but not if the argument
3025 contains the __COUNTER__ macro. */
3026 TokenString str2;
3027 sym_push2(&s->next, s->v, s->type.t, 0);
3028 tok_str_new(&str2);
3029 macro_subst(&str2, nested_list, st);
3030 tok_str_add(&str2, 0);
3031 s->next->d = str2.str;
3032 }
3033 st = s->next->d;
3034 }
3035 for(;;) {
3036 int t2;
3037 TOK_GET(&t2, &st, &cval);
3038 if (t2 <= 0)
3039 break;
3040 tok_str_add2(&str, t2, &cval);
3041 }
3042 if (str.len == l0) /* expanded to empty string */
3043 tok_str_add(&str, TOK_PLCHLDR);
3044 } else {
3045 tok_str_add(&str, t);
3046 }
3047 } else {
3048 tok_str_add2(&str, t, &cval);
3049 }
3050 t0 = t1, t1 = t;
3051 }
3052 tok_str_add(&str, 0);
3053 return str.str;
3054 }
3055
3056 static char const ab_month_name[12][4] =
3057 {
3058 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3059 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3060 };
3061
3062 static int paste_tokens(int t1, CValue *v1, int t2, CValue *v2)
3063 {
3064 CString cstr;
3065 int n, ret = 1;
3066
3067 cstr_new(&cstr);
3068 if (t1 != TOK_PLCHLDR)
3069 cstr_cat(&cstr, get_tok_str(t1, v1), -1);
3070 n = cstr.size;
3071 if (t2 != TOK_PLCHLDR)
3072 cstr_cat(&cstr, get_tok_str(t2, v2), -1);
3073 cstr_ccat(&cstr, '\0');
3074
3075 tcc_open_bf(tcc_state, ":paste:", cstr.size);
3076 memcpy(file->buffer, cstr.data, cstr.size);
3077 tok_flags = 0;
3078 for (;;) {
3079 next_nomacro1();
3080 if (0 == *file->buf_ptr)
3081 break;
3082 if (is_space(tok))
3083 continue;
3084 tcc_warning("pasting \"%.*s\" and \"%s\" does not give a valid"
3085 " preprocessing token", n, cstr.data, (char*)cstr.data + n);
3086 ret = 0;
3087 break;
3088 }
3089 tcc_close();
3090 //printf("paste <%s>\n", (char*)cstr.data);
3091 cstr_free(&cstr);
3092 return ret;
3093 }
3094
3095 /* handle the '##' operator. Return NULL if no '##' seen. Otherwise
3096 return the resulting string (which must be freed). */
3097 static inline int *macro_twosharps(const int *ptr0)
3098 {
3099 int t;
3100 CValue cval;
3101 TokenString macro_str1;
3102 int start_of_nosubsts = -1;
3103 const int *ptr;
3104
3105 /* we search the first '##' */
3106 for (ptr = ptr0;;) {
3107 TOK_GET(&t, &ptr, &cval);
3108 if (t == TOK_PPJOIN)
3109 break;
3110 if (t == 0)
3111 return NULL;
3112 }
3113
3114 tok_str_new(&macro_str1);
3115
3116 //tok_print(" $$$", ptr0);
3117 for (ptr = ptr0;;) {
3118 TOK_GET(&t, &ptr, &cval);
3119 if (t == 0)
3120 break;
3121 if (t == TOK_PPJOIN)
3122 continue;
3123 while (*ptr == TOK_PPJOIN) {
3124 int t1; CValue cv1;
3125 /* given 'a##b', remove nosubsts preceding 'a' */
3126 if (start_of_nosubsts >= 0)
3127 macro_str1.len = start_of_nosubsts;
3128 /* given 'a##b', remove nosubsts preceding 'b' */
3129 while ((t1 = *++ptr) == TOK_NOSUBST)
3130 ;
3131 if (t1 && t1 != TOK_PPJOIN) {
3132 TOK_GET(&t1, &ptr, &cv1);
3133 if (t != TOK_PLCHLDR || t1 != TOK_PLCHLDR) {
3134 if (paste_tokens(t, &cval, t1, &cv1)) {
3135 t = tok, cval = tokc;
3136 } else {
3137 tok_str_add2(&macro_str1, t, &cval);
3138 t = t1, cval = cv1;
3139 }
3140 }
3141 }
3142 }
3143 if (t == TOK_NOSUBST) {
3144 if (start_of_nosubsts < 0)
3145 start_of_nosubsts = macro_str1.len;
3146 } else {
3147 start_of_nosubsts = -1;
3148 }
3149 tok_str_add2(&macro_str1, t, &cval);
3150 }
3151 tok_str_add(&macro_str1, 0);
3152 //tok_print(" ###", macro_str1.str);
3153 return macro_str1.str;
3154 }
3155
3156 /* peek or read [ws_str == NULL] next token from function macro call,
3157 walking up macro levels up to the file if necessary */
3158 static int next_argstream(Sym **nested_list, TokenString *ws_str)
3159 {
3160 int t;
3161 const int *p;
3162 Sym *sa;
3163
3164 for (;;) {
3165 if (macro_ptr) {
3166 p = macro_ptr, t = *p;
3167 if (ws_str) {
3168 while (is_space(t) || TOK_LINEFEED == t || TOK_PLCHLDR == t)
3169 tok_str_add(ws_str, t), t = *++p;
3170 }
3171 if (t == 0) {
3172 end_macro();
3173 /* also, end of scope for nested defined symbol */
3174 sa = *nested_list;
3175 while (sa && sa->v == 0)
3176 sa = sa->prev;
3177 if (sa)
3178 sa->v = 0;
3179 continue;
3180 }
3181 } else {
3182 ch = handle_eob();
3183 if (ws_str) {
3184 while (is_space(ch) || ch == '\n' || ch == '/') {
3185 if (ch == '/') {
3186 int c;
3187 uint8_t *p = file->buf_ptr;
3188 PEEKC(c, p);
3189 if (c == '*') {
3190 p = parse_comment(p);
3191 file->buf_ptr = p - 1;
3192 } else if (c == '/') {
3193 p = parse_line_comment(p);
3194 file->buf_ptr = p - 1;
3195 } else
3196 break;
3197 ch = ' ';
3198 }
3199 if (ch == '\n')
3200 file->line_num++;
3201 if (!(ch == '\f' || ch == '\v' || ch == '\r'))
3202 tok_str_add(ws_str, ch);
3203 cinp();
3204 }
3205 }
3206 t = ch;
3207 }
3208
3209 if (ws_str)
3210 return t;
3211 next_nomacro_spc();
3212 return tok;
3213 }
3214 }
3215
3216 /* do macro substitution of current token with macro 's' and add
3217 result to (tok_str,tok_len). 'nested_list' is the list of all
3218 macros we got inside to avoid recursing. Return non zero if no
3219 substitution needs to be done */
3220 static int macro_subst_tok(
3221 TokenString *tok_str,
3222 Sym **nested_list,
3223 Sym *s)
3224 {
3225 Sym *args, *sa, *sa1;
3226 int parlevel, t, t1, spc;
3227 TokenString str;
3228 char *cstrval;
3229 CValue cval;
3230 CString cstr;
3231 char buf[32];
3232
3233 /* if symbol is a macro, prepare substitution */
3234 /* special macros */
3235 if (tok == TOK___LINE__ || tok == TOK___COUNTER__) {
3236 t = tok == TOK___LINE__ ? file->line_num : pp_counter++;
3237 snprintf(buf, sizeof(buf), "%d", t);
3238 cstrval = buf;
3239 t1 = TOK_PPNUM;
3240 goto add_cstr1;
3241 } else if (tok == TOK___FILE__) {
3242 cstrval = file->filename;
3243 goto add_cstr;
3244 } else if (tok == TOK___DATE__ || tok == TOK___TIME__) {
3245 time_t ti;
3246 struct tm *tm;
3247
3248 time(&ti);
3249 tm = localtime(&ti);
3250 if (tok == TOK___DATE__) {
3251 snprintf(buf, sizeof(buf), "%s %2d %d",
3252 ab_month_name[tm->tm_mon], tm->tm_mday, tm->tm_year + 1900);
3253 } else {
3254 snprintf(buf, sizeof(buf), "%02d:%02d:%02d",
3255 tm->tm_hour, tm->tm_min, tm->tm_sec);
3256 }
3257 cstrval = buf;
3258 add_cstr:
3259 t1 = TOK_STR;
3260 add_cstr1:
3261 cstr_new(&cstr);
3262 cstr_cat(&cstr, cstrval, 0);
3263 cval.str.size = cstr.size;
3264 cval.str.data = cstr.data;
3265 tok_str_add2(tok_str, t1, &cval);
3266 cstr_free(&cstr);
3267 } else if (s->d) {
3268 int saved_parse_flags = parse_flags;
3269 int *joined_str = NULL;
3270 int *mstr = s->d;
3271
3272 if (s->type.t == MACRO_FUNC) {
3273 /* whitespace between macro name and argument list */
3274 TokenString ws_str;
3275 tok_str_new(&ws_str);
3276
3277 spc = 0;
3278 parse_flags |= PARSE_FLAG_SPACES | PARSE_FLAG_LINEFEED
3279 | PARSE_FLAG_ACCEPT_STRAYS;
3280
3281 /* get next token from argument stream */
3282 t = next_argstream(nested_list, &ws_str);
3283 if (t != '(') {
3284 /* not a macro substitution after all, restore the
3285 * macro token plus all whitespace we've read.
3286 * whitespace is intentionally not merged to preserve
3287 * newlines. */
3288 parse_flags = saved_parse_flags;
3289 tok_str_add(tok_str, tok);
3290 if (parse_flags & PARSE_FLAG_SPACES) {
3291 int i;
3292 for (i = 0; i < ws_str.len; i++)
3293 tok_str_add(tok_str, ws_str.str[i]);
3294 }
3295 tok_str_free_str(ws_str.str);
3296 return 0;
3297 } else {
3298 tok_str_free_str(ws_str.str);
3299 }
3300 do {
3301 next_nomacro(); /* eat '(' */
3302 } while (tok == TOK_PLCHLDR);
3303
3304 /* argument macro */
3305 args = NULL;
3306 sa = s->next;
3307 /* NOTE: empty args are allowed, except if no args */
3308 for(;;) {
3309 do {
3310 next_argstream(nested_list, NULL);
3311 } while (is_space(tok) || TOK_LINEFEED == tok);
3312 empty_arg:
3313 /* handle '()' case */
3314 if (!args && !sa && tok == ')')
3315 break;
3316 if (!sa)
3317 tcc_error("macro '%s' used with too many args",
3318 get_tok_str(s->v, 0));
3319 tok_str_new(&str);
3320 parlevel = spc = 0;
3321 /* NOTE: non zero sa->t indicates VA_ARGS */
3322 while ((parlevel > 0 ||
3323 (tok != ')' &&
3324 (tok != ',' || sa->type.t)))) {
3325 if (tok == TOK_EOF || tok == 0)
3326 break;
3327 if (tok == '(')
3328 parlevel++;
3329 else if (tok == ')')
3330 parlevel--;
3331 if (tok == TOK_LINEFEED)
3332 tok = ' ';
3333 if (!check_space(tok, &spc))
3334 tok_str_add2(&str, tok, &tokc);
3335 next_argstream(nested_list, NULL);
3336 }
3337 if (parlevel)
3338 expect(")");
3339 str.len -= spc;
3340 tok_str_add(&str, -1);
3341 tok_str_add(&str, 0);
3342 sa1 = sym_push2(&args, sa->v & ~SYM_FIELD, sa->type.t, 0);
3343 sa1->d = str.str;
3344 sa = sa->next;
3345 if (tok == ')') {
3346 /* special case for gcc var args: add an empty
3347 var arg argument if it is omitted */
3348 if (sa && sa->type.t && gnu_ext)
3349 goto empty_arg;
3350 break;
3351 }
3352 if (tok != ',')
3353 expect(",");
3354 }
3355 if (sa) {
3356 tcc_error("macro '%s' used with too few args",
3357 get_tok_str(s->v, 0));
3358 }
3359
3360 parse_flags = saved_parse_flags;
3361
3362 /* now subst each arg */
3363 mstr = macro_arg_subst(nested_list, mstr, args);
3364 /* free memory */
3365 sa = args;
3366 while (sa) {
3367 sa1 = sa->prev;
3368 tok_str_free_str(sa->d);
3369 if (sa->next) {
3370 tok_str_free_str(sa->next->d);
3371 sym_free(sa->next);
3372 }
3373 sym_free(sa);
3374 sa = sa1;
3375 }
3376 }
3377
3378 sym_push2(nested_list, s->v, 0, 0);
3379 parse_flags = saved_parse_flags;
3380 joined_str = macro_twosharps(mstr);
3381 macro_subst(tok_str, nested_list, joined_str ? joined_str : mstr);
3382
3383 /* pop nested defined symbol */
3384 sa1 = *nested_list;
3385 *nested_list = sa1->prev;
3386 sym_free(sa1);
3387 if (joined_str)
3388 tok_str_free_str(joined_str);
3389 if (mstr != s->d)
3390 tok_str_free_str(mstr);
3391 }
3392 return 0;
3393 }
3394
3395 /* do macro substitution of macro_str and add result to
3396 (tok_str,tok_len). 'nested_list' is the list of all macros we got
3397 inside to avoid recursing. */
3398 static void macro_subst(
3399 TokenString *tok_str,
3400 Sym **nested_list,
3401 const int *macro_str
3402 )
3403 {
3404 Sym *s;
3405 int t, spc, nosubst;
3406 CValue cval;
3407
3408 spc = nosubst = 0;
3409
3410 while (1) {
3411 TOK_GET(&t, &macro_str, &cval);
3412 if (t <= 0)
3413 break;
3414
3415 if (t >= TOK_IDENT && 0 == nosubst) {
3416 s = define_find(t);
3417 if (s == NULL)
3418 goto no_subst;
3419
3420 /* if nested substitution, do nothing */
3421 if (sym_find2(*nested_list, t)) {
3422 /* and mark it as TOK_NOSUBST, so it doesn't get subst'd again */
3423 tok_str_add2(tok_str, TOK_NOSUBST, NULL);
3424 goto no_subst;
3425 }
3426
3427 {
3428 TokenString str;
3429 str.str = (int*)macro_str;
3430 begin_macro(&str, 2);
3431
3432 tok = t;
3433 macro_subst_tok(tok_str, nested_list, s);
3434
3435 if (str.alloc == 3) {
3436 /* already finished by reading function macro arguments */
3437 break;
3438 }
3439
3440 macro_str = macro_ptr;
3441 end_macro ();
3442 }
3443 if (tok_str->len)
3444 spc = is_space(t = tok_str->str[tok_str->lastlen]);
3445 } else {
3446 if (t == '\\' && !(parse_flags & PARSE_FLAG_ACCEPT_STRAYS))
3447 tcc_error("stray '\\' in program");
3448 no_subst:
3449 if (!check_space(t, &spc))
3450 tok_str_add2(tok_str, t, &cval);
3451
3452 if (nosubst) {
3453 if (nosubst > 1 && (spc || (++nosubst == 3 && t == '(')))
3454 continue;
3455 nosubst = 0;
3456 }
3457 if (t == TOK_NOSUBST)
3458 nosubst = 1;
3459 }
3460 /* GCC supports 'defined' as result of a macto substitution */
3461 if (t == TOK_DEFINED && pp_expr)
3462 nosubst = 2;
3463 }
3464 }
3465
3466 /* return next token with macro substitution */
3467 ST_FUNC void next(void)
3468 {
3469 redo:
3470 if (parse_flags & PARSE_FLAG_SPACES)
3471 next_nomacro_spc();
3472 else
3473 next_nomacro();
3474
3475 if (macro_ptr) {
3476 if (tok == TOK_NOSUBST || tok == TOK_PLCHLDR) {
3477 /* discard preprocessor markers */
3478 goto redo;
3479 } else if (tok == 0) {
3480 /* end of macro or unget token string */
3481 end_macro();
3482 goto redo;
3483 }
3484 } else if (tok >= TOK_IDENT && (parse_flags & PARSE_FLAG_PREPROCESS)) {
3485 Sym *s;
3486 /* if reading from file, try to substitute macros */
3487 s = define_find(tok);
3488 if (s) {
3489 Sym *nested_list = NULL;
3490 tokstr_buf.len = 0;
3491 macro_subst_tok(&tokstr_buf, &nested_list, s);
3492 tok_str_add(&tokstr_buf, 0);
3493 begin_macro(&tokstr_buf, 2);
3494 goto redo;
3495 }
3496 }
3497 /* convert preprocessor tokens into C tokens */
3498 if (tok == TOK_PPNUM) {
3499 if (parse_flags & PARSE_FLAG_TOK_NUM)
3500 parse_number((char *)tokc.str.data);
3501 } else if (tok == TOK_PPSTR) {
3502 if (parse_flags & PARSE_FLAG_TOK_STR)
3503 parse_string((char *)tokc.str.data, tokc.str.size - 1);
3504 }
3505 }
3506
3507 /* push back current token and set current token to 'last_tok'. Only
3508 identifier case handled for labels. */
3509 ST_INLN void unget_tok(int last_tok)
3510 {
3511
3512 TokenString *str = tok_str_alloc();
3513 tok_str_add2(str, tok, &tokc);
3514 tok_str_add(str, 0);
3515 begin_macro(str, 1);
3516 tok = last_tok;
3517 }
3518
3519 ST_FUNC void preprocess_start(TCCState *s1, int is_asm)
3520 {
3521 char *buf;
3522
3523 s1->include_stack_ptr = s1->include_stack;
3524 s1->ifdef_stack_ptr = s1->ifdef_stack;
3525 file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
3526 pp_expr = 0;
3527 pp_counter = 0;
3528 pp_debug_tok = pp_debug_symv = 0;
3529 pp_once++;
3530 pvtop = vtop = vstack - 1;
3531 s1->pack_stack[0] = 0;
3532 s1->pack_stack_ptr = s1->pack_stack;
3533
3534 set_idnum('$', s1->dollars_in_identifiers ? IS_ID : 0);
3535 set_idnum('.', is_asm ? IS_ID : 0);
3536
3537 buf = tcc_malloc(3 + strlen(file->filename));
3538 sprintf(buf, "\"%s\"", file->filename);
3539 tcc_define_symbol(s1, "__BASE_FILE__", buf);
3540 tcc_free(buf);
3541
3542 if (s1->nb_cmd_include_files) {
3543 CString cstr;
3544 int i;
3545 cstr_new(&cstr);
3546 for (i = 0; i < s1->nb_cmd_include_files; i++) {
3547 cstr_cat(&cstr, "#include \"", -1);
3548 cstr_cat(&cstr, s1->cmd_include_files[i], -1);
3549 cstr_cat(&cstr, "\"\n", -1);
3550 }
3551 *s1->include_stack_ptr++ = file;
3552 tcc_open_bf(s1, "<command line>", cstr.size);
3553 memcpy(file->buffer, cstr.data, cstr.size);
3554 cstr_free(&cstr);
3555 }
3556
3557 if (is_asm)
3558 tcc_define_symbol(s1, "__ASSEMBLER__", NULL);
3559
3560 parse_flags = is_asm ? PARSE_FLAG_ASM_FILE : 0;
3561 tok_flags = TOK_FLAG_BOL | TOK_FLAG_BOF;
3562 }
3563
3564 /* cleanup from error/setjmp */
3565 ST_FUNC void preprocess_end(TCCState *s1)
3566 {
3567 while (macro_stack)
3568 end_macro();
3569 macro_ptr = NULL;
3570 }
3571
3572 ST_FUNC void tccpp_new(TCCState *s)
3573 {
3574 int i, c;
3575 const char *p, *r;
3576
3577 /* might be used in error() before preprocess_start() */
3578 s->include_stack_ptr = s->include_stack;
3579 s->ppfp = stdout;
3580
3581 /* init isid table */
3582 for(i = CH_EOF; i<128; i++)
3583 set_idnum(i,
3584 is_space(i) ? IS_SPC
3585 : isid(i) ? IS_ID
3586 : isnum(i) ? IS_NUM
3587 : 0);
3588
3589 for(i = 128; i<256; i++)
3590 set_idnum(i, IS_ID);
3591
3592 /* init allocators */
3593 tal_new(&toksym_alloc, TOKSYM_TAL_LIMIT, TOKSYM_TAL_SIZE);
3594 tal_new(&tokstr_alloc, TOKSTR_TAL_LIMIT, TOKSTR_TAL_SIZE);
3595 tal_new(&cstr_alloc, CSTR_TAL_LIMIT, CSTR_TAL_SIZE);
3596
3597 memset(hash_ident, 0, TOK_HASH_SIZE * sizeof(TokenSym *));
3598 cstr_new(&cstr_buf);
3599 cstr_realloc(&cstr_buf, STRING_MAX_SIZE);
3600 tok_str_new(&tokstr_buf);
3601 tok_str_realloc(&tokstr_buf, TOKSTR_MAX_SIZE);
3602
3603 tok_ident = TOK_IDENT;
3604 p = tcc_keywords;
3605 while (*p) {
3606 r = p;
3607 for(;;) {
3608 c = *r++;
3609 if (c == '\0')
3610 break;
3611 }
3612 tok_alloc(p, r - p - 1);
3613 p = r;
3614 }
3615 }
3616
3617 ST_FUNC void tccpp_delete(TCCState *s)
3618 {
3619 int i, n;
3620
3621 /* free -D and compiler defines */
3622 free_defines(NULL);
3623
3624 /* free tokens */
3625 n = tok_ident - TOK_IDENT;
3626 for(i = 0; i < n; i++)
3627 tal_free(toksym_alloc, table_ident[i]);
3628 tcc_free(table_ident);
3629 table_ident = NULL;
3630
3631 /* free static buffers */
3632 cstr_free(&tokcstr);
3633 cstr_free(&cstr_buf);
3634 cstr_free(&macro_equal_buf);
3635 tok_str_free_str(tokstr_buf.str);
3636
3637 /* free allocators */
3638 tal_delete(toksym_alloc);
3639 toksym_alloc = NULL;
3640 tal_delete(tokstr_alloc);
3641 tokstr_alloc = NULL;
3642 tal_delete(cstr_alloc);
3643 cstr_alloc = NULL;
3644 }
3645
3646 /* ------------------------------------------------------------------------- */
3647 /* tcc -E [-P[1]] [-dD} support */
3648
3649 static void tok_print(const char *msg, const int *str)
3650 {
3651 FILE *fp;
3652 int t, s = 0;
3653 CValue cval;
3654
3655 fp = tcc_state->ppfp;
3656 fprintf(fp, "%s", msg);
3657 while (str) {
3658 TOK_GET(&t, &str, &cval);
3659 if (!t)
3660 break;
3661 fprintf(fp, " %s" + s, get_tok_str(t, &cval)), s = 1;
3662 }
3663 fprintf(fp, "\n");
3664 }
3665
3666 static void pp_line(TCCState *s1, BufferedFile *f, int level)
3667 {
3668 int d = f->line_num - f->line_ref;
3669
3670 if (s1->dflag & 4)
3671 return;
3672
3673 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_NONE) {
3674 ;
3675 } else if (level == 0 && f->line_ref && d < 8) {
3676 while (d > 0)
3677 fputs("\n", s1->ppfp), --d;
3678 } else if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_STD) {
3679 fprintf(s1->ppfp, "#line %d \"%s\"\n", f->line_num, f->filename);
3680 } else {
3681 fprintf(s1->ppfp, "# %d \"%s\"%s\n", f->line_num, f->filename,
3682 level > 0 ? " 1" : level < 0 ? " 2" : "");
3683 }
3684 f->line_ref = f->line_num;
3685 }
3686
3687 static void define_print(TCCState *s1, int v)
3688 {
3689 FILE *fp;
3690 Sym *s;
3691
3692 s = define_find(v);
3693 if (NULL == s || NULL == s->d)
3694 return;
3695
3696 fp = s1->ppfp;
3697 fprintf(fp, "#define %s", get_tok_str(v, NULL));
3698 if (s->type.t == MACRO_FUNC) {
3699 Sym *a = s->next;
3700 fprintf(fp,"(");
3701 if (a)
3702 for (;;) {
3703 fprintf(fp,"%s", get_tok_str(a->v & ~SYM_FIELD, NULL));
3704 if (!(a = a->next))
3705 break;
3706 fprintf(fp,",");
3707 }
3708 fprintf(fp,")");
3709 }
3710 tok_print("", s->d);
3711 }
3712
3713 static void pp_debug_defines(TCCState *s1)
3714 {
3715 int v, t;
3716 const char *vs;
3717 FILE *fp;
3718
3719 t = pp_debug_tok;
3720 if (t == 0)
3721 return;
3722
3723 file->line_num--;
3724 pp_line(s1, file, 0);
3725 file->line_ref = ++file->line_num;
3726
3727 fp = s1->ppfp;
3728 v = pp_debug_symv;
3729 vs = get_tok_str(v, NULL);
3730 if (t == TOK_DEFINE) {
3731 define_print(s1, v);
3732 } else if (t == TOK_UNDEF) {
3733 fprintf(fp, "#undef %s\n", vs);
3734 } else if (t == TOK_push_macro) {
3735 fprintf(fp, "#pragma push_macro(\"%s\")\n", vs);
3736 } else if (t == TOK_pop_macro) {
3737 fprintf(fp, "#pragma pop_macro(\"%s\")\n", vs);
3738 }
3739 pp_debug_tok = 0;
3740 }
3741
3742 static void pp_debug_builtins(TCCState *s1)
3743 {
3744 int v;
3745 for (v = TOK_IDENT; v < tok_ident; ++v)
3746 define_print(s1, v);
3747 }
3748
3749 /* Add a space between tokens a and b to avoid unwanted textual pasting */
3750 static int pp_need_space(int a, int b)
3751 {
3752 return 'E' == a ? '+' == b || '-' == b
3753 : '+' == a ? TOK_INC == b || '+' == b
3754 : '-' == a ? TOK_DEC == b || '-' == b
3755 : a >= TOK_IDENT ? b >= TOK_IDENT
3756 : a == TOK_PPNUM ? b >= TOK_IDENT
3757 : 0;
3758 }
3759
3760 /* maybe hex like 0x1e */
3761 static int pp_check_he0xE(int t, const char *p)
3762 {
3763 if (t == TOK_PPNUM && toup(strchr(p, 0)[-1]) == 'E')
3764 return 'E';
3765 return t;
3766 }
3767
3768 /* Preprocess the current file */
3769 ST_FUNC int tcc_preprocess(TCCState *s1)
3770 {
3771 BufferedFile **iptr;
3772 int token_seen, spcs, level;
3773 const char *p;
3774 char white[400];
3775
3776 parse_flags = PARSE_FLAG_PREPROCESS
3777 | (parse_flags & PARSE_FLAG_ASM_FILE)
3778 | PARSE_FLAG_LINEFEED
3779 | PARSE_FLAG_SPACES
3780 | PARSE_FLAG_ACCEPT_STRAYS
3781 ;
3782 /* Credits to Fabrice Bellard's initial revision to demonstrate its
3783 capability to compile and run itself, provided all numbers are
3784 given as decimals. tcc -E -P10 will do. */
3785 if (s1->Pflag == LINE_MACRO_OUTPUT_FORMAT_P10)
3786 parse_flags |= PARSE_FLAG_TOK_NUM, s1->Pflag = 1;
3787
3788 #ifdef PP_BENCH
3789 /* for PP benchmarks */
3790 do next(); while (tok != TOK_EOF);
3791 return 0;
3792 #endif
3793
3794 if (s1->dflag & 1) {
3795 pp_debug_builtins(s1);
3796 s1->dflag &= ~1;
3797 }
3798
3799 token_seen = TOK_LINEFEED, spcs = 0;
3800 pp_line(s1, file, 0);
3801 for (;;) {
3802 iptr = s1->include_stack_ptr;
3803 next();
3804 if (tok == TOK_EOF)
3805 break;
3806
3807 level = s1->include_stack_ptr - iptr;
3808 if (level) {
3809 if (level > 0)
3810 pp_line(s1, *iptr, 0);
3811 pp_line(s1, file, level);
3812 }
3813 if (s1->dflag & 7) {
3814 pp_debug_defines(s1);
3815 if (s1->dflag & 4)
3816 continue;
3817 }
3818
3819 if (is_space(tok)) {
3820 if (spcs < sizeof white - 1)
3821 white[spcs++] = tok;
3822 continue;
3823 } else if (tok == TOK_LINEFEED) {
3824 spcs = 0;
3825 if (token_seen == TOK_LINEFEED)
3826 continue;
3827 ++file->line_ref;
3828 } else if (token_seen == TOK_LINEFEED) {
3829 pp_line(s1, file, 0);
3830 } else if (spcs == 0 && pp_need_space(token_seen, tok)) {
3831 white[spcs++] = ' ';
3832 }
3833
3834 white[spcs] = 0, fputs(white, s1->ppfp), spcs = 0;
3835 fputs(p = get_tok_str(tok, &tokc), s1->ppfp);
3836 token_seen = pp_check_he0xE(tok, p);
3837 }
3838 return 0;
3839 }
3840
3841 /* ------------------------------------------------------------------------- */
Tiny C Compiler main repository